Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Utilization management of acute care services : evaluation of the SWITCH index system Wiggins, Sandra 1988

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1988_A6_7 W54.pdf [ 11.81MB ]
Metadata
JSON: 831-1.0097839.json
JSON-LD: 831-1.0097839-ld.json
RDF/XML (Pretty): 831-1.0097839-rdf.xml
RDF/JSON: 831-1.0097839-rdf.json
Turtle: 831-1.0097839-turtle.txt
N-Triples: 831-1.0097839-rdf-ntriples.txt
Original Record: 831-1.0097839-source.json
Full Text
831-1.0097839-fulltext.txt
Citation
831-1.0097839.ris

Full Text

U T I L I Z A T I O N M A N A G E M E N T O F A C U T E C A R E SERVICES: E V A L U A T I O N O F T H E S W I T C H INDEX SYSTEM By Sandra Wiggins B A . , York University, 1977 M . A . , York University, 1978 A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T O F T H E REQUIREMENTS FOR T H E D E G R E E O F M A S T E R O F SCIENCE (Health Services Planning and Administration) i n T H E F A C U L T Y O F G R A D U A T E STUDIES Department of Health Care and Epidemiology We accept this thesis as conforming to the required standard T H E UNIVERSITY O F BRITISH C O L U M B I A November 1988 © Sandra Wiggins, 1988 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Health Care & Epidemiology The University of British Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 DE-6(3/81) i i A B S T R A C T In recent years, concern about the rising costs of health care has prompted the development of programs aimed at reducing utilization of hospital services and facilities while maintaining an acceptable standard of care. One of the major strategies that has emerged in the effort to accomplish these dual objectives, is utilization management. Al though there are a number of different approaches, the primary aim of all utilization management programs is to identify and eliminate unnecessary and inappropriate hospital use. To date, most of the utilization research and program development has taken place in the United States. To a great extent, this effort has focussed on the development and use of norms for utilization based on a breakdown of length of stay data by diagnostic-related groups (DRG's). Canadian interest in this type of approach is reflected in the recent development of data bases defined by case-mix groups (CMG's) . However, while continued efforts are being made to refine these schemes, they have been vulnerable to the criticism that they do not provide adequately objective criteria for establishing what constitutes appropriate patterns of hospital use. In addition, because they are based on statistically derived norms, they have been criticized as lacking sufficient clinical relevance to encourage physician support. Since hospital utilization is largely determined by the medical staff, utilization management programs that fail to obtain physician support are unlikely to succeed. A n alternative approach, which appears to be gaining in popularity, involves the formulation of criteria which can be used to determine what constitutes appropriate and necessary hospital use. Essentially, it is argued i i i that by directly identifying the source and nature of misuti l ization, it should be possible to develop more effective strategies for the resolution of identified problems. Th e A m e r i c a n Appropriateness Evaluat ion Protocol designed by Ger tman & Restuccia (1981) is one of the earliest and most highly tested examples of a criterion-based system. In C a n a d a , interest in this type of approach is more recent and, consequently, little attention has as yet been focussed on the development and use of clinical criteria in ut i l izat ion review and management. O n e exception, however, is the S W I T C H Index System. This system, which was developed and implemented in 1984 by the Peace A r c h District Hosp i ta l (White Rock, B.C.) , makes a direct attempt to identify and eliminate days of hospital stay dur ing which no appropriate acute care services are being prov ided . T h e criteria used in this system are classified under the headings Signs, W i n d , Intramuscular Therapy , Tubes , Consultant , and Hospice . Patients are considered to be appropriately placed in the hospital if, o n any given day, at least one of the specified criteria are met. Otherwise they are classified as Off-Index and action is taken to identify the source of the prob lem and to initiate corrective action. Since a major objective of the S W I T C H system is to identify and eliminate inappropriate use, an observable outcome, if the p r o g r a m is successful, should be a reduction in length of stay. The present study investigated this hypothesis by comparing pre- and post- intervention length of stay trends at the Peace A r c h District Hospita l . In addit ion, to take into account any general secular trends in length of stay over time, the Peace A r c h length of stay was compared to the length of stay observed for a control group of three peer-group member hospitals. i v A l t h o u g h data covering the four year per iod 1982 to 1985, indicated that the length of stay at the Peace A r c h District Hospi ta l h a d been decreasing over time, no component of this general decline could be attributed to the S W I T C H Index System. T i m e series regression analyses failed to detect changes in either the slope or the height of the estimated response curve. H o w e v e r , limitations in the study design do not permit any conclusions regarding the potential effectiveness of this system. Characteristics specific to the Peace A r c h District Hospi ta l m a y have prevented the detection of an effect. In addit ion, because it is l ikely that there w o u l d be a lag between w h e n the program was implemented and w h e n it might be expected to effect a reduction in length of stay, the follow-u p per iod of eleven months may have been too short for the determination of the program's effectiveness. V T A B L E O F C O N T E N T S Page N o . Abstract i i Table of Contents v List of Tables ix List of Figures x A c k n o w l e d g e m e n t xi C H A P T E R I. Introduction 1 1.1 Statement of the Problem 1 1.2 Thesis Scope and Organizat ion 4 C H A P T E R II. A m e r i c a n Perspectives on Hospi ta l Ut i l izat ion and Ut i l i za t i on M a n a g e m e n t 6 n . l O v e r v i e w 6 n.2 Evidence of Excessive Hospi ta l Use 8 11.2.1 Comparat ive Geographic Studies 10 11.2.2 Comparat ive Cost Reimbursement and Prepa id Practice Studies 11 n.3 Perspectives o n Hospi ta l Ut i l izat ion 12 n.3.1 General Conceptua l and N o r m a t i v e Perspectives 12 U.3.2 M o d e l l i n g the Hosp i ta l Ut i l izat ion Process 16 II.4 Ear ly Developments in Ut i l izat ion Review and Management 21 II.4.1 T h e Introduction of Restrospective Ut i l i zat ion Review 21 n.4.2 A N e w A p p r o a c h : Concurrent Uti l izat ion Review 26 U.4.2.1 Objectives of the PSRO-Prototypes 27 11.4.2.2 Characteristics of the PSRO-Prototypes 28 11.4.2.3 Effectiveness of the PSRO-Prototypes 31 U.4.3 The Transi t ion to PSRO's 39 n.4.4 Major Characteristics of P S R O Review 42 n.4.4.1 A d m i s s i o n Certification 42 U.4.4.2 C o n t i n u e d Stay Certification 43 JJ.4.5 T h e D H E W Evaluat ion of P S R O Review 44 n.4.5.1 Status of P S R O Program Development 44 U.4.5.2 Effectiveness of P S R O Review* 45 (cont...) vi Table of Contents (cont). n.4.5.3 Impact of P S R O Review on Heal th Care Expenditures 49 1X4.6 Other PSRO-Related Research 52 n.4.7 T he Development of D R G ' s 63 n.4.7.1 The Construct ion of D R G ' s 63 II.4.7.2 Implications for Ut i l izat ion Review 65 U.4.7.3 Initial Reactions to D R G ' s 66 II.4.7.4 T he Reformulat ion of D R G ' s 69 n.4.7.5 Suggested Appl icat ions for D R G ' s 70 II.4.8 A n Alternative A p p r o a c h : Assessing A p p r o p r i a t e U t i l i z a t i o n 71 U.4.8.1 Measur ing Appropr ia te Use 74 II.4.8.2 Assessing the Reasons for Inappropriate Use 84 U.4.8.3 E m p i r i c a l Findings on the A m o u n t and Nature of Inappropriate Use 86 II.4.8.4 Implications of the Empir i ca l Studies 93 II.5 Current Trends in Uti l izat ion Management: The E r a of Prospective Payment 95 n.5.1 Th e Transit ion to the Federal Prospective Payment System 97 n.5.2 Major Objectives of the PPS 99 n.5.3 Major Characteristics of the Medicare PPS 101 n.5.4 Implications of PPS for the Cost, Use and Qual i ty of Hospi ta l Care 102 n.5.4.1 Potential Problems Associated with PPS Incentives 102 n.5.4.2 Concerns Regarding the A d e q u a c y of D R G ' s 107 n.5.5 T h e Role of External Review Organizat ions U n d e r P P S 115 n.5.5.1 Goals and Objectives of PRO's 116 n.5.5.2 Characteristics of the P R O Program 116 H.5.6 T h e Role of Internal Review Committees U n d e r PPS 119 n.5.7 Current Trends in Uti l izat ion Review Procedures 120 n.5.8 Ear ly Effects of the PPS on Hospi ta l Uti l izat ion, Costs, and Qual i ty 124 n.5.8.1 Nat iona l Trends in Uti l izat ion, Costs, a n d Qual i ty 125 H.5.8.2 T h e Results of the N e w Jersey Experiment: Implications for the Medicare PPS 127 H.5.9 Comments 131 (cont...) Table of Contents (cont). vii C H A P T E R III. Ut i l izat ion Management in Canada: T h e S W I T C H Index System and Other Developments 134 111.1 T h e Conceptual Context of Ut i l izat ion Management in C a n a d a 134 111.2 Current Status of Ut i l izat ion Management in C a n a d a 142 UI.2.1 Retrospective Review Programs 143 JH.2.1.1 Diagnosis-Specific Profile Analyses 143 ITJ.2.1.2 Service-Specific Profile Analyses .150 HI.2.2 Concurrent Review Programs 151 m.2.2.1 The Hospi ta l for Sick C h i l d r e n 151 m.2.2.2 T h e V i - C a r e Program 153 111.3 T h e S W I T C H Index System 154 111.4 Study Purpose and Rationale 160 111.5 Hypotheses 161 C H A P T E R IV. Methods 163 I V . l Study Design 163 IV.2 Sample 164 IV.3 Data 166 IV.4 Length of Stay Measures 169 IV. 5 Data Analyses 171 C H A P T E R V . Results 174 V . l Graph ica l Analyses 174 V . l . l Unadjusted Average Length of Stay 174 V . l . 2 C a s e - M i x Adjusted Average Length of Stay 174 V.2 Regression Analyses 177 V . 3 Supplementary Analyses 187 C H A P T E R VI . Discussion 191 V I . 1 T h e S W I T C H Index System 191 VI.2 Alternative Explanations for the Fai lure to Detect a Program Effect 195 VI.3 Methodolog ica l Issues a n d Study Limitations 198 (cont...) viii Table of Contents (cont). VI.4 Implications for Future Research and Program Planning 200 VI.5 Policy Implications 204 VI.6 Conclusion 205 References 206 Appendices: Appendix A . Medicare Levels of Care Criteria for Medical/ Surgical Patients 221 Appendix B. Pediatric-Modified Appropriateness Evaluation Protocol 223 Appendix C. List of Potential Barriers to Appropriate Utilization 224 Appendix D. Data Elements Included on the FIMRI Hospital Discharge Abstract 227 Appendix E. Release of Data Form 231 ix LIST O F T A B L E S Page No. Table I. Appropriateness Evaluat ion Protocol: Day-of -Care Criteria 77 Table II. Appropriateness Evaluat ion Protocol: A d m i s s i o n Criteria 83 Table III. Studies of Inappropriate Hospi ta l Ut i l izat ion 87 Table IV. Studies Indicating Reasons for Inappropriate "Day-of-Care" Use 92 Table V . Studies Indicating Reasons for Inappropriate A d m i s s i o n s 94 Table V I . S W I T C H Index System Criteria 156 Table VII. Characteristics of P A D H and N o n - S W I T C H Hospitals: 1982-1985 165 Table VIII. Patient Characteristics of P A D H and N o n - S W I T C H Hospitals: 1982-1985 168 Table IX. Regression Coefficients U s i n g A L S R as the Dependent Variable: P A D H and N o n - S W I T C H Hospitals - Standard Set of C M G ' s 179 Table X . Regression Coefficients U s i n g A L S R as the Dependent Variable: P A D H and N o n - S W I T C H Hospitals - O C C R e m o v e d - Standard Set of C M G ' s 185 Table XI. Regression Coefficients U s i n g A L S R as the Dependent Variable: P A D H and N o n - S W I T C H Hospitals - Patients A g e d 18 to 69 - Standard Set of C M G ' s 190 LIST O F F I G U R E S X Page N o . Figure 1. Subgroup Partit ioning of M D C 55 - U r i n a r y Calculus 66 Figure 2. Kardex A c u i t y Index 159 Figure 3. Hypothet ica l Trends Detectable with T i m e Series Regression Analys i s 172 Figure 4. Unadjusted Average Length of Stay by Hospi ta l G r o u p 175 Figure 5. C a s e - M i x Adjusted Average Length of Stay by Hosp i ta l G r o u p 175 Figure 6. C a s e - M i x Adjusted Length of Stay Ratio by Hospi ta l G r o u p 176 Figure 6. C a s e - M i x Adjusted Adjusted Length of Stay Ratio by Hospi ta l G r o u p - Patients A g e d 18 to 69 189 xi A C K N O W L E D G E M E N T T h e research reported in this thesis was supported by a grant f r o m the Bri t i sh C o l u m b i a H e a l t h C a r e Research F o u n d a t i o n (Grant # 87 (86-1). I gratefully acknowledge the members of m y thesis committee and, in particular, Dr . S.B. Sheps for his patient support and excellent supervis ion. Thanks are also extended to the Peace A r c h District H o s p i t a l a n d the peer g r o u p member hospitals w h i c h k i n d l y permit ted access to their patient data files, a n d to the H M R I for their t imely assistance in p r o v i d i n g the requested data. 1 CHAPTER I Introduction 1.1 Statement of the Problem Historically, the financial and structural mechanisms created in the development of Canada's universal health insurance program did little to encourage economy or efficiency in the delivery of health care services. Rather, they favoured the use and expansion of the most expensive forms o f health care. The growth and prominence of the hospital sector, which currently accounts for over 40% of total health care expenditures (National Health Expenditures, 1987), has been a prime example (Soderstrom, 1978; Iglehart, 1986a, 1986b). During the past decade, however, as federal and provincial governments have become increasingly concerned about the continued affordability of Canada's publicly financed health insurance program (Canadian Comprehensive Auditing Foundation, 1987), economic restraint policies enacted at both levels of government have challenged hospitals to find new ways to improve efficiency and contain costs while maintaining an acceptable standard of care. In response, a few hospitals have begun to develop and implement innovative utilization management strategies designed to identify and eliminate inappropriate and unnecessary use of hospital facilities and services. Although utilization management is a recent innovation in Canadian hospitals, American hospitals have been experimenting with such programs for over twenty years. Utilization review, which involves "the assessment of the appropriateness and efficiency of hospital care through review of the medical record" (Payne, 1987, p. 109), has, since 1966, been a 2 mandatory requirement for hospitals participating in government-funded health insurance programs. In subsequent years, addit ional legislation p r o v i d e d the impetus for the development of the more comprehensive cost-containment programs k n o w n as uti l ization management. In essence, uti l ization management involves a "deliberate action by payers or hospital administrators to influence providers of hospital services to increase the efficiency and effectiveness with which services are provided" (Payne, 1987, p.709). M o r e specifically, as defined by the A m e r i c a n Hosp i ta l Associat ion, uti l ization management is: T h e p lanning , organiz ing, directing, and controll ing of the health care product in a cost-effective manner whi le mainta in ing h igh quality care and contributing to the overall goals of the institution. This is accomplished through the judicious use of resources to control inappropriate inpatient admiss ion, lengths of stay, and use of ancillary services (Connor , M a c k , & H a n d l e m a n , 1983). T h u s , whi le uti l ization review typically involves the measurement a n d evaluation of hospital use against some pre-established criteria (Dodds , 1974), ut i l ization management incorporates this information into the development of a systematic action p lan designed to correct identified problems . A s noted by A n d e r s o n & Lomas (1988) and Rachlis & Fooks (1988), a l though most C a n a d i a n hospitals have implemented some form of uti l ization review, in response to accreditation requirements, few of these programs invo lve more than measurement. A n absence of defined standards for use have, unti l recently, inhibited the development of evaluation strategies and, consequently, remedial action plans have been almost nonexistent. 5 In recent years, however, a few hospitals have begun to implement more comprehensive uti l ization management programs. For the most part, these programs have been designed, in concert wi th their A m e r i c a n predecessors, to facilitate reductions in length of hospital stay. In contrast t o the U n i t e d States, however, few attempts have been made to quantify the amount of inappropriate use reflected in C a n a d i a n hospitals' length of stay experience. In addit ion, to date few systematic attempts have been made t o determine whether current uti l ization management programs are effecting reductions in length of hospital stay nor have any studies been undertaken to compare the cost-effectiveness of the different p r o g r a m initiatives. Research conducted in the Uni ted States suggests that ongoing evaluation of new util ization management programs is critical because such programs can become extremely costly and t ime-consuming without p r o d u c i n g sufficient benefit to warrant their costs. Therefore, since it does appear that interest in util ization management is growing in C a n a d a , it is becoming increasingly important for evaluation studies to be initiated. The study reported in this thesis thus represents a pre l iminary exploration of the effects of one recently implemented uti l ization management program. This p r o g r a m , called the S W I T C H Index System, was implemented in 1984 at the Peace A r c h District Hospi ta l in Whi te Rock, British C o l u m b i a . A s wi l l be detailed in Chapter JJL the S W I T C H system was designed to facilitate the identification and el imination of inappropriate and unnecessary days of hospitalization. A s s u m i n g that such days exist, and that they are controllable by the hospital a n d / o r its medical staff, then if the system is effective an observable outcome ought to be a reduction in the hospital's 4 average length of stay. The present study was designed to explore this hypothesis . i.2 Thesis Scope and Organizat ion A s noted in the A m e r i c a n Hosp i ta l Association's definition of uti l ization management, resource management and quality assurance are both important components of the uti l ization management process. Clear ly , if hospitals are to begin emphas iz ing cost-consciousness in the ' del ivery of hospital services, a system for monitor ing the quality of care is essential. H o w e v e r , a l though resource management and quality assurance are in m a n y respects inseparable (Donabedian, 1969), the focus of this thesis is o n the development of systems a imed specifically at i m p r o v i n g the efficiency a n d appropriateness of hospital resource util ization. N o attempt is made to prov ide a comprehensive review of the large and growing b o d y of quality assurance literature. T o date, most of the literature on uti l ization (resource) management has been based on the A m e r i c a n experience. F e w publications on the nature and status of util ization management in C a n a d a have as yet begun to appear in the academic literature. Therefore, literature relevant to both C a n a d a and the U n i t e d States is discussed. In describing the A m e r i c a n experience, however, attention is l imited to developments that have occurred in response to federal and state initiatives. Innovations in uti l ization management occurr ing as a result of the actions of private health insurance agencies are not discussed because 1) publ ic sector developments are likely to be more relevant to C a n a d a , and 2) very little information o n private sector developments has been found in the academic literature. 5 This thesis, w h i c h provides a history of uti l ization management and describes an exploratory evaluation of the Canadian-based S W I T C H Index System, is organized as follows: Chapter II: details the evolut ion of uti l ization management in the Un i t ed States from 1966 to the present, describes the factors believed to influence hospital uti l ization, in particular, hospital length of stay, and discusses the current status of theoretical models of hospital uti l ization Chapter UI: examines the factors prompt ing Canada's awakening interest in the uti l ization management process, describes the various p r o g r a m initiatives undertaken in hospitals across C a n a d a , and provides a detailed description of the S W I T C H Index System. In addit ion, the purpose and rationale for the evaluative study are delineated. Chapter IV: describes the methods used in evaluating the S W I T C H Index System. Chapter V : describes the results of the study. Chapter VI: discusses the results, the limitations of the study, and the pol icy implications. In addit ion, recommendations for future research are made . 6 C H A P T E R II A m e r i c a n Perspectives on Hospi ta l Ut i l izat ion  a n d Ut i l i zat ion Management U . l O v e r v i e w T h e U n i t e d States, l ike most western industrial nations, has experienced rap id growth in health care expenditures d u r i n g the past twenty-five years. In 1965, for example, total health care expenditures accounted for only 5.9% of the nation's Gross Nat ional Product (GNP) ; by 1987, this total had c l imbed to 11.4%. Moreover , it has been estimated that by 1990, if expenditures continue to grow at current rates, as m u c h as 12% of the G N P wi l l be devoted to the health care sector (Ginzberg, 1987). T w o of the factors identified as major contributors to the observed increases in health care expenditures include: 1) the establishment, in 1965, of publ ic ly financed health insurance for the elderly (Medicare) and the indigent (Medicaid); and 2) the high, and still r is ing costs associated with treating patients in hospitals (Lave & Leinhardt , 1976a; Bean, 1988). A l t h o u g h concerns about the costs and efficiency of publ ic ly financed hospital care remain current, they are not new. For over twenty years, government authorities, health care professionals, a n d health service researchers have been searching for cost-effective ways to contain the cost a n d improve the efficiency of the publ ic ly financed hospital sector (Brown, 1986; Dans , Weiner , & Otter, 1985; Lantos, 1984; M c M a h o n , 1984). C o m p l i c a t i n g the situation, however, has been the prob lem of h o w to curtail costs whi le at the same time maintaining the quality of care at an acceptable standard (Boaz, 1979; H u b e r , Wolfe & H a r d w i c k , 1974; Kel ley, 7 W e n g , & Watson , 1979; M c M a h o n , 1984; Sanazaro, Goldste in , Roberts, Maglott , and McAl l i s t er , 1972). In the ongoing effort to achieve the dua l objectives of cost containment and quality assurance, federal and state authorities have launched a variety of regulatory programs a imed at reducing over-util ization of hospital services and facilities (Brown, 1986). For years, it has been assumed that a substantial port ion of hospital use is unnecessary or i n a p p r o p r i a t e 1 (Dans et al., 1985; G o r a n , 1979). Thus , d u r i n g the late 1960's and throughout the 1970's, it was thought that hospital costs could be contained, without jeopardiz ing the quality of care prov ided , s imply by el iminating excessive use. Accord ing ly , in 1966, legislation was enacted w h i c h called for the establishment of hospital uti l ization review committees and reimbursement controls (Dans et al. , 1985). In the m i d -1970's, more stringent uti l ization controls were implemented a n d final authority for uti l ization review and management was delegated to external review agencies called Professional Standards Review Organizat ions (PSRO's) . T h r o u g h o u t this per iod , the pr imary a im of uti l ization review was to reduce the average length of hospital stays. This emphasis was prompted by a widespread belief that most instances of excessive use occurred dur ing the 1 Unnecessary utilization is defined, in this context, as the provision of care that has no demonstrable effect on patients' outcomes. For the most part, this term is used in reference to hospital stays that are longer than necessary. It can also refer, however, to the use of services or procedures that are ineffective. Inappropriate utilization, on the other hand, can refer either to the admission (treatment) of patients to (in) hospitals for conditions not requiring that intense a level of care (Goran, 1979) or to the failure to provide needed services (Payne, 1987). Thus, inappropriate use can refer to situations reflecting either over- or under-utilization. Historically, inappropriate overutilization has attracted the most attention. More recently, as quality of care has become a more salient issue, problems associated with underutilization have also increased in importance. 8 latter stages of a hospital episode (Goldberg & Hol loway , 1975). In addition, because each day of stay is associated with hotel costs, at the m i n i m u m , it was thought that reductions in length of stay w o u l d be accompanied by proportionate reductions in hospital costs (Lave & Leinhardt , 1976b). By the 1980's, however, it h a d become clear that, on their own, util ization review procedures were not sufficient to contain costs (Averi l l & M c M a h o n , 1977; Boaz, 1979; Dobson et al., 1978). Therefore, in 1983, the federal government enacted legislation which called for a massive restructuring of the Medicare hospital re imbursement system. Specifically, the re imbursement method was changed from cost-based, retrospective reimbursement to case-based, prospective reimbursement (Iglehart, 1982a, 1982b; M a y , 1985). W i t h the advent of the new prospective payment system (PPS), uti l ization review has evo lved into uti l ization management; i.e., it has developed from a medical review p r o g r a m into a strategy employed by federal officials and hospital administrators to effect changes in provider and physic ian behaviour (Payne, 1987). In the transition, the utilization review process has taken on a new focus, emphas iz ing the appropriateness of hospital level care rather than the length of the hospital episode. The basic assumption, however, is the same as before, namely, that a substantial portion of hospital use is unnecessary or inappropriate (Hughes et al., 1984). n .2 Evidence of Excessive Hospi ta l Use D u r i n g the 1950's, private health insurance companies h a d begun to promulgate the not ion that insured hospital services were being "abused" (i.e., overuti l ized) by patients a n d physicians (Shain & Roemer, 1959). A w a r e of such claims, a number of researchers began trying to identify and measure the factors that were contributing to overall patterns of hospital 9 util ization and misuti l ization. In one of the most important of these early studies, Shain & Roemer (1959) found that hospital uti l ization (measured as hospital days per thousand population) was strongly related to bed supply (beds per thousand population). W h i l e acknowledging that util ization tended to increase wi th capacity, these investigators d i d not concur wi th the general op in ion that such a relationship indicated abuse. Instead, they suggested that in the presence of excess supply, physicians' perceptions of what constituted appropriate hospital use were modif ied; i.e., that as capacity increased, the range of medical conditions deemed appropriate for hospitalization widened. A s noted by Evans (1984) and M c L u r e (1982), subsequent studies have so consistently supported the original f indings of Shain & Roemer, that their conclusion ("beds that are built tend to be used") has become k n o w n as "Roemer's Law". In fact, to date, Roemer's L a w remains one of the most c o m m o n l y accepted generalizations about hospital use (Evans, 1984). Moreover , because it suggests that there is, to some extent at least, a discretionary component in hospital util ization decisions, it has itself been cited as one of the earliest indications of excessive hospital use ( M c L u r e , 1982). Since the publ icat ion of Shain & Roemer's classic study, numerous other investigations have been undertaken in an effort to identify and measure unnecessary and inappropriate hospital use. T h e studies that have been pr imar i ly responsible for promot ing the notion that a considerable port ion of hospital use is unnecessary have been grouped into two categories: 1) Comparat ive geographic studies, and 2) Studies comparing fee-for service wi th prepaid practice groups. A l t h o u g h both of these groups of studies prov ide only indirect evidence of excessive hospital use, support ing 10 evidence from a third group, one that attempts to define and measure inappropriate use more directly, has recently begun to appear in the literature. H o w e v e r , since these studies have been done wi th in the context of refining and modi fy ing current util ization review procedures, they are not inc luded here. Instead, they are described, in section 11.4.8. II.2.1 Comparat ive Geographic Studies D u r i n g the past twenty years, there has beer, a proliferation of uti l ization studies based on geographic comparisons. For the most part, these studies have emphasized length of hospital stay as the chosen measure of hospital util ization. H o w e v e r , other measures that have been employed inc lude patient days per thousand populat ion, rates of surgical procedures, and admission rates. A detailed discussion of the problems associated w i th the definition and measurement of hospital uti l ization is p r o v i d e d in a section II.3.2. A s noted by the Congressional Office of Technology Assessment ( O T A ) in an advisory publ i shed in 1983, the results of studies compar ing average length of hospital stay have been consistent and persistent over time. T h e Northeast region of the U n i t e d States has tended to have the longest average length of hospital stay while the West has had the shortest ( O T A , 1983). A c c o r d i n g to O T A data, the eastern hospital lengths of stay have been approximately forty percent higher than those in the west. T h e reported variations in length of stay among geographic regions over time do not, in and of themselves, prov ide evidence to support the assumption of unnecessary use. T h e y could , for example, reflect differences i n regional demographics (eg. age distributions) or severity of illness. In addit ion, it could be that regions wi th shorter lengths of stay have poorer 11 patient health outcomes. In rev iewing the literature on regional lengths of stay, however, neither Evans (1984), G o r a n (1979), M c L u r e (1982), nor the O T A (1983) was able to f ind evidence which suggested that the observed variations in length of stay could be adequately accounted for by differences in demographic or severity of illness variables. Similarly , as noted by G o r a n (1979) and M c L u r e (1982), populat ion studies standardized for age and sex, compar ing regional admiss ion rates, and days of hospital care per thousand populat ion consistently reported wide variations in uti l ization rates. Likewise , W e n n b e r g and Gitt lesohn (1982), w h o found substantial variations in surgical rates across 193 geographic areas in six N e w E n g l a n d states, were unable to attribute their results to differences in demographic , economic or patient condit ion variables. Rather, such differences were deemed to be more closely associated with the supply of physicians within a given area and their preferred patterns of practice. Final ly , a l though there is general agreement that the evidence on health status outcomes associated wi th uti l ization has been somewhat equivocal , there is a consensus that h igh levels of hospital use are not a guaranty for quality (Enthoven, 1983; Evans , 1984; M c L u r e , 1982; O T A , 1983; Wennberg & Gittlesohn, 1982). n.2.2 Comparat ive Cost Reimbursement and Prepaid Practice Studies The costs and utilization rates associated wi th prepaid practice groups, such as Hea l th Maintenance Organizat ions ( H M O ' s ) , have frequently been cited to support claims that expenditures and util ization by patients with traditional insurance coverage (i.e., cost-based hospital re imbursement and fee-for-service phys ic ian reimbursement) m a y be unnecessarily h igh (Enthoven, 1978, 1983; Luft , 1978). T o examine the val idity of such claims, Luft (1978) conducted a major review of comparative studies performed 12 between 1958 and 1971. W i t h respect to costs, he found that, in general, the studies he reviewed d i d support claims that expenditures were lower for H M O ' s . In particular, he noted that the Kaiser Permanente plan, one of the oldest and most wel l k n o w n of the H M O ' s , had total per capita costs that ranged from 10 to 40 percent lower than those with traditional insurance coverage. In addit ion, he observed that most of the difference in costs was attributable to reductions in hospital use. Prepaid practice groups were found to have between 15 and 40 percent fewer admissions and between 25 and 45 percent fewer hospital days per thousand populat ion. A l t h o u g h questions about the comparabil i ty of the populat ions and the quality of care p r o v i d e d can be raised, little evidence has been prov ided to suggest that either concern can account totally for the observed results. M a n y of the studies reviewed by Luft had incorporated populat ion standardization procedures. In addit ion, as noted by Evans (1984), observed reductions in use apparently have not affected mortality or morbidi ty , two c o m m o n l y used measures of quality. Despite these observations, however, the effect of prepaid group practice on quality of care remains a matter of debate (Ginzberg, 1985; Luft , 1978). Furthermore, as Luft (1978) and Enthoven (1983) have indicated, the issue of whether prepa id practice group achieve reductions in costs a n d use through selection of healthier patients, has not yet been resolved. U.3 Perspectives on Hospi ta l Ut i l izat ion U.3.1 General Conceptual and N o r m a t i v e Perspectives W h e n the U n i t e d States Congress enacted the 1965 Social Securities A c t it p r o v i d e d a mandate for the establishment of two publicly-f inanced health insurance programs, Medicare and M e d i c a i d . Medicare , a federal 13 program, was to pay for hospital and medical services p r o v i d e d to the elderly and the disabled; M e d i c a i d , a joint federal-state program, was to pay for services p r o v i d e d to the indigent (Enthoven, 1983). T h e passage of this legislation clearly reflected the conceptual and normative perspectives that dominated the health care scene at the time. A c c o r d i n g to B r o w n (1986), the development of the A m e r i c a n health care system, prior to the 1970's, was guided by six beliefs or propositions. These he s u m m a r i z e d as follows: First, curing illness is the pr imary function of the health care system. A n i m p r o v e d health care system w i l l improve health status outcomes. Second, one of the m a i n problems with the health care system is that there is too little of it. T h i r d , the only other serious prob lem is l imited financial access to mainstream care among disadvantaged citizens, notably the poor and the elderly. Fourth , payments for professional a n d institutional services should be based on the actual incurred costs of hospitals and the usual and customary charges of physicians, so long as these are reasonable. Fifth, in health care, markets do not and cannot work. Sixth, in health care, regulation is the w r o n g answer to market failures. (Adapted from B r o w n , 1986, p. 572). A s original ly formulated, the Medicare and M e d i c a i d legislation embraced the proposit ions out l ined above. W i t h i n this framework, the only role envis ioned for publ ic pol icy intervention was in the removal of financial barriers to care (Brown, 1986; Evans , 1984). A c c o r d i n g l y , the p r i m a r y a i m of the legislation was to increase the quality a n d quantity of health services available, and to ensure that the elderly and the indigent had access to these services. Moreover , reimbursement strategies were to 1 4 inc lude actual cost (retrospective) reimbursement for hospitals and fee-for-service payments for physicians. In line wi th the dominant views on reimbursement a n d regulation, the only constraint p laced on reimbursement was that submitted claims reflect 'reasonable' costs (Gr imald i & Michelett i , 1984). W i t h i n this framework, referred to as the "traditional" (Ro, 1969) or "Medico-Technical" (Evans, 1984) mode l , uti l ization controls were not . considered necessary because the uti l ization of hospital and other health care services were assumed to be a direct response by providers to the medical needs of patients, subject only to currently available medical technologies (Evans, 1984; Ro, 1969). A n y observed variations in use were explained solely in terms of differing medical philosophies about the "technologically-prescribed method of treatment" (Ro, 1969, p. 298) and , -occasionally, 'extra-medical' factors such as the age and sex of the patient. W h e n the legislation was enacted, however, the experience of private insurers wi th "overutilization" of insured hospital services h a d already begun to suggest that uti l ization decisions were not a "micromanifestation of objective, scientific, professional judgement" (Brown, 1986, p. 574). Instead, it appeared that such decisions were subject to the influence of such non-cl inical factors as medical practice pattern preferences and the availabil ity of facilities and services (Shain & Roemer, 1959), insurance coverage (Studnicki & H o n e m a n n , 1983), income-increasing objectives ( M c L u r e , 1982; G o r a n , 1979), and hospital organizational procedures (Studnicki & H o n e m a n n , 1983; Q u e r i d o , i963). Therefore, the inclusion of open-ended reimbursement strategies and the absence of uti l ization controls in the M e d i c a r e / M e d i c a i d legislation were interpreted by some as reflecting 1 5 a rather naive acceptance, o n the part of the government, of the "traditional" v iewpoint (Brian, 1972, 1973). Consequently , critics of the federal p r o g r a m began to raise serious questions about the impact public f inancing w o u l d have on the uti l ization and, thereby, the costs of insured hospital services (Fulchiero, Mi l l er , Foley, Ballantine, & A m o r o s i n o , 1978). In particular, it was feared that the provis ion of publ ic health insurance w o u l d promote overuti l ization of hospital services and, consequently, inflate hospital costs. In essence, it was argued that with the provis ion of publ ic insurance, the function of price, as a stabilizer between supply and d e m a n d , w o u l d be negated (Evans, 1984). Thus , by removing price as a barrier to care, it was anticipated that util ization w o u l d increase excessively, subject only to the constraint of available supply. Moreover , because reimbursement was to be p r o v i d e d on an actual cost basis, it was feared that hospitals w o u l d be motivated to promote overuti l ization. U n d e r the federal reimbursement scheme, hospitals could increase their revenues by admitt ing more patients and by keeping them in hospital longer. In addit ion, physicians, w h o were responsible for admitt ing and discharging patients, received a higher fee for treating patients in hospitals than in less costly settings. Hence , incentives favouring excessive hospital ization were bel ieved to be operating o n both the hospital and its physicians (Enthoven, 1983; M a y , 1985). T o counteract this possibility, the federal government a m e n d e d the original M e d i c a r e / M e d i c a i d legislation to inc lude utilization review requirements. T h e amended legislation was enacted i n 1966 and p r o v i d e d a major impetus for the subsequent proliferation of util ization review a n d management programs a n d hospital uti l ization research. 16 In the fo l lowing pages, the evolut ion of uti l ization review from its initial formulat ion under the 1966 amendment to its present status wi l l be described. In addit ion, the literature describing the effectiveness of the various uti l ization strategies wi l l be reviewed. Before turning to that discussion, however, it is important to examine the literature that has attempted to define, measure, and mode l the factors that influence hospital uti l ization decisions. For the most part, the regulatory mechanisms that' have been implemented, over the years, in the effort to reduce excessive hospital use, have focussed on length of hospital stay as the pr imary measure of hospital utilization. Consequently, a considerable b o d y of literature has grown up in an attempt to identify and quantify the factors that are important in determining the length of a hospital episode. A knowledge of such factors, and h o w they interact, is necessary if predictions about the effects of pol icy initiatives are to be made, and if evaluations of the effectiveness of uti l ization rev iew/management programs are to be meaningful . Therefore, the remainder of this section wi l l be devoted to the general hospital uti l ization literature. In particular, studies that have attempted to m o d e l length of stay wi l l be reviewed. II.3.2 M o d e l l i n g the Hosp i ta l Ut i l izat ion Process H o s p i t a l uti l ization is a complex, interactive process invo lv ing patient, phys ic ian, a n d hospital characteristics. In addit ion, it is vulnerable to the influence of policies a imed at effecting changes in hospital and phys ic ian behaviour. Thus , for example, it is subject to the influence of changes i n reimbursement strategies and insurance coverage. A mode l of hospital ut i l ization that fails to include a description of the impl ic i t or explicit m o d e l of the hospital to wh ich it refers wi l l , therefore, be l imited in 1 7 its usefulness. A s noted by Evans (1984), however, most of the literature that has been accumulated on hospital util ization "has been at best loosely l inked to theoretical models of hospital behaviour" (p.185). Generalizations induced from empir ica l research have been far more c o m m o n . Moreover , the attempts that have been made, to date, to develop formal models of hospital behaviour have not been very successful (Evans, 1984). T h e numerous processes, factors, and interrelationships that govern hospital behaviour are extremely complicated and, often, idiosyncratic and culture-b o u n d . Consequently , as Evans states, "[r]ealistic description and analysis of such complex processes can lead to very useful generalizations about h o w hospitals are l ikely to behave, and to respond to changes in their external environments, but they m a y never be expressed in a formal analytic framework which is either realistic or useful" (p. 173). T h u s , a l though as yet lacking a formal theoretical framework, the literature o n hospital uti l ization has p r o v i d e d some valuable insights into the factors that influence patterns of uti l ization and misuti l ization. For example, on the basis of the research done by Shain and Roemer (1959), it is n o w wide ly accepted that physicians' util ization decisions can be modif ied by changes in bed capacity. Likewise , in contrast to the "traditional" v iew of uti l ization, an accumulat ion of research has contributed to widespread acceptance of the v iew that uti l ization decisions, "though predominant ly based o n clinical considerations, are also influenced by nonmedica l factors" (Boaz, 1979, p. 316). T h u s , for example, studies demonstrating w i d e variations in surgical rates across geographical regions (Wennberg & Gitt lesohn, 1982), and in admiss ion rates wi th in diagnosis-defined patient groups (Wennberg, McPherson , & Caper , 1984), have contributed to 1 8 acceptance of the notion that utilization is affected by factors such as physic ian supply and practice pattern preferences. A s ment ioned previously , the major variable e m p l o y e d in the formulat ion of uti l ization policies, the development a n d evaluation of uti l ization management programs, a n d the conceptualization of hospital uti l ization models , has been length of hospital stay. This emphasis o n length of stay has occurred for several reasons. First, w h e n the initial uti l ization review regulations were enacted in 1966, a measure of util ization was needed that could be obtained, objectively and easily, nation-wide. Length of stay satisfied these requirements (Goldberg & H o l l o w a y , 1975). Second, w h e n uti l ization review procedures were first being developed, "no definitive standard separating necessary from unnecessary use [had] been established" (Goran, 1979, p. 7). Evidence available from comparative studies, however, suggested that most instances of misuti l izat ion were occurring d u r i n g the latter stages of inpatient episodes. A l t h o u g h this belief has since been dispel led, it was a major factor g iv ing rise to an emphasis on hospital lengths of stay (Goldberg & H o l l o w a y , 1975; Z i m m e r , 1974). T h i r d , length of stay has traditionally been accepted as an important measure of hospital performance a n d resource consumpt ion (Berki, Ashcraft , & Ne wbr ande r , 1984; L a v e & Leinhardt , 1976b; Ro, 1969). Final ly , length of stay has been identified as one of the most important variables contributing to hospital costs, since the hotel costs associated with each bed day are, on average, the single most costly component of each hospital episode (Knapp, Speedie, Yaeger, & K n a p p , 1980; L a v e & Leinhardt , 1976b). W i t h the advent of uti l ization control policies, a major impetus was p r o v i d e d for hospital util ization research. Previous studies, such as those 1 9 described above, in the section on unnecessary util ization, had generated questions about the premise that util ization decisions were based solely on medical and technological imperatives. But, little was as yet k n o w n about the relative importance of non-cl inical variables on outcome measures such as length of hospital stay. A s one of the major aims of the newly enacted uti l ization review requirements was the el imination of unnecessary or inappropriate use, as measured by reductions in length of stay, a better understanding of the factors, other than uti l ization review, that cou ld influence the length of a hospital episode was becoming increasingly important. In particular, information concerning the variables that needed to be taken into account w h e n attempting to evaluate the effectiveness of uti l ization review strategies was sought. In response to the need for more information, several researchers attempted to identify variables that could , potentially, influence length of stay ( Becker, Shortell, & Neuhauser, 1980; Berki et al., 1984; Boaz, 1979; Boone, C o u l t o n , & Keller, 1981/82; L a v e & Leinhardt, 1976b; M c L u r e , 1982; Ro , 1969). It was hoped that the results of such studies w o u l d eventually lead to the development of a formal m o d e l of hospital uti l ization (based on length of hospital stay). T o date, however, no comprehensive and realistic m o d e l of hospital length of stay has been developed. T h e major factors inhibit ing the development of such a m o d e l have been the lack of appropriate data and, accordingly, the inability to study more than a few variables at a time (Berki et al . , 1984; Studnicki & H o n e m a n n , 1983). It has become clear, however, that two of the most important determinants of length of stay are the "patient's health status at entry and the medica l condit ion p r o m p t i n g the admission" (Lave & Leinhardt , 1976b, 20 p. 329). In addit ion, a positive relationship between the patient's age on admiss ion and the length of the inpatient episode has consistently been reported in the literature (Berki et al., 1984; Goldfarb, H o r n b r o o k , & Higgins , 1983; L a v e & Leinhardt , 1976b; Ro, 1969). M o r e specifically, clinical variables such as diagnosis, admiss ion status, severity of illness, complications, and number of surgical procedures, as well as patient age, have repeatedly been found to be more important determinants of length of stay than non-clinical variables - for example, occupancy rate, income, insurance coverage, distance from hospital, size of the hospital, teaching status of hospital , day of the week of admiss ion, type of accommodat ion in hospital , phys ic ian age, physic ian specialty - (Berki et al., 1984; Boaz, 1979; Goldfarb, H o r n b r o o k , & Higg ins , 1983; Goldfarb , Hornbrook , & Rafferty, 1980; Lave & Leinhardt , 1976b; Ro , 1969). This is not to say, however, that non-clinical variables are unimportant . Significant and, often, complex relationships have been identified between each of the above variables and length of stay. But, in comparison to the importance of the clinical variables, their overall influence has been small . For example, in one of the most comprehensive studies (Lave & Leinhardt , 1976b), variables such as the patient's pr imary diagnosis, number of diagnoses, number of surgical procedures, admission a n d discharge status accounted for 40 per cent of the variability in log length of stay. T h e addit ion of hospital variables such as occupancy rate and day of week admitted plus patient variables such as age, marital status, race and sex increased R 2 by only 2 per cent. T h u s , whi le m a n y of these variables were f o u n d to have significant beta coefficients, their overal l contribution to the explanatory power of the regression m o d e l was small . H o w e v e r , because only a few of the potentially important non-clinical variables have as yet 21 been inc luded in any one study, the simultaneous effects and interactions a m o n g these variables have not been adequately tested (Berki et al., 1984; L a v e & Leinhardt , 1976b). Nonetheless, whi le efforts to develop a realistic m o d e l of hospital length of stay continue, it has become clear that any studies that attempt to assess the effects of non-clinical variables on length of hospital stay must control, at the m i n i m u m , for differences in clinical attributes. T h u s , for example, efforts to evaluate the effectiveness of uti l ization review or management programs on length of hospital stay wi l l be meaningful only if the important effects of patient age and medical condit ion have first been taken into account. In addit ion, because complex relationships have been found to exist between length of stay and variables such as hospital size, bed capacity, occupancy rate and teaching status, an attempt should be made to ensure the comparabil i ty of hospitals inc luded in the analysis. II.4 E a r l y Developments in Ut i l izat ion Review and Management II.4.1 T h e Introduction of Retrospective Ut i l i zat ion Rev iew A s ment ioned in section n.2, one of the earliest indications of hospital overut i l izat ion came from private health insurance companies (Shain & Roemer, 1959). In response, a number of hospitals began to develop and implement uti l ization review programs. In addi t ion , the insurance companies themselves began to institute retrospective claims review procedures. By 1960, trustees of the A m e r i c a n M e d i c a l Associat ion h a d a p p r o v e d in principle the idea of utilization review. In 1963, it was r e c o m m e n d e d by the Joint C o m m i s s i o n o n Accreditat ion of Hospita ls and, by 1966, it h a d become a standard for accreditation (Foster, 1968). 22 Despite these private sector initiatives, the major impetus for what has since become the proliferation of uti l ization review and management programs was the 1965 enactment of the Medicare and M e d i c a i d legislation. In response to concerns that publ ic insurance w o u l d promote overuti l ization of hospital services, and lead to an unacceptable cost burden , the federal government decided to follow the lead p r o v i d e d in the private sector a n d pursue cost containment by invok ing uti l ization controls (Fulchiero et al., 1978; Lantos, 1984). Accord ing ly , in 1966, util ization review was made a mandatory requirement for hospitals participating in the Medicare and M e d i c a i d programs (Knowlton, 1975; Lantos, 1984; Schmitz & Schoenhard, 1976). In addit ion, the fiscal intermediaries responsible for administer ing the plans, were mandated to conduct retrospective claims review and to deny payment for care that was deemed to reflect unnecessary or inappropriate use of hospital facilities or services. In particular, claims were to be denied if it was concluded that the services p r o v i d e d cou ld have been p r o v i d e d in a less costly setting (eg. outpatient clinic) or wi th fewer days stay in hospital (Flashner, Reed, White , & Norr i s , 1972; G o l d b e r g & H o l l o w a y , 1975; Snider, 1977). U n d e r the terms of the amended legislation, each participating hospital was required to establish a Ut i l izat ion Rev iew (UR) committee. This committee was to be comprised of physicians w h o were to be responsible for ensuring that hospital util ization was both appropriate and medical ly necessary. In addit ion, the committee was to review the uti l ization of professional services, g iv ing particular emphasis to the ascertainment of overuse or underuse of such services (Lantos, 1984). The p r i m a r y objective of the committee was to be "the maintenance of h igh 23 quality patient care, and an increase in the effective util ization of hospital services" (Health Insurance for the A g e d , 1966). A s initially enacted, the regulations governing the uti l ization procedures to be implemented by the hospital U R committees p r o v i d e d few guidelines wi th respect to when and h o w the reviews should be done. A c c o r d i n g l y , the uti l ization review programs that were implemented tended to emphasize retrospective analysis of inpatient episodes (Flashner, Reed, White , & Norr i s , 1972; Snider, 1977).. That is, patient stays were, for the most part reviewed long after the actual patients invo lved had been discharged f rom the hospital. A n exception to this process, however, occurred with patient stays of extended duration. U n d e r the terms of the uti l ization review regulations such cases had to be reviewed on the 12th and 18th days of hospitalization and every 30 days thereafter (Schmitz & Schoenhard, 1976). This emphasis on extended stays der ived from a general belief that most instances of misuti l ization w o u l d occur d u r i n g extended hospital stays. A l t h o u g h this assumption w o u l d later be challenged, at the time it was widespread and pervasive (Zimmer, 1974). In addi t ion to uti l ization review, the newly mandated U R committees were also required to conduct, on an ongoing basis, medical care evaluation studies ( M C E ' s ; K n o w l t o n , 1975). These studies, essentially medical audits, were designed to assure that the quality of care p r o v i d e d was of an adequate professional standard. In particular, they were to "emphasize qual i ty improvement through cont inuing medical education targeted specifically to correct identified problems" (Goran, Roberts, Kel logg , F ie ld ing , & Jessee, 1975, p. 15). A t the time, however, a lack of adequate tools for quality of care assessment discouraged the effective development of 24 M C E ' s (Goran et al., 1975). Moreover , because the federal regulations inc luded the possibility of retroactive claims denial , more emphasis was placed on uti l ization review than on quality assessment (Flashner, Reed, Whi te , & N o r r i s , 1972). O v e r the next five years m a n y U R committees were established and a variety of uti l ization review programs were developed and implemented. A s noted above, with the exception of extended stay cases, almost all of the patient reviews were conducted retrospectively. Likewise , the reviews conducted by the fiscal intermediaries were retrospective. Moreover , when instances of misuti l izat ion were identified, hospital claims were denied retroactively. This procedure put the hospitals at considerable financial risk for situations over which they h a d little control. Physicians were responsible for admitt ing and discharging patients from the hospital; the hospital administrat ion cou ld do little to intervene in this process. Consequent ly , the threat of retroactive denials became a cont inuing source of tension between hospital administrators and their medical staffs (Flashner, Reed, White , & Norr i s , 1972; L a v e & Leinhardt , 1976a; T o m , 1976). B y 1970, it was generally conceded that the federal effort to control uti l ization was not work ing (Dans et al., 1985; Flashner, Reed, C o b u r n , & Fine, 1973; Flashner, Reed, White , & Norr i s , 1972; Fulchiero et al., 1978; T o m , 1976). In a review of several studies which attempted to evaluate the effectiveness of retrospective review procedures, Foster (1968) was unable to f ind any evidence to support claims that uti l ization review reduced length of stay or admiss ion rates. It should be noted, though, that all of the studies reviewed by Foster suffered from major methodological problems. Thus , in 25 the absence of any well designed studies, Foster's conclusions may have been premature. In a later review of the literature, however, C lendenning , Wol fe , S h u m a n , & H u b e r (1976) found only one significant change that cou ld be attributed to the federal util ization review effort; namely, that "hospitals with h igh lengths of stay in 1966-1967 h a d a larger reduction in lengths of stay by 1969-1970 than hospitals which started with lower lengths of stay" (p. 753). The possibility that this reduction may s imply reflect a regression to the mean effect, however, cannot be discounted. A l t h o u g h empir ic evidence on the effectiveness, or lack of effectiveness, of retrospective review was l imited, the fact that hospital costs were continuing to increase p r o m p t e d the conclusion that the federal effort h a d been a failure (Brian, 1972; Dans eta l . , 1985; Fulcheiro et al., 1978). M a n y reasons were cited for the program's lack of success. O n e of the most frequent criticisms was the retrospective nature of most of the review procedures. Since these procedures d i d not identify util ization problems unti l the patient h a d already left the hospital, the U R committee could do little to effect change in phys ic ian behaviour. Moreover , the threat of retroactive denial of claims, and the tensions this generated, d i d little to encourage cooperation between the hospital and its medical staff (Flashner, Reed, White , & Norr i s , 1972; Fulchiero et al., 1978). Consequently , support for the p r o g r a m was difficult to maintain. Other reasons cited to explain the program's failure include the inabil i ty or unwil l ingness of U R committees to u p h o l d decisions that had a negative impact o n their peers, haphazard enforcement of the regulations, poor coordination between the activities of Medicare a n d M e d i c a i d , and the lack of appropriate norms for evaluating the medical necessity and 26 appropriateness of care (Flashner, Reed, White , & Norr i s , 1972; Flashner, Reed, C o b u r n , & Fine, 1973; Fulcheiro et a l , 1978). II.4.2 A N e w A p p r o a c h : Concurrent Ut i l izat ion Review By the late 1960's, acknowledgement of the failure of the federal review strategy to control either hospital costs or use had prompted several responses. First, state M e d i c a i d authorities had begun reporting that, at current rates of cost increases, M e d i c a i d trust funds w o u l d soon be insufficient to meet demands for medical care (Flashner, Reed, Whi te , & Norr i s , 1972). Consequently , they began to appeal to regional Foundations for M e d i c a l Care ( F M C ' s ) 1 for assistance in developing more effective uti l ization review strategies (Flashner, Reed, Whi te , & N o r r i s , 1972; Frederick, 1975). Second, in 1970, Senator Wallace Bennett (R., Utah) introduced a proposed amendment to the federal Medicare and M e d i c a i d legislation which called for the establishment of Professional Standards Rev iew Organizat ions (PSRO's). Final ly , and also in 1970, the Nat iona l Center for Hea l th Services Research and Development launched the Experimental M e d i c a l Care Review ( E M C R O ) p r o g r a m (Sanazaro et al. , 1972). T h e combined impact of these initiatives was the development of new approaches to util ization review which emphasized: 1) the use of admiss ion screening or certification procedures to set initial expected lengths of stay (target dates) for each hospital admission, 2) the establishment of concurrent rather than retrospective review procedures, and 3) the institution of extended stay certification procedures. U n t i l the FMC's are voluntary medical review organizations which represent a large number of physicians practicing within a defined geographic region. 27 mid-1970's, however, most of the new programs were developed as a result of the state M e d i c a i d and E M C R O initiatives. The enactment of the P S R O legislation, introduced by Senator Wallace in 1970, d id not take place until 1972, a n d the regulations delineating P S R O responsibilities were not established unt i l 1974. Nonetheless, by 1975, the P S R O s h a d gradual ly begun to replace the state and E M C R O programs and, by 1981, they had been established in 187 of 195 geographically designated areas (Dans et al., 1985). T o a large extent, the major components of the federal P S R O programs were der ived directly from the p r e - P S R O M e d i c a i d p r o g r a m strategies. Moreover , because the transition to PSRO's took so m a n y years, most of the evaluative studies done d u r i n g the 1970's concentrated on the M e d i c a i d concurrent review programs. Some of the major p r e - P S R O programs, developed between 1970 and 1973, that were later used as prototypes for the development of the P S R O program initiatives inc luded the Certi f ied Hosp i ta l Admini s tra t ion P r o g r a m ( C H A P ) , the Hosp i ta l A d m i s s i o n and Surveil lance P r o g r a m ( H A S P ) , the Predischarge Ut i l izat ion Rev iew P r o g r a m ( P D U R ) , the N e w Mexico Experimental M e d i c a l Care Rev iew Organ iza t ion ( N M E M C R O ) , and the C o m m o n w e a l t h Hea l th Agencies M o n i t o r i n g Program ( C H A M P ) . T h e objectives a n d major characteristics of these programs are described below. n.4.2.1 Objectives of the PSRO-prototvpes A l t h o u g h the various prototype programs listed above differed to some extent in administrative details, they all h a d the same objective. The p r i m a r y goal was cost containment, wh ich was to be achieved through reductions i n inappropriate a n d unnecessary hospital use. L i k e the federal retrospective review program, however, this goal was tempered, at least 28 implic i t ly , by the caveat that such reductions were to be obtained without jeopardiz ing the quality of care prov ided . The E M C R O programs, in fact, were established with quality assurance as their pr imary objective (Sanazaro et al., 1972). Moreover , important developments in quality assurance d i d take place as a result of the E M C R O initiatives (Goran et al., 1975). Nonetheless, cost concerns were so widespread at the time that all of the programs, the E M C R O ' s inc luded, tended to emphasize cost containment objectives over quality assurance activities. W i t h the advent of concurrent review, the major strategy e m p l o y e d to achieve cost containment invo lved the development and use of norms for util ization based on length of hospital stay. A s with the federal retrospective review program, it was still wide ly assumed that most cases of misuti l izat ion w o u l d be found dur ing the latter days of an inpatient episode. Therefore, because each extra day in hospital added hotel costs at the m i n i m u m , such days were targeted for el imination. But, whereas the federal p r o g r a m required extended stay reviews of all patients, regardless of diagnosis, at certain specified times, the new concurrent review programs began to use diagnosis-specific length of stay norms (Goldberg & H o l l o w a y , 1975). H.4.2.2 Characteristics of PSRO-prototypes A s noted above, the 1966 federal util ization review regulations placed responsibil ity for the review process was placed in the hands of hospital ut i l izat ion rev iew committees. W i t h the new approach to uti l ization review, this responsibility was transferred to either the state M e d i c a i d authority, a regional F M C , or some other designated external reviewer. This change was initiated in an effort to avo id the criticism that hospital U R 29 committees were often reluctant to "uphold difficult and unpopular decisions which could have an adverse effect on their peers" (Flashner, Reed, White , & Norr i s , 1972, p.1154). In addit ion, the previous emphasis o n retrospective review, wh ich had been found to be both ineffective and contentious, was replaced, in all of the new programs, wi th concurrent review procedures. Typica l ly , these procedures inc luded two components: 1) admiss ion screening and the approval of a specified length of hospital stay, and 2) extended stay certification for cases that could not be discharged wi th in the time frame approved u p o n admission. W h i l e all the new programs incorporated extended stay certification, the extent to which admiss ion screening was carried out varied from one p r o g r a m to another. T h u s , C H A P (Brian, 1972, 1973) and the N M E M C R O (Brook & Wi l l i ams , 1976; Brook, Wi l l i ams & R o l p h , 1978) required preadmiss ion certification for all elective hospital admissions and length of stay certification, wi thin 24 to 48 hours after admiss ion, for emergency cases. T h e H A S P (Flashner, Reed, Whi te , & N o r r i s , 1972) and the C H A M P (Fulcheiro et al., 1978) required screening of admissions for certain diagnoses and for physicians wi th a history of hospital misuti l ization, as wel l as length of stay certification for all cases, emergent and elective, wi th in 24 hours of admiss ion. The P D U R (Clendenning et al., 1976; L a v e & Leinhardt , 1976a), o n the other h a n d , h a d no admission screening procedure, only a prospectively determined length of stay certification. For programs with admiss ion screening procedures, whether pre-admiss ion or at the time of admiss ion, the purpose was to identify and eliminate inappropriate hospitalizations. T h e length of stay certification process, based o n diagnosis-specific length of stay norms calculated for the region or for each indiv idual 30 hospital , was intended to eliminate days of hospitalization which could not be substantiated on the basis of medical need. W i t h i n 24 hours before the expiration date for the initial length of stay certification, all of the programs required that the attending physic ian request extended stay certification for those patients w h o were not to be discharged. T h e diagnosis-specific norms used to set the initial target-date, however, differed f rom one program to the other. The C H A P , H A S P , N M E M C R O , and C H A M P all used data compi led by the Professional Activit ies Studies (PAS) for the state in which the p r o g r a m was located. W i t h the exception of the C H A M P , all of these programs set their initial target dates us ing the 50th percentile length of stay for the diagnostic groupings recorded by the P A S ; the C H A M P used the 75th percentile. In contrast, the database used by the P D U R to establish initial lengths of stay was the Hospi ta l Ut i l izat ion Project ( H U P ) . The H U P recorded hospital data, by county, for the State of Pennsylvania. The P D U R , therefore, used 50th percentile norms for the county, rather than the state, in w h i c h the p r o g r a m h a d been launched. Typica l ly , nurse coordinators or physic ian consultants were hired by the external reviewer to perform the necessary reviews and to grant the necessary approvals . These reviewers were located in each hospital; off-site rev iew was on ly required in situations where there were conflicting views or physician-init iated appeals. W h e n the reviewer was a registered nurse, rather than a physic ian, admiss ion or length of stay requests cou ld not be denied without authorization from a designated physic ian consultant. A s an incentive for hospitals and physicians to cooperate wi th the new approach, some of the programs (i.e., C H A P , H A S P , C H A M P ) 31 abandoned the previous ly used control mechanism of retroactive claims denial . By requir ing that admissions and length of stay be prospectively certified such a punit ive mechanism was deemed unnecessary (Flashner, Reed , Whi te , & N o r r i s , 1972; Flashner, Reed, C o b u r n , & Fine, 1973). Despite these major changes in the approach to uti l ization review, few attempts were made, prior to 1972, to determine whether the new concurrent review programs were any more effective in containing costs or reduc ing uti l ization than were the federal retrospective review programs. Thus , the subsequent passage of the P S R O legislation was based, for the most part, on anecdotal reports and the intuitive appeal of the review mechanisms that were being employed (Sanazaro, 1977). The ongoing deve lopment of P S R O pol icy formulat ion, however, depended heavi ly on the perceived effectiveness of the prototype programs (Sayetta, 1976). Consequent ly , between 1972 and 1979, a number of evaluative studies began to appear in the literature. n.4.2.3 Effectiveness of the PSRO-Prototypes In early evaluations of the effectiveness of two of the p r e - P S R O programs, Brian (1972,1973) and Flashner, Reed, C o b u r n a n d Fine (1973) maintained that significant reductions in average length of stay h a d been obtained wi th the C H A P and H A S P , respectively. Brian further c la imed that admissions h a d decl ined by 11.4% in the eight month per iod fo l lowing implementat ion of C H A P . In addit ion, both sets of investigators estimated potential cost savings, as a result of the observed reductions in admiss ion rates a n d length of stay, that ranged from $9 mi l l ion to $250 mi l l ion . A s these were a m o n g the earliest attempts to evaluate the effectiveness of concurrent review programs their initial impact o n P S R O 32 activity was considerable. Subsequent critiques of the studies (Davidson, Wacker , & K l e i n , 1973; Sayetta, 1976; C lendenn ing et al., 1976), however, cited weaknesses i n the evaluation methods that were sufficient to invalidate the purported program successes. Specific concerns inc luded the absence of appropriate comparative data, the use of baseline periods that were too short to establish appropriate util ization trends, and the failure to report the methods used to compute expected util ization rates in the follow-u p periods. The major criticism, however, was that in none of the investigations d i d the researchers attempt to adjust for differences in the case mix a m o n g the hospitals studied, or wi th in the hospitals over time. Hospi ta l util ization research had , by this time, already begun to suggest that one of the most important factors contributing to hospital length of stay was the patient's medical condit ion or diagnosis (Lave & Leinhardt , 1976b; Ro , 1969). It was concluded, therefore, that no evaluations of program effectiveness a imed at reducing length of stay w o u l d be va l id if, at the m i n i m u m , they failed to control for this important variable. In an attempt to address some of these criticisms, W y l i e & Flashner (1979) reevaluated the H A S P p r o g r a m using data from the six month period pr ior to the program's implementation (July to December, 1971) and the comparable six m o n t h per iod fo l lowing its implementat ion (July to December, 1972). A total of 46 Illinois hospitals, covered by H A S P , were inc luded in the study (for a total of 327,943 discharges). T o adjust for possible case mix variations in the two time periods, each M e d i c a i d patient discharge was matched to a comparison group of n o n - M e d i c a i d patients according to the fol lowing characteristics: final diagnosis, age, presence of more than one diagnosis, and whether or not surgery h a d been involved. 33 T h e comparison group was formed using P A S data for all hospitals in the Centra l Region of the Uni ted States d u r i n g the calendar year 1971. This group was used to compute 'expected lengths of stay' for each Illinois study patient in each of the pre- and post- intervention periods. In analyz ing the data, M e d i c a i d patients in the H A S P hospitals were compared to n o n - M e d i c a i d patients in the H A S P hospitals. (The H A S P procedure was used with M e d i c a i d patients only). For each group, the . observed average length of stay in the before and after periods were computed. In addit ion, expected values for each group, in each per iod, were computed us ing the larger Central Region hospital statistics. T h e authors reported a reduct ion in per patient length of stay in the fol low-up per iod that was significantly greater for the M e d i c a i d (0.8 day reduction) than the n o n - M e d i c a i d (0.2 day reduction) group (p_<.01).. T h e y concluded, therefore, "that concurrent review, as implemented by H A S P , helped shorten [length of stay] in 1972" (Wylie & Flashner, 1979, p. 607). In a change from the study done i n 1973, no claims were made regarding the program's cost-reducing effects. T h e authors d i d maintain, however, that the p r o g r a m h a d contributed to a slow d o w n in growth. A l t h o u g h this s tudy represented a methodological improvement over the earlier studies, it was still problematic. In particular, the use of the calendar year 1971 for the calculation of expected rates, instead of the comparable six month periods from 1971 and 1972, raises questions about the val id i ty of the authors reported conclusions. By not us ing the 1972 regional data, the possibility of decreasing trends in the comparison group were not taken into account. It was also unfortunate that they d i d not expand the pre- and post- study periods beyond the reported six months in 34 each direction. Longer time periods w o u l d have he lped to establish clearer trends over time and, also, to show whether the observed effects were temporary, endur ing or s imply regression to the mean. Final ly , their comments regarding the program's effects on costs were based solely on impressions der ived from the observed reduction in length of stay. Since no attempt was made to analyze the program's effect on admiss ion rates, and given the lack of any analysis of cost data, their contention that the • p r o g r a m h a d s lowed growth was unsubstantiated. Subsequent studies, designed to evaluate the effectiveness of the P D U R (Clendenning et al., 1976; L a v e & Leinhardt , 1976a), the N M E M C R Q (Brook & Wi l l iams , 1976; Brook et al., 1978), and the C H A M P (Fulcheiro et al., 1978), were more thorough. W i t h the exception of the C H A M P study, however, none of these evaluative efforts were able to demonstrate a reduct ion in length of stay specifically attributable to the introduct ion of concurrent, target based review. C l e n d e n n i n g et al. in their comparison of 11 P D U R and 11 n o n - P D U R hospitals (matched for b e d size, location, teaching status, a n d number of services offered) examined length of stay data o n M e d i c a i d patients, all of w h o m were under the age of 65. Thus , in contrast to the H A S P study, M e d i c a i d patients were used in both the P D U R a n d n o n - P D U R groups. This is an important distinction because, in this study, the M e d i c a i d patients inc luded in the n o n - P D U R group were also subject to review. Th e difference between the two groups, however, was that the P D U R group was subject to the new target-date based concurrent review whereas the n o n - P D U R group was subject to retrospective review a n d retroactive claims denial . 35 In analyz ing the data, to control for differences in case mix, 14 of the most c o m m o n M e d i c a i d diagnoses were examined. Data encompassing eleven months prior to the implementat ion of the P D U R was compared to data encompassing eleven months after implementation (Feb 1, 1972 - Dec. 31,1972 and Feb 1,1973 - Dec. 31,1973, respectively). W i t h the exception of maternity cases (i.e., two of the fourteen diagnoses), no differences in average length of stay could be attributed to the P D U R . The authors concluded, therefore, that, the concurrent review process h a d not been any more effective than the pol icy of retroactive denials in reduc ing length of stay, with the possible exception of maternity cases. In a second evaluation of the P D U R (Lave & Leinhardt , 1976a), essentially the same conclusion was reached. In this study, M e d i c a i d patients in 9 P D U R hospitals were compared to M e d i c a i d patients in 8 non-P D U R hospitals. Th e n o n - P D U R hospitals had all implemented the federally required post-discharge util ization review programs and, in contrast to P D U R hospitals, were subject to retroactive claims denial. T h u s , the p r i m a r y purpose of the evaluation was to determine whether pre-discharge uti l ization review was any more effective than post-discharge review in reduc ing length of hospital stay. In addit ion, comparisons were m a d e between M e d i c a i d patients in the P D U R and n o n - P D U R hospital and Blue Cross patients in the same hospitals. This was done to determine whether the P D U R or post-discharge review programs differed in effectiveness f rom the non-regulated review programs used by private insurers . T o analyze the data, regressions were done us ing institutional characteristics such as occupancy rate, bed capacity, teaching status and 36 percentage of hospital populat ion covered by M e d i c a i d as control variables. A s wel l , the dependent variable, length of stay, was adjusted to control for case mix. The results indicated that under both the P D U R and n o n - P D U R programs, significant reductions in length of stay h a d been incurred over time. Moreover , these reductions were, for the most part, more significant than those observed for Blue Cross patients. In compar ing the P D U R to the n o n - P D U R hospitals, however, no difference in length of stay was observed. D o w n w a r d trends had started prior to the introduction of the P D U R and, therefore, the authors concluded that the observed reductions in length of stay could not be attributed to the post-discharge review program. Instead, it appeared that both pre-discharge and post'discharge programs were equally effective. Consequently , the authors suggested that, prior to further pol icy development , more attention should be given to the determination of wh ich method was more cost-effective and, also, to the ascertainment of the impact of different util ization review strategies on the quality of care p r o v i d e d . In 1978, Brook and his associates publ i shed the findings from a major s tudy that h a d been undertaken to ascertain the impact of the N M E M C R O o n M e d i c a i d hospital costs, use, and quality (Brook et al., 1978). These investigators reviewed data accumulated over a four year per iod (1971-1975) relating to the use of ambulatory services, hospital admissions, hospital days bi l led per 100 eligible persons, average length of stay, hospital costs, and cost savings. In all cases, the pr imary method of analysis used was time series regression. W h e r e possible, comparisons to national trends were used in order to account for secular trends. H o w e v e r , no pre-program baseline was established. Instead, it was argued that because the p r o g r a m h a d been 3 7 implemented s lowly (between September 1971 and M a r c h , 1972) and because there were no observed discontinuities fo l lowing its implementat ion, it was appropriate to compare the later months to the earlier months of the study per iod . T h e major findings of this study are summar ized as follows: 1) W i t h the exception of medical ly reviewed injections, ambulatory service use (Prescriptions, office visits, and laboratory visits) increased over time. In contrast, util ization of reimbursable injections decreased by 75% between 1971 and 1975. The largest decreases were observed among antibiotic injections such as short-acting penici l l in , l incomyc in , ampic i l l in , and tetracycline. It was concluded that, since all of these injections were used pr imar i ly to treat viral infections, the observed "decreases [reflected] cl inically significant improvements in the quality of care" (Brook, et al., 1978, p. 257). The authors attributed these reductions directly to NMEMCRO activities. 2) Nei ther admiss ion certification nor extended stay certification h a d any demonstrable impact o n length of hospital stay (whether adjusted for case mix or not), total hospital days, or the proport ion of stays exceeding the 50th P A S percentile. Moreover , there was evidence to suggest that admiss ion rates increased, rather than decreased, d u r i n g the study per iod. 3) The percentage increase in M e d i c a i d expenditures per eligible beneficiary per year was significantly greater than the percentage increase in services bi l led per eligible person. Consequently , Brook et al. suggested that, even if peer review was successful in curtail ing util ization, it w o u l d be ineffective as a cost containment device because of cost increases associated wi th price inflation, technological innovations, and changes in case mix. 38 4) T h e costs of operating the concurrent medical care review program were between $750,000 and $1,000,000 per year whereas total savings attributed to this process averaged only $700,000 per year. T h u s , the savings incurred d i d not cover the costs of operating the program. Substantial cost savings (approximately $16,000,000) were realized as a result of N M E M C R O administrative activities (eg. d o w n w a r d adjustments of hospital rates, comparisons of claims to fee schedules, and payment denials for administrative or medical reasons). H o w e v e r , none of these activities required professional medical review. A l t h o u g h these findings are interesting, it should be recalled that they refer to the experience of only one state, the State of N e w Mexico . Consequently , it is impossible to say whether similar results w o u l d be obtained in other jurisdictions. In addit ion, the lack of both a pre-program baseline and a comparable control group make the results difficult to interpret, a n d the authors' conclusions open to debate. In the final evaluative study to be considered here, Fulchiero, et al., (1978) compared the length of stay experience of M e d i c a i d patients to that of n o n - M e d i c a i d patients, in the same 57 hospitals. M e d i c a i d hospitalizations were subject to C H A M P review whereas n o n - M e d i c a i d hospitalizations were not reviewed. A total of seven six month periods were studied; two periods before the implementation of C H A M P and five periods after implementation. In each period, the data was standardized for age and case m i x . T h e results of the C H A M P study indicated a consistently decreasing trend in length of stay for M e d i c a i d patients that began in the pre-program per iod and cont inued after the implementat ion period. In compar ing the 3 9 average of all p r e - C H A M P periods to the average of all p o s t - C H A M P periods, the age and case-mix adjusted length of stay for the M e d i c a i d sample decreased by 11.9%. A similar trend was observed for non-Medicare patients, but the magnitude of the decrease was smaller for this group (6.6%). T h e authors concluded, therefore, that a 5.3% differential decrease cou ld be attributed to the C H A M P . Moreover , because the two groups studied were from the same hospitals, the authors suggested that the decrease observed in the non-Medica id group may , at least in part, have been due to an educational effect of the C H A M P that was not l imited to the M e d i c a i d patients. It should be emphasized, however, that this evaluation used non-M e d i c a i d patients for its length of stay comparisons. Since these patients were not subject to systematic review, this study does not address the issue of whether concurrent review is any more effective than post-discharge review. In fact, the observed decrease in length of stay d u r i n g the pre-C H A M P per iod , when only the federal retrospective review program was in effect, suggests that the results attributed to the C H A M P m a y need to be interpreted wi th caution. n.4.3 T h e Transi t ion to PSRO's It should be recalled at this point that the above mentioned studies were performed in response to, rather than as precursors of, the 1972 P S R O legislation. T h u s , the implementat ion of the P S R O review system was already wel l u n d e r w a y by the time evaluators first began to suggest that concurrent uti l ization review, a major component of the P S R O system, was no more effective than the previous ly mandated retrospective review strategy. G i v e n the magnitude of the P S R O program and the substantial 40 costs associated wi th its operation (Burford & A v e r i l l , 1979) it is somewhat surpris ing that a more intensive effort to establish the effectiveness of the program's major components d i d not precede its implementation. Despite the absence of objective data, however, at the time the legislation was passed, conventional w i s d o m held that concurrent review programs w o u l d be effective in reduc ing unnecessary hospital use and, thereby, in containing hospital costs (Sanazaro, 1977). Consequently , the implementation of the p r o g r a m was initiated, and only later d i d questions concerning the effectiveness and cost-effectiveness of concurrent review begin to attract attention. A s initially enacted, the P S R O program, like its prototypes, was intended to have two basic objectives: "1) to assure the quality and 2) to assure the appropriate util ization of health care services" (Goran et al., 1975, p. 2). These objectives were to be met through the implementat ion and integration of three review mechanisms: 1) concurrent review, w h i c h was to inc lude admiss ion and continued stay certification procedures similar to those developed by the prototypes; 2) M C E ' s (medical care evaluation studies), w h i c h were first introduced with the 1966 M e d i c a r e / M e d i c a i d amendments a n d were to continue to be concerned wi th the assessment of the quality of care prov ided; and 3) profile analysis, a new procedure which was to be used to identify potential problems associated wi th either the uti l ization or quality of services prov ided . F r o m the outset, however, the P S R O p r o g r a m was regarded, by the health care community , as pr imari ly a cost containment strategy (Berry, 1977; Dans et a l , 1985; G o r a n , 1979). This perception was reinforced when , in 1977, the responsibility for the PSRO's was transferred from the Publ ic Hea l th Service to the Hea l th Care Financing 41 Admini s tra t ion (Dans et al., 1985). Because of this perception, most P S R O resources were used to implement the labor intensive concurrent review procedures (Burford & A v e r i l l , 1979; G o r a n et al., 1975). M C E ' s , wh ich had been required under the previous federal review system, continued to be deve loped a n d implemented. A s with the previous system, however, they still received relatively less attention than the uti l ization review procedures (Dobson et al., 1978). T h e profile analysis component of the P S R O program was, for the most part, ignored (Dobson et al., 1978; G o r a n , 1979). Thus , throughout the 1970's the implementation of the P S R O p r o g r a m was predominant ly characterized by the development and instigation of admiss ion and extended stay certification procedures. A l t h o u g h the util ization review procedures employed b y the PSRO's were similar to the procedures used b y the earlier state-initiated review systems, the PSRO's themselves had a number of distinctive characteristics. These have been summarized by Dobson et al. (1978) as follows: 1 P S R O s are structured organizations created solely for the purpose of implement ing the law. Previously , uti l ization review h a d been left to i n d i v i d u a l hospitals, wi th supervisory responsibil ity d i v i d e d between state certification survey agencies and fiscal intermediaries. 2 P S R O s are physician-sponsored organizations, in contrast to state agencies or fiscal intermediaries. 3 P S R O m e m o r a n d a of understanding with state agencies and fiscal intermediaries assure that P S R O review decisions o n medical care necessity are binding on the fiscal agent, thereby prec luding any retrospective denial of the c la im by the fiscal agent o n the grounds the stay is not medical ly necessary. Prior to P S R O , util ization review called for decisions by fiscal intermediaries o n all issues of patient eligibility, coverage, a n d medica l necessity. 4 P S R O implementat ion has been formally guided through a structure of transmittals, support centers, a n d technical 42 assistance. This has tended to produce greater uniformity than other uti l ization strategies. 5 P S R O concurrent review employs un i form norms and standards established on an areawide basis. In contrast, norms for other forms of concurrent review have varied from hospital to hospital, (p. 117). T h u s , the P S R O system represented a major effort o n the part of the federal government to develop a "truly national uti l ization review program" (Dans et al. , 1985, p. 1131). W i t h i n this national system, however, each P S R O was to develop its o w n standards and criteria for determining appropriate and necessary hospital use (Berry, 1977). Thus , although it was h o p e d that the P S R O program w o u l d reduce the observed variabil ity in regional length of stay patterns, no specific targets were set for the reduction of excessive hospital use (Goran, 1979). II.4.4 Major Characteristics of P S R O Review U.4.4.1 A d m i s s i o n Certification U n d e r the P S R O mandate, all Medicare a n d M e d i c a i d admissions were to be reviewed to assure that a hospital level of care was medical ly necessary. Each admiss ion deemed appropriate after P S R O review was then certified and assigned an initial length of stay. In most instances, the admiss ion certification procedure was performed by a specially trained review coordinator. Typica l ly , the reviewer was a registered nurse w h o was h ired by the P S R O to conduct reviews in a particular hospital. Admiss ions deemed appropriate cou ld be certified by the review coordinator without P S R O consultation. H o w e v e r , decisions concerning questionable admissions cou ld only be made a P S R O physic ian consultant. In general, the criteria used to determine the appropriateness of admissions cou ld be of two types: "1) criteria specific to a particular problem, 4 3 diagnosis, or procedure, and 2 ) criteria which specify the types of services w h i c h should be prov ided at a hospital level of care" (Goran et al., 1975, p. 9). The actual criteria to be used were to be developed by each P S R O and appl ied , o n a u n i f o r m basis, to each hospital wi thin the PSRO's jurisdiction. O n c e a patient's admiss ion was certified, the assignment of the initial length of stay checkpoint could be made on the basis of 1) regional length of stay norms for patients wi th similar characteristics such as age and diagnosis, or 2 ) hospital level of care criteria or standards, as defined by the P S R O (Goran et al., 1975). However , regional norms, such as those p r o v i d e d by P A S , were to be used whenever possible and, in general, the initial checkpoint was not to exceed the 50th percentile of such norms. n . 4 . 4 . 2 C o n t i n u e d Stay Certification C o n t i n u e d stay review was to be performed on all patients still in the hospital on the assigned target date. It was designed to assure the necessity of an extended stay and, like the admiss ion certification procedure, was to be performed by the review coordinator. The certification of an extended stay was to be based o n either hospital level of care criteria, P S R O defined discharge criteria, or diagnosis specific criteria (Goran et al., 1975). W h e n continued stay review indicated that an extended stay was necessary the review coordinator was to certify the extension and assign a new length of stay checkpoint. In questionable cases, the decision to certify or deny an addit ional stay was to be made by the P S R O physician consultant. This process of cont inued stay review w o u l d then be repeated each time the length of stay checkpoint was reached unti l the patient was ready for discharge. 44 W i t h the implementation of the two components of the concurrent review process, admission certification and continued stay certification, retroactive denials of claims were eliminated. The certification process was to be binding on the fiscal intermediary and, therefore, was to ensure that, once certified, the claims of eligible beneficiaries for services covered under Medicare and Medicaid would be paid (Dobson et al., 1978). For the most part, conclusions about the effectiveness of the PSRO strategy in reducing the utilization and costs of publicly financed hospital care were based on the results of the studies which evaluated the PSRO prototype programs. However, in accordance wi th a requirement of the 1972 PSRO legislation, a formal evaluation of the PSRO program was undertaken, in 1977, by the US Department of Health, Education and Welfare (DHEW). According to Dobson et al. (1978), this study had three major objectives. First, it was to document the status of PSRO implementation. Second, it was to determine whether the concurrent review procedures employed in the PSRO program were more effective than other forms of review (retrospective and prototype) in reducing length of hospital stay. Third , it was to determine what impact the PSRO review system was having on health care expenditures. H.4.5. The D H E W Evaluation of PSRO Review 1 JJ.4.5.1 Status of PSRO Program Development As noted by Dobson et al. (1978), the planning, development and funding of PSRO programs was based on a three stage process. During the The DHEW evaluation was summarized in a paper published by Dobson, et. al. (1978). All of the material contained in this section (II.4.5.1 - II.4.5.3) is therefore based on the Dobson et. al. publication. 4 5 initial phase, w h e n review systems and organizational structures were being developed, the P S R O was considered to be in the 'planning stage'. Once the necessary systems and structures were in place, the P S R O was considered to be in the 'conditional stage'. The 'conditional stage' was to last a m i n i m u m of 24 months and was designed to ensure that the P S R O was operationally viable. A t the end of the 'conditional stage', by which time the P S R O was expected to be performing both short-term and long-term care review, the P S R O was to be designated as 'operational'. U n d e r the federal mandate, a total of 203 geographically defined areas were identified for P S R O d e v e l o p m e n t . A s indicated by Dobson et al. (1978), when the D H E W ' s evaluation of the P S R O system was undertaken in 1977, 108 PSRO's were operating at the operational stage and 64 were in the p lanning stage. N o n e h a d as yet been designated fully operational. The implementation of review activities had typically begun wi th 100% concurrent review in short-term hospitals. M C E studies were implemented next, usual ly about one year after the concurrent review procedures. Profile analyses, on the other hand , had not yet received m u c h attention. E.4.5.2 Effectiveness of P S R O Review A c c o r d i n g to Dobson et al. (1978), several different strategies were e m p l o y e d in the effort to ascertain the effectiveness of P S R O concurrent review in reduc ing hospital util ization. O n e approach invo lved the assessment of the impact of P S R O review on three measures of utilization: 1) days of care per thousand Medicare patients, 2) admissions per thousand Medicare patients, and 3) average length of stay of Medicare patients. For these analyses, hospital util ization data from 18 active P S R O areas were 46 compared with data from 26 non-active areas. In order to isolate the effects of P S R O review, active and non-active PSRO's were matched o n 15 demographic and health system variables (the actual variables used for matching were not delineated). In addit ion, statistical adjustments were made to control for: baseline util ization levels, populat ion density, percent of families wi th income under $5000 per a n n u m , hospital beds per 100,000 populat ion , weighted P S R O - w i d e hospital occupancy rates, and prior M e d i c a i d concurrent review. - The calendar year 1974 was used to establish the baseline uti l ization rates for both active and non-active areas; data from the calendar year 1976 were used to establish utilization rates in the post-P S R O period. Data from the baseline per iod was used to forecast expected util ization rates and these expected rates were compared to the actual rates observed in 1976. The database used to perform the analyses was derived from M e d i c a r e claims files. O n an aggregate level, no significant differences were found between the active and non-active areas on any of the three uti l ization variables (days of care, admissions, length of stay). H o w e v e r , a significant interaction was observed between bed capacity and P S R O impact. Greater reductions in days of care per thousand Medicare patients were observed among PSRO's wi th a higher bed capacity than among those with fewer beds per thousand populat ion . This interaction could , perhaps, be explained through appl icat ion of Roemer's L a w . That is, pr ior to P S R O review, hospitals with greater bed capacity m a y have been p r o v i d i n g a larger amount of discretionary care. If so, then wi th the advent of P S R O review, these hospitals w o u l d have h a d more discretionary days of care available for e l iminat ion than the hospitals wi th more l imited bed capacity. 4 7 When the data were disaggregated by active PSRO area, 7 of the 18 active PSRO's were found to be more effective in reducing hospital utilization than were their matched comparisons. The other 11 areas showed either no systematic variation in utilization rates or higher utilization rates than their matched comparison. In addition, although Dobson et al. (1978) did not provide any data, they reported that "of the seven PSRO's showing lower actual rates of days of care than expected, six had imputed dollar savings greater than the incremental costs of PSRO review, i.e., six were found to be cost-effective." (p. 119). These results were not given much weight, however, as the overall proportion of programs exhibiting an effect was no different than would be expected by chance. Although one of the main purposes of the above analyses was to ascertain the effectiveness of the concurrent utilization review component of the PSRO program, this objective was hampered by the fact that even in the non-active PSRO areas some forms of concurrent review were being employed. Since no data were available which would enable ascertainment of the extent to which concurrent review was being used in non-active PSRO areas, Dobson et al. (1978) maintained that the results concerning concurrent review were inconclusive. In the comparison of PSRO versus non-PSRO review, however, it was concluded that no evidence of a differential effect had been found. The second approach used in the evaluation involved an analysis of 1.19 million hospital records derived primarily from the PAS discharge abstract data service. For this analysis, 92 PSRO hospitals were compared to 42 non-PSRO hospitals. The two groups of hospitals were examined for differences in length of hospital stay and case mix severity for both Medicare 4 8 a n d M e d i c a i d patients. In performing the length of stay analyses, the data were adjusted to control for case mix, geographic region, major hospital and medical resources characteristics. For the case mix analysis, length of stay trends in the P S R O and n o n - P S R O groups were observed for patients matched on diagnosis, age, number of diagnoses, and whether or not a surgical procedure took place. In addition, the severity of case mix "was examined using the SysteMetrics 'Staging Methodology' for 23 specific diagnoses" (Dobson et al., 1978, p. 121). T h e results of the length of stay comparisons indicated that, for Medicare patients under P S R O review, average length of stay fell by .23 of a day more than for Medicare patients not under P S R O review. Conversely , it was found that for M e d i c a i d patients under review the average length of stay fell, but by .20 of a day less than for M e d i c a i d patients not under review. T h e combined impact of P S R O review, for both Medicare and M e d i c a i d patients, was a .09 day reduction. Final ly , a significant (p_<.05) decrease in the percentage of Medicare patients staying beyond the 75th percentile of matched comparisons in the P A S data base was associated with P S R O review. A similar but nonsignificant decrease was observed for M e d i c a i d patients. Since the actual percentage decreases were not reported, however, it is impossible to tell whether the observed reductions represented a meaningfu l effect. W i t h respect to the case mix analysis, it was observed that in both P S R O a n d n o n - P S R O hospitals there was a trend towards increasing severity of case mix. This observation was based on the f inding that expected lengths of stay for all matched patient groups increased over time. H o w e v e r , using a severity of illness staging method developed by Systemetrics, Inc., it was 4 9 observed that among Medicaid patients, the trend towards increasing case mix severity was more pronounced in the PSRO than the non-PSRO hospitals. Problems with the reliability of the severity index and the coding of diagnostic data, however, rendered these findings suggestive rather than conclusive. 13.4.5.3 Impact of PSRO Review on Health Care Expenditures On the basis of the utilization data, Dobson et al., concluded that, at best, PSRO review had produced only minimal reductions in hospital utilization rates. As one of the major implicit goals of the PSRO system was cost containment, however, the next step taken in the evaluation was to determine 1) how the PSRO system could be expected to affect expenditures and 2) what influences it would have to have on utilization in order to be considered an effective cost containment device. In addition, given that one of the primary assumptions prompting the development of the PSRO system was that inappropriate hospital utilization was at least in part responsible for increasing expenditures, an attempt was made to determine what impact PSRO review had on the appropriateness of inpatient care. Finally, an effort was made to estimate the costs associated with PSRO review and to perform a cost-benefit analysis. In considering the extent to which the PSRO system could be expected to affect health care expenditures, Dobson et al. list six factors over which the system would have to be able to exert some control over in order to have a significant effect on costs. These factors include: 1) Price of materials and wages; 2) Rate increase in complexity of health care services provided due to technological advancement and shifts in medical care practice; 5 0 3) T h i r d party coverage patterns and controls on reimbursable costs; 4) Populat ion (size, age, distribution, and health status); 5) Rate of increase in the supply of health care resources caused by increased fixed costs to be amortized and induced by demand; 6) Ut i l izat ion and case mix. (Adapted from Dobson et al., 1978, p. 122). The P S R O system, however, had no legislated mandate to exert control over any of the these factors, except utilization. Consequently , w h e n v iewed in the context of all of the factors that influence hospital expenditures, the impact of P S R O review on uti l ization w o u l d have to have been considerable in order for it to perform effectively as a cost containment strategy. Since the observed impact o n util ization was negligible, the evaluators suggested that the P S R O program lacked leverage as a cost containment device. T o support this contention, they p r o v i d e d a break d o w n of the unit costs associated wi th P S R O review, national p r o g r a m costs, and incremental costs associated wi th the 'new' concurrent versus the 'old' retrospective review systems. F r o m these data they estimated the amount by w h i c h util ization w o u l d have to be reduced in order for costs to contained. In comput ing these estimates, two sets of assumptions were used: 1) that a system invo lv ing 100% concurrent review was employed and 2) that a 20% targeted review system was employed. U n d e r a targeted review system, M C E and profile analysis data w o u l d be used to identify potential problem cases, diagnoses, or areas. T h u s , rather than reviewing all cases, only those cases most likely to be problematic w o u l d be subject to concurrent review. The decision, by Dobson et al., to use estimates based on 51 an assumption of 20% of cases reviewed was arbitrary and was intended only to prov ide a basis for comparison. U s i n g cost data constructed from hospital, P S R O , Medicare , and M e d i c a i d accounting records, the unit costs associated with P S R O review were estimated to be $16.10 (+- $.77) per eligible beneficiary. The estimated national costs for a fully implemented 100% P R S O review system in 1976 were $268.5 mi l l i on for all eligible beneficiaries. A s s u m i n g only 20% review, the estimated costs for 1976 w o u l d have been $184 mi l l ion. In comput ing the incremental costs associated with 100% P S R O review, the P S R O system was estimated to cost $187.2 mi l l ion more than the previous retrospective system. A s s u m i n g 20% review, the incremental costs associated wi th P S R O review were estimated to be $76.4 mi l l ion. A l t h o u g h one of the major intents of the D H E W evaluation was to compute a cost-benefit ratio for the 18 active P S R O areas, the lack of a significant reduction in uti l ization rates (i.e., benefit) prevented the pursuit of this objective. Instead, two alternate approaches were taken. First, a break-even analysis was performed on the aggregate data. Second, cost-benefit ratios were computed for each of the seven i n d i v i d u a l PSRO's in w h i c h reductions were observed. In the break-even analysis, it was determined that a 100% review p r o g r a m w o u l d pay for itself if it could achieve and maintain reductions in uti l ization rates of between 1.6% and 2.05%. A 20% review program w o u l d pay for itself if reductions o n the order of .65% to .83% could be maintained. Recall , however, that the aggregate reduction observed from 100% review was on ly .09%. Therefore, in aggregate, it was concluded that, in 1976, the system was not breaking even. 5 2 The cost-benefit analysis of the i n d i v i d u a l PSRO's that had effectively reduced uti l ization revealed that for six of these seven PSRO's the cost-benefit ratio was greater than one. In estimating the impact of these achievements on health care expenditures, however, it was determined that if all of the PSRO's had been able to achieve cost-benefit ratios of 2 the reduct ion in expenditures w o u l d have been less than 1%. Since none of the PSRO's had ratios as h igh as 2, the effect on expenditures w o u l d have been considerably less than 1%. Moreover , in order to achieve ratios of 2 it was determined that PSRO's w o u l d have to generate uti l ization rate reductions on the magnitude of 3.2% to 4.1%. N o n e of the PSRO's inc luded in this evaluation were found to achieve such levels of reduction. T h e final question addressed in the D H E W evaluation was that of the impact of P S R O review on the appropriateness of hospital care. A t the time the evaluation was performed, however, no empir ical ly der ived measures of appropriateness were available. In addit ion, Dobson et al. indicated that atterripts to define measures that could serve as proxies for identifying changes in appropriateness of care were frustrated by the nature of the available data. Consequently, it was stated that no conclusions could be d r a w n wi th respect to whether the reductions in uti l ization observed in some PSRO's were the result of decreases in inappropriate use or, alternatively, the result of a failure to prov ide necessary care. It was suggested, therefore, that m u c h addit ional work was needed in terms of def ining a n d measuring the appropriateness of hospital uti l ization. H.4.6 Other PSRO-Related Research A l t h o u g h the D H E W study was the most comprehensive evaluation of the P S R O review system, other studies performed in the late 1970's 5 3 tended to confirm its findings. A v e r i l l and M c M a h o n (1977), for example, deve loped a theoretical m o d e l of the Cont inued Stay Certification (CSC) process a n d used it to test the potential cost-effectiveness of the system. In deve lop ing the model , a number of different assumptions were made about the distribution of length of stay, the length of delay between the time of C S C review a n d implementat ion of appropriate action, the proport ion of patients w h o experience unnecessary days of hospitalization, the number of days of unnecessary hospitalization, and the distribution of unnecessary hospital stays as a function of length of stay. A l t h o u g h the authors conceded that the results of their analysis were vulnerable to criticism on the grounds that they were based on theoretically derived, rather than empirical ly der ived, assumptions, they maintained that a lack of empir ical data necessitated the acceptance of theoretical assumptions. T o compensate for this, whenever an assumption was required, a sensitivity analysis was performed to test over the range of possible and reasonable values. In general, the basic steps in the mode l l ing procedure were invoked as follows. First, the number of patients with each possible length of stay was calculated. Second, the number of patients that w o u l d be reviewed was calculated. T h i r d , the number of patients not discharged before the results of the review cou ld be implemented was calculated. Fourth , the proport ion of patients f rom step three w h o w o u l d experience unnecessary hospital days was calculated. Fifth, the proport ion of patients from step four-who w o u l d experience a sufficiently large number of unnecessary days of hospital care, relative to their length of stay, to be unnecessarily hospital ized at the time of the review was calculated. Sixth, the ratio of the number of patients in step five to the total number of patients reviewed was calculated. This ratio, 5 4 w h i c h the authors referred to as the Y i e l d from Review (YR), p r o v i d e d the fraction of patients w h o could have their length of stay directly affected by C S C review. T h e n , by estimating the cost of a C S C review and using the Y R computed in step six, the financial benefits that w o u l d have to accrue in order for the process to be cost-effective could be determined. U s i n g length of stay data for all patients discharged from the Yale-N e w H a v e n hospital between A p r i l 1971 and September 1973, the authors determined that, for each case of inappropriate use discovered, savings of between $500 and $10,000 w o u l d have to be realized in order for the C S C process to break even. O n average, the amount of savings required per case was $2000. The authors concluded, therefore, that as it was unl ikely that any reductions in length of stay attributed to C S C w o u l d produce such savings, the cost of the review process w o u l d most likely outweigh its benefits. T h e y then suggested that the C S C process be de-emphasized and that more attention be p a i d to the development of less costly review mechanisms such as Profile Analys i s and M e d i c a l Care Evaluat ion Studies. H i r s c h , R i m m , and Welsch (1977) expressed concerns that the P S R O strategy for cont inued stay review d i d not take into consideration whether extension requests could have been avo ided through the provis ion of better care or better p lanning (eg. shorter preoperative stays). In addit ion, they were concerned that the use of diagnoses to establish length of stay norms might not be the most effective and efficient w a y of relating the reason for hospitalization to the process of care. Consequently, they proposed, as an alternative to diagnoses der ived norms, the use of norms based on the Reason for A d m i s s i o n ( R F A ) . 5 5 The PvFA categories used in the study were developed by H i r s c h et al. w i th the help of medical records librarians from four communi ty hospitals. A total of 2000 consecutive l ive adult discharges were categorized according to the reason for the admission. A s well , the length of stay associated with each discharge was recorded. The reason for admission was to reflect specifically w h y the physic ian admitted the patient to the hospital. Orig ina l ly , 100 categories were defined but these were subsequently reduced to a final set of 25 PvFA categories. Example R F A ' s include M i n o r Surgery (with subgroups defined), Major Elective Surgery (with subgroups) , Major Emergency Surgery (with subgroups), Acute Nonsurg ica l Illness (with subgroups) , Management of C h r o n i c Illness, Psychiatric, Support ive and T e r m i n a l , and W o r k u p s . The length of stay norms assigned to each R F A category were computed from the length of stay data collected from the 2000 records used to define the categories. For most R F A categories, the modal length of stay was used as the norm. For the rest, the n o r m was taken as the mean or, in some cases, it was based on special considerations (eg. empir ica l ly defined norms found in the medical literature). Unfortunately , no attempt was made to ascertain the reliability or val idity of the R F A categories. Furthermore, as noted by the authors, the R F A designation was arbitrary and based on a small and geographically contained sample. T h e p r i m a r y purpose of R F A review was to identify three types of patient charts: 1) charts with a complication or other patient factor (P) indicat ing that length of stay could not be reduced without adversely affecting the patient's health; 2) charts indicating overuti l ization (U) of hospital resources; and 3) charts indicating a delay in discharge due to less than opt imal quality (Q) of care. T h e purpose of the study was to compare 5 6 the efficiency of the R F A strategy with that of the diagnosis-based strategy in identifying patients whose length of stay could have been reduced without adversely affecting the quality of care. T o accomplish this objective, 500 hospital charts were categorized by R F A and then screened according to the R F A length of stay norms. O f these, 106 charts were found to have lengths of stay that exceeded the established norms. F r o m the remaining 394 charts, 106 that d i d not exceed the norms were selected to serve as controls. The two groups of charts were then screened according to diagnosis-based norms using both the 50th and 75th percentiles as target dates for review. T h e charts were also subjected to i n d i v i d u a l review by the hospital's medical director and the senior author of the research paper. Th e purpose of these reviews was to categorize the charts according to the P, U , and Q , designations. Each designation decision represented a consensus on the part of the two reviewers. T h e results of the study indicated that the use of R F A and diagnosis der ived norms were equally effective in identifying hospital episodes that cou ld have been shortened without adversely affecting quality. However , the use of diagnostic norms selected out 76% more charts for i n d i v i d u a l review than d i d the use of R F A norms. Therefore, a l though recognizing the l imitations of the study methods, the authors suggested that the use of diagnosis determined norms may not be the most efficient method of continued stay review and that the R F A procedure might prov ide a viable alternative. In another study concerned about the relevance of diagnosis as the basis for establishing norms for continued stay review, Burford and A v e r i l l (1979) examined the relationship between diagnostic information available 5 7 at admission and discharge. These authors were interested in determining to what extent incorrect assignments, made on the basis of information available at admission, would affect the cost-effectiveness of the continued stay certification (CSC) process. In particular, it was hypothesized that length of stay norms applied at time of admission would result in one of the following situations: 1) Review conducted at the correct time, 2) Review conducted earlier than it should have been, 3) Review conducted later than it should have been. In addition, it was predicted that there would be errors of inclusion and omission. Errors of inclusion would occur in situations where patients who were reviewed too early, would not have been subject to review at all, had the appropriate norms been employed. Errors of omission would occur among the patients who were discharged before being reviewed because the review date established on admission was later than it should have been. The database used for the study was the PSRO Hospital Discharge Data Set. The study data set consisted of 52,210 patient records collected during the ten month period ending June 1976 for 68 hospitals in one PSRO area. A comparison of admitting and discharge diagnoses revealed that, of the 52,210 records, 41.3% had a change in primary diagnosis between admission and discharge. Furthermore, although the majority of the 52,210 patients were found to have been reviewed on time (58.5%), 27.6% were reviewed earlier than they should have been, on the basis of their discharge diagnosis, and 13.8% were reviewed late. With respect to errors of inclusion and omission, 6.6% of the 52,210 patients were unnecessarily reviewed, and 2.6% were inappropriately excluded from review. Thus, in total, 41.5% of the patient 5 8 records examined were found to have problems in the t iming and performance of C S C review. The authors concluded, therefore, that since early and unnecessary reviews increase the number of unproduct ive reviews, a n d missed and late reviews decrease the potential for length of stay reductions, all such problems w o u l d decrease the effectiveness of C S C review. Because they also found that some diagnostic categories were more variable than others, they suggested that, for those categories, alternative methods for initiating the C S C review process should be explored. They indicated that one possibility w o u l d be to set a routine time per iod that w o u l d prov ide time for diagnostic work-up . Once the work-up was. completed, the diagnosis could be confirmed and an appropriate length of stay n o r m assigned. H o w e v e r , as this w o u l d result in an extra review for m a n y patients it w o u l d likely increase the costs of the C S C process. Burford and A v e r i l l (1979), not ing that the cost-effectiveness of concurrent review was already questionable, concluded that it might be more product ive to l imit C S C review to those diagnostic categories identified as problematic through the less costly mechanism of Profile Analys is . Boaz (1979), o n the other hand , found little evidence to support the use of the C S C process at all. In a study designed to explore the implications for hospital uti l ization of p r o v i d i n g medical care in the most appropriate (i.e., least cost) setting, this investigator concluded that pre-admiss ion certification was more l ikely than C S C to have an overall impact o n hospital uti l ization. This conclusion was based o n a sample of 8, 936 nonmaternity patients discharged from 60 short stay general hospitals d u r i n g the year ended o n June 30, 1970. U s i n g this data, a regression mode l incorporating medical , social, and economic variables was developed in an attempt to 5 9 measure the differential effects of these variables o n p r e - P S R O length of hospital stay. Since one of the major objectives of P S R O review was to eliminate hospital uti l ization attributable to the systematic influences of nonmedica l factors, Boaz argued that the impact of P S R O review on p r o v i d i n g care in the most appropriate setting could be assessed through the determination of the magnitudes of the effects of social and economic factors on hospital episodes. She also attempted to predict the impact o n hospital util ization of a pre-admiss ion certification process that w o u l d screen out all patients w h o could be treated effectively in an ambulatory setting. Final ly , she attempted to determine to what extent hospital stays w o u l d be shortened if C S C restricted all hospital stays to the "critical m i n i m u m required by medical standards" (p. 317). Three regression models , each using length of stay as the dependent variable a n d a variety of clinical, economic, and social variables as predictors, were defined: 1) for all patients, 2) for urgent admissions, and 3) for nonurgent admissions. The overall predictive power of each mode l was small but significant (R 2=.22, .21, .22, respectively; p_<.0001 in all cases). In all three models , the medical variables (eg. type of admiss ion, number of diagnoses, discharge status, and surgical status) and age were more strongly associated wi th length of stay than were the social (eg. l i v ing arrangements, occupation) and economic (eg. payment source) variables. W h e r e effects of social a n d economic variables were observed, they tended to be concentrated a m o n g patients whose diagnoses w o u l d have been unl ikely to qualify them for hospital ization under a preadmiss ion screening process. Because of these f indings, the author concluded that C S C w o u l d be unl ike ly to have an appreciable effect o n length of stay. She suggested instead that future P S R O 6 0 activity should focus on the pre-admission screening process as this w o u l d be more l ikely to lead to substantive reductions in hospital utilization. In addit ion, she suggested that future evaluations of continued stay review should place less emphasis o n measuring reductions in length of stay, and more on determining to what extent the review process effects changes in hospital case mix. In particular, she indicated that a more relevant measure of the procedure's success w o u l d be the extent to which hospital case mix shifts towards those diagnoses that, by any criteria, can only be treated in an acute care hospital setting. N o n e of the above studies employed a r a n d o m i z e d controlled trial design and, therefore, they cannot be said to have p r o v i d e d a definitive answer to the question of P S R O effectiveness or cost-effectiveness. H o w e v e r , the accumulat ion of evidence from these as wel l as the p r e - P S R O p r o g r a m studies was sufficient to motivate the federal government to begin questioning the efficiency wi th which the program was being del ivered (Sanazaro, 1977). Observ ing that some PSRO's had been more effective than others in reduc ing uti l ization (Dobson et.al.,1978), the government attempted to impose greater efficiency in the operation of P S R O review by reduc ing fund ing levels and by requiring that PSRO's begin to develop measures b y w h i c h to demonstrate that the p r o g r a m was hav ing a positive and documentable impact (Adler & Milste in , 1983; B r o w n & L e v y , 1980). T h u s , b y 1979 the average reimbursement rate for P S R O review was reduced from $12 to $8.70 per inpatient admission and considerable pressure was exerted o n the PSRO's to m o v e away from the labour intensive and costly mechanism of 100% concurrent review and to begin to f ind ways of 61 focussing the review process on the cases and/or hospitals that were most likely to be problematic (Goran, 1979; Mullin, 1983). One of the most frequently suggested methods of improving PSRO efficiency and cost-effectiveness was through the expanded use of Profile Analysis (Averill & McMahon, 1977; Burford, & Averill, 1979; Dobson et al., 1978; Goran, 1979; Mullin, 1983). Profile Analysis is a form of retrospective review in which aggregated hospital, practitioner, and patient care data are subject to pattern analysis (Goran, 1979). As mentioned earlier, from the inception of the PSRO review system, Profile Analysis had been included as one of the system's three major components. It failed to gain much attention in the early years, however, because of a lack of appropriate data. As noted by Goran (1979), it took "a considerable amount of time and effort for PSRO's to establish data collection systems, procure data processing services and collect information for a long enough period to develop profiles of practice, all of which are prerequisites to profile analysis" (p.17), In addition, a standard method of defining and measuring hospital case mix was required in order to assure that comparisons based on the development of institutional and patient profiles would be appropriate and valid (Dobson et al., 1978; Doremus, 1980; Goran, 1979). As noted in an earlier section, hospital case mix had long been acknowledged as a major factor contributing to patterns of hospital utilization. Until the late 1970's, however, a case mix measurement system applicable to any hospital, region, or PSRO population had not been developed (Doremus, 1980; McMahon, 1984). Consequently, in order to comply with PSRO regulations many hospitals and PSRO's developed length of stay norms using diagnostic categories as defined by the 6 2 International Classification of Diseases (ICD) or P A S . Both of these classification systems were criticized, however, for their failure to categorize patients into homogeneous groups defined i n terms of hospital resource consumpt ion (Fetter, Shin, Freeman, A v e r i l l , & T h o m p s o n , 1980; Jenkins & Cole , 1984; M c M a h o n , 1984). Thus , there was concern that attempts to compare uti l ization patterns across hospitals and PSRO's w o u l d be inval id; i.e., that any observed variability in util ization w o u l d be due to differences in case mix rather than patient management. Similar concerns were expressed by researchers w h o were attempting to study and m o d e l hospital uti l ization as wel l as by those w h o were attempting to study relationships between hospital util ization and costs (Lave & Leinhardt , 1976b; L u k e , 1979; M c M a h o n , 1984). In response to such concerns, a group of researchers from Yale Univers i ty (Fetter et al., 1980) began work ing on a patient classification system called Diagnosis Related G r o u p s (DRG's). O n the basis of research initiated in the early 1970's, this group of investigators sought to develop a management tool that w o u l d relate "the demographic , diagnostic and therapeutic characteristics of patients to the output they are p r o v i d e d so that cases are differentiated by only those variables related to the condit ion of the patient (eg. age, pr imary diagnosis) and treatment process (eg. operations) that affect his uti l ization of the hospital's facilities" (Fetter et al., 1980, p. 2). In other words , the a im was to produce a system of categories wi th in which the observed variability in util ization w o u l d be pr imari ly attributable to differential management of similar patient types rather than to differences in case mix ( M c M a h o n , 1984; Y o u n g , Swinkola , & Hut ton , 1980). If successful, this system cou ld then be employed as a tool for util ization 6 3 review, uti l ization research and, ultimately, for hospital p lanning , budget ing, and cost control (Fetter et al., 1980; M c M a h o n , 1984). n.4.7 The Deve lopment of D R G ' s A s described by Fetter et al. (1980), the pr imary purpose of the D R G approach to patient classification was to "identify in the hospital acute-care setting a set of case types, each representing a class of patients with similar processes of care and a predictable package of services (or product) from .an institution" (p.3). In addit ion, the D R G developers wanted to ensure that the case type definitions w o u l d be meaningful to both medical and non-medical users and easy to implement in a variety of settings. T o achieve these objectives, the D R G system was developed in accordance with the fo l lowing criteria: 1. Class definitions must be medical ly meaningful . 2. Class types should be defined o n the basis of variables that are relevant to output util ization and readily available on hospital abstracts. 3. T h e number of classes must be manageable as wel l as mutual ly exclusive a n d exhaustive. 4. T h e classes should contain patients with similar expected measures of output uti l ization. (Adapted from Fetter et. al., 1980, p. 5). U.4.7.1 T h e Construct ion of D R G ' s T h e construction of the initial set of D R G ' s invo lved two steps. In the first step, a committee of physicians was asked to identify diagnostic groups that were 1) consistent in terms of their anatomic, physiopathologic classification, or in their clinical management, 2) mutual ly exclusive and exhaustive, and 3) representative of a sufficient number of patients (Fetter et al. , 1980). F r o m this effort, a total of 83 Major Diagnostic Categories (MDC's) were defined. The diagnostic coding system chosen as the standard for the 64 construction of the D R G system was the I C D A 8 1 . However , since diagnostic data in the database were coded using different classification schemes (eg. P A S , I C D A 8 , and H I C D A 2 2 ) , this step also invo lved translating the various coding systems into their I C D A 8 equivalent. In the second step, an iterative statistical analysis was undertaken to determine whether any of the 83 M D C ' s cou ld be further subdiv ided into groups that w o u l d add to the homogeneity of the classification scheme. T h e database used to construct the initial D R G ' s consisted of 500,000 hospital records from 118 hospitals in N e w Jersey, 150,000 records from one Connecticut hospital, and 52,000 records of federally funded patients from 50 hospitals in one P S R O region (South Carolina). Once the 83 M D C ' s had been defined, an interactive computer system called A U T O G R P was e m p l o y e d to conduct the subgroup partitioning. This system was designed to facilitate r a p i d analysis of large volumes of complex data and, through the use of the C L A S S I F Y facility, p r o v i d e d an algorithm for determining partitions o n the basis of interactively specified clinical and statistical criteria (Fetter et al . , 1980). The uti l ization output measure (i.e., the dependent variable) used in the analysis was length of stay. This measure was selected because of its historical acceptance as a proxy measure of resource consumpt ion , its easy availability, and its reliability. T h e final set 3 of input variables (i.e., independent variables) used in defining the class types International Classification of Diseases, Adapted for Use in the United States, Eighth Revision. Hospital Adaptation of ICDA, Second Edition. Other variables such as sex, tertiary diagnosis and tertiary surgical procedure were examined but were not found to be important in explaining output utilization (i.e., length of stay). 6 5 inc luded pr inc ipal diagnosis, presence of secondary diagnosis, principal surgical procedure, secondary surgical procedure, and age. T h u s , wi thin each M D C , the C L A S S I F Y algorithm was invoked to identify groups of patients that differed with respect to length of stay. The various independent variables were examined and weighed simultaneously in each analysis. The decision to accept, reject, or modify a suggested subgroup was based on both the statistical evidence and the clinical committee's medical knowledge. The iterative process continued unti l the introduction of addit ional variables d i d not significantly reduce the variance in length of stay or the subgroup was too small to warrant further partit ioning (i.e., w h e n there were fewer than 100 observations in the subgroup). Th e result of the analysis was the definition of 383 D R G ' s which the system's developers c laimed were clinically interpretable, reflected similar patterns of util ization, made use of routinely collected data, and were mutual ly exclusive and exhaustive (Fetter et al., 1980). A n example of the parti t ioning of one M D C is shown in Figure 1. U.4.7.2 Implications for Ut i l izat ion Rev iew W i t h the development of D R G ' s , Fetter et al. (1980) maintained that they h a d p r o v i d e d an "effective mechanism [for use] in profile analysis by p r o v i d i n g a structure of consistent patient-class definitions wi th in which institutional performance [could] be compared based o n similar types of patients" (p. 27). T h e y thus believed that wi th the accumulat ion of sufficient data, D R G ' s w o u l d permit the development of a more timely and cost-effective system of retrospective review than the costly P S R O concurrent review process that was in use at the time. Hospitals or PSRO's w o u l d be able to monitor patterns of uti l ization over time and , through the 6 6 F I G U R E 1. Subgroup Partitioning of M D C 55 - U r i n a r y Calculus Major Diagnostic Category 55: Urinary Calculus Urinary Calculus Without Suigery Urinary Calculus -no suigery -no secondary diagnosis Urinary Calculus -no suigery -with secondary diagnosis Urinary Calculus with cystoscopy passage of catheter to kidney, other minor operations Urinary Calculus with nephrotomy, cystotomy; ureter-otomy, other major operations (Source: Adapted from Fetter et a l . , 1980) use of standard statistical techniques, identify the occurrence of deviant cases. Once identified, these "outliers" could then be investigated to determine whether they were the result of errors, unusual events in the ut i l izat ion patterns of one or more patients, an overal l change in treatment procedures, or misuti l ization. Moreover , by us ing D R G ' s to adjust for case mix differences a m o n g hospitals a n d / o r PSRO's , inter-organizational comparisons of uti l ization patterns w o u l d be facilitated (Fetter, et al., 1980). E.4.7.3 Initial Reactions to D R G ' s Initial reactions to the D R G classification system were mixed. It was generally agreed that the D R G system was an improvement over P A S and I C D taxonomies (Berki et al., 1984; Doremus , 1980; Iglehart, 1982b; Simborg, 6 7 1981). H o w e v e r , it was also recognized that the system suffered from a number of significant limitations. O f particular concern was the fact that the D R G ' s were developed using data from only N e w Jersey, Connecticut, and South Caro l ina (Doremus, 1980; Y o u n g et al., 1980). A s noted by Doremus (1980), variations in length of stay by geographic region had long been observed and documented. T h u s , the geographically l imited data set used by the D R G developers led to questions concerning the external val idity of the partit ioning process. T o investigate this issue, Y o u n g et al. (1980) appl ied the D R G grouping process on data from Western Pennsylvania. Starting wi th approximately 690,000 patient records, a prel iminary analysis was undertaken to identify 10 M D C ' s for inclusion in the study. The patient records in each of these M D C ' s were then submitted to the A U T O G R P C L A S S I F Y algorithm for subgroup partitioning. A c c o r d i n g to Y o u n g et al., none of the terminal groups produced as a result of this analysis were the same as the D R G ' s identified by Fetter et al. (1980). Specifically, they found that the independent variables (eg. surgery, diagnosis, etc.) that produced the opt imal grouping wi th the Pennsylvania data often differed from those used in the D R G system. A s well , the number of terminal groups generated wi th in each of the 10 M D C ' s often differed from the number of D R G ' s generated from the same M D C ' s . O n the basis of these findings, the authors conc luded that the D R G system was va l id only for the patient populat ion f rom w h i c h it was developed, and thus any attempt to generalize the results f rom one patient populat ion to another w o u l d be inappropriate. Another important issue raised wi th respect to D R G development concerned the accuracy of the data used in their construction. Doremus (1980) noted that in a study conducted by the Nat ional A c a d e m y of Sciences 6 8 to determine the reliability of Medicare patient abstract data, the principal diagnosis was found to be reliable only 57.2 percent of the time, in one sample, and 65.2 percent of the time in a second sample. Furthermore, the presence of a secondary diagnosis was noted correctly in only 74.5 percent of the cases studied. Similarly, the notation of principal surgical procedure was correct in only 78.9 percent of the cases. It was concluded, therefore, that the data discrepancies were of sufficient magnitude to be considered a major p r o b l e m in the construction of D R G ' s . F inal ly , questions regarding the clinical homogeneity of the D R G defined patient groups were raised. Berki et al. (1984) investigated this issue by examining the extent to which variations in length of stay (LOS) within D R G ' s were associated with clinical variables not inc luded in the D R G partit ioning process. It was argued that "to the extent that D R G ' s as currently defined capture the LOS- inf luenc ing clinical factors of case complexity and severity, observed variations of L O S with in D R G ' s should be attributable to nonclinical factors" (Berki et al., 1984, p. 126); i.e., no systematic association wi th clinical factors should be observed. T o test this proposi t ion, Berki et al. examined the length of stay distributions of 7 D R G ' s . U s i n g a regression m o d e l that inc luded both clinical (eg. total number of diagnoses, intensity of nurs ing services, t iming of radiologic and laboratory services) and nonclinical (age, sex, distance to hospital , method of payment , day admitted, type of admission, length of preoperative stay) variables as predictors of length of stay, within each D R G , these investigators found that in all cases the clinical variables significantly reduced the observed variabil ity i n length of stay. For some of the DRG's studied, significant relationships between length of stay and non-clinical 6 9 variables were also identified. H o w e v e r , none of these variables reached the significance of the clinical variables. T h e y concluded, therefore, that D R G ' s had not yet been developed to the point where they represented sufficiently precise categories for purposes of either util ization research, review or hospita l re imbursement . II.4.7.4 T h e Reformulat ion of D R G ' s W h e n Fetter et al., (1980) described the method used to construct the D R G ' s , they indicated that the system w o u l d require continuous development and revision as coding changes occurred and more data became available. Accord ing ly , w h e n an updated version of I C D codes (ICD-9 - C M ) was implemented in 1979, the D R G system h a d to be reevaluated and revised. A l t h o u g h the reformulation of the D R G ' s invo lved the use of the same A U T O G R P algorithm as h a d been used previously , an attempt was made to address concerns about the external val idi ty and homogeneity of the system (Fetter, 1985). Thus , in 1981, a new database was developed using 1979 data from a sample of 332 P A S member hospitals. This new database was selected to be as representative of the national populat ion as possible, w i t h in the P A S membership criteria, and consisted of a total of 1.4 mi l l ion patient records. F r o m this database, "a sample of 394,814 records was selected according to a stratified r a n d o m sampl ing procedure, with M D C ' s as the strata" (Hornbrook, 1985, p. 301). It should be noted, however, that these were not the same M D C ' s as had been used in the previous D R G development process. Orig inal ly , 83 M D C ' s h a d been defined. In the revised D R G system, the number of M D C ' s was reduced to twenty-three. T h e partit ioning of all principal diagnoses into these 23 M D C ' s was based on either the organ system affected or the specialist w h o w o u l d typically 7 0 provide care (Fetter, 1985). In most cases, each M D C was then disaggregated according to the presence or absence of a procedure performed in the operating room. Subsequent partitions were based on either the principal diagnosis or procedure category. F r o m these categories, addit ional partitions were identified according to age, the existence of comorbidities a n d / o r complications, and , occasionally, discharge status. The result of this process was the definition of 467 D R G ' s in contrast to the 383 D R G ' s identified in the o l d system (Fetter, 1985; H o r n b r o o k , 1985; Wi l l iams , K o m i n s k i , D o w d , & Soper, 1984). II.4.7.5 Suggested Appl icat ions for D R G ' s A s indicated above, D R G ' s were developed initially in response to a need for a standard patient classification system that could be used to i m p r o v e the cost-effectiveness of util ization review procedures employed at local, regional , or national levels. Accord ing ly , in the mid-1970's "when D R G ' s were first designed, they were intended as a management system for hospitals, as a tool for util ization review, and as a research tool" (Smits, Fetter, & M c M a h o n , 1984, p. 72). However , by 1980, when the article describing the method used to construct the initial D R G ' s was publ i shed, they were also being promoted as a potential mechanism for case mix accounting in hospital budget ing and prospective reimbursement (Fetter, et. al. , 1980). In fact, a major experiment wi th D R G - b a s e d reimbursement was already underway in the state of N e w Jersey (Iglehart, 1982b; M a y , 1985). Moreover , w i th the reformulation of D R G ' s in 1981, it was generally acknowledged that it w o u l d only be a matter of time before a D R G - b a s e d prospective payment system w o u l d be introduced by the federal government (Iglehart, 1982a; S imborg, 1981). This predict ion was borne out 71 in 1983 w h e n legislation call ing for the implementat ion of D R G - b a s e d prospective payment for Medicare hospital services was passed (Iglehart, 1983). A l t h o u g h the details of the Medicare prospective payment system (PPS) wi l l be discussed later, it is mentioned here for two reasons. First, the federal P P S was implemented so quickly fo l lowing the reformulation of the D R G ' s that almost all of the research concerning the adequacy of the D R G system was done after the enactment of the PPS legislation. Second, although most of the concerns regarding the adequacy of the new D R G ' s have important implications for util ization review and research, critical appraisals of the D R G system have focussed most of their attention on the implications of us ing D R G ' s as a mechanism for hospital reimbursement. Therefore, the specific concerns that have been raised, wi th respect to both applications (utilization review a n d hospital reimbursement) w i l l be discussed in the PPS section. Before turning to that discussion, however, another uti l ization review procedure that began to emerge in the early 1980's, as an adjunct or alternative to D R G - b a s e d review, w i l l be described. II.4.8 A n Alternative A p p r o a c h : Assessing A p p r o p r i a t e Ut i l izat ion Since the mid-1960's, the major factor motivat ing the federal government to implement uti l ization control strategies h a d been the assumption that a considerable amount of hospital use was inappropriate a n d unnecessary. U n t i l the mid-1970's, however, few attempts h a d been m a d e to objectively measure and quantify inappropriate use or to establish a definitive standard of what w o u l d constitute inappropriate use (Goran, 1979). Rather, the government's decision to mandate uti l ization review was based p r i m a r i l y o n indirect evidence p r o v i d e d through comparative geographic and prepaid practice studies; wide variations in regional 72 uti l ization rates, after standardization for age and sex, were assumed to reflect inappropriate use. Therefore, because there was no established standard for defining appropriate use, and because it was assumed that most instances of misuti l izat ion were occurring in the latter port ion of inpatient episodes, the p r i m a r y intent of the early federal util ization control strategies was s imply to reduce the observed regional variability in hospital lengths of stay (Goran, 1979; Go ldberg & H o l l o w a y , 1975; Z i m m e r , 1974). Thus , concurrent review, profile analysis, and D R G ' s were all developed with length of stay as the pr imary variable of interest. By the mid-1970's, however, when Fetter and his associates were beginning to work o n the development of D R G ' s , another group of investigators was beginning to question the merits of uti l ization review procedures that relied solely u p o n statistically der ived length of stay norms as the basis for reducing inappropriate util ization (Gertman & Restuccia, 1981; G o l d b e r g & H o l l o w a y , 1975; Restuccia & H o l l o w a y , 1976; Rothberg, & G e r t m a n , 1981; Sieverts, 1978; Z i m m e r , 1974). A s wi l l be discussed in more detail below, studies done by members of this group had begun to show that a h igh percentage of misuti l ization occurred not only a m o n g the longer stay cases but a m o n g the shorter stay cases as well (Restuccia & H o l l o w a y , 1976; Z i m m e r , 1974). It was suggested, therefore, that target date concurrent review procedures might fail to identify the cases of misuti l ization that "occurr[ed] d u r i n g hospital stays that violate[d] no length of stay norms" (Sieverts, 1978, p. 602). If so, the effectiveness of concurrent review w o u l d be seriously l imited. Moreover , from a quality of care perspective, it was feared that cont inued emphasis o n length of stay target dates a n d norms w o u l d foster "an attitude that if a patient has reached the average length of stay that 73 patient 'should' or 'must' be ready to leave the hospital , irrespective of i n d i v i d u a l medical realities" (Goldberg & H o l l o w a y , 1975, p. 478). A s noted by G e r t m a n & Restuccia (1981), although total util ization measures, such as length of stay, can be useful for describing general trends in hospital use, they prov ide no way to determine whether observed increases, or decreases, in util ization are the result of changes in inappropriate use or changes in appropriate use. Moreover , they provide no information as to the reasons for inappropriate use, the nature a n d amount of inappropriate use, or the extent to which inappropriate use can be controlled (Gertman & Restuccia, 1981; G o l d b e r g & H o l l o w a y , 1975; Restuccia & H o l l o w a y , 1976; Rothberg & Gertman, 1981; Sieverts, 1978; Z i m m e r , 1974). G i v e n such potential shortcomings in the normative length of stay approach to util ization review, it was suggested that PSRO's w o u l d be more effective if they began to focus their attention on actual patient needs for hospitalization (Dobson et al., 1978; G o l d b e r g & H o l l o w a y , 1975; Sieverts, 1978). Th e promulgat ion of this approach was impeded , however, by the lack of a reliable and va l id method for evaluating appropriate and inappropriate hospital use. A s stated by Rothberg and G e r t m a n (1981), "[m]easuring unnecessary hospital use in an 'objective' fashion is a difficult, expensive and sensitive task. M e r e l y defining what constitutes 'unnecessary' (the precondit ion for measurement) has been a prob lem of major proportions" (p. 48). T h u s , to date, most of the research o n the appropriateness of hospital uti l ization has focussed on measurement issues. A t the same time, however, attempts have been made to begin quantifying inappropriate use and to document the reasons for it. A l t h o u g h 74 methodological problems render most of these studies exploratory at best, consistency in the reported results suggests that they may have important implications for both uti l ization management and quality assurance p r o g r a m efforts. II.4.8.1 M e a s u r i n g Appropr ia te Use Since the mid-1970's, a number of different methods have been deve loped to measure the appropriateness of hospital util ization. Z i m m e r (1974), for example, had physic ian reviewers examine patient records and talk to staff and patients in order to make a decision about the patient's need for hospitalization. T h e subjective nature of this approach, however, rendered it particularly vulnerable to criticisms concerning its reliability (Gertman & Restuccia, 1981; Go ldberg & Holloway,1975; Restuccia & H o l l o w a y , 1976; Rosser, 1976). Thus , Go ldberg & H o l l o w a y (1975) suggested that, unti l more explicit criteria for appropriate uti l ization review could be deve loped, the broad service criteria for hospital "level of care" as generated by the Medicare program or the M c K a y - D e e Hospi ta l Center should be used. In a subsequent study, Restuccia and H o l l o w a y (1976) adopted this approach a n d used the Medicare "level of care" criteria (See A p p e n d i x A ) to identify instances of misuti l izat ion in one hospital setting. A l t h o u g h this approach was considered an improvement over the more subjective approach used by Z i m m e r (1974), it was generally considered only the first step in the development of explicit util ization criteria (Goldberg & H o l l o w a y , 1975; G e r t m a n & Restuccia, 1981). A completely different, albeit interim, approach was suggested by Rothberg & G e r t m a n (1981). These authors maintained that, since defining and measur ing inappropriate uti l ization is a time consuming a n d difficult 75 process, attitudinal surveys designed to assess perceptions of inappropriate use cou ld be used pending the development of more objective measures. This method received little attention, however, because in the same year that it was proposed an article was publ ished describing what has been, to date, the most significant advancement in the measurement of inappropriate util ization; namely, the Appropriateness Evaluat ion Protocol (Gertman & Restuccia, 1981). T h e Appropriateness Evaluat ion Protocol (AEP) was designed in response to the need for a "valid, reliable and simple technique for determining the appropriateness of hospital use, in terms of the necessity for a patient being in a hospital bed on a given day" (Gertman & Restuccia, 1981, p. 867-868). Because the developers also wanted the A E P to be "applicable to as m a n y patients as possible" (p. 857), it was decided that the instrument should be "diagnosis-independent" and applicable to all patients admitted into "adult medicine, surgery and gynecology services" (p.857). In addit ion, it was decided that the A E P should consist of "a set of explicit criteria which , if any single one was met, w o u l d indicate that a day of care was appropriate" (p. 857). Final ly , i n order to improve the efficiency and reliability of the A E P review process, an upper l imit of 30 criteria items was established. After several years of development and testing, 27 "day-of-care" criteria were adopted. These criteria items were organized into three categories: "medical services, nurs ing / l i f e support services, and patient condit ion factors" (Gertman & Restuccia, 1981, p. 857). T h e first two categories contained items that assessed whether or not, o n a given day of hospital ization, the patient was receiving a service or services that w o u l d 76 typically only be p r o v i d e d in an acute care setting. The third category assessed factors that w o u l d generally be considered to indicate that the patient's condit ion was sufficiently unstable to require hospitalization, whether or not they were receiving medical or nurs ing services on the day of review. A list of all the criteria developed by Gertman and Restuccia is p r o v i d e d in Table I. O n e of the major objectives of Gertman & Restuccia (1981), in deve lop ing the A E P , was to obtain higher levels of inter-rater reliability than h a d been obtained with previous measurement techniques. O f particular concern was the level of reviewer agreement that w o u l d be obtained among the cases deemed inappropriate or uncertain (specific agreement). A s indicated by G e r t m a n & Restuccia, relatively h igh levels of overal l agreement between reviewers (73-89%) h a d been fairly c o m m o n in previous studies, even in the studies that used subjective criteria. In these same studies, however, the rate of agreement obtained among the cases deemed inappropriate , by at least one reviewer, was m u c h lower (0-53%). G i v e n that one of the pr imary functions of hospital uti l ization review, or P S R O review, is the identification of cases that require further study, G e r t m a n a n d Restuccia argued that, a l though h igh levels of overall agreement are important, they are less important, for the purposes of uti l ization review, than are h igh levels of specific agreement. That is, unless the review instrument can identify reliably the cases that are inappropriate , the review process wi l l be both ineffective and inefficient. In order to evaluate the reliability (overall and specific) and the val idi ty of the A E P , Ger tman & Restuccia (1981) conducted two trials. In the first trial, two nurse reviewers and one physic ian reviewer appl ied the A E P 77 T A B L E I. Appropriateness Evaluat ion Protocol: "Day-of-Care" Criteria A . M e d i c a l Services 1. Procedure in operating room that day 2. Scheduled for procedure in operating room the next day, requiring preoperative consultation or evaluation 3. Cardiac catheterization that day 4. Angiography that day 5. Biopsy of internal organ that day 6. Thoracentesis or paracentesis that day 7. Invasive central nervous system diagnostic procedure (eg., lumbar puncture, cysternal tap, ventricular tap, penumoencephalography) that day 8. Any test requiring strict dietary control, for the duration of the diet 9. New or experimental treatment requiring frequent dose adjustments under direct medical supervision 10. Close medical monitoring by a doctor at least three times daily (observations must be documented in record) 11. Postoperative day for any procedure covered in number 1 or 3-7 above B. N u r s i n g / L i f e Support Services 1. Respiratory care - intermittent or continuous respirator use and/or inhalation therapy (with chest PT, IPPB) at least thrice daily 2. Parenteral therapy - intermittent or continuous IV fluid with any supplementation (electrolytes, protein, medications) 3. Continuous vital sign monitoring - at least every 30 minutes, for at least 4 hours 4. Intramuscular and/or subcutaneous injections at least twice daily 5. Intake and output measurement 6. Major surgical wound and drainage care (chest tubes, T-tubes, hemovacs, Penrose drains) 7. Close medical monitoring by nurse at least three times daily, under doctor's orders C . Patient C o n d i t i o n Factors Within 24 hours before day of review 1. Inability to void or move bowels (past 24 hours) not attributable to neurologic disorder Within 48 hours before day of review 2. Transfusion due to blood loss . 3. Ventricular fibrillation or ECG evidence of acute ischemia, as stated in progress note or in ECG report 4. Fever at least 101 rectally (at least 100 orally), if patient was admitted for reasons other than fever 5. Coma - unresponsiveness for at least one hour 6. Acute confusional state, not due to alcohol withdrawal 7. Acute hematologic disorders, significant neutropenia, anemia, thrombocytopenia, leukocytosis, erythrocytosis, or thrombocytosis yielding signs or symptoms 8. Progressive acute neurologic difficulties Within 14 days before day of review 9. Occurrence of a documented, new acute myocardial infarction or cerebrovascular accident (stroke) (Source: G e r t m a n & Restuccia, 1981) 78 retrospectively to 100 randomly selected patient records. T h e sample consisted of approximately 50% of the patients that were in the medical and surgical units of one hospital on the specific day chosen for review. In the second trial, the same reviewers appl ied the A E P to a new sample of 100 patients. Th i s time, however, various combinations of concurrent and retrospective review were used by the three reviewers. T o assess the inter-rater reliability of the A E P , overall agreement and specific agreement rates were computed. In addit ion, overall agreement rates between nurse reviewers, and between each nurse reviewer and the phys ic ian reviewer were evaluated. Th e investigators also compared the rates obtained w h e n the A E P was appl ied retrospectively as opposed to concurrently and from trial one to trial two. A l t h o u g h no data regarding these comparisons was presented, the authors c laimed that the rates were very similar. Hence, in the reported analyses, the data from both trials were poo led . T o evaluate the val idi ty of the A E P , several methods were used. First, the agreement between the physician-nurse pairs was used as an estimate of predictive validity. Second, critical appraisal of the A E P criteria b y four P S R O committees was used to assess face validity. Final ly , the use of 'overrides' was used as an indication of content validity. T h e use of 'override' options were inc luded in the A E P in order to account for any defects in the established criteria. Essentially, overrides "allowed a reviewer to 1) judge a day appropriate w h e n no objective criteria were met, usually because the list of criteria was not sufficiently comprehensive; or 2) judge a day as inappropriate even w h e n an objective criteria was met, because the reviewer felt the service or condit ion was not sufficient in the particular 79 case to justify use of an acute hospital bed" (Gertman & Restuccia, 1981, p. 865). O n the basis of the 200 patients and patient records reviewed in the study, overall agreement rates, each evaluated using a K a p p a statistic, between pairs of reviewers ( R N a / R N b , R N a / M D , R N b / M D ) ranged from a l ow of 92% to a high of 94% (p_<0.0001 for each comparison). Thus , approximately 93% of the time, paired reviewers agreed on whether or not a given day in a patient's hospital stay was appropriate or inappropriate. The agreement rate among all three reviewers was 88.1% (Kappa statistic; p_<0.001). Specific agreement rates (i.e., the percentage of paired agreements o n cases identified as inappropriate by at least one reviewer) between pairs were lower, ranging from 73% to 79%, but notably higher than the rates obtained in previous studies. The specific agreement rate a m o n g all three reviewers was 63.8%. The "override" options were used by the R N a / R N b pair 21 times (7 by R N a and 14 by R N b ) to make a day appropriate and 19 times (7 and 12) to make a day inappropriate. For the M D / R N a and M D / R N b pairs, each had 24 overrides m a k i n g a day appropriate and 17 overrides m a k i n g a day inappropriate . A n examination of the override cases indicated that, for each pair of reviewers, the use of the override options "improved the level of agreement between reviewers" (Gertman & Restuccia, 1981, p. 865). M o r e o v e r , w h e n both reviewers invoked an override opt ion, "in no case d i d [the] two reviewers make oppos ing override decisions" (p. 865). Another aspect of the G e r t m a n & Restuccia study inc luded an assessment of the appropriateness of admiss ion to the hospital. U n l i k e the criteria deve loped for the day-of-care assessment, however, no explicit 80 criteria were used in the admiss ion assessment. Instead, the reviewers were asked to make decisions about the appropriateness of admiss ion on the basis of their professional judgement. This approach was adopted for two reasons. First, in an appraisal of existing admiss ion review procedures, the authors h a d been unable to identify a set of criteria that were reliable and va l id . Second, by hav ing the reviewers use subjective professional judgement in determining the appropriateness of admiss ion, a comparison of agreement rates obtained using this approach cou ld be made to the agreement rates obtained using explicit criteria. In examining the appropriateness of admiss ion decisions, G e r t m a n & Restuccia found that the overall agreement rates between reviewer pairs ranged from 89.3% to 91.8% (Kappa statistics; p_<0.01 in each case) and among all three reviewers was 87.1% (Kappa statistic; p_<.001). Thus , in terms of overal l agreement on the appropriateness or inappropriateness of admiss ion, the rates were very similar to those obtained in the criteria-based days-of-care assessment. W i t h respect to specific agreement rates, however, the picture was quite different. A m o n g the admissions deemed inappropriate by at least one reviewer, the level of agreement ranged from 30.5% to 40.0%; rates m u c h lower than the rates computed for the days-of-care review. A l t h o u g h the authors acknowledged that the admiss ion and days-of care reviews were not directly comparable, they nonetheless believed that the observed differences in specific agreement rates were most l ikely due to the differential use of subjective a n d objective criteria. T h e final measurement issue addressed by Gertman & Restuccia was to what extent higher levels of inter-rater reliability could, wi th appropriate adjustments to the A E P , be obtained. T o investigate this issue, they 81 examined all of the reasons for disagreements among reviewers. This analysis was l imited, however, by the small number of disagreements that had actually occurred and by the variability in the problems cited. A l though no f irm conclusions cou ld thus be drawn, three sets of factors were found to have contributed most frequently to the occurrence of inter-rater disagreements. First, over half of the disagreements were associated with patients "for w h o m there was also disagreement over the appropriateness of the admiss ion (p. 867). Since most of these patients were invo lved in non-routine treatment regimens, the authors conc luded that the formulat ion of explicit criteria for such cases w o u l d be difficult and, therefore, they w o u l d be "more conducive to being handled on an exception basis rather than by the application of a general set of criteria such as the A E P " (p. 867). Second, a number of the disagreements centered around the mental status of the patients. A n d , once again, the authors felt that only in a very few cases w o u l d more stringent criteria reduce the frequency of disagreements. Final ly , it was observed that a number of disagreements related to one of the nurs ing service criteria (i.e., intake and output measures). O n occasions w h e n this was the only criterion met in a review, it could be problematic. Specifically, it was noted that physicians tended to overuse this service - by requesting it w h e n it was not necessary or by fail ing to discontinue it at the appropriate time. Hence , the authors suggested that this criterion needed further study. Overa l l , G e r t m a n & Restuccia estimated that interrater reliability o n the A E P could possibly be increased "from the 92-94 per cent level to the 95-96 per cent level" (p. 867). Increases b e y o n d that, however, were considered unl ikely . 82 T h r o u g h subsequent development and testing, G e r t m a n & Restuccia expanded the A E P to include explicit criteria for admiss ion review (Siu et al., 1986; Walter, 1983). A total of 18 admiss ion criteria, organized into severity of illness and intensity of service categories, were added. In a study done by S iu , et al. (1986), overall inter-rater agreement rates for these criteria were reported to be between 84% and 92% (Kappa statistics; p_<0.001 in all cases). H o w e v e r , no data on specific agreement rates were prov ided . The complete list of A E P admission criteria is prov ided in Table II. In addit ion to the adult version of the A E P , a modif ied A E P has recently been developed for application to pediatric inpatients (Kemper, 1987a). A l t h o u g h modif ied , the structure and content of the pediatric A E P is very similar to the adult A E P . Hence, no further description wi l l be p r o v i d e d herein. A list of the pediatric "day-of-care" criteria, however, are p r o v i d e d in A p p e n d i x B. T o m y knowledge, admiss ion criteria for pediatric patients have not yet been developed. A s wi th D R G ' s , reactions to the A E P have been mixed. O n the positive side, it has been praised for being quick and easy to apply (Walter, 1983). In addit ion, in a survey of over 100 hospitals that were using the A E P , it was found that "both U R officials and physicians pre ferred] it to other forms of review" (Walter, 1983, p. 177). H o w e v e r , although the A E P is clearly one of the most reliable and va l id instrument that has yet been deve loped to assess appropriateness of hospital use, there are potential problems associated wi th its employment as a tool for uti l ization management. O n e particular concern is that, in the effort to contain costs, insurers might implement the A E P criteria alone, without the override options (Walter, 1983). Since most of the overrides in the G e r t m a n & 83 T A B L E II. Appropriateness Evaluat ion Protocol: A d m i s s i o n Criteria A . Severity of Illness 1. Sudden onset of unconsciousness or disorientation (coma or unresponsiveness) 2. Pulse rate of less than 50 or greater than 140 3. Systolic blood pressure of less than 90 or greater than 200; diastolic of less than 6 0 or greater than 120 4. Acute loss of sight or hearing 5. Persistent fever equal to or greater than 100 (orally) or greater than 101 (rectally) for more than five days. 6. Acute loss of ability to move a body part 7. Active bleeding 8. Severe electrolyte or blood gas abnormality 9. Acute or progressive sensory, motor, circulatory, or respiratory embarrassment sufficient to incapacitate the patient (One of the seven intensity-of-service criteria listed below must also be met) 10. ECG evidence of acute ischemia; there must be suspicion of a new MI 11. Wound dehiscence or evisceration B. Intensity of Service 1. Use of IV medications and/or fluid replacement (does hot include tube feedings) 2. Surgery or procedure scheduled within 24 hours requiring general or regional anesthesia or equipment and facilities available only in a hospital 3. Vital-sign monitoring every two hours or more often (may include telemetry or bedside cardiac monitoring) 4. Use of chemotherapeutic agents that require continuous observation for a life-threatening toxic reaction 5. Treatment in an ICU 6. Use of intramuscular antibiotics at least every eight hours 7. Intermittent or continuous respirator use at least every eight hours (Source: Walter , 1983) Restuccia (1981) study resulted in a designation of the reviewed day as "appropriate", failure to include the override options cou ld lead to inappropriate reimbursement denials (Walter, 1983) A second concern is that strict adherence to appropriateness criteria might raise the proport ion of acute hospital days to a point where most are "filled with h igh anxiety and low satisfaction for the patient, the physician, and other providers of care" 84 (Dans et al., 1985, p. 1135). E v e n the developers themselves have recommended caution in the application and interpretation of A E P criteria (Gertman & Restuccia, 1981). Specifically, they have indicated that the A E P does not adjust for situations "where absolute efficiency is not feasible" (p. 869) nor does it take into account whether or not the "causes of inappropriate use are wi th in the control of the hospital and the patient's physician" (p. 869). Accord ing ly , the identification of inappropriate use, by itself, could be of l imited value, unless it is accompanied by an understanding of the reasons for it. In fact, if examined in isolation from its causes, it cou ld have expensive detrimental effects. A s Ger tman & Restuccia have stated, "optimizing one objective of the medical care system, such as appropriateness of hospital use, may result in suboptimizat ion of other objectives, such as appropriateness of use of all levels of care. Careful consideration of such trade-offs must be made by health care decision makers" (p. 869). In the fo l lowing section, therefore, the methods that have been used to identify the reasons for inappropriate use are described. n.4.8.2 Assess ing the Reasons for Inappropriate Use Since most of the studies on inappropriate use have been concerned pr imar i ly wi th identifying and measuring its occurrence, few attempts have been made as yet to develop a systematic method for assessing w h y inappropriate use occurs. T o date, the most comprehensive attempt to study a n d classify the reasons for inappropriate use has been undertaken by Restuccia & H o l l o w a y (1976). In the only previous attempt, Z i m m e r (1974) s imply asked physic ian reviewers to indicate whether an observed inappropriate day was the result of: 1) an unnecessary admiss ion, 2) an unnecessary delay in discharge, or 3) an unnecessary delay in performing 85 surgery or other in-hospital procedure. In addit ion, he had the reviewers indicate.what level of care w o u l d have been appropriate, assuming that al l levels were available; i.e., on the day of the review, w o u l d the patient have been more appropriately placed in a "long-term hospital, nurs ing home, organized home care wi th mult iple services, at home wi th one or two home care services, or at home wi th no home care services" (p. 454). In contrast to Zimmer's (1974) approach, Restuccia & H o l l o w a y (1976) interviewed 9 uti l ization review nurse coordinators, 11 discharge planners, 3 physicians, and 3 administrators, f rom five different hospitals, in order to develop an a priori list of factors that were k n o w n , by these indiv iduals , to have caused inappropriate use at their facility. The final list (called "barriers to appropriate utilization") consisted of 63 factors organized into four categories "according to 'areas of responsibility'; i.e., to whose action or inaction a particular barrier could be attributed" (p. 562). The four categories were thus defined as "Physician Responsibility", "Hospital Responsibility", Patient or F a m i l y Responsibil ity" a n d "Environmental Responsibility". Barriers attributable to the Physic ian were those relating to the physician's professional knowledge or availability to render services; Hosp i ta l barriers related to operational systems and to the organization's employees; Patient barriers related to the patient's (or family's) participation in the treatment process; and Environmenta l barriers related to factors beyond the immediate control of the hospital , physic ian, or patient (eg. insurance coverage, availability of appropriate alternative services, etc). A list of the specific barriers inc luded in each category is prov ided in A p p e n d i x C . G e r t m a n & Restuccia (1981) employed a shortened version of this list in order to assess the reasons for inappropriate day-of-care use among the 86 patients inc luded in their study. In addit ion, for patients designated as inappropriate admissions, G e r t m a n & Restuccia had their reviewers indicate whether the admiss ion was considered unnecessary because it invo lved: 1) the provis ion of services that could have been done on an outpatient basis or in a lower level facility, 2) no clear need for medical care of any k i n d , or 3) some other specified reason. H .4.8.3 E m p i r i c a l F indings on the A m o u n t and Nature of Inappropriate. Use For the most part, the researchers that have attempted to quantify and describe patterns of inappropriate use have focussed their attention on one or more of the fo l lowing four questions: 1) to what extent does inappropriate util ization occur?; 2) at what point in an inpatient episode is inappropriate uti l ization most l ikely to occur?; 3) what is the relationship between inappropriate use and length of hospital stay?; and 4) what are the major factors that contribute to inappropriate use? T h e findings relevant to each of these questions wi l l be discussed in turn. 1. A m o u n t of Inappropriate Use. A l m o s t all of the "appropriateness of utilization" studies have attempted, in the process of methodological development , to quantify the amount of inappropriate use observed in their samples. Some of these studies have assessed only the occurrence of inappropriate day-of-care use. Others have examined both the inappropriateness of day-of-care use and hospital admissions. A summary of the major findings of these studies, as wel l as some important characteristics relating to their research methods, are p r o v i d e d in Table III. 00 T A B L E III. Studies of Inappropriate1 Hospital Utilization A u t h o r S t u d y S t u d y P o p u l a t i o n (Year) P e r i o d a n d D e s c r i p t i o n D a y - o f - C a r e R e v i e w s A d m i s s i o n R e v i e w s T o t a l N o . % I n a p p r o p r i a t e T o t a l N o . % I n a p p r o p r i a t e Z i m m e r 1968-70 O n e h o s p i t a l ; A l l C l i n i c a l S e r v i c e s ; (1974) 3,369 p a t i e n t s ; S u b j e c t i v e a s s e s s m e n t ; 2,695 pat ients r e v i e w e d t w i c e o n s a m e d a y ; 674 r e v i e w e d o n c e . 6,064 9.1 R e s t u c c i a &: 1973-74 O n e h o s p i t a l ; M c d / S u r g i c a l S e r v i c e s ; H o l l o w a y 218 pat ien ts ; A s s e s s m e n t b y M e d i c a r e C r i t e r i a ; (1976) A l l pa t ien ts r e v i e w e d o n each d a y o f stay. 1,902 10.6 G e r t m a n & n o t O n e h o s p i t a l ; M e d / S u r g i c a l S e r v i c e s ; R e s t u c c i a g i v e n 192 pat ien ts ; A E P a s s e s s m e n t ; m o s t (1981) p a t i e n t s r e v i e w e d three t i m e s o n s a m e d a y . 496 25.0 496 10 S i u et. a l . , (1986) 1974-82 100 h o s p i t a l s ; 6 c i t i c s ; M e d / S u r g S e r v i c e s ; 1,132 p a t i e n t s ; A E P a s s e s s m e n t ; E a c h pa t ien t r e v i e w e d once . 1,132 35.0 1,132 2 3 * A n a d d i t i o n a l 1 7 % w o u l d h a v e been i n a p p r o p r i a t e h a d a m b u l a t o r y se rv i ces b e e n a v a i l a b l e i n the h o s p i t a l s s t u d i e d . A to ta l o f 4 0 % o f the a d m i s s i o n r e v i e w s w e r e t h u s p o t e n t i a l l y i n a p p r o p r i a t e . 88 A s can be seen from this table, the amount of inappropriate "days-of-care" use identified in these studies ranges from 9% to 35%, with the higher rates obtained in studies that employed explicit criteria (i.e., the A E P ) . The proport ion of inappropriate admissions ranges from 10% to 23%. A g a i n , the higher rates are associated with the use of A E P criteria. Gertman & Restuccia (1981) have suggested that higher rates of inappropriate use should be obtained with the A E P because it was designed to prov ide an "absolute type of efficiency standard" (p.869). Subjective techniques, on the other hand , permit reviewers to modi fy their judgements according to whether or not absolute efficiency is feasible. H o w e v e r , the higher rates m a y also be a function of temporal differences, differences in study methods, or differences in the study populations. For these reasons, the studies shown in Table III are not directly comparable. Moreover , since all of these studies were done wi th in the context of one hospital or wi th in narrowly defined geographic regions, little can be said about expected rates of inappropriate hospital use across the country. N o r has sufficient data been accumulated to indicate trends in misuti l ization over time or h o w such trends have been effected, if at all , by the institution of pol icy interventions such as PSRO's . Nonetheless, it w o u l d appear reasonable to suggest that, despite such interventions, inappropriate use of hospital facilities and services has persisted. Certainly, the prob lem has remained of sufficient magni tude to warrant, as Rothberg & Gertman (1981) suggested, continued vigi lance in the moni tor ing and management of hospital uti l ization. 2. T i m i n g of Inappropriate Use. O n l y one of the studies mentioned i n Table HI attempted to determine at what point in an inpatient episode, an inappropriate day was most likely to occur. T o address this question, 89 Restuccia and H o l l o w a y (1976) classified the 201 inappropriate days identified in their study according to the "stage of stay"; i.e., according to the day o n which the inappropriate day(s) occurred relative to the overall sequence of days defining each inpatient episode. Thus , inappropriate days classified as "initial-stage" were those that occurred prior to the recording of the first appropriate day or when no appropriate day was recorded prior to discharge; "midstage" referred to an inappropriate day or days that occurred between two series of appropriate days; and "end-stage" referred to inappropriate days that immediately preceded discharge but fol lowed a consecutive series of appropriate days. U s i n g this classification scheme, Restuccia & H o l l o w a y found that, of all 201 inappropriate days identified, 10% were classified as initial-stage, 25% were midstage, and 65% were end-stage. Thus , for the hospital in which this study was done, the c o m m o n assumption that most misuti l ization occurs in the latter stages of inpatient episodes appeared to be confirmed. Whether or not this was a general or hospital-specific phenomenon was not investigated. 3. T h e Relationship Between Misut i l izat ion and Length of Stay. T o explore the nature of the relationship between inappropriate use and length of stay Z i m m e r (1974) examined the percentage of inappropriate days that occurred among patient stays of 1-7 days, 8-14 days, 15-21 days, 22-30 days, 31-60 days, and over 60 days. H e found that this percentage increased from 4.9% for patients hospital ized from 1-7 days, to 24.6% for patients with stays exceeding 60 days. In order to examine this trend further, misutilizers were reclassified according to whether their hospital stay was 21 days or less or 90 greater than 21 days 1 . W i t h this dichotomizat ion of inpatient lengths of stay, it was found that 52.6% of the misutilizers had stays of 21 days or less while only 47.4% had stays over 21 days. Thus , whi le the overall percentage of misuti l ization was found to increase wi th length of stay, a higher proport ion of the identified misutilizers had lengths of stay that, at the time the study was conducted, w o u l d have failed to trigger an extended stay review. A l t h o u g h these findings were important in the sense that they p r o m p t e d consideration of misuti l ization in shorter stay cases, a different pattern w o u l d l ikely have been obtained had a different target date been used. In a more detailed study, however, Restuccia & H o l l o w a y (1976) reported f indings that conf irmed Zimmer's contention that the prob lem of misuti l ization among shorter stay cases was worthy of note. These investigators used the diagnosis-related length of stay norms developed by P A S to identify the 50th and 75th percentiles for the patients inc luded in their study. These percentiles were chosen because they reflected the most c o m m o n l y used target dates in P S R O review. Accord ing ly , of the 201 inappropriate days identified in their study, 21 % occurred before the 50th percentile and 37% occurred before the 75th percentile. Moreover , when the Medicare m a x i m u m time l imit for certification was used (i.e., 12 days), 55% of the inappropriate days were found to occur before this checkpoint. A l t h o u g h the instrument used by Restuccia & H o l l o w a y (1976) to assess inappropriate use (i.e., Medicare level of care criteria) has been criticized as being overly general a n d unreliable (Goldberg & H o l l o w a y , 1975; Gertman & Restuccia, 1981), the implications of such findings were considered to 1 The use of 21 days to separate short from long stays was based on the fact that, for the patients reviewed, this was the target used by the state fiscal intermediaries to instigate recertification review. 91 p rov id e reasonable justification for continued w o r k in the area (Sieverts, 1978). In addit ion, although Restuccia and H o l l o w a y ' s f inding that the greatest percentage of misuti l ization occurred in the "end-stage" of patient episodes, the application of commonly used length of stay criteria w o u l d have missed not only a significant proport ion of the "initial" and "midstage" misuti l ization, but a sizeable proport ion of the "end-stage" misuti l izat ion as well . This tended to support the notion that, for their study populat ion at least, util ization reviews based o n length of stay norms w o u l d be of l imited effectiveness. 4. Reasons for Inappropriate Use. In order to develop a better understanding of the factors that were contributing to the occurrence of inappropriate days, a few of the investigators mentioned in Table III attempted to document the reasons for each of the inappropriate days identified in their studies. T h e findings of these studies are s u m m a r i z e d in Table IV. A l t h o u g h the lack of a standard instrument for assessing the reasons for inappropriate use makes comparisons from one study to another difficult, certain trends do emerge. For example, across all the studies, the most c o m m o n l y cited reasons for inappropriate days were discharge-related. Moreover , discharge problems were found to be pr imar i ly attributable to a lack of appropriate lower level care alternatives or a failure to access lower level alternatives at the appropriate time. The second major factor found to contribute to inappropriate use was physician-related; i.e., there was a tendency o n the part of some physicians to adopt overly conservative treatment management practices. The percentage of inappropriate days attributed to problems in hospital operational procedures, o n the other h a n d , was relatively small . ON TABLE IV. Studies* Indicating Reasons for Inappropriate "Dav-of'-Care" Use Author No. of Inappropriate (Year) Days Identified No. of Reasons Identified Breakdown of Inappropriate Days Reason n % Zimmer 517 517 Admission Unnecessary 50 9.7 (1974) Delay in hospital procedure 42 8.1 Delay in discharge 425 82.2 Restuccia & 201 229** Physician Responsibility 95 41.5 Maior Reasons Holloway (related to discharge) (26) (11.4) (1976) -too conservative 16.0% Hospital Responsibility 17 7.4 -delays transfer 7.5% (related to discharge) (8) (3.5) Environment -terminal patients Patient Responsibility 22 9.6 not transferred 6.5% (related to discharge) (6) (2.6) -nursing facility not available 23.6% Environmental Factors 95 41.5 (related to discharge) (86) (37.6) Gertman & 124 124 Inappropriate level of care 53 43.0 Restuccia Physician too Conservative 30 24.0 (1981) Delay in hospital procedure 14 11.5 Delay in Transfer 9 7.0 Reason not indicated 18 14.5 For descriptive information concerning these studies refer to Table III. 28 of the Inappropriate days had 2 reasons listed. 93 A s can be seen in Table V , the reasons cited to account for inappropriate hospital admissions reflect a tendency to admit patients that could have been treated on an outpatient basis and a failure to use lower levels of care appropriately. It should be noted, however, that at least in the case of the patients inc luded in the Siu et al. (1986) study, the failure to treat patients on an outpatient basis was due to the fact that, in all of the hospitals s tudied, ambulatory services were not available. Thus , although 17% of the cases studied d i d not require acute level care, the lack of the appropriate outpatient services rendered their admissions necessary. II.4.8.4 Implications of the E m p i r i c a l Studies A l l of the findings mentioned above demonstrate the importance of deve loping an understanding of the factors that impede appropriate util ization. W h i l e the results of these studies must be considered prel iminary, since they are based largely on case studies and, therefore, lack generalizability, they have important implications for the development of policies a imed at reducing inappropriate utilization. A s noted by Restuccia & H o l l o w a y (1976), although most of the identified inappropriate use was l ikely to occur d u r i n g the latter stage of an inpatient episode, it was pr imar i ly attributable to the l imited availability of appropriate alternatives. T h u s , whi le normative uti l ization review procedures m a y facilitate the identification of such misutil izers, the ability of physicians a n d hospitals to reduce this form of misuti l ization w o u l d be l imited because, in the short r u n at least, it lies beyond the scope of their control. For similar reasons, it w o u l d be difficult to effect reductions in inappropriate use resulting from the second most c o m m o n cause of misuti l izat ion, namely, conservative medical practice. W i t h the use of instruments such as the A E P and the T A B L E V . S t u d i e s * I n d i c a t i n g R e a s o n s for I n a p p r o p r i a t e A d m i s s i o n s . a\ Author No. of Inappropriate Breakdown of Inappropriate Admissions (Year) Admissions , Reason n % Gertman & Restuccia (1981) 49 Services could have been provided on outpatient basis. Lower level of care needed. No need for any care Other Missing 33 67.0 12.0 2.0 8.0 10.0 Siu et. al. (1986) 452 Lower level of care needed. Services could have been provided on outpatient basis. 174 192 38.0 42.5 Other 86 19.5 * For descriptive information concerning these studies refer to Table 111. 9 5 "Barriers to Inappropriate Util ization", however, hospitals could begin to identify inefficiencies or lapses in operational procedures that contribute to misuti l izat ion and, within that realm of responsibility at least, begin to take corrective action. Whether the use of such techniques, either alone or as a supplement to length of stay reviews, w i l l facilitate increased reductions in inappropriate hospital utilization remains to be seen. T o m y knowledge, no studies have as yet attempted to evaluate the effectiveness of the A E P approach in reducing misuti l ization. N o r have any studies been done w h i c h compare the general effectiveness, or cost-effectiveness, of the A E P approach versus the normative approach. Since the A E P was developed, however, a major reformulation of federal health care pol icy has resulted in the abandonment of the P S R O concurrent review strategy (Dans et al., 1985). U n d e r federal legislation enacted in 1983, hospitals are no longer retrospectively re imbursed for the actual costs incurred in treating Medicare patients. Instead, they are pa id on the basis of a prospectively determined rate calculated for each of the 467 defined D R G ' s (Iglehart, 1982a). Accord ing ly , the nature and objectives of uti l ization rev iew and management have undergone important changes. T h e remainder of this chapter wi l l , therefore, be devoted to the new prospective payment system, wi th particular emphasis g iven to its impl icat ions for hospital uti l ization and uti l ization management. n.5 Current Trends in Uti l izat ion Management: T h e E r a of Prospective  P a y m e n t A l t h o u g h the development of D R G ' s and the A E P were important refinements in uti l ization review technology, by the late 1970's pol icy analysts a n d health service researchers h a d begun to question the merits of 96 rely ing u p o n regulatory mechanisms, such as the PSRO's , to contain publ ic ly financed hospital costs (Brook et al., 1978; Dobson et al., 1978; Goran,1979). A l t h o u g h decreases in hospital length of stay h a d been observed throughout the 1970's (Pokras & Kubishke , 1985), little evidence h a d been p r o v i d e d to suggest that the P S R O p r o g r a m h a d been effective in containing costs (Dans et al., 1985; Ginzberg , 1985; G o r a n , 1979; Lantos, 1984). Moreover , G o r a n (1979) argued that since PSRO's h a d no mandate to affect costs directly, "they should be held accountable for the el imination of unneeded services and for the shifting of services from inappropriate ly intense levels of care to lesser ones, but not for the cost and reimbursement consequences of such actions" (p.3). Instead, it was suggested that, in order to contain costs, the P S R O system w o u l d have to be supplemented with major statutory changes in the method used to reimburse hospitals (Brook et al. , 1978; G o r a n , 1979); that new financial incentives favouring efficiency w o u l d have to be developed to replace the inflationary incentives p r o v i d e d under the traditional method of cost-based retrospective reimbursement (Iglehart, 1982a). In fact, d u r i n g the latter half of the 1970's, a number of states h a d already begun to experiment wi th alternative reimbursement methods (Iglehart, 1982a). A l t h o u g h the details of the various methods differed from one state to another, they all shared one characteristic in common; i.e., they were all based o n a prospective system of payment. T h u s , rather than retrospectively re imburs ing hospitals o n the basis of average per d i e m costs incurred, these states h a d begun to develop systems based either on cost center budgets, global budgets, or prospectively determined rates-per-case (Iglehart, 1982a). Regardless of the method used, the basic intent was the same; namely , to impose front e n d limits o n 9 7 hospital reimbursement (Iglehart, 1982a) and, thereby, reverse previous inflationary trends. n.5.1 T h e Transi t ion to the Federal Prospective Payment System In 1982, the annual increase in total hospital expenditures was determined to be almost three times the national inflation rate (Iglehart, 1983) and estimates of future expenditures predicted that Medicare costs w o u l d double by 1987, if no changes in trends could be effected (Iglehart, 1982a). Concerns about the continuing viability of the Medicare trust fund (Dans et al. 1985; Iglehart, 1983; Pettingill & Westell , 1984) and the variable performance of the PSRO's (Brook et al. , 1978; Dans et al., 1985; Dobson et al., 1978) led to the 1982 enactment of the Tax Equi ty and Fiscal Responsibil ity Act ( T E F R A ) and the Peer Review Improvement A c t (PRIA; Dans et al., 1985; L i p p , 1984). W i t h the passage of T E F R A , "two broad measures were approved: an interim program of tighter Medicare controls on hospitals..., a n d a directive to the Department of Hea l th and H u m a n Services ( D H H S ) to develop a prospective payment system (PPS) that, w h e n implemented w o u l d effectively impose a spending ceiling for Medicare o n hospitals" (Iglehart, 1982a, p. 1288). P R I A called for the creation of "one professional review organization (PRO) in each state to replace the P S R O program" (Lipp , 1984, p. 283). A l t h o u g h the federal administration wanted to abolish the PSRO's immediate ly , the congress elected to delay taking action unti l the new P P S could be implemented and the new system of external review cou ld be established (Dans et al., 1985). In contrast to the traditional method of per d i e m cost reimbursement, the tighter Medicare controls mandated by T E F R A were to be appl ied on the basis of cost per discharge (Iglehart, 1982a; Pointer & Ross, 1984). Thus , "to 98 implement the new regulations and to prepare for the eventual use of a prospective payment system, Congress directed the D H H S to develop a case mix index for every hospital" (Iglehart, 1982a, p. 1289). In order to comply wi th this directive, however, a method of defining and quantifying the mul t iproduct nature of the hospital was needed. In the State of N e w Jersey, an experimental prospective rate-per-case system suggested that D R G ' s cou ld be used for this purpose (May, 1985). A l t h o u g h a number of concerns were expressed about the appropriateness of us ing D R G ' s for reimbursement purposes (Doremus, 1980; Iglehart, 1982b; S imborg, 1981), they were quickly adopted by the federal government as the "only applicable a n d acceptable approach to product-oriented payment" (Johnson & A p p e l , 1984, p, 128). Thus , in October 1983, the federal government passed the Deficit Reduct ion A c t w h i c h p r o v i d e d the mandate for the implementat ion of a D R G - b a s e d prospective payment system (PPS) for the hospital inpatient care of Medicare beneficiaries (Dans et al., 1985; Iglehart, 1983). O n e of the major assumptions p r o m p t i n g this initiative was that "differences in efficiency rather than differences in characteristics of the patient populat ion or in the quality of care [were] responsible for interinstitutional differences in the cost of p r o v i d i n g care" (Stern & Epstein, 1985, p. 622). It was believed, therefore, that through the implementat ion of a system of un i form payments, quality of care cou ld be maintained while hospitals w o u l d be p r o v i d e d wi th financial incentives to i m p r o v e efficiency a n d to"provide care at the lowest possible cost for each patient admitted" (Stern & Epstein, 1985, p. 622). A s wi l l be discussed in more detail below, however , the incentives p r o v i d e d under the new P P S were complex and, consequently, it was anticipated that they might result in a n u m b e r of 9 9 unintended and deleterious effects (Gr imald i & Michelett i , 1984; W i l s o n , 1984). Therefore, in recognition of the potential for problematic responses, all hospitals participating in Medicare were required to enter into a contract wi th an approved external review organization, i.e., P R O , by October 1984, or risk the termination of their Medicare payments (Gr imald i & Michelett i , 1984). n . 5 . 2 Major Objectives of the PPS Since its inception in 1983, the pr imary goal of the federal PPS has been to "moderate the growth of government spending for health care and to force hospitals to assume greater financial risk for the care del ivered wi th in their institutions" (Iglehart, 1982b, p. 1655). In order to achieve this goal , hospital reimbursement is p r o v i d e d on the basis of a defined "clinical product" (Gertman & Lowenstein, 1984). The basic objectives of the PPS are thus: "1) to relate the prices pa id for caring for illnesses to the actual costs on an illness-specific basis; 2 ) to reward efficient hospitals and penalize inefficient ones; and 3) to encourage the involvement of key medical decision-makers (the physicians) in decisions regarding costs and resource allocation" (May, 1985, p. 17c). T h e first objective is to be met by establishing a national price per case that reflects the actual average cost of treating patients wi th in each of the 467 D R G ' s (Gertman & Lowenste in , 1984; M a y , 1985). Hospitals are p a i d the established D R G price, for each patient that falls into a g iven D R G category, "regardless of what it costs the hospital to care for that patient" (May , 1985, p. 17c). T h u s , the hospital's total reimbursement w i l l be determined by their case mix rather than their resource consumption (May , 1985; G e r t m a n & Lowenste in , 1984; Pointer & Ross, 1984). 100 T h e second objective, namely, rewarding efficiency and penaliz ing inefficiency, is expected to be achieved through the payment process (May, 1985). A s noted above, hospitals are pa id a set rate per case, regardless of the actual resources consumed or the patient's length bf stay. Thus , in contrast to the prev ious reimbursement method, w h i c h rewarded hospitals for keeping patients as long as possible and for p r o v i d i n g as m a n y services as possible, the new system is designed to reward hospitals financially for reduc ing the length of hospital stay, the intensity of service, and the costs of supplies and labour (May, 1985). In addit ion, because hospitals are permitted to keep any difference between their actual costs and the D R G rate, they are rewarded for increasing the efficiency of their operational procedures, and motivated to ensure the completeness and accuracy of their records (Iglehart, 1983; M a y , 1985). Conversely , since no addit ional funds wi l l be p r o v i d e d , in the event of deficits, hospitals that fail to prov ide necessary care wi th in the established rates wi l l be penalized (Iglehart, 1983; M a y , 1985). F ina l ly , the third objective, i.e., increased phys ic ian involvement , is expected to be achieved as physicians become cognizant of the effects their prescribed inpatient treatment patterns have on the financial solvency of the hospital (Iglehart, 1982a; M a y , 1985). T h e necessity for increased phys ic ian involvement is based o n three factors. First, since physicians have the p r i m a r y responsibility for admitt ing, treating a n d discharging patients, their involvement is crucial if the new reimbursement strategy is to succeed (May , 1985). Second, physicians are directly invo lved , through their record keeping behaviour, i n determining h o w m u c h their hospital is p a i d ( M a y , 1985). Thus , the accuracy and timeliness of their medical record keeping w i l l be an important issue (May, 1985). T h i r d , long term reductions 10 1 in hospital util ization require a change in medical practice patterns. In particular, as noted by T h u r o w (1985), treatment practices that conform with the traditional phi losophy of "Do no harm" need to be replaced by practices that are gu ided by the precept "Employ a treatment only w h e n y o u are sure that it w i l l make a noticeable improvement" (p. 613). Increased involvement in the management of hospital resources m a y facilitate such a change. n.5.3 Major Characteristics of the Medicare PPS A l t h o u g h the PPS was initiated in 1983, it w i l l not be fully implemented unti l the end of 1988. Thus , whi le the ultimate a im was to develop a single national price per D R G ( M c M a h o n , 1984), an effort was made, d u r i n g the early years of the PPS, to protect hospitals from u n d u e financial penalties by calculating the basic D R G rate us ing a b lend of hospital , regional, and national costs (Davis, A n d e r s o n , & Steinberg, 1984; M a y , 1985). W h e n implementat ion is completed, however, the calculation of the price per case wi l l be based on six variables: 1) the basic D R G rate (calculated us ing 100% national historical costs for persons wi th in each D R G ) ; 2) the urban or rural location of the hospital; 3) regional labor costs; 4) indirect teaching costs (direct educational expenses are re imbursed at cost); 5) the proport ion of the hospital's patients that are eligible for M e d i c a i d or Supplementa l Security Income; and 6) the patient's outlier status - i.e., whether the patient has an unusual ly long length of stay or h igh cost of care (Davis et al. , 1984; Jencks & Dobson, 1987). A t the present time, the PPS applies only to the payment of hospital inpatient care p r o v i d e d to Medicare beneficiaries. "Hospital outpatient services are excluded, as are certain rehabilitative, psychiatric, and long-term 102 care services [provided] in distinct parts of communi ty hospitals" (Davis et al. , 1984, p. 141). In addit ion, the PPS does not apply to the payment of capital costs nor to the "payment of wages and other direct costs of interns and residents" (Jencks & Dobson , 1987, p. 680). n.5.4 Implications of PPS for the Cost, Use and Qual i ty of Hospi ta l Care A s indicated above, the financial incentives p r o v i d e d under the PPS are complex. Thus , whi le it is generally agreed that the PPS provides strong incentives to reduce resource consumpt ion wi th in an inpatient episode (Davis et al., 1984; Gertman & Lowenstein, 1984; Jencks & Dobson , 1987; Pointer & Ross, 1984; Stern & Epstein, 1985; Wyszewiansk i , Thomas , & F r i e d m a n , 1987), critics of the PPS have identified a number of potential problems that cou ld reduce the effectiveness of the system, as it is presently structured (Davis et al., 1984; H o r n , Sharkey, Chambers & H o r n , 1985; Jencks, Dobson , Wi l l i s , & Feinstein, 1984; O m e n n , & C o n r a d , 1984; S imborg, 1981; Speigel & Kavaler , 1985; Stern & Epstein, 1985; Wennberg , M c P h e r s o n , & Caper , 1984). These concerns generally fall into two categories: 1) potential problems associated with the incentive structure of the P P S itself; and 2) potential problems associated with the use of D R G ' s as the method of def ining the "clinical product". In the fol lowing section, the specific problems identified on the basis of theoretical analyses of the PPS incentive structure are delineated. Later, in sections n.5.5 and n.5.8, organizational responses and empir ical data relevant to the hypothes ized problems are discussed. U.5.4.1 Potential Problems Associated wi th PPS Incentives A number of critics have suggested that, in order to increase revenues, hospitals m a y respond to the PPS by reducing quality of care 103 (Iglehart, 1983; M a y , 1985; Stern & Epstein, 1985; Wyszewiansk i , et al., 1987). Clear ly , the financial incentives p r o v i d e d by the PPS favour hospitals that can effect per case reductions in resource use. A s noted by Wyszewiansk i et a l , (1987), if such reductions are achieved through el imination of unnecessary services then quality of care wi l l not be affected adversely. H o w e v e r , if hospitals respond to the PPS by encouraging underuti l izat ion of needed services then quality of care wi l l be d iminished (Iglehart, 1983; Stern & Epstein, 1985; Wyszewiansk i et al., 1987). W h i l e the potential certainly exists for the PPS to achieve its objectives at the expense of quality, Wyszewiansk i et al. (1987) argue that, in fact, it could have the opposite effect. Specifically, they suggest that, in response to the potentially negative effects of the PPS, physicians m a y become more attentive to quality of care issues. In addit ion, as uti l ization goes d o w n , competit ion among hospitals w i l l l ikely increase. If so, hospitals wi l l be motivated to prov ide high quality care because this wi l l become an increasingly important factor in the competit ion for patients. Nonetheless, unti l it can be demonstrated empirical ly that the PPS has not reduced the quality of care p r o v i d e d to Medicare beneficiaries, quality wi l l remain a serious issue (May , 1985). In addi t ion to quality of care concerns, potential problems have been identif ied i n connection wi th the cost containment objectives of the PPS. For instance, since payment is by the case, hospitals m a y attempt to increase their total revenues by treating a larger number of cases (Gertman & Lowenste in , 1984; M a y , 1985; Stern & Epstein, 1985; Wennberg et al. , 1984). In particular, they m a y attempt to increase the admissions most l ikely to be profitable (i.e., the less severely i l l , a n d those that, in general, require fewer 104 services and shorter lengths of stay) while reducing those that are not (Gertman & Lowenstein, 1984; Stern & Epstein, 1985) In addit ion, through a similar selection process, hospitals may begin to restrict, o n the basis of profitability, the mix of services they prov ide (Eastaugh & Eastaugh, 1986; M a y , 1985; O m e n n & C o n r a d , 1984). Smits et al., (1984) suggest that since doctors, not hospitals, make admitt ing decisions, the extent to which hospitals can achieve such selectiveness may be l imited. H o w e v e r , W e n n b e r g et al., (1984), in a study of admiss ion variability wi th in D R G ' s , found evidence to suggest that admission decisions are subject to considerable medical discretion and variations in medical practice preferences. T h e y concluded, therefore, that as capacity increases, in response to PPS- induced reductions in length of stay, hospitals w i l l have no difficulty "recruiting physicians whose clinical interests coincide wi th a profitable mix of services, or 'product line' (p. 299). W h i l e some analysts of the PPS suggest that specialization by hospitals can have positive benefits, in the sense that by treating larger numbers of the same types of patients greater efficiencies m a y be achieved (Eastaugh & Eastaugh, 1986; M a y , 1985; O m e n n & C o n r a d , 1984), others fear that it wi l l result in l imit ing access to care for m a n y beneficiaries or in m a k i n g certain types of care unavailable to Medicare patients (Omenn & C o n r a d , 1984; Stern & Epstein, 1985). The P P S m a y also motivate hospitals to encourage physicians to split the inpatient episodes of patients wi th mult iple problems into two or more admissions, or to discharge patients early, even if it means a subsequent readmiss ion wi l l be required (Davis, et al., 1984; G e r t m a n & Lowenste in , 1984; M a y , 1985; Stern & Epstein, 1985). In this manner they cou ld receive payment for each admiss ion, thereby increasing their total revenues. The 105 possibility that these practices might be encouraged under the PPS has raised serious concern because, even before implementation of the PPS, readmissions were found to have been a major and costly problem for the Medicare p r o g r a m (Anderson & Steinberg, 1984). Since the PPS applies only to inpatient care, there m a y also be a tendency to shift a variety of inpatient procedures to an ambulatory setting (Davis et a l , 1984; Eastaugh & Eastaugh, 1986; Gertman & Lowenstein, 1984; Nathanson , 1984a). A l t h o u g h this is, in fact, one of the objectives of the PPS, in the short r u n it wi l l tend to increase costs because the D R G rates for inpatient services are based on historical costs (Davis et al., 1984; Eastaugh & Eastaugh, 1986). T h u s , by performing these procedures o n an outpatient basis, the hospital w o u l d receive payment twice. W h i l e such"double d ipping" may create problems initially, it has been argued that they wi l l become less of a concern when the D R G rates have been revised using post-PPS data (Nathanson, 1984a). Because the assignment of a D R G depends on the sequence of pr imary and secondary diagnosis, hospitals m a y also seek to increase reimbursement by manipulat ing the order ing of these diagnoses; a process referred to as " D R G creep" (Simborg, 1981). A l t h o u g h the PPS legislation ensures that outright manipulat ion of this sort wi l l be considered fraud, in m a n y cases the resequencing of the diagnoses to favour the more profitable D R G may be medical ly justifiable; i.e., there may be a difference of medica l op in ion as to w h i c h of several diagnoses should be listed first or second (Nathanson, 1984b; S imborg, 1981; Stern & Epstein, 1985). A n o t h e r potential prob lem that could have a significant effect on the cost, use a n d quality of care relates to the fact that surgical D R G ' s receive a 106 higher rate of payment than medical D R G ' s (May, 1985; O m e n n & C o n r a d , 1984)). T h u s , there m a y be a tendency to increase the number of surgical procedures performed. There m a y also be a tendency to delay the discharge of patients who are nearing the "outlier" trim point (Gertman & Lowenstein, 1984). A s noted earlier, hospitals can receive addit ional payments for patients w h o are determined to have unusual ly long lengths of stay or h igh costs of care. T h u s , w h e n it is not possible to discharge a patient within the time frame allotted for h i s / h e r D R G , the hospital m a y attempt to delay discharge until the patient has been in long enough to be eligible for an outlier adjustment. Final ly , because the federal PPS applies to the Medicare p r o g r a m alone, hospitals confronted with reduced Medicare payments could begin shifting costs to private payers (Iglehart, 1982a; Wennberg et al., 1984). In so do ing , the incentives to improve efficiency through reductions in length of stay and the use of ancillary services, w o u l d be considerably weakened (Davis et al., 1984; M a y & Wasserman, 1984). A l l of these potentially negative responses to the PPS could occur regardless of the accuracy of the case mix measurement system u p o n which it is based. H o w e v e r , If the classification of cases is inaccurate and results in unfair payments to hospitals, then addit ional problems cou ld be created. A s stated by Jencks & Dobson (1987), "case-mix adjustment for payment must help the prospective payment system to promote efficiency and contain costs whi le protecting beneficiaries and hospitals" (p. 680). T o protect beneficiaries, the payment rates must be "accurate enough so that hospitals cannot identify a n d avo id patients whose care wi l l cost more than the Medicare payment" (Jencks & Dobson , 1987, p. 680). A t the same time, to 107 protect hospitals, the payment rates must be h igh enough to ensure that Medicare patients receive medical ly appropriate care. Otherwise , "hospitals might face financial hardship or insolvency, and beneficiaries might face poor or inaccessible care" (Jencks & Dobson , 1987, p. 680). The D R G system, w h i c h is the cornerstone of the Medicare PPS, has been criticized for failing to prov ide the necessary level of accuracy (Conkl in , L ieberman, Barnes, & Louis , 1984; Jencks et al., 1984; O m e n n & C o n r a d , 1984; Stern & Epstein, • 1985). In particular, the system has been criticized for failing to account adequately for differences in severity of illness (Berki, 1984; Brewster, K a r l i n , et al., 1985; C o n k l i n et al., 1984; H o r n , Bulkley, et al., 1985; H o r n & H o r n , 1984; H o r n , H o r n & Sharkey, 1984; H o r n , Sharkey, Chambers & H o r n , 1985; M e n d e n h a l l , 1984; Y o u n g , 1984). Thus , although it is generally agreed that case-based reimbursement for hospitals is necessary (Johnson & A p p e l , 1984) the adequacy of the D R G system has become a topic of considerable debate. W h i l e the fo l lowing discussion w i l l focus on the limitations of D R G ' s for the purposes of reimbursement, most of the concerns identified relate equal ly to the use of D R G ' s for other purposes such as uti l ization review and uti l ization research (Conkl in et al., 1984; Wi l l i ams et al. , 1984). U.5.4.2 Concerns Regarding the A d e q u a c y of D R G ' s O n e of the major requirements of any case mix measurement system is that it identify groups of patients that are homogeneous wi th respect to the outcome variable of interest. A l t h o u g h an effort was made by Fetter and his associates to i m p r o v e the clinical homogeneity of the revised D R G ' s , critics of the system contend that excessive variabil ity wi th in D R G case mix groups remains a serious problem. E v e n the developers themselves have indicated that "if all D R G ' s from a large data base are plotted and inspected, 108 some groupings appear more cohesive than others (Smits et al., 1984, p. 72). Furthermore, many critics of the federal DRG-based PPS have pointed out that, as originally constructed, DRG's were intended to be used for the purposes of utilization review, not reimbursement (Doremus, 1980; Hornbrook, 1985; Johnson & Appel, 1984). Thus, they were designed to identify patients that were homogeneous with respect to length of stay, not costs. Although several investigators (Lave & Leinhardt, 1976b; Luke, 1979) have shown that length of stay and hospital costs are positively correlated, Doremus (1980) has indicated that "it would be naive to imagine that length of stay by itself is an accurate indicator of patient treatment costs" (p. 50). H e suggested, therefore, that to improve the accuracy of the prospective reimbursement system, a new set of cost-based DRG's should be developed. He acknowledged, however, that it would take some time for such a reformulation of the DRG system because of problems associated with obtaining the necessary cost data. Many factors have been identified in an attempt to explain the variability observed within DRG's. However, most of the debate surrounding the use of DRG's, for reimbursement as well as for other applications, has centred on the extent to which the DRG system accounts adequately for differences in illness severity. Several critics of DRG-based reimbursement have suggested that, to improve payment equity, the DRG system should either be replaced or supplemented with severity of illness measures (Conklin et al., 1984; Horn & Horn, 1984; Horn, Horn, & Sharkey, 1984; Nathanson, 1985). Accordingly, a number of investigators have proposed alternative or supplementary classification systems that attempt to identify and group patients along dimensions of illness severity. Included 109 among these alternative systems are the Severity of Illness Index (Horn , H o r n , & Sharkey,1984), Patient Management Categories (Young, 1984), A P A C H E II (Knaus, Draper , Wagner & Z i m m e r m a n , 1985), Disease Staging (Conkl in et al. , 1984), and M E D I S G R P S (Brewster, K a r l i n , et al., 1985). Furthermore, in studies done by H o r n a n d her associates, it has been found that the inclusion of a severity of illness adjustment of D R G ' s reduced the variability observed in both costs and length of stay per case by approximately 40% (Horn & H o r n , 1984; H o r n , H o r n & Sharkey, 1984; H o r n , Sharkey, Chambers , & H o r n , 1985). Likewise , C o n k l i n et al. , (1984) found that by using Disease Staging criteria to adjust D R G ' s , w i t h i n - D R G variabil ity in cost per discharge was reduced by 8 to 16%. Similar variance reductions have been attributed to the use of M E D I S G R P S - a d j u s t e d D R G ' s (Jencks & Dobson , 1987). A s indicated by Smits et al. (1984), however, in order to ensure that the D R G system w o u l d be "economical and feasible" only data that were available to all hospitals were used in constructing the D R G ' s . T h u s , it is to be expected that the inclusion of addit ional clinical variables (eg. severity measures) w i l l reduce the variabil ity within D R G ' s . H o w e v e r , since most of the severity measures referred to above require the abstraction of data from patients' medical records, a costly and time consuming process, it is uncertain whether the benefits of incorporat ing such data w o u l d outweigh their costs. Moreover , as Jencks, & D o b s o n (1987) point out, little evidence has been p r o v i d e d to suggest that any of the p r o p o s e d severity of illness measures offers a major improvement over D R G ' s , i n predict ing average cost per case across hospitals, particularly w h e n the P P S adjustments for location, teaching and 'disproportionate share' are taken into account. A s noted by M c M a h o n (1984), recognition of the fact 1 10 that patients may differ in illness severity does not automatically suggest that D R G ' s wi l l be inval id . Patients with the most severe illnesses do not necessarily require more resources than those w h o are less severely i l l . Nonetheless, because the presence of systematic variations in severity of illness, wi th in certain hospitals, may have implications for their ability to remain solvent under the PPS, severity of illness remains an important issue. A l t h o u g h Medicare officials are continuing to study the p r o b l e m . (Nathanson, 1985), it is unl ikely that severity-adjusted D R G ' s wi l l be adopted for the PPS, unti l the relevant data is abstracted on a regular basis (Berki, 1984). Nevertheless, it is recognized that D R G ' s must be updated on a regular basis, and revised and modif ied as new data becomes available (Berki, 1984). A l t h o u g h the proponents of severity adjusted D R G ' s argue that most of the observed w i t h i n - D R G variability is due to differences in severity of illness, others have suggested that severity of illness is not the only factor that contributes to heterogeneity wi th in D R G ' s . A s noted by Jencks and D o b s o n (1987), Smits et al. (1984), and Ger tman and Lowenste in (1984), variations in the appropriateness of care p r o v i d e d wi th in a given D R G , and differences in phys ic ian practice patterns are also l ikely to account for some of the observed variability. Neither of these factors, however, represent problems for the D R G system per se. Instead they are sources of variability that the D R G - b a s e d PPS is designed to identify and reduce. Other factors identified as potential sources of D R G heterogeneity, that m a y be problematic for the D R G system, include: 1) coding errors, 2) methodological problems, a n d 3) misclassification problems. 111 1) C o d i n g errors. O n e concern raised by critics of the revised D R G system relates to the accuracy of the discharge abstract data that are to be used by hospitals in assigning D R G ' s (Gertman & Lowenstein, 1984; Johnson & A p p e l , 1984). T o study this problem, Johnson & A p p e l (1984) compared the Medicare case mix of 26 hospitals, as determined from discharge claims data, to the actual case mix, as determined from each patient's medical record. A total of 75,893 Medicare discharges in 1980 and 77,640 discharges in 1981 were evaluated. In each case, a D R G based on the discharge abstract data was assigned as wel l as a D R G computed from the medical record. F r o m the 1980 data, agreement between the two D R G ' s was observed in only 49.4% of the cases. In an examination of the results by hospital, the range of agreement ranged from 31% to 74% with the higher agreement rates being most c o m m o n l y associated wi th small suburban hospitals. The lowest agreement rates were typically associated with the larger tertiary care hospitals. By 1981, the overall rate of agreement h a d increased to 53.0% with a range from 36% to 71%. N o changes in the relationships between hospital type and agreement rates were observed. T h e major reasons for the observed discrepancies were identified as: incorrect assignment of I C D - 9 - C M code, discharge diagnosis u n k n o w n at the time the claims data was submitted, incorrect procedure codes listed o n the discharge abstract, poor documentation and coding of secondary diagnostic a n d procedure information, and frequent omiss ion of, or errors in , the recording of sex a n d birth date information (Johnson & A p p e l , 1984, p. 130). G e r t m a n & Lowenste in (1984), in a review of studies done by the Institute of Management and Research of the Univers i ty Hospita ls of C leve land , echoed the concerns raised by Johnson & A p p e l (1984). ' 1 1 2 A c c o r d i n g to G e r t m a n & Lowenstein, the Cleve land studies reported error rates ranging from 20 to 40% which were attributed pr imar i ly to "sloppiness and to the use of unqualif ied record abstractors" (p.80). A l t h o u g h both Johnson & A p p e l (1984) and G e r t m a n & Lowenste in (1984) concluded that such coding errors posed a significant problem for all applications of the D R G system which rely u p o n historical discharge data, they also suggested that with time, experience, and the incentives p r o v i d e d under the new reimbursement system, the error rate w o u l d l ikely decline. 2) Methodolog ica l problems. Several concerns have been expressed wi th respect to the methods used to construct D R G ' s . Wi l l iams et al. (1984), for example, argued that the algorithm used to develop the D R G ' s (i.e, the A U T O G R O U P C L A S S I F Y algorithm) has three important limitations. First, it places greater emphasis o n means than on variances in identifying subgroups. T h u s , it "may fail to combine subgroups with slightly different means but similar variances, and it wi l l fail to separate subgroups with identical means but different variances" (p. 19). Second, it examines only one variable at a time and , therefore, could miss important interactions. T h i r d , it requires clinical judgement in the selection of the independent variables to use in the partit ioning process. Thus , subjectivity is built into the system at each level of subgroup formation. These inherent limitations in the C L A S S I F Y algori thm p r o m p t e d Wi l l i ams et al. to conclude that the D R G ' s w o u l d not represent homogeneous patient groups. H o w e v e r , since no attempt was made to evaluate the effect of these limitations o n actual hospital data, they were unable to determine to what extent these methodological l imitations w o u l d translate into actual problems for the D R G system. 1 13 A second methodological concern relates to the fact that DRG's were constructed on the basis of the I C D - 9 - C M coding system. A s indicated by M u l l i n (1985) and Smits et al. (1984), the I C D - 9 - C M coding system lacks specificity in the assignment of procedure codes. In particular, it "does not dist inguish among different surgical approaches to the same problem. This means that surgical procedures of different lengths and magnitude, and wi th dissimilar effects on patients and ultimately on resource use, m a y be coded in an identical fashion" (Smits et al., 1984, p. 75). Moreover , the use of unstandardized "coding conventions" by medical records departments could result in the assignment of I C D codes that are mis leading ( M u l l i n , 1985). The final methodological issue concerns the use of the "principal", as opposed to the "primary", diagnosis in the partit ioning of the D R G ' s . T h e pr inc ipa l diagnosis is the one which , after study, is determined to have been pr imar i ly responsible for triggering the admission. The p r i m a r y diagnosis, o n the other hand , is the one w h i c h is found to have accounted for the bu lk of the patient's hospital episode (Roberts, Re id , and Irwin, 1985; Smits et al., 1984). Since the pr imary diagnosis is identified in terms of total resource use, Smits et al. (1984) argued that its use in the formation of D R G ' s w o u l d l ikely i m p r o v e the classification of some patients. H o w e v e r , because they lacked a N o r t h A m e r i c a n data base that inc luded a code for p r i m a r y diagnosis, they were unable to determine h o w m u c h of an improvement it w o u l d make. Roberts et al. (1985), therefore, investigated this issue in Austra l ia , where the pr imary , rather than the pr inc ipal , diagnosis is the convention. U s i n g medical records data collected over a three m o n t h per iod in 1984, Roberts et al. found that, of 1064 records studied, 6.4% had a change in the sequence of diagnosis, w h e n the data were recoded using the 114 definition for the principal rather than the primary diagnosis. However, only 4.04% of the 1064 recoded records had a change in D R G as a result of the resequencing. These findings support the contention that there w i l l be changes in D R G classification for some patients when a different definition of the pr incipal /pr imary diagnosis is used. However, the magnitude of the changes observed in the Roberts et al. study were small, and no evidence was provided to indicate whether the changes did , in fact, improve the • homogeneity of the affected groups. 3) Misclassification problems. A final issue of importance to the use of DRG's for reimbursement or for utilization review, relates to the fact that several diagnoses may have been underrepresented in the database used in the construction of the DRG's. One example is cystic fibrosis. Of the approximately 1.5 mi l l ion records included in the original database, only 19 cases were found to have a primary diagnosis of cystic fibrosis (Horn, Horn , Sharkey, Beall, et al., 1986). The major reason identified to account for this underrepresentation was the finding that the D R G database d id not include data from hospitals wi th cystic fibrosis treatment centres (Horn, et al., 1986). Consequently, these patients may have been placed inappropriately into DRG's that do not accurately reflect their resource utilization. The possibility that other infrequent cases may be inappropriately allocated has thus raised concerns that there may be a number of "garbage can" DRG's; i.e., DRG's into which "miscellaneous low-volume cases have been dumped" (Mendenhall, 1984, p. 86). To the extent that this actually occurs, it could create problems for utilization review as wel l as for other applications, such as reimbursement. 115 In s u m m a r y , a l though the development of D R G ' s represents a major improvement in case mix measurement for the purposes of reimbursement, uti l ization review and research, it is still not considered to be problem-free. W i t h time and addit ional research, however, it is likely that most, if not all , of the current problems can be resolved. Nonetheless, because of the potential for problems, with both D R G ' s and the perverse incentives p r o v i d e d by the P P S itself, hospitals wi l l have to monitored closely so as to ensure that neither the quality of care nor access to care becomes problematic for Medicare beneficiaries. In the next section, the mechanisms that are currently being used to monitor hospital performance are described. n.5.5 T h e Role of External Rev iew Organizat ions U n d e r PPS In recognition of the potential negative effects of the PPS, a new external review system consisting of statewide PRO'S has been implemented ( G r i m a l d i & Michelett i , 1984). A l t h o u g h m a n y of the P R O ' s are s imply transformed PSRO's (Dans et al. , 1985), their goals and objectives differ considerably from those of their predecessors. U n d e r the new system of review, m u c h greater emphasis is placed o n quality assurance, admiss ion a n d readmiss ion review, and the review of medical appropriateness. Whereas the PSRO's were pr imar i ly concerned wi th reducing length of stay, such an emphasis is no longer necessary under the PPS; the payment system itself provides the necessary incentives to reduce length of stay. Moreover , since the rate of payment is based o n the case, not the length of stay, any failures to achieve reductions in length of stay wi l l result in financial problems for the hospital , not the government ( G r i m a l d i & Michelett i , 1984). 116 H.5.5.1 Goals and Objectives of PRO's T h e pr imary goal of the P R O program is to reduce inappropriate and unnecessary hospital uti l ization whi le maintaining an adequate standard of care. Specific objectives for the P R O program include: 1) shifting care from inpatient to outpatient settings, when it is safe and effective to do so; 2) moni tor ing patterns of admiss ion and val idating D R G assignments so as to prevent " D R G creep"; 3) monitor ing patterns of readmissions so as to detect any increases in readmission rates; 4) reducing the number , and adverse consequences of, invasive procedures; 5) ensuring that medical ly necessary services are prov ided; and 6) ensuring that hospitals are not re imbursed for inappropriate or unnecessary use (Dans et al., 1985). H.5.5.2 Characteristics of the P R O Program T h e P R O p r o g r a m has five major components: admiss ion review, outlier review, admiss ion pattern review, D R G val idat ion, and quality rev iew. 1. A d m i s s i o n review. Because the PPS pays a flat rate per Medicare discharge, and thereby provides hospitals with an incentive to increase admissions, P R O admiss ion review is conducted in order to ensure that unnecessary or inappropriate admissions are not re imbursed. A p p r o x i m a t e l y 10% of Medicare admissions are subject to preadmiss ion review (Lipp , 1984). For these patients, once the admiss ion has been approved , no further review is required (unless the patient becomes a "day" or "cost" outlier) and reimbursement is guaranteed. T h e remaining 90% of M e d i c a r e admissions are reviewed retrospectively. Consequently , in the majority of cases, hospitals wi l l face the risk of retroactive denials of re imbursement if admissions are found, after the fact, to have been 117 unnecessary or inappropriate (Lipp, 1984). T o prevent one form of "double dipping", PRO's are directed to give special attention to admissions that are transferred from acute care beds to beds which are exempt from the PPS (eg. rehabilitation units, psychiatric units, or swing beds). In addit ion, to ensure that patients are not being discharged prematurely, special attention is to be p a i d to patients readmitted to an acute care hospital wi thin seven days of a previous admiss ion (Grimald i & Micheletti^. 1984). v 2. Out l ier review. Since hospitals can receive addit ional payment for "day" or "cost" outliers, PRO's are responsible for ensuring the val idi ty of the outlier c laim (Gr imald i & Michelett i , 1984). For "day" outliers, the medica l reviewer is required to evaluate the appropriateness and necessity of all inpatient days prior to the outlier threshold. If a number of these days are found to have been unnecessary, no addit ional payment wi l l be made; i.e., the hospital w i l l receive the set D R G rate and no more. For "cost" outliers, the medical reviewer is required to review the necessity of the admiss ion and the services rendered over the course of the inpatient episode. If any of these services are found to have been unnecessary, bi l led twice, not actually performed, or not ordered by a physic ian, addit ional payment m a y be denied (Gr imald i & Michelett i , 1984). 3. A d m i s s i o n pattern review. Since hospitals can increase their revenues b y increasing their Medicare admissions, PRO's are required to monitor admiss ion patterns. Some of the potential problems that the P R O w i l l be seeking to identify include: 1) a tendency to treat outpatients o n an inpatient basis; and 2) an increase in readmission rates resulting f rom patients being discharged prematurely or from a tendency to "use more than one stay to care for patients wi th mult iple problems" (Grimald i and 118 Michelett i , 1984, p. 31). A d m i s s i o n pattern review is conducted on a quarterly basis. A screening mechanism (eg. profile analysis) is used to identify hospitals with unusual changes in length of stay, number of admissions or readmissions, and case mix. Once aberrant hospitals have been identified, then the P R O is required to take "necessary corrective action, w h i c h m a y involve education activities, denial of payments , or expanded review" (Grimald i and Micheletti , 1984, p. 32). 4. D R G val idation. Because reimbursement is p r o v i d e d on the basis of D R G assignment, the PRO'S are required to review, on a quarterly basis, the accuracy and completeness of the clinical information used by the hospital in assigning patients to D R G ' s . In particular, the reviewer is to pay attention to "whether hospitals are coding clinical information in a way that i m p r o p e r l y maximizes reimbursement rather than accurately reflects a patient's medica l condition" (Gr imald i and Michelett i , 1984, p. 33). If inaccuracies in the clinical information are found, or i m p r o p e r patterns of D R G assignment - " D R G creep" - are discovered, financial penalties m a y be i m p o s e d ( G r i m a l d i and Michelett i , 1984). 5. Qua l i ty review. There is considerable uncertainty about what effect the P P S w i l l have o n the quality of inpatient care (Jencks et al. , 1984; O m e n n & C o n r a d , 1984; Stern & Epstein, 1985). In particular, there are fears that hospitals, in the effort to economize, w i l l encourage underut i l izat ion of necessary services a n d premature discharges of some patients (Gr imald i & Michelett i , 1984). If so, the PPS could have a negative effect on quality of care. T o counteract this possibility, PRO's are required to conduct quality reviews a n d to participate in national studies of abusive practices. A t the present time, however, no specific criteria or guidelines for quality review 119 have been established (Grimaldi & Micheletti, 1984). Hence, until such criteria and quidelines can be developed, the PRO's are expected to use, as indicators of a reduction in quality, unprecedented reductions in a hospital's average length of stay, and/or an increase in complaints from nursing > homes and home health care agencies that patients are being discharged from the acute care setting in worse condition than they had been previously (Grimaldi & Micheletti, 1984). If such changes are observed, PRO'S will be required to investigate further to determine whether patients are being denied essential services or are being discharged prematurely. If patterns of abuse are detected, the hospital can be excluded from participation in Medicare or financial penalties can be imposed on the physician(s) involved and/or the hospital (Lipp, 1984). U.5.6 The Role of Internal Review Committees Under PPS Under the PPS, a hospital utilization review committee is required only for those hospitals that are not yet under contract with a PRO. Once, the PRO program has been implemented nationwide, hospital utilization review activities will cease to be mandatory requirement (Grimaldi & Micheletti, 1984). It has been argued, however, that with the advent of the PPS, the importance of the utilization review committee, to the hospital itself, will increase substantially (Grimaldi & Micheletti, 1984; Owens & Averill, 1984). In order to remain financially solvent, hospitals will need to establish procedures for identifying inappropriate and unnecessary utilization of hospital days and ancillary services. In addition, they will need to integrate their utilization review and quality assurance activities in order to avoid expensive duplication (Grimaldi & Micheletti, 1984). Most importantly, utilization review committees will need to be empowered to 120 identify the reasons for misuti l ization and to implement appropriate corrective action (Gr imald i and Micheletti , 1984; O w e n s & A v e r i l l , 1984). II.5.7 C u r r e n t Trends in Ut i l izat ion Review Procedures W i t h the new emphasis o n appropriateness of acute inpatient care, PRO's and hospitals need tools that wi l l enable them to identify, in an efficient and cost-effective manner, inappropriate admissions, days-of-care use, a n d ancillary service utilization. Historical ly , concerns about inappropriate uti l ization have emphas ized overuse. H o w e v e r , inappropriate use also refers to underuti l ization (Payne, 1987). W i t h the advent of the PPS, and its strong incentives to reduce util ization, concerns about underuse, and its close associate, quality of care, are increasing (Dans et a l , 1985; Wyszewiansk i et al., 1987). T o date, however, most of the uti l ization review procedures developed to identify inappropriate use still remain focussed o n the prob lem of overuti l ization (Payne, 1987). A n d most of this effort has been directed at the identification of unnecessary or inappropriate admissions and days-of-care use; little effort has as yet been appl i ed to the problem of assessing inappropriate util ization of ancillary services (Hughes et al., 1984). Consequently , only the methods that have been developed to assess unnecessary or inappropriate admissions and days-of-care use are discussed herein. A s has already been stated, one of the most c o m m o n l y used procedures to identify unnecessary util ization is profile analysis; i.e, the retrospective evaluation of changes in length of stay patterns by D R G ' s , physicians, and inpatient service units. By establishing length of stay norms by D R G ' s , patients wi th unusual ly short or unusual ly long stays can be identified. V e r y short stays may reflect inappropriate admissions; 121 exceptionally long stays m a y reflect inappropriate days-of-care (Goran, 1979). Length of stay profiles are l imited in usefulness, however, because they prov ide no information as to the source or nature of the prob lem (Payne, 1987). In addit ion, they are vulnerable to the coding conventions used by specific hospitals. For example, under PPS there is a strong incentive for hospitals to record mult iple diagnoses, complications a n d / o r comorbidities in the sequence that maximizes reimbursement (Simborg, 1981). But, "if a patient is placed in a category with a higher severity level due to coding practices designed to maximize reimbursement, he or she wi l l probably have a lower length of stay than others w h o are proper ly in the category" (Payne, 1987, p. 739). Consequently, to the extent that such coding practices occur, length of stay profiles could fail to identify patients that, had they been proper ly classified, w o u l d be in need of further review. In addit ion, since profile analysis is based o n the use of D R G ' s , it is vulnerable to all of the problems discussed earlier in relation to D R G heterogeneity. Final ly , as was also ment ioned previously , since empirical evidence suggests there is no direct relationship between length of stay and inappropriate use (Restuccia & H o l l o w a y , 1976; Z i m m e r , 1974), length of stay profiles may miss uti l ization problems that occur d u r i n g inpatient episodes that do not violate the established norms (Payne, 1987). A more promis ing approach to the identification of overuti l ization, employs explicit criteria to assess the appropriateness of inpatient days and to identify the reasons for misutil ization. A t present, the A E P , wh ich was described in section n.4.8, supplemented by a method for recording the reasons for inappropriate use, is considered the "state of the art" in appropriateness of care assessment (Payne, 1987). Since the development of 122 the A E P , however, other criterion-based review instruments have been developed and are currently being used by hospitals and PRO's to assess the appropriateness of acute level care. For example, Interqual, Inc. has deve loped a review system called the Intensity of Service, Severity of Illness, a n d Discharge Screens wi th an Appropriateness component (the ISD-A; L a m p r e y & Jacobs, 1987). Similarly, Systemetrics, Inc. has developed a tool called the Standardized M e d r e v i e w Instrument (SMI; Payne, 1987). Both of these instruments, like the A E P , can be appl ied in both admiss ion and day-of-care review. There is considerable variabil ity, however, a m o n g the three tools in the number of criteria employed , the complexity of their decision rules, and the extent to which they have been tested for reliability and val id i ty . A s indicated earlier, the A E P consists of 27 day-of-care criteria and 18 admiss ion criteria. In contrast, the I S D - A consists of 69 generic criteria plus an addit ional 20-60 criteria which are appl ied on the basis of the b o d y system affected. Specific criteria have been developed for a total of 12 b o d y systems as wel l as for psychiatric patients (Payne, 1987). The S M I , on the other hand , consists of 117 admiss ion criteria (ordered from the most to the least objective), 30 level of care criteria, and 26 continued stay criteria (Payne, 1987). T h e decision rule for the A E P is relatively straightforward. For an admiss ion or day of care to be considered appropriate, on ly one of the relevant criteria needs to be met (Gertman & Restuccia, 1981). W i t h the ISD-A , "any patient admitted to the hospital must [meet] either one severity of illness (SI) or one intensity of service (IS) criterion o n admiss ion and must have met both an SI and an IS criterion by the first review fo l lowing the 123 complet ion of 24 hours in the hospital" (Lamprey & Jacobs, 1987, p. 6). Final ly , for the S M I , if one of the 117 admission criteria are met, the admiss ion is considered appropriate; if one of the level of care criteria and one of the continued stay criteria are met, the day of care is considered appropriate (Payne, 1987). W i t h respect to the reliability and val idi ty of the three instruments, results have been reported in the literature for the A E P and the S M I , only. T h e I S D - A is proprietary and, to date, no results f rom the systematic testing of the instrument, if it has even been conducted, have been publ i shed (Payne, 1987). In contrast, the A E P has been subjected to extensive testing and has been found to be reliable and va l id by both the instrument's developers and others (Gertman and Restuccia, 1981; Siu et al., 1986; Stumwasser, Paranjpe, et al., as cited in Payne, 1987). T h e S M I , o n the other hand , has been found to have low reliability and, as a result, it has been suggested that it no longer be used for the purposes of uti l ization review (Stumwasser, Paranjpe, et al., as cited in Payne, 1987). Appropriateness measures have two major limitations. Firstly, as wi th the length of stay profiles, appropriateness measures are intended to be screening devices. Thus , any problems identified through these procedures must be considered tentative unti l a more indepth review of the particular case can be conducted. In particular, since the initial reviewers are usually nurses, questions concerning the appropriateness of an admiss ion, . cont inued stay or reimbursement must be referred to a designated physic ian advisor before any decisions are made (Payne, 1987). Secondly, all three measures inc lude some component of subjective judgement. T h u s , there is a potential for reviewer bias. T o min imize this potential, reviewers must by 124 thoroughly trained, supervised, and monitored so as to ensure that the criteria are appl ied appropriately and effectively (Payne, 1987). The application of tools such as the A E P and the I S D - A in util ization review and uti l ization management is a recent development. Thus , at the present time, it is u n k n o w n whether such procedures are effective in reduc ing inappropriate use. In addit ion, it is u n k n o w n whether they are more or less effective or cost-effective than profile analyses. Intuitively, it w o u l d seem reasonable to assume that by targeting all inappropriate use, rather than just unnecessarily long stays, greater reductions in inappropriate use could be achieved. Whether these reductions wi l l be sufficiently large to warrant the further development and implementat ion of these techniques wi l l depend on the extent to which the inappropriate use they identify can be controlled by the hospital and its medical staff. A s noted earlier, pre l iminary research suggests that the major reasons for inappropriate use are related to physic ian practice patterns and the lack of appropriate alternatives to acute care. In both cases, time wi l l be required before the necessary corrective action can be taken; neither is immediately under the control of the hospital or its medical staff. Therefore, the development, testing, and evaluation of appropriateness measures wi l l have to be conducted over a sufficiently long per iod of time to permit the development a n d implementat ion of the appropriate long term strategies for change. n.5.8 Ear ly Effects of the PPS on Hospi ta l Uti l ization, Costs, and Qual i ty G i v e n that the PPS is still i n the process of being phased in , it is too early to determine whether or not it has been effective in reducing the rate of increase in hospital expenditures. In fact, it wi l l be a number of years before sufficient stability is achieved within the system to permit researchers 125 to draw any f irm conclusions about its effects on hospital costs, utilization or quality (May, 1985; M a y & Wasserman, 1984). In the interim, a few pre l iminary studies of national trends in hospital costs, use, and quality have begun to appear in the literature. For the most part, however, the studies that have been undertaken, to date, to evaluate the effectiveness of D R G - b a s e d prospective payment have focussed on the experience of the experimental program that was implemented in the State of N e w Jersey in 1980. Since this system was used as the mode l for the Medicare PPS, the findings from the N e w Jersey experiment are expected to have important implications for the Medicare system(Hsiao & D u n n , 1987; M a y , 1985; M a y & Wasserman, 1984). H o w e v e r , there is one important difference between the two systems that wi l l l imit the generalizability of the N e w Jersey studies. In contrast to the federal PPS, wh ich applies only to Medicare beneficiaries, the N e w Jersey system applies to all payers (May, 1985; M a y & Wasserman, 1984; Iglehart, 1982b). Consequently , there is a greater potential for cost shifting under the federal system (Iglehart, 1982b; Wennberg et al., 1984). Therefore, the f indings concerning the effectiveness of the N e w Jersey system are likely to prov ide an overestimate of what can be achieved under the federal system ( M a y & Wasserman, 1984). n.5.8.1 Nat iona l Trends in Uti l izat ion, Costs and Qual i ty For the most part, the f indings from the available national trend studies (Des Harnais et al., 1987; K e l l y & Bankhead, 1985; Shortell & Hughes , 1988) are i n accordance wi th the objectives of the PPS. A l t h o u g h the data inc luded i n these studies relate only to the first year fo l lowing implementat ion of the PPS, comparisons to pre-PPS trends indicate that the average length of stay h a d decreased significantly (Des Harnais et al., 1987; 126 Ke l ly and Bankhead, 1985), and that total hospital costs were approximately "$2 bi l l ion dollars less than Medicare officials had expected to pay for hospital charges [in 1984]" (Kelly & Bankhead, 1985). Moreover , despite fears that the P P S w o u l d generate increases in admiss ion and readmission rates, no significant increases in either rate were observed in 1984 (Des Harnais et al. , 1987; Ke l ly & Bankhead, 1985). In fact, Des Harnais et al. (1987) reported that, for the 729 hospitals inc luded in their data base, Medicare admissions decl ined by 5.4% in 1984. Similarly, but in a more specific study, a recent examination of readmission rates among Medicare patients treated for congestive heart failure revealed no significant increases in readmissions before and after implementation of the PPS (Rich & Freedland, 1988). These f indings are of l imited usefulness, however, because they are based on the experience of only one hospital, and are relevant to only one D R G . A l t h o u g h the above mentioned findings m a y be suggestive of PPS effects, none of the studies referred to above have been able to attribute their results directly to the influence of the PPS. Most of the observed trends are continuations of trends that h a d begun prior to the implementat ion of the PPS (Des Harnais et al., 1987; Ke l ly & Bankhead, 1985). In addit ion, no effort has as yet been made to take into account the effects of other factors such as P R O ' s , or the growth of prepaid practice organizations, on hospital uti l ization a n d total health care expenditures (Kelly & Bankhead, 1985). F inal ly , no studies have as yet begun to examine the cost-effectiveness of the PPS. W i t h respect to the effect of the P P S on the quality of care, the situation is even more uncertain. Des Harnais et al. (1987) examined trends i n mortal ity rates, comorbidity / compl i ca t ion rates, and severity of illness 127 from 1980 to 1984 and found no significant increases in any of these measures following implementation of the PPS. However, in a more detailed study of the effects of regulatory constraints, payment constraints, and competition on inpatient mortality rates, Shortell and Hughes (1988) found significant associations between high mortality rates and the stringency of prospective rate-review programs. In addition, hospitals in highly competitive markets (as measured by enrollment in HMO's) were found to have higher mortality rates than hospitals in less competitive markets. While the authors acknowledged that their findings were exploratory, and limited to 16 medical conditions, the fact that they were based on data from 981 hospitals across 45 states led them to conclude that there is a "need for improved monitoring of the issue of the quality of care and patients' outcomes as regulatory and competitive approaches to hospital cost containment continue to become more stringent" (p. 1106). n.5.8.2 The Results of the New Tersey Experiment: Implications for the  Medicare PPS In one of the earliest studies to examine the effects of the New Jersey DRG-based prospective payment system, May & Wasserman (1984) compared a matched sample of D R G and non-DRG hospitals. A l l eight of the D R G hospitals included in the study had entered the prospective payment system during the first year of its operation in 1980. Data from 1979 to 1983 was examined to determine whether the D R G system had had a measurable effect on organizational and economic variables. The relevant findings of this study are summarized as follows: 1. The amount and quality of data collected by the D R G hospitals improved following implementation of prospective payment. The number 128 of incomplete discharge abstracts dropped from 22.8% to 15.8%. However , the time taken by medical records departments to complete the data abstraction process increased from 4.5 days to 5.3 days. 2. Physicians were more invo lved in managerial decis ion-making in the D R G hospitals than in the n o n - D R G hospitals. 3. O n average, it cost each of the eight D R G hospitals studied an addit ional $91,092 per a n n u m to prepare for and implement the new prospective payment system. 4. In 1980, each of the hospitals that entered the new system received an average of $2.3 mi l l i on more in reimbursement than they w o u l d have under the state's previous reimbursement plan. 5. In large part due to the influx of addit ional revenues, which were p r o v i d e d to hospitals in order to offset start-up costs, there was no measurable reduction in the rate of cost increases that cou ld be attributed to the new reimbursement system. O n the basis of these findings, M a y & Wasserman (1984) concluded that, in terms of data quality and physic ian involvement, the N e w Jersey system was hav ing the anticipated effects. W i t h respect to cost containment, however, they were more cautious. A l t h o u g h none of the data they examined suggested that the system h a d been effective in containing costs, they noted that it w o u l d be some years before it w o u l d be possible to determine whether or not the expected gains in efficiency w o u l d be sufficient to offset the substantial start-up costs that were required for the system's implementat ion . In a more recent study, Hs iao & D u n n (1987) performed time series analyses on data from 80 of the 97 hospitals that had entered the N e w Jersey 129 system by the end of 1982. The dependent variables used in their analyses inc luded total per capita expenditures, total costs per case, average length of stay a n d inpatient admissions. The time per iod covered in the study extended from 1971 to 1984. It should be noted that, d u r i n g this per iod, and prior to the introduction of the D R G - b a s e d prospective payment system, a prospective per d i e m system of reimbursement had been implemented by the State of N e w Jersey. This system was implemented in 1976 and remained in effect until its replacement by the D R G system in 1980 (Hsaio & D u n n , 1987). Thus , the time trend analysis conducted by H s a i o & D u n n covered three different reimbursement periods: the pre-prospective payment per iod , 1971-1975; the per d iem prospective payment period, 1976-1979; and the D R G - b a s e d prospective payment per iod, 1980-1984. For the purposes of the analyses of expenditures, all cost data was converted to 1969 constant dollars. H o w e v e r , no control group was inc luded in the analysis nor was any attempt made to include additional factors, other than time and intervention, that might affect the study's dependent variables. A l t h o u g h these limitations in the study's design make it impossible to attribute any of the observed trends directly to the new payment system, the implications of the f indings for the Medicare PPS merit their discussion. T h e major f indings are s u m m a r i z e d as follows: 1. A significant reduction in the rate of growth of per capita expenditures was associated wi th the introduction of the prospective per d i e m system. Prior to the implementation of this system, per capita costs had been increasing at a rate of 8% per year; fo l lowing its implementation the rate was 2%-3% per year. This trend continued fol lowing implementat ion of the D R G system. 130 2. Total costs per admission increased at an average annual rate of 3%-4% prior to 1976. Following implementation of the per diem system, the rate declined to 2% per annum. With the implementation of the DRG system, the rate declined to less than .5% per annum. For both of these systems, the decrease was significant. 3. Average length of stay changed little between 1971 and 1975. With the introduction of the per diem system, however, a significant increase in length of stay was observed. In contrast, following implementation of the DRG system, length of stay dropped significantly. 4. Prior to the implementation of the per diem system, admissions were increasing at a rate of 4% per year. After the per diem system was implemented this rate slowed to .7% per year. This change was significant. With the advent of the DRG system, admissions started to increase slightly, at a rate of 1 % per year, but this increase was not statistically significant. As noted by Hsiao and Dunn (1987), these results suggest that although the DRG system appears to have been effective in reducing length of stay and, consequently, the costs per admission, it has been no more effective than the per diem prospective system in controlling the rate of growth of overall per capita expenditures. In attempting to explain this result, the authors suggested that the increase in the number of admissions observed, following implementation of the DRG system, may have been sufficient, even though not statistically significant, to offset the cost savings generated by the reductions in length of stay and costs per case. However, it should be emphasized that both prospective payment systems were more effective than the retrospective payment system in containing costs and reducing utilization. Although Hsiao & Dunn did not attempt to measure 131 the effect of the N e w Jersey system on the quality of care prov ided , they concluded that because of the system's effect on length of stay, there is a need to apply vigilance in monitor ing discharge practices. In addit ion, because of the slight increase in admissions observed under D R G - b a s e d prospective payment, they suggested that there is "a need to regulate and moni tor the appropriateness of clinical services" (p. 219). II.5.9 C o m m e n t s T h e evolut ion of uti l ization review a n d uti l ization management in the U n i t e d States has been a dynamic process spanning more than two decades of study, debate, and, more often than not, expensive trial and error. A n d , even though knowledge has been gained, m u c h is still u n k n o w n about the factors that affect hospital utilization. Clearly , util ization is a complex process that is not easily addressed by simplistic assumptions and pol icy initiatives. A mult ipl ic i ty of factors contribute to the pattern of uti l ization observed in any given hospital, and , consequently, attempts to effect changes in util ization have been hampered by a lack of data and an inabil ity to study more than a few of the important variables at a time. This basic lack of understanding has contributed to the numerous unanticipated effects of the government's attempts to regulate hospital uti l ization and to impose limits o n expenditures for hospital-based medical care. The most salient example of unanticipated effects is the P S R O p r o g r a m that dominated the publ ic ly f inanced hospital sector throughout the 1970's. A l t h o u g h at the time it appeared to have merit intuitively, the expenses incurred in implement ing , maintaining, a n d moni tor ing the p r o g r a m far outweighed any of the observable benefits. Yet, it took almost a decade for the limitations of the P S R O program to be recognized and addressed. A n d 132 even though, by the early 1980's, it was widely acknowledged to have been a failure, the subsequent development of the PPS, the implementation of D R G ' s , and the development of the PRO's , has proceeded once again without adequate pilot testing and evaluation. In particular, one of the last topics to be addressed in any of the pol icy initiatives that have been enacted in the U n i t e d States, has been their cost-effectiveness. W i t h the advent of the D R G - b a s e d PPS, it is somewhat surpris ing that more time was not taken to study its effects before engaging in nationwide implementation. A l t h o u g h , clearly, this was a move born of necessity, given the state of the Medicare trust fund , one might have expected a little more diligence in v iew of the expensive and t ime-consuming failure of the P S R O p r o g r a m effort. For C a n a d i a n pol icy analysts, hospital administrators, health service researchers, a n d government officials, thoughtful consideration of the A m e r i c a n experience is important for a number of reasons: 1. It reveals the enormous amount of time and money that can be spent on unproduct ive strategies w h e n p r o g r a m 'development and pol icy formation is based on trial and error rather than on empirical ly der ived knowledge . 2. It underscores the importance of pilot projects, complete with formal evaluat ion plans, in p r o g r a m development. 3. It highlights the need for patience before implement ing program strategies that invo lve major publ ic expenditures. 4. It provides a major rationale for the promot ion of health services research, an area of research that, to this day, remains h ighly undeve loped in both the U n i t e d States and Canada . 133 A l t h o u g h the C a n a d i a n and A m e r i c a n health care systems differ in m a n y important ways, an understanding of the A m e r i c a n experience could help C a n a d i a n p r o g r a m developers avo id some of the more costly errors incurred in the U n i t e d States. Thus , for example, it is hoped that through this rather long description of the history of util ization review and management, readers of this document wi l l have gained an increased appreciation for the importance of administrative and organizational research in the hospital environment. A s wel l , since uti l ization review and uti l ization management are topics that are only just beginning to attract the attention of C a n a d i a n hospital administrators and other health service professionals, it is hoped that a sincere effort wi l l be made to incorporate p r o g r a m evaluation into the design, implementat ion and maintenance of any uti l ization management p r o g r a m efforts. In so do ing it m a y be possible to avo id some of the more deleterious effects of such programs and maintain the C a n a d i a n hospital-based care at the h igh standard that it has historically enjoyed. 134 C H A P T E R HI Ut i l i zat ion Management in C a n a d a :  T h e S W I T C H Index System and Other Developments HI . l T h e Conceptual Context of Ut i l izat ion Management in C a n a d a A s indicated previous ly , hospital uti l ization management has only recently begun to attract attention in Canada . Consequently , little information on the nature and status of uti l ization management in the C a n a d i a n hospital sector has as yet been published. Nonetheless, there are indications that government officials, health service researchers, hospital administrators and other health care professionals are beginning to recognize and show an interest in techniques designed to monitor and i m p r o v e the management of hospital resource use (Anderson & L o m a s , 1988). For example, at a national level, interest in promot ing improvements in resource management methodologies has been reflected in the ongo ing deve lopment and implementat ion of the Management Information Systems (MIS) Project (Bolley, 1987; L o w r y , 1983). The M I S Project, w h i c h "is a co-operative effort of the C a n a d i a n Hospi ta l Associat ion, the federal a n d provinc ia l governments and the provinc ia l hospi ta l /heal th associations" (MIS Project, 1986, p. 27), is intended "to make data management and collection comparable" (p.27) and to prov ide a "standard and potentially powerfu l tool for moni tor ing resource use" (Anderson & L o m a s , 1988, p.17). Likewise , at the hospital level, uti l ization management p r o g r a m descriptions and case studies have begun to appear in the literature (Coombs & Richter, 1986; El iasoph & Hassen, 1986; 1987; M c D o n o u g h & V a z , 1987; M c G e o r g e , de M o r a , & Legros, 1985; N u s b a u m , 1988; Winche l l , 1985). In addit ion, health service professionals have begun to identify 135 appropriate util ization as one of the major challenges facing hospitals in the 1990's (Lazare, 1987); at least one health association has recently held a w o r k s h o p on uti l ization management (British C o l u m b i a Heal th Associat ion, 1987); and papers call ing attention to the need for improved hospital resource management have become c o m m o n (Canadian Hosp i ta l Associat ion, 1984; Clemenhagen, C h a m p a g n e , Contandr iopoulos , & Pineault, 1985; Kl ic ius & Sawyer, 1986; Komes , 1988; Mar ies , 1984; M a c K e n z i e , M a r k l e , & Croke , 1987; Schultz, 1986). Final ly , at a conference he ld by the Centre for Health Economics and Policy Analys is ( C H E P A ) in M a y , 1988, four papers were presented, each of wh ich dealt exclusively with the topic of uti l ization analysis in C a n a d a (Anderson & Lomas , 1988; Rachlis & Fooks, 1988; Roos & Roos, 1988; Suttie, 1988). Despite these indications of interest, uti l ization management has been slower to emerge as a priority for C a n a d i a n hospitals than for their A m e r i c a n counterparts. It wi l l be recalled that, in the U n i t e d States, uti l ization review and management was mandated by the federal government in an attempt to contain rap id ly escalating costs in the publ ic ly f inanced hospital sector. A l t h o u g h it was subsequently recognized that, in the absence of direct financial controls, util ization management was not p r o d u c i n g the desired effect, the mixture of publ ic and private funding sources and del ivery systems that comprise the A m e r i c a n health care scene prohibi ted the implementat ion of the kinds of controls, such as prospective global budget ing, deemed necessary to achieve cost containment (Evans, 1983; Iglehart, 1982a). W i t h the advent of D R G ' s , however, a tool was p r o v i d e d for the introduction of prospectively appl ied financial constraints and, consequently, the Un i t ed States is currently attempting to contain 136 hospital expenditures through a combinat ion of financial and uti l ization controls. In contrast, Canada's federal government has not responded to cost concerns by mandat ing uti l ization controls. A l t h o u g h it has been recognized that inappropriate and unnecessary util ization of hospital facilities and services contribute to overall hospital expenditures (Bennett & K r a s n y , 1977; C a n a d i a n Hospital Association, 1984; Evans , 1983; 1984; Soderstrom, 1978), and thereby to total health care expenditures, constitutional arrangements that place responsibility for all health and hospital services, health education, and regulation of the health professionals wi th the provinc ia l governments, prohibit the C a n a d i a n federal government from enacting legislation a imed directly at controll ing hospital uti l ization (Evans, 1983; LeCla i r , 1975;.Soderstrom, 1978). H o w e v e r , because C a n a d a has a national health insurance p r o g r a m w h i c h is financed, in part, by the federal government, it can attempt to influence the way in w h i c h prov inc ia l governments structure and administer prov inc ia l health services by changing the financial incentives p r o v i d e d under the terms of Federal -Provincia l cost-sharing agreements (Evans, 1983) Consequently , w h e n concerns about ris ing health care costs began to surface in the m i d -1970's, the federal government responded by passing the Federal-Provincial Fiscal Arrangements a n d Established Programs Financing A c t ( E P F A ) . W i t h the enactment of the E P F A , the previously open-ended cost-sharing formula , wh ich essentially w o r k e d out to a 50-50 federal and provincial contribution, was replaced wi th block grants and tax transfer payments des igned to l ink "the annual increase in the federal contribution...to the growth of the G N P " (Iglehart, 1986a, p. 207). In addit ion, annual grants of $20 per capita (indexed to the G N P ) were introduced in an attempt to 137 promote the development of less costly alternatives to hospital care (Iglehart, 1986a). In essence, the a im of this legislation was to transfer responsibil ity for cost containment to the provinces by placing them at greater financial risk "when aggregate outlays for health care [grew] at a faster rate than the economy as a whole" (Iglehart, 1986a, p. 207). In so do ing , prov inc ia l governments were p r o v i d e d wi th greater incentives than previous ly to f ind ways to improve the efficiency and reduce the costs of hospital (and other) services. A t the provinc ia l level, governmental responses to the changes in the federal fund ing formula have s imilarly been aimed pr imar i ly at changing traditional financial incentives. But in this case, the change in incentives has been directed main ly at the hospital sector. Thus , in most provinces, hospitals are n o w funded on the basis of prospectively determined global budgets (Evans, 1983). This contrasts wi th the earlier system of line-item budgets and actual cost reimbursement which p r o v i d e d hospitals with little flexibility in the way in which funds could be used and few incentives to contain costs, since budget deficits were typically absorbed by the government (Soderstrom, 1978). W i t h the introduction of global budgets, however , most prov inc ia l governments have introduced a pol icy whereby hospitals that overrun their budget do not automatically receive deficit funding . Moreover , in some instances, hospitals that manage to end the year wi th a surplus are permitted to retain at least a part of that surplus (Soderstrom, 1978; British C o l u m b i a Min i s t ry of Heal th , 1985; M c D o n o u g h & V a z , 1987). T h u s , as a result of these changes, hospitals have been p r o v i d e d w i th greater incentives to contain the costs of their operations 138 since any deficits they incur may have to be made u p through sources outside the prov inc ia l insurance program. A second response of provincia l governments has been to curtail hospital expansion (Bennett & Krasny , 1977). Such a response is easier to implement , in C a n a d a than in the U n i t e d States because of the separation, in C a n a d a , of operating expenses from capital spending (Evans, 1983; Iglehart, 1986b). Hospital's must apply to their provincia l ministry of health in order to obtain approval a n d / o r funding for capital expansion. Consequently , although as Evan's (1983) states "the process is not perfect , and is often rather ad hoc, ... on balance it has been substantially more successful in constraining hospital investment than the U .S . free-for-all" (p. 13). U n t i l recently, at the hospital level, the most typical response of administrators to provinc ia l economic restraint policies has been to close hospital beds and lay off staff (Barrable, 1987; Mercer, 1986;. A s noted by Mercer (1986), however, such responses are essentially "short-term solutions to what was v iewed as a short-term problem" (p. 39). H o w e v e r , since there are no signs that economic restraint policies are l ikely to be abandoned in the near future, new management strategies are required to assist hospitals i n deve loping effective long-term responses to conditions of cost constraint. Furthermore , the need for such strategies is becoming more acute since most hospitals, after more than a decade of economic restraint, are currently facing one or more of the fol lowing problems: "long wait ing lists, bed shortages, forced reduction of beds, need for n e w and expanded services, discharge problems and the lack of communi ty services, and h igh occupancy rates" ( M c D o n o u g h & V a z , 1987, p. 42). In addit ion, m a n y of Canada's 139 hospitals are getting o ld and are in need of restoration and updat ing (Canadian Hospi ta l Association, 1984; Iglehart, 1986a). A s a result of the various cost containment initiatives described above, C a n a d a has apparently been more successful than the U n i t e d States in containing total health care expenditures (Evans, 1983; Iglehart, 1986a). T h u s , whi le the U n i t e d States is currently spending more than 11% of its G N P on health care (Ginzberg, 1987), C a n a d a has managed to ho ld expenditures relatively stable at approximately 8.6% of its G N P (National Hea l th Expenditures , 1987). Moreover , "Canada's control over health care expenditures has been achieved wi th comprehensive coverage both of people and of services" (Evans, 1983, p. 4). In contrast, over 30 mi l l ion A m e r i c a n citizens currently have no insurance coverage at all (Iglehart, 1986a). H o w e v e r , the problems that are arising in response to continued economic restraint and, in particular, the growing concern about its eventual impact on the quality and accessibility to care, have led some analysts to suggest that a new spate of increases, wi th in the hospital sector at least, m a y wel l be just around the corner (Evans, 1983;1984; Iglehart, 1986a). A t the same time, however, it continues to be assumed by the federal government , prov inc ia l governments a n d a number of health service professionals that a considerable port ion of current hospital uti l ization is inappropriate a n d / o r unnecessary (Anderson & Lomas , 1988; C a n a d i a n Hospi ta l Associat ion, 1984; Iglehart, 1986a; Roos & Roos, 1988; Suttie, 1988). A l t h o u g h few studies have as yet been undertaken to quantify the extent of inappropriate or unnecessary use in C a n a d i a n hospitals, research o n variations in the use of surgical procedures such as cesarean sections, hysterectomies, tonsillectomies a n d adenoidectomies, and studies 140 compar ing prov inc ia l and international variations in length of hospital stay suggest that there is r o o m for improvement in the util ization of hospital services and facilities (Anderson & Lomas , 1984, 1985; Bennett & Krasny , 1977; D y c k , et al., 1977; Evans, 1983,1984; Roos, 1983; Roos, Henteleff, & Roos, 1977; Roos, 1979; Roos & Roos, 1988). A s indicated by Roos & Roos (1988), "[r]esearch us ing util ization data has challenged the assumption that the knowledge base under ly ing conventional medical practice produces a rational allocation of resources among clinical priorities a n d limits demand to those services k n o w n to be effective" (p.3). Consequently , there is growing acceptance of the belief that "at least some of the things that happen to patients a n d some of the time patients spend in hospital are controllable by the hospital a n d / o r its medical staff" (Halperin & Neuhauser , 1976, p 63). T o deal with this situation, a number of remedial programs have been suggested. These include the development of a regional referral and admiss ion scheduling process (Canadian Hospi ta l Associat ion, 1984), the development of external uti l ization review committees (Suttie, 1988), expanded emphasis o n prevention and the deve lopment of alternatives to hospital care (Bennett & K r a s n y , 1977; Roos & Roos , 1988; Soderstrom, 1978), and intensification of government intervention in health care del ivery, through uti l ization analysis and management (Evans, 1983; Suttie, 1988). Thus , as one analyst suggests, an appropriate government intervention w o u l d be the "requirement that, before requests f rom hospitals for capital and deficit funding be considered, active uti l ization analysis be demonstrated in the subject area of the request" (Suttie, 1988, p. 18). 141 W h i l e there is at present no mandate for hospitals to develop comprehensive uti l ization management programs, a few hospitals have begun to experiment with uti l ization management procedures. F r o m the perspective of these organizations, such procedures are necessary for the purposes of both cost-containment and quality assurance (Clemenhagen et al. , 1985; C o o m b s & Richter, 1986; El iasoph & Hassen, 1986; 1987; M c D o n o u g h & Vaz/1987; M c G e o r g e et al. , 1985; N u s b a u m , 1988; Winche l l , 1985). A l t h o u g h most hospitals have implemented some form of uti l ization review, since uti l ization review is a requirement for hospital accreditation (Canadian Comprehens ive A u d i t i n g Foundat ion , 1987; Rachlis & Fooks , 1988), the usual emphasis has been on reviewing util ization for the purpose of medical audit rather than for efficiency or cost-effectiveness (Bennett & Krasny , 1977; Rachlis & Fooks, 1988). This is perhaps understandable g iven that uti l ization review falls under the jurisdict ion of physicians who , traditionally and appropriately, have been more concerned wi th clinical* issues than with issues of cost (Bennett & Krasny , 1977; C lemenhagen et'al., 1985). G i v e n the current era of cost restraint, however, a few hospital boards and hospital administrators are beginning to p lay a more d y n a m i c role in managing the uti l ization of the hospital's resources. T h u s , whi le uti l ization review still remains the responsibil ity of the hospital's medica l staff, administrators are becoming more invo lved in the process b y w o r k i n g together wi th physicians to f ind ways to eliminate unnecessary a n d inappropriate use. Clear ly , the e l imination of misuti l izat ion is a desirable goal. F r o m a quality of care perspective, beneficial effects w o u l d l ikely include a reduction in the potential psycho-social and iatrogenic effects of unnecessary services. F r o m a cost perspective, 142 beneficial effects w o u l d l ikely include the el imination of the expense of unnecessary services and, therefore, the freeing u p of resources which could be used to service unmet needs (Restuccia, 1982). In order to work, however, a level of cooperation and communicat ion is needed between administrators and their medical staffs that has been u n c o m m o n in previous years Bennett & Krasny, 1977; Evans, 1983, 1984; Iglehart, 1986a; 1986b; Suttie, 1988). W i t h the pressures of reduced bed capacity, h igh occupancy rates, and long wait ing lists, all of wh ich make it more difficult for physicians to admit their patients into the hospital, the opportunity to develop this needed level of cooperation has presented itself to a number of receptive administrators. In response there has been a growth in the development of new and innovative strategies a imed, at least in part, at effecting changes in the way in w h i c h providers util ize the hospital's services and facilities. In the fol lowing section, the major developments that have occurred, other than the S W I T C H Index System, wi l l be described. Unfortunate ly , few studies have as yet been done to determine whether the programs that have been implemented have achieved their objectives. H o w e v e r , to the extent that any data is available it w i l l be discussed. F o l l o w i n g this review, a detailed description of the S W I T C H Index System wi l l be p r o v i d e d and the purpose of the present study wi l l be described. UI.2 Current Status of Ut i l izat ion Management in C a n a d a O n the basis of the available literature, it w o u l d appear that the ut i l izat ion r e v i e w / m a n a g e m e n t programs implemented in C a n a d i a n hospitals tend to be retrospective and diagnosis- or service-specific. O n l y three hospitals appear to have as yet begun to experiment wi th concurrent, diagnosis- independent programs comparable to the A m e r i c a n A E P . 143 H o w e v e r , as noted by Rachlis & Fooks (1988), to obtain a more comprehensive picture of the state of uti l ization management in Canada , a survey of i n d i v i d u a l hospitals w o u l d have to be conducted. Such a survey was b e y o n d the scope of this thesis and, therefore, it is possible that some equal ly innovative programs, other than those described below, may currently be underway. III.2.1 Retrospective Rev iew Programs III.2.1.1 Diagnosis-Specific Profile Analyses A s in the U n i t e d States, interest in uti l ization management in C a n a d a has been given impetus, in part, by the development of a patient classification system analogous to D R G ' s This system, called Case M i x G r o u p s ( C M G ' s ) was developed in the early 1980's by the Hosp i ta l M e d i c a l Records Institute ( H M R I ) . The H M R I is a not-for-profit, federally chartered c o m p a n y that specializes in information processing for health care institutions ( H M R I , 1983; H o d g s o n & O r m e r o d , 1985). C a n a d i a n hospitals can submit discharge abstract data to the H M R I (or to the A m e r i c a n P A S service) for processing and analysis. Information extracted from the medical record a n d submitted to the H M R I includes province and institution, patient sex, date of birth, length of stay, admission date, discharge date, admiss ion status, discharge status, most responsible physic ian code, diagnoses, and procedures ( H M R I , 1983). A complete list of the data elements inc luded on the H M R I Discharge Abstract is p r o v i d e d in A p p e n d i x D . In deve loping C M G ' s , the H M R I used essentially the same procedure as was used in developing the D R G ' s , with two exceptions: 1) in concert wi th C a n a d i a n practice, the diagnostic and procedure codes used in 144 constructing C M G ' s were based o n the ICD-9 rather than the I C D - 9 - C M coding conventions; and 2) C M G ' s were constructed using the pr imary diagnosis (i.e., the one most responsible for the patient's pattern of utilization) rather than the principle diagnosis (i.e., the one most responsible for prompt ing the decision to admit) (MacKenz ie et al., 1987). L i k e D R G ' s , the dependent variable used to reflect patterns of util ization was length of hospital stay. In addition, age and sex were inc luded as predictors of length of stay when these variables were found to affect the observed pattern of utilization. The result of the classification process was the development of 465 C M G ' s which , l ike their A m e r i c a n counterparts, were supposed to categorize patients according to their resource util ization (as measured by length of stay) into relatively homogeneous groups (Botz & Singh, 1985; H o d g s o n & O r m e r o d , 1985; M a c K e n z i e et al., 1987) W i t h the development of C M G ' s , a number of analysts began to promote their use in applications such as cost-accounting, budgeting, resource management, and quality assurance (Botz & Singh, 1985; H o d g s o n & O r m e r o d , 1985; Macies , 1984; M a y , 1983; M c G e o r g e et al., 1985; M a c K e n z i e et al., 1987; Stoughton, 1983; Z u c k e r m a n , 1983). T o facilitate their use in uti l ization review a n d management, the H M R I routinely provides to each member hospital quarterly and annual reports o n the hospital's ind iv idua l performance and the performance of a peer g r o u p 1 of hospitals. A variety of reports are p r o v i d e d inc luding length of stay profiles by C M G , by physician, a n d b y patient service ( H M R I , 1983). Despite their availability, profile 1 HMRI peer groups are defined according to bed size and type. Thus, hospitals are placed into one of the following groups: 1-99 bed hospitals; 100-199 bed hospitals; 200-399 bed hospitals; 400+ bed hospitals; Teaching hospitals; Paediatric hospitals. Hospitals can be in more than one group; for example, a 300 bed teaching hospital (HMRI, 1984). 145 analyses based o n C M G ' s have not yet become, to any great extent, a major component of most hospital's uti l ization review or management activities (Suttie, 1988). O n e notable exception, is the Kings ton General Hospi ta l in Kingston , Ontario. Orig inal ly a test site for the C M G system, this hospital has devoted considerable effort to the task of developing and implement ing product- l ine costing a n d util ization review systems based o n C M G ' s (McGeorge et al., 1985). T w o reasons have been identified for the apparent reticence, on the part of hospital administrators and util ization review committees, to incorporate C M G - b a s e d length of stay review. First, according to the results of a survey conducted by Suttie (1988), m a n y hospital util ization review committees lack the necessary technical expertise to understand and apply profile analysis der ived from H M R I (CMG-based) length of stay reports. Second, a number of writers have suggested that there is a reluctance to use C M G ' s because of the criticisms, discussed in Chapter n, that have been raised in connection wi th the D R G system. O f particular concern are the criticisms related to the failure of D R G ' s , and presumedly C M G ' s , to account adequately for variations in severity of illness (Hodgson & O r m e r o d , 1985; M c G e o r g e et al., 1985; Mackenz ie et al., 1987; Stoughton, 1983). A s noted by M c G e o r g e et. al. (1985), s imilar concerns about the variability wi th in D R G / C M G categories were raised d u r i n g the development of the C M G -based management system's at Kings ton General Hospi ta l . In response, the system was modi f i ed to incorporate the Severity of Illness Index developed by H o r n and her associates ( H o r n , H o r n , & Sharkey, 1984). Likewise , the H M R I continues to refine C M G ' s (MacKenzie et al. , 1987). Consequently , some analysts expect that implementation of C M G - b a s e d systems wi l l 146 expand w h e n C M G ' s have been reformulated to include a severity of illness measure. T h e inclusion of such a measure is anticipated to occur with the next generation of C M G ' s . (MacKenzie et al., 1987). It is to be hoped that w h e n and if this expanded use occurs, evaluations of the effectiveness and cost-effectiveness of emerging C M G - b a s e d resource management strategies w i l l begin to appear in the literature. W h i l e C M G - b a s e d length of stay profiles are, at present, the most c o m m o n l y available screening mechanisms for use in uti l ization management programs, six hospitals have begun to develop and implement an alternative form of diagnosis-specific profile analysis called the V a l u e Improvement Process (VIP). These hospitals include: the Foothil ls Hosp i ta l in Ca lgary , Alberta , the Vancouver General Hosp i ta l in Vancouver , British C o l u m b i a , the Toronto Genera l Hosp i ta l in Toronto , Ontar io , the Montrea l General Hosp i ta l in Montrea l , Quebec, the Roya l Victoria Hosp i ta l in Montrea l , Quebec , and the Victoria General Hospi ta l in Hali fax, N o v a Scotia. T h e V I P was developed by the Baxter Corporat ion (previously k n o w n as Traveno l , Inc.), an American-based consulting f irm. It is a proprietary system and , as such, there appears to be no background information in the academic literature that details the history of its development or provides any data wi th respect to its reliability, validity, or effectiveness. H o w e v e r , a few articles have been publ i shed w h i c h describe, and prov ide pre l iminary data on , the V I P as it has been implemented in C a n a d i a n hospital settings. L i k e other uti l ization management programs, the p r i m a r y objective of the VTP is to reduce costs whi le maintaining, or i m p r o v i n g , quality (Winchel l , 1985). H o w e v e r , the approach taken to achieve this objective is 147 quite unique. In essence, the program is conducted in two phases. Phase I involves a detailed review of patient's medical records to identify "homogeneous groups of patients wi th similar conditions, operations, and physical attributes" (Eliasoph & Hassen, 1986). The end result of this review is the selection of a sample of prototypical patients, for a given therapy, (eg. myocard ia l infarction) wi th no residual complications that w o u l d alter the treatment regimen (Eliasoph & Hassen, 1986). Once this sample has been identified, "Cost Profiles" are developed on the basis of six cost-component categories, namely: b lood and therapy, lab and tests, pharmacy, supply , O.R. , a n d R o o m & Care (Winchell , 1985). The process is repeated at each of the six participating hospitals in order to compile "Cost Profiles" that are comparable from one hospital to the other. Once all six Cost Profiles have been deve loped the information is used to construct a "Best Demonstrated Cost Profile" (BDC) . This B D C reflects the "lowest cost performance of each category f rom all participating hospitals. In other words , the lowest R o o m -and-Care amongst all six hospitals combined with the lowest Blood-and-T h e r a p y cost, lowest Pharmacy, etc." (Winchel l , 1985, p. 85). This information is then p r o v i d e d to a special hospital Task Force composed of physicians, nurses, support staff, and administrators (Eliasoph & Hassen, 1987). B y compar ing their hospital's cost performance to the other hospitals' performance and to the B D C , the Task Force can begin to target areas, wi thin specific medica l therapies, where opportunities for improvements in the cost and quality of patient care can be made (Eliasoph & Hassen, 1986). Phase II then involves the development, wi th in each hospital , of specific recommendat ions for improvement , and the formulat ion of a detailed implementat ion p l a n (Winchel l , 1985). 148 T o date, the Foothills Hospi ta l has progressed the furthest of the six hospitals in implement ing the VTP. O n e of the factors that has been suggested to account for the faster response at the Foothills Hospi ta l is the fact that, since 1981, it has also been a participant in an Alberta per-case f u n d i n g experiment (Coombs & Rich ter, 1986). T h u s , s imilar to the D R G -based reimbursement system appl ied in the Un i t ed States, the Foothills Hosp i ta l has a greater incentive to reduce util ization than hospitals that are re imbursed on the basis of global budgets (Winchel l , 1985). Moreover , since the V I P process is expensive and t ime-consuming, some of the original hospitals have h a d to delay further implementat ion unt i l addit ional resources can be made available (Winchel l , 1985). A c c o r d i n g to El iasoph & Hassen (1986; 1987), a total of ten V I P studies have been completed at the Foothills Hospi ta l on procedures such as myocard ia l infarction, total h ip replacement, cholecystectomies, hysterectomy, and cataract surgery. A s w o u l d be expected, a c o m m o n theme emerging from all of the studies was the importance of length of stay in determining each therapy's total costs. Consequently , length of stay became a major target for reduction. Thus , for example, o n the basis of Phase I investigations triggered by the inter-hospital comparison of "Cost-Profiles" for total h ip replacements, it was determined that two factors were contributing to unnecessarily long hospital stays: 1) an overly conservative physiotherapy protocol, and 2) the lack of physiotherapy on weekends. T h r o u g h implementat ion of a new protocol and by initiation of weekend physiotherapy, the length of stay for total hip replacement (simple cases) was observed to decrease from 21.3 days in the month prior to implementation of Task Force Recommendations to 14.8 days, 6 months later (Eliasoph & 149 Hassen, 1986). Similarly, in a study of the impact of the V I P on cataract surgery, a comparison of the "Cost Profile" from the Foothills Hospi ta l to that of another major C a n a d i a n teaching hospital and five A m e r i c a n hospitals, indicated that the Foothills Hospi ta l was spending approximately $300 more per patient than were the comparison hospitals (Eliasoph & Hassen , 1987). Once again, the major factor accounting for the cost difference was found to be length of stay and, in particular, the pre-operative length of stay. Another major factor found to be contributing to the cost difference was the lack of availability of day care beds, which thereby l imited the number of patients that could be treated on an outpatient basis. Therefore, the recommendations formulated and implemented were a imed pr imari ly at increasing the treatment of patients on an outpatient basis (by lifting restrictions on the use of day care beds for cataract surgery) and at decreasing the pre-operative length of stay (by making better use of coordinated pre-op work-ups and the hostel facility). F ive months after implementat ion of the recommendations, it was reported that the proport ion of patients treated on an out-patient basis h a d risen from 4 5 % to 74% and that the per patient length of stay had been reduced from 3.1 to 2.2 days. Moreover , these results, combined wi th reduced lens and supply costs, are expected to generate annual savings of $350,000 (Eliasoph & Hassen, 1987). W h i l e these results appear promis ing , they must be interpreted wi th caution because they are based on a before-after study design wi th no control group. In addit ion, the t ime-period over which the data have been tabulated is relatively short. Consequently , there is no way to determine to what extent the observed reductions reflect s imply a regression to the mean 150 effect, secular trends in length of stay, or an actual reduction. In addition, no attempt has as yet been made to determine whether the benefits, in terms of reduced costs a n d / o r increased quality, outweigh the reportedly substantial costs of implement ing and maintaining the system (Winchel l , 1985) nor has an attempt been made to compare the VIP to other programs for its cost-effectiveness. Final ly , as noted by Rachlis & Fooks (1988), V I P comparisons are based o n communi ty norms and "it is not clear h o w the appropriateness of the norm[s] are evaluated" (p.24). O n a more positive note, however, the VIP's team approach to uti l ization management, reflected in the work of the mult i -d isc ip l inary Task Force, has been wel l received by the hospital's medical staff (Eliasoph & Hassen, 1987). Moreover , the emphasis on i m p r o v i n g the quality of patient care first, wi th emphasis on cost-reduction coming second, has been an important factor in promot ing communicat ion and cooperation a m o n g the various personnel invo lved in the process (Eliasoph & Hassen, 1986; Winche l l , 1985). Since these are both critical features to the success of any uti l ization management process, they ought not to be ignored in considering the merits of the system. JJJ.2.1.2 Service-Specific Profile Analyses Hospi ta ls that have begun to implement uti l ization management programs that assess a n d respond to util ization problems identified on a service or department level include Saint Michael's Hosp i ta l , in Toronto ( M c D o n o u g h & V a z , 1987), a n d the C h e d o k e - M c M a s t e r Hosp i ta l , in H a m i l t o n (Canadian Comprehens ive A u d i t i n g Foundat ion , 1987). A l t h o u g h the two programs differ in detail, wi th the Saint Michael's p r o g r a m being more centralized a n d the C h e d o k e - M c M a s t e r program being 151 highly decentralized, the basic approach is similar; util ization profiles are deve loped for each department, using information such as length of stay and occupancy rate by service. W h i l e the C h e d o k e - M c M a s t e r p r o g r a m relies u p o n internally deve loped criteria for prob lem identification, the Saint Michael's Hosp i ta l uses both internal and peer group hospital comparative data in its analyses. O n c e recommendations for improvement have been made, support staff at Chedoke-McMaster , and special Task Forces at Saint Michael 's , are p r o v i d e d to assist the clinical services in implement ing corrective action plans. T o date, no studies have been found in the literature assessing the extent to which either of these programs have been effective in reduc ing inappropriate, inefficient, or unnecessary util ization. HI.2.2 Concurrent Rev iew Programs A s noted above, information has been located on only three examples of concurrent review programs implemented in C a n a d i a n hospitals. These inc lude the p r o g r a m implemented at the Hosp i ta l for Sick C h i l d r e n , in Toronto; the V i - C a r e p r o g r a m that is currently being implemented by the Victor ia Hosp i ta l Society, in Victoria; and the S W I T C H Index System which has been implemented at the Peace A r c h District Hosp i ta l , in Whi te Rock. A l l of these programs make use of diagnosis- independent uti l ization criteria. H o w e v e r , only the programs f rom the Hospi ta l for Sick C h i l d r e n a n d the Victor ia Hosp i ta l Society wi l l be described in this section. The S W I T C H Index System, which is the subject of the evaluation reported in this thesis, w i l l be described separately. m.2.2.1 T h e Hospi ta l for Sick C h i l d r e n A l t h o u g h the focus of this thesis is on util ization r e v i e w / m a n a g e m e n t in the general acute care hospital setting, a brief 152 description of the program implemented at the Hosp i ta l for Sick C h i l d r e n is inc luded for completeness. A c c o r d i n g to information compi led by the C a n a d i a n Comprehens ive A u d i t i n g Foundat ion ( C C A F ; 1987), Toronto's Hosp i ta l for Sick C h i l d r e n has introduced "a detailed concurrent audit of all patient charts" (p.58). The review is conducted by a team of non-medical "quality assurance screeners" selected by the hospital administration and the medical staff. These "screeners" review each patient's record every three days, assessing the care p r o v i d e d for incidents, util ization problems, and documentat ion problems. T o assist in problem identification, the care is assessed against 22 generic criteria. [Unfortunately, no information on the nature of these criteria nor the util ization problems they address was p r o v i d e d in the C C A F report] W h e n problems are identified that require immediate attention, they are related either to the attending physic ian, the nurs ing staff, or the "physician advisor" for the particular clinical service. In addit ion, summaries of all identified problems are compi led by the screeners for month ly reviews wi th the physic ian advisors and for follow-u p action. Problems are discussed by the physician advisors at divisional rounds and remedial action strategies are developed. S u m m a r i z e d reports of the problems and the actions taken are then submitted to the hospital's mult i -d isc ipl inary quality assurance committee, and , finally, to the Qual i ty Assurance Commit tee of the board. A l t h o u g h no data have been p r o v i d e d to indicate whether the p r o g r a m has been effective, the "approach is bel ieved to have greater educational impact (resulting in behavioural change) than a retrospective process in which the information m a y be several weeks or months o ld by the time it is brought forward for review" ( C C A F , 1987, p. 59). In light of the A m e r i c a n experience wi th similarly 153 detailed a n d labour-intensive concurrent review programs, however, it w o u l d be interesting to see if this expectation w o u l d be met if it could be compared , for its cost-effectiveness, with other retrospective or, less labour-intensive concurrent review strategies. m.2.2.2 T h e V I - C A R E Program T h e Greater Victoria Hosp i ta l Society in Victoria , British C o l u m b i a has recently begun implementat ion of a uti l ization management /qua l i ty assurance p r o g r a m called V I - C A R E (Victoria Integrated Care Alternatives Rev iew & Evaluation). This program, which at present applies only to patients classified as medical emergencies, is initiated u p o n a patient's admiss ion, a n d continues wi th periodic concurrent moni tor ing throughout each inpatient episode (Nusbaum, 1988). W h e n fully implemented , the V I -C A R E p r o g r a m wi l l be comprised of two components: 1) application of the I S D - A appropriateness criteria developed by L a m p r e y and Jacob (1987); and 2) application of the M E D I S G R P S severity of illness criteria (Brewster, Jacobs, & Bradbury , 1984; Brewster, K a r l i n , et al., 1985). Since the I S D - A review criteria have already been discussed (Chapter II section II.5.7) no further elaboration o n them wi l l be p r o v i d e d here. The M E D I S G R P S criteria were also mentioned in Chapter II, but they were not described. In brief, the M E D I S G R P S severity of illness index classifies patients into one of five groups based on the fo l lowing criteria: G r o u p 0 - patients wi th no key clinical findings; G r o u p 1 - patients wi th m i n i m a l findings, indicating a low potential for organ failure; G r o u p 2 - patients wi th either acute findings connoting a short time course wi th an unclear potential for organ 154 failure, or severe findings wi th h igh potential for future organ failure; G r o u p 3 - patients with both acute and severe findings indicat ing a h igh potential for imminent organ failure; G r o u p 4 - patients with critical findings indicating the presence of organ failure. (Source: Brewster, Jacobs & Bradbury, 1984; Brewster, Kar l in , et a l , 1985) T o date, on ly the concurrent I S D - A review component of the V I - C A R E p r o g r a m has been implemented. Implementation of the M E D I S G R P S system is targeted for the fall of 1988. In the meantime, a retrospective review of patients' charts has been initiated in order to develop norms, wi th in C M G ' s , for use wi th this system (Sheps, personal communicat ion , A u g u s t 1988). Later, w h e n the M E D I S G R P S severity of illness index has been implemented , and relevant norms have been established, patients will be assigned a M E D I S G R P S code u p o n admission and again u p o n discharge. T h e m a i n objective of this process wi l l be to ensure that patients leave the hospital at a lower level of severity than the level at wh ich they were admitted. Thus , by apply ing the I S D - A appropriateness criteria and by moni tor ing changes in the M E D I S G R P classifications, it should be possible to identify and correct problems related to both util ization and quality (Nusbaum, 1988). H o w e v e r , since the p r o g r a m is still being phased-in, it w i l l be some time before information on its effectiveness and , ultimately, its cost-effectiveness wi l l be available. UI.3 T h e S W I T C H Index System In 1984, Peace A r c h District Hospi ta l ( P A D H ) , a communi ty hospital located in W h i t e Rock, Brit ish C o l u m b i a , developed a n d implemented one 155 of the first d y n a m i c util ization management programs to appear in a C a n a d i a n acute care hospital setting. Several hospital-specific factors motivated this action: 1) consistently h igh occupancy rates over a per iod of years; 2) a concomitant shortage of acute care beds; and 3) the infeasibility of routine patient transfers due to the relative isolation of the hospital from other acute care facilities. These factors coupled with province-wide economic restraint policies and the h igh costs associated wi th m o d e r n medica l technologies, p r o m p t e d the hospital's administration to begin seeking ways to eliminate inappropriate and unnecessary uti l ization of the hospital's acute care beds (Frobb, personal communicat ion, October, 1985). T h e result was the development of a comprehensive uti l ization management p r o g r a m invo lv ing admiss ion and discharge p l a n n i n g and inpatient util ization review. In addit ion, a close w o r k i n g relationship was established between the Ut i l izat ion Management Commit tee and the M e d i c a l A u d i t Commit tee to ensure that the quality of service p r o v i d e d at the hospital was maintained at an acceptable professional standard. T h e cornerstone of the P A D H util ization management program, and the component that sets it apart f rom other util ization management efforts, is a uti l ization review process called the S W I T C H Index System ( S W I T C H ) . Similar in intent to the A E P (Gertman & Restuccia, 1981) and the I S D - A (Lamprey & Jacob, 1987), the S W I T C H uses explicit criteria to identify hospital days d u r i n g w h i c h no appropriate treatment or diagnostic activity is taking place. In contrast to the numerous criteria e m p l o y e d in the I S D - A system, the S W I T C H system employs only six categories of criteria aggregated under the headings Signs, W i n d , Intramuscular Injections, 156 Tubes, Consultant (Active), and Hospice . A description of the specific criteria inc luded under each heading is p r o v i d e d in Table VI. T A B L E VI . T h e S W I T C H Index System Criteria S I G N S - (Q6H) (Q4H): - V i ta l Signs - N e u r o Signs - This index factor to be m a r k e d w h e n a phys ic ian specifically orders that vital signs be recorded at 6 hourly intervals or less. - Implies that nurs ing observation is required in anticipation of a possible deterioration in patient status. - Dependent on patterns of practice, the clinical care team m a y opt to use a Q 4 H description of this factor. W I N D (Respiratory Therapy: - Cont inuous oxygen - Regular ly administered sidestream medications - Chest physiotherapy - Th i s index factor describes only regularly administered therapies and does not refer to p r n orders. - If the patient has stabilized at their best pre-hospitalization status, d i r e c t i o n s h o u l d be a t ta ined f r o m the h e a d n u r s e a n d Convalescence Response T e a m as to cont inuing the care of the patient at home wi th outpatient therapy and homecare nurs ing . I N T R A M U S C U L A R I N T E C T I O N S : - Antibiot ics - Analgesics - A d m i n i s t e r e d more frequently then Q 8 H . - If therapy is expected to continue for a period longer than 5 days (Table V I cont...) 157 Table VI (continued). and patient status is otherwise stable; consideration should be directed to ut i l iz ing Outpatient Services or H o m e C a r e N u r s i n g for administrat ion of medications. - Does not describe administration of subcutaneous insul in. T U B E S : - I V . T h e r a p y (medications and fluids) - Nasogastric Tubes - Foley catheters - Epidurals - Chest Tubes /Endotrachea l Tubes - Drains to continuous or intermittent suction - In most cases, the use of invasive tubes and catheters is self-l imited. If use is anticipated for greater than 10 days, the charge nurse should be informed. - Does not include T K V O I. V . Lines. C O N S U L T A N T (Active): - Impl ies d a i l y visits b y a consultant w i t h c h a n g i n g orders , indicating a change in the pattern of care. - Implies that the changing status of patient condit ion necessitates da i ly visits b y consultants for further invest igations a n d / o r treatment. H O S P I C E : - Indicates the palliative care patient. - Impl ies that the patient is u n l i k e l y to recover to a status permitt ing discharge home or to a L o n g T e r m Care Facility. - Denotes involvement of Palliative C a r e Team. (Source: Peace A r c h District Hospi ta l , 1985) 158 Potential cases of inappropriate or unnecessary bed use are identified, on a dai ly basis, by means of an On-Index/Off-Index classification system. Patients classified as Off-Index are those who fail to meet at least one of the S W I T C H criteria. The process of O n / O f f - I n d e x classification is initiated with the complet ion of the Kardex A c u i t y Index (See Figure 2) at the time of the patient's admission. Thereafter, d u r i n g each nurs ing shift, the A c u i t y Index is updated to reflect any changes in the patient's status. This function is usual ly performed by the supervis ing R . N . although it m a y be delegated to the W a r d C l e r k under the direction of the supervis ing R . N . D u r i n g each day shift, the H e a d N u r s e produces a N o m i n a l Rol l S W I T C H w h i c h is s imply a listing of all the patients o n the w a r d indicating their O n / O f f Index status. This N o m i n a l Rol l is reviewed by the evening Charge N u r s e , for accuracy and updat ing , and later is finalized by the N i g h t Supervisor. The N i g h t supervisor is also responsible for compi l ing an OFF-Index census list. This list consists of all patients designated as Off-Index at the time of census collection. In addit ion, at the time the census is taken, each Off-Index day is assigned to the "most responsible" service; i.e., the service most responsible for the patient's Off-Index status. This information is taken from the A c u i t y Index (see Figure 2) and indicates whether, at the time it was made, the off-index assignment was due pr imar i ly to a delay in physic ian communicat ion , delays in in-house procedures, or delays in outside services. A t 0800 every day, the H e a d Nurse , the T e a m Leader, the Director of Social Services and the phys ic ian uti l ization review coordinator review the Off-Index patients for accuracy a n d initiate the distribution of the Off-Index Census list to hospital service departments for their attention and , if appropriate, corrective action. A s wel l , the T e a m Leader initiates discussion with 159 Figure 2. Kardex A c u i t y Index A C U I T Y I N D E X M . D . C O M M U N I C A T I O N # • S I G N S Wai t ing Direction Vita l Signs 0 6 H I N - H O U S E S E R V I C E S : N u r s i n g Observat ion Radiology (Indicative of possible Laboratory deterioration) Out-Patients • W I N D T H E R A P Y O.R. 0 2 Sidestream M e d s Social Services Chest Physiotherapy Convalescence • I . M . M E D I C A T I O N S O r t h o • T U B E S N e u r o I.V. or I.V. M e d s Obstetrical N . G . / F o l e v / E p i d u r a l Other O U T S I D E S E R V I C E S : • C O N S U L T (Active) Diagnostic Dai ly Visits B . C . C . I C h a n g i n g Orders H o l v Fami ly • H O S P I C E Other T O T A L # O F D A Y S O F F I N D E X : (Source: Peace A r c h District Hospi ta l , 1985) physicians regarding discharge plans. A n exception occurs wi th patients designated as Off-Index due to convalescence. These patients are referred, o n the first day of being so designated, to a Convalescence Response T e a m ( C R T ) for initiation of discharge planning. Final ly , w h e n a patient is discharged, the Hea l th Records Department abstracts information concerning the number of days off-index, the specific reasons for each off-index day, and the C R T assessment are entered into the hospital's computer system. Thi s information is used to compile regular historical summaries so that trends in Off-Index use can be identified. 160 A s with the A m e r i c a n A E P , one of the most interesting characteristics of the S W I T C H is that it focuses the attention of the medical , nursing, administrative, and other professional staff on the prob lem of identifying what is important about acute care hospital beds. In addit ion, it provides the hospital with a mechanism for increasing the involvement of the medical staff in the management of the hospital's resources. A s with any util ization management program, the success of the S W I T C H is dependent on the involvement and cooperation of the medical staff. A previous attempt to implement a uti l ization review procedure that employed length of stay target dates had met with failure because it was v iewed by the medical staff as punit ive and arbitrary (Frobb, personal communicat ion, October, 1985). T h e S W I T C H , on the other hand , provides information that is more meaningful and acceptable to clinical staff than the statistical and normative approach of length of stay based util ization review. Final ly , unl ike some of the other methods of concurrent or in-depth review (eg. the VIP) , the S W I T C H is relatively easy to implement and maintain. m.4 Study Purpose and Rationale T h e general purpose of the present study was to prov ide pre l iminary data o n the effectiveness of the P A D H util ization management program. Specifically, the study was designed to determine whether the P A D H system of uti l ization management h a d been effective, over time a n d in comparison to a control group of communi ty hospitals, in reducing the length of adult, acute-care hospital stays. T h e decision to use length of stay as the outcome variable of interest was based on two factors: 1) since the P A D H program was designed to eliminate inappropriate and unnecessary days of hospitalization, an observable outcome, assuming 161 that changes in case mix have been controlled, should be a reduction in length of stay, and 2) since comparison hospitals were to be inc luded in the evaluation, a measure of effectiveness was needed that w o u l d be obtainable across all participating hospitals. T r y i n g to determine whether the S W I T C H h a d an effect o n length of stay is, however, a complicated problem. A s noted by Lave & Leinhardt (1976a), "over a per iod of time, exogenous events unrelated to the review p r o g r a m can influence length of stay. Changes in medical technique, therapy, and hospital staffing m a y each play a part" (p.961), as m a y variations in case-mix. Th e P A D H was thus compared to a group of peer hospitals on the assumption that secular events w o u l d be l ikely to affect the length of stay patterns of both the P A D H and the peer group in a similar manner. T h e changing case-mix problem was dealt with in two ways, as w i l l be described in Chapter IV. Thus , although the present study is exploratory, it is important because it represents the first attempt to incorporate both a "control" group bf hospitals and an adjustment for case-mix changes over time, in evaluating the effects of a uti l ization management p r o g r a m implemented i n a C a n a d i a n hospital setting. UI.5 Hypotheses 1) A s s u m i n g that secular trends have affected both the P A D H and n o n - S W I T C H hospitals in similar ways, then it is expected that d u r i n g the p e r i o d pr ior to the implementat ion of the S W I T C H , average length of stay trends for the P A D H wi l l not differ significantly from those of the N o n -S W I T C H hospitals. 162 2) In the per iod fol lowing the introduction of S W I T C H , it is expected that the average length of stay for the P A D H wi l l fall relative to the N o n -S W I T C H hospitals. 163 C H A P T E R IV M e t h o d s I V . l Study Design A quasi-experimental time-series design wi th a control group (Campbel l & Stanley, 1969) was used to examine trends in the average length of hospital stay at the P A D H . Twe lve periods of four months each were s tud ied 1 : eight periods before the implementation of the S W I T C H (February, 1982 through September, 1984), one per iod d u r i n g w h i c h the S W I T C H was implemented (October, 1984 through January, 1985) and three periods d u r i n g which the S W I T C H was fully operational (February, 1985 through December, 1985). Patients under the age of 18 and patients w h o d ied whi le in the hospital were excluded from the study. A l t h o u g h the inclusion of a control group provides a baseline for compar ing secular changes in average length of stay, secular trends represent only one potential explanation of any observed changes in average length of stay over time. T h e clinical characteristics of patients and hospital characteristics such as teaching status, occupancy rate, and bed capacity have also been found to affect length of stay (Lave & Leinhardt , 1976a; 1976b). Hence , in designing the study, an effort was made to select hospitals comparable to the P A D H in terms of bed size and teaching status (i.e., non-teaching). Potential effects due to hospital occupancy were taken into account in the analytic phase of the investigation. Final ly , since the basic unit of observation for this evaluation is the hospital, not a patient case, differences in patient characteristics were dealt wi th by us ing the C M G An exception is period 12 which is based on the three month period October, 1985 to December, 1985. 164 patient classification system to ascertain, and to adjust for, differences in , hospital case mix. rv.2 S a m p l e a) T h e S W I T C H Hospi ta l . Because the S W I T C H has only been implemented in one hospital , namely the P A D H , this one hospital alone constitutes what is usual ly referred to as the intervention group. The P A D H is a publ ic , general, communi ty hospital located in White Rock, B . C . Information concerning the hospital's bed size and occupancy is p r o v i d e d in Table VII. b) T h e C o m p a r i s o n ( N o n - S W I T C H ) G r o u p . In order to provide comparabi l i ty to the P A D H , at least in terms of bed size and non-teaching status, the hospitals inc luded in the N o n - S W I T C H group were initially identified according to the fo l lowing criteria: i) A member of the B . C . Min i s t ry of Health's defined Peer G r o u p for the P A D H ( M O H Peer G r o u p #15)1. ii) A member of the B . C . Heal th Association's ( B C H A ) defined Peer G r o u p for the P A D H ( B C H A Peer G r o u p #12).2 iii) A member of H M R I since at least January 1, 1982. 1 MOH Peer Group 15 includes the following hospitals: Cowichan District Hospital, Duncan; Maple Ridge Hospital, Maple Ridge; Mount Saint Joseph Hospital, Vancouver; Peace Arch District Hospital, White Rock; Saint Joseph's General Hospital, Comox; Saint Mary's Hospital, New Westminster; Saint Vincent's Hospital, Vancouver; Trail Regional Hospital, Trail. 2 BCHA Peer Group 12 includes the following hospitals: Chilliwack General Hospital, Chilliwack; Cowichan District Hospital, Duncan; Langley Memorial Hospital, Langley; Maple Ridge Hospital, Maple Ridge; MSA General Hospital Abbotsford; Mount Saint Joseph Hospital, Vancouver; Peace Arch District Hospital, White Rock; Penticton Regional Hospital, Penticton; Richmond General Hospital, Richmond; Saint Mary's Hospital, New Westminster; Saint Vincent's Hospital, Vancouver; Trail Regional Hospital, Trail; Vernon Jubilee Hospital, Vernon. 165 O n the basis of these criteria, a total of five hospitals were determined to be eligible for inclusion in the study. In order to obtain the data for the evaluation, however, written permission had to be obtained from each of the hospitals. Four of the five eligible hospitals agreed to participate and provided the necessary "Release of Data Forms" (See Appendix E). Thus, when the study began, four hospitals comprised the N o n - S W I T C H group. Subsequently, however, one of these four hospitals failed to provide Table VII. Characteristics of P A D H and N o n - S W I T C H Hospitals: 1982-1985 P A D H N o n - S W I T C H * Hospitals Number of Hospitals 1 3 Number of Acute Care Beds per quarter M e a n 114.9 121.3 per Hospital M e d i a n 112.0 122.8 St. Dev. 5.6 5.4 M i n i m u m 112.0 114.3 M a x i m u m 129.0 134.5 Quarterly Occupancy Rates** M e a n 89.6 % 83.2 % per Hospital M e d i a n 91.0 83.1 St. Dev. 5.3 2.0 M i n i m u m 78.0 79.6 M a x i m u m 98.0 87.2 Statistics reflect the average across hospitals. Occupancy rates were computed on the basis of beds available for acute care use (i.e., set up and staffed) rather than on the basis of each hospital's rated bed complement. 166 . necessary hospital data and was dropped from the study. The final N o n -S W I T C H group, therefore, consisted of three P A D H peer-group-member hospitals. S u m m a r y statistics on the bed size, and occupancy rates of the N o n - S W I T C H group of hospitals, are p r o v i d e d in Table VII. It is u n k n o w n whether any systematically vary ing characteristic(s) of the excluded hospitals contributed to their non-participation. If so, then it is possible that the length of stay results m a y have been affected by a self-selection bias. O n the basis of statistics reported in the C a n a d i a n Hospi ta l v Directory (Canadian Hospi ta l Association, 1982; 1983; 1984; 1985) it was estimated that, on average, the excluded hospitals had a higher number of acute care beds (Mean per Quarter = 180) than d i d the three N o n - S W I T C H hospitals (Mean per Quarter = 121.3). However , in the absence of length of stay and occupancy data, it is impossible to determine whether the two excluded hospitals differed sufficiently from the others to affect the average length of stay patterns observed in the participating comparison group of hospitals . IV.3 Data Data for the study were obtained from two sources: H M R I and each participating hospital. T o obtain patient data, requests were made to each hospital for release of H M R I records pertaining to all of the hospital's discharges occurring between January, 1982 and December, 1985. A total of 101,858 patient records were received from H M R I . F r o m this database, 21,512 records were removed because they pertained to the hospital that was d r o p p e d from the study. In addit ion, because of the decision to group the patient data into quarterly intervals, beginning February, 1982, the data from January, 1982 were removed from the database. These two deletions left a 167 total of 78,674 patient records. Next, since the data-base also inc luded patients under the age of 18 and patients w h o h a d died whi le in the hospital, both of these groups of records had to be removed. This left a final data base of 59,259 patient records; 14,762 from P A D H and 44,497 f rom the N o n -S W I T C H group of hospitals. The patient data obtained from the H M R I records included: age, sex, length of stay, admission status (emergent, urgent, elective), discharge status (dead or still born, discharged, signed out), and C M G designation. Descriptive statistics on each of these variables (except discharge status, since all deaths were excluded) for the P A D H and the group of N o n - S W I T C H hospitals are p r o v i d e d in Table VIII. Hospital-specific data (eg., monthly occupancy rates, and bed size) were obtained f rom each of the participating hospitals (See Table VII). In addit ion, each hospital was asked to prov ide an indication of their beds per 1000 populat ion. In 1982, the P A D H had a bed per 1000 populat ion ratio of 2.6 wh ich , by 1985 had been reduced to 1.6 beds per 1000 populat ion. Unfortunate ly , comparative ratios for the N o n - S W I T C H group of hospitals cou ld not be computed as two of the three hospitals in this group were unable to p r o v i d e the information. Final ly , each of the N o n - S W I T C H hospitals was asked to prov ide a description of its uti l ization review activities d u r i n g the per iod of the study. T h e purpose of this request was to determine: 1) whether any of the comparison hospitals h a d uti l ization r e v i e w / m a n a g e m e n t programs that were similar in nature and intensity to the S W I T C H program; and 2) whether any changes in the review procedures fo l lowed by the comparison hospitals h a d occurred at or around the time the S W I T C H was implemented at the P A D H . The information obtained indicated that, while 168 T A B L E Vm. Patient Characteristics of P A D H and N o n - S W I T C H Hospitals:  1982-1985 P A D H N o n - S W I T C H Hospitals N u m b e r of Hospita ls Tota l N u m b e r of A d u l t Discharges* N u m b e r of M e a n Discharges per M e d i a n quarter St. Dev. M i n i m u m M a x i m u m 14,762 1,230.2 1,238.0 130.4 855.0 1359.0 44,497 1,236.0** 1,276.0 103.0 976.0 1331.0 N u m b e r (%) of Patients by A g e group N u m b e r (%) of Patients by Sex N u m b e r (%) of Patients by A d m i s s i o n Status 18-29 30-49 50-69 70 + M a l e F e m a l e Emergent U r g e n t Elect ive U n k n o w n 2,680 (18.2%) 2,920 (19.8%) 3,542 (24.0%) 5,620 (38.1%) 5,767 (39.1%) 8,995 (60.9%) 6,302 (42.7%) 3,224 (21.8%) 5,236 (35.5%) 0 10,632 (23.9%) 11,193 (25.2%) 11,759 (26.4%) 10,913 (24.5%) 17,654 (39.7%) 26,838 (60.3%) 16,707 (37.5%) 10,586 (23.8%) 17,201 (38.7%) 3 ( 0.0%) » Excluding Deaths »* Average across hospitals. 169 each of the comparison hospitals had engaged in some form of utilization review, d u r i n g the study per iod, the S W I T C H p r o g r a m was the most systematic and the only one initiated o n admiss ion, and activated on a daily basis, throughout each inpatient episode. Moreover , the S W I T C H program was the only one to incorporate a documented set of "appropriate day-of-care use" criteria. In contrast, the procedures fol lowed by the N o n - S W I T C H hospitals were typically retrospective and tended to focus pr imar i ly on assessing the need for continued hospitalization of long-stay cases. O n e of the three hospitals had also incorporated a system of concurrent discharge p lanning , us ing length of stay target dates defined by diagnosis, into its uti l ization review process. T h u s , a l though there was some variat ion in the intensity and rigour wi th which util ization review procedures were appl ied i n the comparison hospitals, none of these procedures approached the stringency and frequency of application typical of the S W I T C H program. Final ly , wi th the exception of one hospital which , in June 1984, changed its definition of a long-stay case from 10 to 7 days a n d changed its meeting schedule from weekly to monthly , no other changes in the uti l ization procedures employed by the N o n - S W I T C H hospitals d u r i n g the study per iod were reported. IV.4 Length of Stay Measures Three measures were constructed, us ing a m o d e l developed by Lave & Leinhardt (1976a), to describe the length of stay behaviour of the two hospital groups over the four-year study period. T h e first, the unadjusted average length of stay ( U A L S ) , was computed s imply as the mean length of stay for each hospital in each time period. H o w e v e r , because the U A L S does not control for differences in the average case mix either over time or across 170 hospitals, comparisons based on this measure could be misleading. Therefore, to control for inter-hospital variations in case-mix, a standard set of CMG's was defined for use in computing the remaining two length of stay measures. This standard set of CMG's consisted of all the CMG's that appeared in the PADH data set during period 1, except C M G 900 (any unrelated OR procedure) and CMG 901 (missing data necessary for C M G assignment). Across all time periods, the standard set of CMG's accounted for 90% of the PADH patient discharges and 88% of the Non-SWITCH hospitals' discharges. Next, to control for intra-hospital changes in case-mix over time, the second length of stay measure, an adjusted length of stay (ALS), was constructed. To compute the ALS, the frequencies associated with the standard set of CMG's in time period 1 were determined and were used to weight each CMG-specific average length of stay in each subsequent period, for each hospital1. Thus, the ALS reflects the average length of stay each hospital would have had in some period j, if it had the same case mix as the first time period but the length of stay of period j (Lave & Leinhardt, 1976a). Once calculated, the ALS was then used to compute the actual dependent variable used in the length of stay analyses. This variable, the adjusted length of stay ratio (ALSR) was computed as the ratio of the ALS in period j divided by the ALS in period 1. The ALSR was computed for each hospital, For example, if a hospital had 50 cases of CMG 14 in period 1 with an average length of stay of 5.0 days and 75 cases of CMG 75 with an average length of stay of 6.2 days, while in period 2 it had 60 cases of CMG 14 with an average stay of 4.5 days and 50 cases of CMG 75 with an average stay of 6.0 days, then the actual average length of stay in period 1 is [(50 x 5) + (75 x 6.2)1/125 = 5.72 and in period j is [(60 x 4.5) + (50 x 6)]/110 = 5.18. The adjusted length of stay for period j, however, would be [(50 x 4.5) + (75 x 6)]/125 = 5.40. 171 in each time period and reflects the change in length of stay over time, relative to the length of stay in period 1. IV.4 Data Analyses To determine whether the S W I T C H was effective in reducing the average length of stay of the P A D H , time series analyses were performed using graphical and ordinary least squares regression techniques (Gillings, Makuc, & Siegel, 1976; Neter & Wasserman, 1974; Veney & Kaluzny, 1984). The graphical procedures were used to examine general trends in length of stay over time and to provide a preliminary indication of the appropriateness of applying linear regression techniques. The regression analyses were used to assess the comparability of the P A D H and the Non-S W I T C H group of hospitals in the pre-intervention period, and to determine whether, in the post-intervention period, there was any evidence of a reduction in the P A D H length of stay which could be attributed to the introduction of the S W I T C H program. In general terms, the application of regression techniques to time series data (interrupted by a program intervention) are used to detect the following trends, any one of which, if observed relative to the pre-intervention period and to the comparison group, would be suggestive of a program effect: 1) A change in the slope of the regression line following program implementation. (See Figure 3a); 2) N o change in slope, but a change in the height of the response curve (i.e., the intercept) following program implementation. (See Figure 3b); or 172 Figure 3. Hypothet ica l Trends Detectable with T i m e Series Regression  A n a l y s i s 01 02 x 03 04 x = P r o g r a m Intervention 0 1 - 0 4 = T i m e Series Observations 01 02 x 03 04 1 01 02 x 03 04 173 3) A change in both the slope and the height of the response curve following program implementation. (See Figure 3c). In order to determine whether any of these changes have occurred, regression equations are defined to include all of the major dependent and independent variables under study plus a number of special indicator variables designed to permit testing of hypotheses relevant to changes in slope and intercept. The specific equations used in the present study - to ascertain whether there had been any reduction, attributable to the implementation of the SWITCH, in the length of stay at the PADH - are defined in Chapter V. All of the analyses reported in this thesis were performed using the Statistical Package for the Social Sciences (SPSSX). In addition, although primary interest was in the detection of a reduction in length of stay attributable to the SWITCH, because of the exploratory nature of the study, all tests of the regression coefficients were performed using two-tailed tests of significance. 174 CHAPTER V Results V.I Graphical Analyses V . l . l Unadjusted Average Length of Stay Between Period 1 and Period 12, the unadjusted average length of stay of PADH patients fell from 8.74 to 6.45 days, for a total decrease of 26 per cent. For patients in the non-SWITCH hospitals the unadjusted average length of stay also declined, falling from 8.34 to 7.97 days, for a total decrease of 4 per cent. Thus, as shown in Figure 4, the unadjusted average length of stay for the PADH started at a slightly higher level and decreased more than that of the Non-SWITCH hospitals. However, as these numbers do not control for differences in the average case mix of the hospital groups, they may be misleading. V.1.2 Case-Mix Adjusted Average Length of Stay After performing the necessary procedures to identify a standard set of CMG's and to compute the ALS, it was determined that the case-mix adjusted average length of stay for the P A D H fell from 8.63 days in Period 1 to 5.22 days in Period 12, a decline of 40 percent. For the Non-SWITCH group of hospitals the case-mix adjusted length of stay, averaged across the hospitals, dropped from 8.06 days in Period 1 to 7.36 days in Period 12, for a total decrease of 9 per cent. Thus, as shown in Figure 5, after adjusting for differences in case-mix, the average length of stay of P A D H patients still started somewhat higher and fell further than the average case-mix adjusted length of stay of the Non-SWITCH group of hospitals. To examine the change in average length of stay over time, relative to the length of stay in Period 1, the ALSR was computed and plotted, as shown in Figure 6. From Figure 4. Unadjusted Average Length of Stay by Hospi ta l G r o u p 9 o , Unadjusted Average Length of Stay 8 7 •• 6 •• 5 •• 4 •• 3 •• 2 •• 1 •• 0 •• 1 ••- Non-SWITCH •°- PADH •+- •+- -+- -4-5 6 7 8 Time Period T 10 11 12 SWITCH Initiated Figure 5. Case -Mix Adjusted Length of Stay by Hospi ta l G r o u p 9 T o. 8 7 •• 6 •• Adjusted 5 " Length of Stay 4 .. 3 •• 2 •• 1 •• 0 •• fv 1 -+-••- Non-SWITCH •O- PADH H H H 1 1 h 2 3 4 5 6 7 8 9 1 0 1 1 1 2 Time Period T SWITCH Initiated 176 Figure 6. C a s e - M i x Adjusted Length of Stay Ratio by Hospi ta l G r o u p Adjusted Length of Stay Ratio i'.oo< 0.8 •• 0.7--0.6--0.5 •• 0.4--0.3--0.2--0.1 •• 0.0--1 SN>x/^>N<vb o — o < ••- Non-SWITCH •O- PADH -+• -+-5 6 7 8 Time Period t SWITCH Initiated 10 11 12 this figure it can be seen that, whi le there appears to have been no systematic change in the A L S R over time for the N o n - S W I T C H group of hospitals, the A L S R of the P A D H has clearly declined, in what could appropriately be described as a linear pattern, over the time per iod studied. Because of the variabil ity in the A L S R , however, it is difficult to tell whether the observed decrease was more, or less, pronounced in the post-intervention per iod . Moreover , the graphs do not take into account the possible effect of changing occupancy rates on the observed length o f stay trends. Consequently , in the next phase of the analysis, regression equations 177 were constructed to control for the effects of occupancy a n d to test for changes in A L S R trends fo l lowing implementat ion of the S W I T C H . V .2 Regression Analyses T o determine general linear trends, and to ascertain whether there h a d been a change in the slope of the regression line fol lowing implementat ion of the S W I T C H program, the fo l lowing equation was estimated for each of the P A D H and N o n - S W I T C H hospitals. A l t h o u g h a significant regression relation was not expected in the latter case, the analysis was inc luded for comparative purposes. (1) ALSRjt = b 0 + ^ P E R I O D + b 2 T S + b 3 O C C i t + e i t In equation (1), A L S R j t is the case-mix adjusted length of stay ratio for hospital i in per iod t. P E R I O D is a linear time variable for the quarterly time intervals a n d assumes the values of 1 through 12. The coefficient of P E R I O D wi l l indicate what the average quarterly rate of decline was over the per iod of time studied. T S is a linear spline (Lave & Leinhardt , 1976a), or indicator variable (Neter & Wasserman, 1974), wh ich takes o n the value of 0 for Periods 1 through 9 and the values 1 to 3 for Periods 10,11, and 12, respectively. Essentially, T S provides the information needed to construct two regression lines: one in the per iod prior to ful l implementat ion of the S W I T C H , and one in the per iod after implementation. By inc lud ing T S in the regression equation, it becomes possible to test whether the slope of the regression line changed fo l lowing implementat ion of the S W I T C H . M o r e specifically, the coefficient of T S wi l l indicate whether there was a change in the rate of decline after the introduction of the S W I T C H . A positive coefficient wi l l suggest that the A L S R began to increase fo l lowing implementat ion of S W I T C H ; a non-significant coefficient wi l l suggest that 178 there was no change in slope attributable to S W I T C H ; and a negative coefficient wi l l suggest that, fo l lowing S W I T C H , the A L S R started to decline at a more r a p i d rate than previously. The final independent variable inc luded in equation (1), O C Q t is the occupancy rate of hospital i in per iod t a n d is inc luded in order to control for length of stay effects associated with differences in hospital occupancy. T h e results of comput ing equation (1) are presented in Table IX. A s expected, a significant regression relation was observed for the P A D H data (F= 15.818, p_ <.001), with the mode l accounting for 80% of the variabil ity in A L S R . In contrast, but also as expected, no significant regression relation was observed for the N o n - S W I T C H hospitals (F= 2.000, p > .05). Moreover , for this latter group, the defined regression m o d e l accounted for only 8% of the variabil ity in A L S R . W h i l e this lack of fit suggests that, for this group, there was no linear decline (or increase) in A L S R over the time per iod studied, it does not preclude the possibility that some other type of relationship might exist. A n examination of Figure 6, however, suggests that the most l ikely explanation for the observed lack of fit is that there has been no systematic change in A L S R over time for the N o n - S W I T C H group of hospitals. Similarly , a comparison of the coefficients of P E R I O D for the P A D H and N o n - S W I T C H hospitals, reveals that whi le the A L S R decreased b y .05 per per iod in the P A D H , no significant rate of decline was observed for the N o n - S W I T C H hospitals. [In fact, the only significant relationship identified for the N o n - S W I T C H group, was that of a .005 increase in A L S R wi th each unit increase in O C C rate.] Next , to determine whether there had been a 179 Table IX. Regression Coefficients 1 us ing A L S R as the Dependent Variable:  P A D H and N o n - S W I T C H Hospitals- Standard Set of C M G ' s Independent Variables P A D H Equations (1) (2) A L S R A L S R N o n - S W I T C H Equat ions (1) (2) A L S R A L S R Cons tant -.263 -.208 .527 .529 Per iod -.049 (-4.650)** -.050 (-5.316)*** -.005 (-1.168) -.005 (-1.124) T S .033 (1.527) .009 (.532) D T V .089 (2.019) .019 (.495) O C C .015 (2.230) .014 (2.367)* .005 (2.063)* .005 (2.043)* F-Ratio 15.818*** 18.935*** 2.000 1.981 R 2 - A d j u s t e d .80 .83 .08 • .08 Ordinary least-squares estimation: t-statistics appear in parentheses (* indicates p < .05; ** indicates p<.01; *** indicates p<.001). The number of observations for PADH is 12; 1 hospital and 12 time periods. The number of observations for the Non-SWITCH group is 36; 3 hospitals and 12 time periods. 180 change in linear trends, (i.e., a change in slope), fo l lowing implementat ion of the S W I T C H , the coefficients of T S were examined. In accordance with expectations, the coefficient of T S for the N o n - S W I T C H group was not significant. Contrary to expectations, however, the coefficient of T S computed for the P A D H was likewise non-significant. T h u s , it w o u l d appear that the general rate of decline in A L S R observed in the pre-intervention per iod cont inued into the post-intervention per iod. M o r e specifically, on the basis of the observed coefficient of TS , there is insufficient evidence to suggest that, fo l lowing implementat ion of the S W I T C H , the A L S R decl ined more rapidly . The absence of a change in the slope of the regression line, however, does not preclude the possibility of a change i n the height of the response curve. Therefore, to test for this possibility, the fo l lowing equation was computed for each hospital group: (2) A L S R j t = a 0 + - a 1 P E R I O D + a 2 D T V + a 3 O C C i t + e i t Equat ion (2) differs from (1) only in the way that the time variable is entered. T S has been d r o p p e d from the equation, since no evidence was obtained to suggest a post-intervention change in slope, and D T V has been added. D T V is a d u m m y variable that takes on a value of 0 for Periods 1 through 9, a n d a value of 1 otherwise. In contrast to the change in slope measured by the T S variable in equation (1), however, equation (2) is designed to determine only if there has been a discontinuity (i.e., a change in height) in the regression line fo l lowing p r o g r a m implementat ion. T h u s , a posit ive D T V coefficient wi l l indicate whether, o n average, the A L S R was higher in the post-intervention per iod than in the pre-intervention period; a non-significant coefficient w i l l indicate that there was no change in the height of the response curve fo l lowing p r o g r a m implementat ion; and a 181 negative coefficient wi l l indicate that, on average, the A L S R was lower in the post- intervention per iod . T h e results of comput ing equation (2) are also shown in Table IX. Once again, the model prov ided a good fit to the P A D H data (F=18.93, p_ <.001), accounting for 83% of the variability in A L S R . Furthermore, as expected, no evidence of a significant regression relation was observed for the N o n - S W I T C H group. Similarly, as indicated by the coefficient of P E R I O D , the same trends showed u p with respect to the slope of the regression line. Final ly , an examination of the coefficient of D T V indicated that there h a d been no significant change in the height of the response curve, fo l lowing implementat ion of the S W I T C H , for either the P A D H or the N o n - S W I T C H group of hospitals. Thus , o n the basis of equations (1) a n d (2), there is little evidence to suggest that, for the P A D H , there has been a reduct ion in length of stay attributable to the introduction of the S W I T C H program. Before such a conclusion can be drawn, however, certain factors relevant to the aptness of the regression models must be taken into account. T o determine the aptness of the regression mode l , for each hospital group, residual analyses were performed in order to detect problems pertaining to non-normal i ty , autocorrelation, a n d heterogeneity among the error terms. A s in any application of regression analysis, certain assumptions must be met in order for the application of time series regression to be appropriate. Specifically, it is assumed that the error terms are independent , normal ly distributed r a n d o m variables wi th equal 182 variance across all levels of the independent var iable 1 . T o determine whether these assumptions were reasonable in the present case, residual plots were constructed and examined for patterns reflecting a violation of the assumptions of normali ty and variance homogeneity. In addit ion, because non-independence of the error terms can be a prob lem with time series data, the Durb in-Watson test for autocorrelation (Neter & Wasserman, 1974) was performed. The results of these analyses indicated that, for the P A D H , none of these potential problems were important issues. T h e residual plots p r o v i d e d no indication that the assumptions of normal i ty and homogeneity h a d been violated, and the computed D u r b i n -Watson statistics were 2.55 and 2.61, for the two equations respectively. Both of these values exceeded the critical value of 1.75 and , therefore, reflected no autocorrelation a m o n g the error terms. Similarly , for the N o n - S W I T C H group, there was no evidence of autocorrelation (Durbin-Watson = 2.20 and 2.55, respectively; exceeding critical value of 1.65) among the error terms in either equation. H o w e v e r , the residual plots p r o v i d e d some evidence to suggest that variance heterogeneity m a y have been a problem. This observation was not unexpected since it was clear from Figure 6 that, for this group, the A L S R h a d been more variable in the earlier periods of the study than in the later periods. H o w e v e r , no attempt was made to transform the data, in order to correct this problem, because such action w o u l d have rendered the models constructed for this group non-comparable to those constructed for the P A D H . In addit ion, given the apparent lack of a In the case of a multiple regression model, the distributional properties attributed to the error terms for a given level of the independent variable assume that all other variables in the model are held constant. 183 systematic relation between A L S R and time, for the N o n - S W I T C H group, such a refinement was considered unwarranted. T h e next potential problem considered was the nature of the relationship a m o n g the independent variables, in particular between P E R I O D and O C C . In the application of mult iple regression techniques, it is desirable to include, as independent variables, only those variables that are h ighly correlated wi th the dependent variable and uncorrelated wi th each other. A l t h o u g h in most research endeavors it is unl ike ly that all of the independent variables wi l l be completely unrelated to each other, the inclus ion of h ighly inter-correlated independent variables can cause a prob lem referred to as multicoll inearity (Neter & Wasserman, 1974). In essence, w h e n there is a h igh degree of association among the independent variables, computed regression coefficients become unstable and prov ide only imprecise information regarding the relative importance of each independent variable. T h u s , on the basis of the computed regression mode l , the "estimated regression coefficients ind iv idua l ly m a y not be statistically significant even though a definite statistical relation exists between the dependent variable and the set of independent variables" (Neter & Wasserman, 1974, p. 339). Accord ing ly , an examination of the correlation coefficients computed among the independent variables inc luded in the present study, suggested that, for the P A D H , the regression coefficients presented in Table LX m a y have been affected by multicoll inearity among the independent variables. Specifically, a h igh positive correlation observed between O C C and P E R I O D (r = +.881; p_ < .001; two-tailed) m a y have caused problems. A significant negative correlation between O C C a n d A L S R (r = -.64; p_ = .024; two-tailed) suggested that O C C was an important explanatory 184 . variable in the observed length of stay trends. In the computation of equation (1), however, the coefficient of O C C failed to reach significance. G i v e n these apparently contradictory findings, consideration was given to the possibility that the multicoll inearity between O C C and Period was mask ing an O C C effect. Thus , equations (1) and (2) were re-estimated without O C C and then equation (1) was re-calculated, d r o p p i n g T S and us ing O C C as a proxy for P E R I O D . For comparative purposes, these same equations were computed for the N o n - S W I T C H hospitals, a l though multicoll inearity between O C C and P E R O D was not an issue for this group. Furthermore, for this group of hospitals, O C C was the only variable found to be correlated with the A L S R . H o w e v e r , in contrast to the P A D H , but in accordance wi th the regression results, this relationship was positive (r = +.340; p_=.042; two-tailed). T h e results presented in Table X conf irmed the previous findings that, for the P A D H , the A L S R decreased with time (.03 per period, with O C C excluded), a n d further that, w h e n no time variables were inc luded in the mode l , the A L S R decreased by .014 per each unit increase in O C C . Moreover , an examination of the goodness of fit measures (i.e., F-Ratio and Readjusted) indicated that, w h e n only O C C was used in the equation, the regression equation retained its significance and the m o d e l accounted for 36% of the variability in A L S R . Since, at the P A D H , O C C tended to increase over the study per iod , these findings prov ide support for the assumption that at least some port ion of the overall decline in A L S R observed for the P A D H was associated wi th the increase in the hospital's occupancy rate over time. F o r the N o n - S W I T C H hospitals, the results presented in Table X s imilarly concur wi th those presented in Table IX; i.e., there was no Table X . Regression Coefficients 1 us ing A L S R as Dependent Variable: P A D H and N o n - S W I T C H Hospitals - O C C R e m o v e d - Standard Set  of C M G ' s Independent Variables P A D H Equations (No OCC) (No OCC) (OCC Only) A L S R A L S R A L S R N o n - S W I T C H Equat ions (No OCC) (No OCC) (OCC Only) A L S R A L S R A L S R Constant .920 .946 2.046 .961 .961 .493 Per iod -.029 (-4.241)*** -.032 (4.723)** -.006 (-1.234) -.006 (-1.245) T S .019 (.790) .013 (.747) D T V .070 (1.314) .031 (.771) O C C -.014 (-2.658)** .005 (2.110)* F - R a t i o 14.736** 16.942 7.067* .788 .807 4.452* R 2 - A d j u s t e d .71 .74 .36 .00 .00 .09 Ordinary least-squares estimation: t-statistics appear in parentheses (* indicates p < .05; ** indicates p<.01; *** indicates p<.001). The number of observations for PADH is 12; 1 hospital and 12 time periods. The number of observations for the Non-SWITCH group is 36; 3 hospitals and 12 time periods. 1 86 significant regression relation between A L S R and the time variables but a rise in A L S R was associated with an increase in hospital occupancy rates. Res idual analyses computed for each of these models indicated that, once again, none of the potential problems of non-normali ty , variance heterogeneity or autocorrelation among the error terms were important issues. A l t h o u g h it was clear from the preceding analyses, that in the pre-intervention per iod the A L S R computed for the P A D H had already begun to decline, at a rate of approximately .05 per quarter, a further analysis was performed to determine whether the P A D H pre-intervention A L S R was lower, o n average, than that of the N o n - S W I T C H group of hospitals. A s hypothesized, it was expected that if the two hospital groups were comparable , then prior to the introduction of S W I T C H there should have been no significant differences in trend or average A L S R d u r i n g the pre-intervention per iod. It has already been established that there was a difference in trend. T o determine whether the two hospital groups differed wi th respect to their average pre-intervention A L S R , the data f rom the first 8 time periods for both the P A D H and N o n - S W I T C H hospitals were merged a n d the fo l lowing equation was computed: (3) A L S R i t = O C C i t + H G R P +e i t In this equation, A L S R ^ is the A L S R in Per iod i of hospital group t. O C Q t is the occupancy rate in Per iod i of hospital group t. H G R P is a d u m m y variable w h i c h defines the hospital group. T h u s , H G R P takes o n the value of 1 for the P A D H and the value of 0 for each of the N o n - S W I T C H hospitals. T h e coefficient of H G R P w i l l indicate whether the pre-intervention A L S R differed, on average, across hospital groups. The results of this analysis 187 indicated: 1) there was a significant regression relation between the dependent variable and the independent variables (F = 8.415; p_<.01); 2) the m o d e l accounted for 32% of the variabil ity in the pre-intervention A L S R (R 2 -Adjus ted = .323); 3) no significant effect of O C C was observed (bi = .001; t=.463; p_>.05); and 4) on average, the pre-intervention A L S R was .14 lower for the P A D H than for the N o n - S W I T C H group of hospitals (b 2 = -.140; t=-3.917; p_<.001). Thus , it w o u l d appear that in addit ion to the observed difference in trend between the two hospital groups, the P A D H h a d a significantly lower pre-intervention A L S R than the N o n - S W I T C H group of hospitals. V.3 Supplementary Analys i s A n examination of the age distributions of the P A D H and N o n -S W I T C H data bases (see Table VII) indicated tha t , while the N o n - S W I T C H patients were fairly evenly distributed among the four defined age groups, the P A D H patients were more heavily weighted in the upper age categories (X 2 = 1,051.3; df = 3; p_<.001). A l t h o u g h A g e was not inc luded as a variable in the above regression analyses (since A g e is one of the variables that was considered in the development of C M G ' s ) , to determine whether this higher proport ion of older patients m a y have masked an effect of the S W I T C H 1 , equations (1) and (2) were re-estimated using only those patients i n the P A D H and N o n - S W I T C H data bases that were under the age of 70 at the time of their discharge from the hospital. The assumptions made here are that older patients may tend to have longer stays, and that they may have less access to appropriate alternatives to acute care. Hence, even if they are identified as ready for discharge, discharge may have to be delayed until an appropriate alternative can be arranged. 188 T h e quarterly A L S R ' s computed for this group of patients, from both the P A D H and N o n - S W I T C H hospitals, are shown graphical ly in Figure 7 and the results of the regression procedures are presented in Table XI. A s can be seen from the regression results, the exclusion of the older age group of patients h a d little effect on the general trends in A L S R over time, for either the P A D H or the N o n - S W I T C H hospitals. H o w e v e r , the graph suggests that, by excluding older patients, the post-intervention length of stay trend observed at the P A D H has become more consistent, wi th a greater divergence from the N o n - S W I T C H group, than has been observed previously . If this trend continues, then it m a y be that wi th an extension of the fo l low-up per iod a reduction in the A L S R , attributable to the S W I T C H , could be detected. Figure 7. Case -Mix Adjusted Length of Stay Ratio by Hospital  G r o u p - Patients A g e d 18 - 69 0.8 0.7--0.6--Adjusted Length of Stay 0.5 •• Ratio 0.4 •• 0.3--0.2--0.1 •• 0.0 1 •O-Non-SWITCH PADH •+- -4- •4- -+- -»-6 7 8 Time Period y T SWITCH Initiated 10 11 12 190 Table XI. Regression Coefficients 1 us ing A L S R as Dependent Variable:  P A D H and N o n - S W I T C H Hospitals - Patients A g e d 18 to 69 - Standard Set of. C M G ' s Independent Variables P A D H Equations (1) (2) (No OCC) A L S R A L S R A L S R N o n - S W I T C H Equat ions (1) (2) (No OCC) A L S R A L S R A L S R Constant -.449 -.501 .935 .520 .548 .937 Per iod -.040 (-3.303)* -.041 (-3.519)** -.017 (-2.199) -.005 (-1.030) -.007 (-1.404) -.006 (-1.103) T S -.014 (-.564) -.029 (-1.022) .008 (.469) .012 (.673)) D T V -.022 (-.387) .039 (.970) O C C .017 (2.223) .018 (2.347)* .005 (1.887) .005 (1.803) F -Rat io 11.09** 10.873** 9.85** 1.637 1.902 .628 R 2 - A d j u s t e d .73 .73 .62 .05 .07 .00 Ordinary least-squares estimation: t-staristics appear in parentheses (* indicates p < .05; ** indicates p<.01; *** indicates p<.001). The number of observations for PADH is 12; 1 hospital and 12 time periods. The number of observations for the Non-SWITCH group is 36; 3 hospitals and 12 time periods. 191 C H A P T E R V I Discuss ion V I . l T h e S W I T C H Index System D i d the S W I T C H Index System effect a reduction in length of stay at the P A D H ? T h e evidence gathered in the present study w o u l d suggest that it d i d not. O n the basis of regression equations computed for both the P A D H a n d the N o n - S W I T C H group of hospitals, it was determined that while there was a general decline in the P A D H length of stay over the time per iod studied, there was no change in the rate of decline, fo l lowing the advent of the S W I T C H , nor was there a change in the average pre- and post-intervention lengths of stay. It is suggested, therefore, that although the P A D H length of stay d i d decline, over the time per iod studied, no component of this decrease could be attributed to the initiation of the S W I T C H . T o determine whether the observed decline cou ld be explained by general secular trends, the regression equations computed for both the P A D H and the N o n - S W I T C H hospitals were compared. The results of this comparison indicated that whi le the P A D H length of stay decreased over the study per iod , the length of stay of the comparison hospitals d i d not. Thus , it does not seem l ikely that general secular trends were responsible for the observed P A D H length of stay trend. H o w e v e r , a l imitation in the study methods prohibits the exclusion of secular trends as a possible explanatory variable. T o be reasonably confident that the observed decrease was not due to secular trends, it must be assumed that no biases were introduced in the selection of the comparison group of hospitals. That is, it must be assumed that the length of stay trend observed for the N o n - S W I T C H group of 192 hospitals was representative of the trend that w o u l d have been observed h a d all of the eligible P A D H peer group member hospitals participated in the study. A s noted in Chapter IV, it was not possible to determine whether the hospitals that d i d not participate in the study differed sufficiently from the participating hospitals to affect the observed length of stay trend. Thus , the possibility that the length of stay results were affected by self-selection biases cannot be discounted. A s s u m i n g the observed P A D H length of stay trend was not attributable to secular trends, nor to the S W I T C H , what other factors could have contributed to the general decline? W h i l e no direct attempt was made in the present study to answer this question, the results suggest that at least one of the potentially important variables is hospital occupancy. For the P A D H , hospital occupancy was found to be negatively associated with length of stay, whi le increasing over time. These findings suggest that, for this hospital, the observed decline in length of stay was at least partially associated wi th an increase in occupancy rate over time. In contrast, for the N o n - S W I T C H group of hospitals, there was a positive association between hospital occupancy and length of stay and no linear relationship between occupancy and time. These apparently contradictory findings are difficult to explain but a consideration of the distributions of hospital occupancy (See Table VII in Chapter IV), for the two hospital groups, provides a potential clue. D u r i n g the time per iod under study, the occupancy rate of the P A D H was, o n average, considerably higher than that of the N o n - S W I T C H group of hospitals (Medians = 91% and 83%, respectively). In fact, the m e d i a n quarterly occupancy rate observed at the P A D H even surpassed the m a x i m u m quarterly occupancy rate observed for the N o n - S W I T C H group 193 (i.e., 87%). These observations suggest that, perhaps, there is a different functional relationship between occupancy and length of stay depending on the level of occupancy under consideration. M o r e specifically, it is suggested that unti l a hospital's occupancy rate consistently approaches m a x i m u m capacity, there m a y not be sufficient incentive for physicians to ensure that patients are discharged as soon as is medical ly appropriate. In fact, when occupancy rates consistently fall in the 80-85% range, suggesting moderate rather than intense pressure on bed availability, there may be an incentive for physicians to keep patients in the hospital longer, so as to ensure that they wi l l have access to a bed when it is needed for a subsequent patient. O n the other hand , as occupancy approaches capacity, peer pressure from other admitt ing physicians m a y be sufficient to motivate more t imely discharge. O f course, such an explanation is pure ly conjecture at the present time. The relationship between hospital occupancy and length of stay is complex, and not yet adequately understood. Nonetheless, the results of the present study concur wi th previous research which has identified occupancy as a potential confounder in length of stay analyses.(Lave & Leinhardt , 1976a, 1976b). A l t h o u g h the general decline in length of stay at the P A D H m a y be at least partially attributable to the effects of hospital occupancy, it is clear from the regression results that other factors must also have been involved . W h e n occupancy alone was used as a predictor variable, it was found to account for only 36% of the variability in the adjusted length of stay ratio. O n the other hand , w h e n time was the only variable considered, 72% of the variabil ity was accounted for and, w h e n both time and occupancy were considered, the predictive power of the m o d e l increased to 80%. T h e inference d r a w n from these results is that, in addit ion to hospital occupancy, 194 some other, as yet undef ined, variable(s) were exerting an influence on the observed trend in P A D H length of stay. Unfortunately, because the present study was designed to prov ide only a global indication of length of stay trends over time, it is not possible to identify what these, addit ional confounders might have been. H o w e v e r , us ing a similar rationale to that discussed above, it is possible that because of the high and increasing occupancy rate observed over time at the P A D H there m a y have been an increased awareness, o n the part of the hospital's medical staff and prior to the development and implementat ion of the S W I T C H , of the need to monitor and ensure efficient and appropriate util ization of an increasingly scarce hospital resource, namely the hospital bed. A s is suggested by the f indings of previous investigators (Averi l l , & M c M a h o n , 1977; D o b s o n et al., 1978; L a v e & Leinhardt , 1976a), there m a y be reason to believe that one of the most important features of an effective uti l ization management p r o g r a m is not the specific mechanics that are used but, rather, the increased status a n d priori ty accorded the util ization process wi th in the hospital setting. Consequent ly , if in the p r e - S W I T C H per iod both the management a n d medica l staff of the P A D H h a d begun to perceive uti l ization management as a priority issue for the hospital, then it is possible that the observed reduction in length of stay may , at least in part, have reflected an informal response to this perception (i.e., a H a w t h o r n e effect). In this context, the subsequent initiation of the S W I T C H m a y s imply have reflected the formalizat ion of a util ization management process that was already well underway . Certainly , it was clear from conversations with hospital personnel that uti l ization problems h a d been under discussion for some 195 time prior to the introduction of the S W I T C H . T h u s , the possibility of an informal response w o u l d appear to have been quite likely. VT.2 Alternative Explanations for the Failure to Detect a Program Effect Despite the evidence gathered in the present study, it is too early to state conclusively that the S W I T C H Index System d i d not effect a reduction in length of hospital stay. W h i l e this possibility cannot be discounted, there are several reasons w h y the present study m a y have failed to detect a reduct ion in length of stay that could have been attributable to the S W I T C H . 1) Pr ior to the initiation of the S W I T C H , concerns about uti l ization problems m a y already have begun to have an effect on the hospital's length of stay. Because the S W I T C H Index System was developed and implemented prior to undertaking the present evaluation, it was not possible to get a first-h a n d assessment of the nature a n d intensity of the uti l ization review procedures employed prior to the advent of the S W I T C H . Consequent ly , as discussed above, it is not possible to discount the possibility that, g iven the consistently h igh occupancy rate experienced by the hospital throughout the study per iod, the effectiveness of the S W I T C H m a y have been attenuated by earlier responses to perceived uti l ization problems. Clear ly , uti l ization issues h a d been of sufficient concern to p r o m p t the development a n d implementat ion of what was, at the time, a h ighly unique and innovative uti l ization management program. It is possible, therefore, that h a d the p r o g r a m been implemented in a hospital for w h i c h occupancy was less of a prob lem a n d for w h i c h util ization was only beginning to emerge as a priority, a different pattern of results m a y have been observed. 196 2) T h e post-intervention per iod m a y have been too short (i.e., too few observations) to detect a reduction in length of stay. G i v e n the small number of observations u p o n w h i c h the regression analyses were based, particularly in the post-intervention per iod, there m a y have been insufficient power to detect anything more than a major change in the slope a n d / o r intercept of the regression line fo l lowing the implementat ion of the S W I T C H . In addit ion, since the S W I T C H was designed to identify and eliminate inappropriate days of care resulting from inefficiencies in organizational procedures, clinical practice patterns, and external services - all of w h i c h could require a considerable investment of time to correct- there m a y be a lag between the program's implementat ion a n d the time w h e n it cou ld be expected to have a discernable effect o n length of stay. Specifically, as each Off-Index case is identified, and as patterns of Off-Index use begin to emerge, time is required to ascertain the source of the prob lem, to formulate and prov ide relevant feedback, and to devise and implement appropriate corrective action. This w o u l d be part iculary true in the case of Off-Index days which occurred because of a lack of appropriate alternatives. A s noted in Chapter II (section U.4.8.3), studies examining the reasons for inappropriate use have typically identified a lack of appropriate alternatives as one of the pr imary factors contributing to inappropriate acute care bed uti l ization (Gertman & Restuccia, 1981; Restuccia & H o l l o w a y , 1976; Z i m m e r , 1974). H o w e v e r , the development of the alternative sevices needed to alleviate this prob lem requires system-wide changes over which an i n d i v i d u a l hospital m a y have little immediate control. Hence , if a lack of appropriate alternatives was a significant problem for the P A D H , a longer 197 fo l low-up per iod w o u l d l ikely be needed before a length of stay effect could be expected to occur. A t the same time, however, if too m u c h time has to pass before an effect is observed, it wi l l be difficult to assert that the effect was due to the S W I T C H and not to other intervening variables. 3) It could be that as the length of stay at the P A D H declined, its case mix severity increased. If so, and if this increased severity was found to be related to an increase in length of stay, then it is possible that an effect of the S W I T C H m a y have been missed. A l t h o u g h C M G ' s were used in an effort to control for changes in hospital case mix over time, it may be that the failure, documented by previous investigators (Brewster et al. , 1985; H o r n & H o r n , 1984; H o r n , H o r n , & Sharkey, 1984; Y o u n g , 1984), of C M G ' s to adequately account for changes in severity m a y have masked a length of stay reduct ion attributable to the S W I T C H . 4) It is difficult to discern a length of stay effect in a hospital that is in the midst of a secular decline (for whatever reason). In order to have a better chance of detecting a p r o g r a m effect, the S W I T C H Index System should have been implemented in a hospital k n o w n to be in a length of stay steady state. T h e fact that the P A D H d i d not have a stable length of stay in the pre-intervention per iod makes it very difficult to identify any component of the general decline that m a y have been attributable to the S W I T C H . This is a major prob lem for evaluations 198 initiated after the development and implementat ion of the p r o g r a m intervention, a n d one that is not amenable to post hoc control. VI.3 Methodolog ica l Issues and Study Limitat ions A l t h o u g h the study is clearly exploratory, it represents one of the first attempts to evaluate a util ization program, implemented in a C a n a d i a n acute care setting, us ing a mult iple time series design wi th a "control" group. W h i l e this design represents an improvement over the more typical single observation, pre- a n d post-intervention case study designs, the study still suffers from a number of serious limitations and methodological problems. O n e of the most serious limitations of the present study, is that the S W I T C H Index System was implemented in only one hospital . Furthermore , the selection of this one hospital was not r a n d o m but was dec ided o n the basis of a recognized concern, on the part of the hospital's administrative a n d medica l staff, about hospital uti l ization. A l t h o u g h an effort was made to take into account the potential effects of secular trends a n d certain hospital-specific factors (eg., occupancy rate, bed size, and teaching status, and case-mix), the potential effects of other unidentif ied, hospital-specific characteristics on the observed length of stay trends at the P A D H are u n k n o w n . Consequently , there is no assurance that the pattern of results seen at the P A D H w o u l d be repeated if the S W I T C H was implemented in another hospital , or group of hospitals. Th i s l imitation w o u l d have been even more problematic h a d an effect of the S W I T C H been found. In that event, it w o u l d have been very difficult to determine with any certainty whether the observed effect was due to the S W I T C H or to other non-related, hospital-specific factors. 199 Similarly , the selection of the comparison group of hospitals was not r a n d o m but was determined o n the basis of specified eligibility criteria and each hospital's wil l ingness to participate in the study. A l t h o u g h the criteria used were reasonable and necessary (in order to increase the comparabil ity of the hospitals), reliance u p o n voluntary participation limits the extent to w h i c h the length of stay trends exhibited by the comparison hospitals can be considered representative of the trends that w o u l d have been observed h a d all eligible hospitals (or a r a n d o m selection of eligible hospitals) participated. A second limitation relates to the fact that no attempt has as yet been made to establish the reliability and val idity of the S W I T C H Index criteria. Consequently , there is a possibility that the lack of an observed effect could be due to the fact that the defined criteria may be lacking in reliability, val idity, or both. Clearly , an attempt to establish the reliability and val idi ty of the p r o g r a m criteria should have been undertaken prior to the implementat ion of the program. A third l imitation relates to the fact that no attempt was made to ensure o n an ongoing and formal basis, that the system was being appl ied consistently and in accordance wi th defined procedures. The operation of the system was not moni tored because the evaluation was initiated after the final time per iod inc luded in the study. T h u s , the possibility that the lack of an effect was due to incomplete or inconsistent application of the system cannot be discounted. Because of these problems, the results of the present s tudy prov ide little basis for d r a w i n g conclusions about the effectiveness or ineffectiveness of the S W I T C H , as a mechanism for reducing length of stay. Nonetheless, sufficient evidence has been gathered to merit caution in p r o m o t i n g the 2 0 0 S V V T T C H Index System as a tool designed to facilitate reductions in hospital length of stay. VI .4 Implications for Future Research and P r o g r a m Planning T h e results of the present study prov ide only global and pre l iminary f indings relevant to a determination of the effectiveness of the S W I T C H Index System of util ization management. The study has focussed on only one of the variables considered important i n the uti l ization management process, namely length of hospital stay. The potential effects of the S W I T C H o n other important variables such as quality of care, intra-hospital communicat ion , the efficiency a n d / o r effectiveness of the hospital's operational procedures, or the util ization of ancil lary services, have not been considered. A l l of these issues merit further study. In addit ion, although the results of present study prov ide some indirect support for the assumption that inappropriate and unnecessary util ization does occur - since, for a standard set of C M G ' s , the P A D H length of stay decreased - no formal attempt has as yet been made to quantify inappropriate use. The S W I T C H Index could be used to facilitate this process. In addit ion, although the present study was unable to detect a length of stay effect of the S W I T C H , the system still has intuitive appeal and greater clinical relevance than other, more statistically oriented, programs. For these reasons, it is l ikely to be more acceptable to medical personnel than programs based on systems such as C M G profile analyses. Since no program wi l l succeed without the support of physicians, this is an important consideration. T h u s , g iven the limitations of the present study, the potential of the S W I T C H to be an effective mechani sm for reduc ing inappropriate uti l ization shou ld not be 2 0 1 discounted too quickly. Moreover , since the S W I T C H Index System provides a mechanism for identifying the reasons for inappropriate days of care, it could also be used to begin developing a better understanding of the factors that contribute to inappropriate bed use. This is important because it w o u l d promote a better appreciation of the extent to which inappropriate use is or is not under the control of the hospital a n d / o r its medical staff. C lear ly , neither hospital administrators nor physicians should be held -accountable for uti l ization problems over which they have no control. For example, lack of available alternatives to acute care hospitalization is a prob lem of the health care system, not of ind iv idua l hospitals. Thus , if a sizeable proport ion of inappropriate bed use is attributable to a lack of alternatives, then this is a problem that needs to be addressed at the prov inc ia l rather than the hospital level. By us ing the S W I T C H Index, or a similar tool such as the A E P , objective and quantifiable data o n the reasons for inappropriate acute care utilization could be gathered and employed in deve lop ing new strategies for reducing excessive hospital use. A more immediate concern, however, is the need for studies that focus o n criteria development and testing. Neither the S W I T C H criteria, nor any other criterion-based util ization index, should be used as a tool for research or uti l ization management unti l it has been demonstrated to be both reliable a n d val id . For this reason, hospitals and researchers interested in us ing appropriateness-of-care measures, might be advised to begin wi th indices that have already been subject to extensive testing. For example, the A E P cou ld be employed . H o w e v e r , since this index wvas developed in the U n i t e d States, it w o u l d still be desirable to conduct reliability and validity 202 tests to determine whether the results of the A m e r i c a n studies could be replicated in C a n a d i a n hospital settings. O n c e it has been established that a reliable and va l id measure of appropriate use has been developed, then to demonstrate its effectiveness as a uti l ization management tool, evaluative projects should be undertaken to determine not only its effect on util ization measures such as length of stay, but also o n quality of care indicators such as readmission and mortality rates a n d changes in illness severity from admission to discharge. Clear ly , one of the most c o m m o n l y expressed fears about the implementat ion of uti l ization management programs has been that reductions in use m a y be achieved at the expense of the quality of care prov ided . Since a pr imary objective of uti l ization management is to reduce excessive uti l ization, while maintaining an adequate standard of care, it is important to be able to demonstrate that both components of this objective have been met. In addit ion, although the present study has focussed on a criterion-based uti l ization management program, it is apparent from the discussion of the literature in Chapter HI, that a number of C a n a d i a n hospitals have begun to experiment wi th other types of programs. Since these different approaches to uti l ization management can vary considerably in cost and labour requirements, it is important that studies be initiated to compare the cost-effectiveness of these, a n d other, newly developed strategies. Final ly , it must be recognized that evaluating the effectiveness of a uti l ization management p r o g r a m is a complex process. A s discussed in Chapter II, m a n y variables have been found to have an effect o n util ization measures such as length of stay. Moreover , because of the complexity and mult ipl ic i ty of factors associated wi th the hospital util ization process, a 2 0 3 comprehensive m o d e l of uti l ization which could be used as a framework for research or p r o g r a m evaluation, has not yet been developed. Since, to date, most of the research directed at hospital util ization has been performed wi th in the A m e r i c a n hospital sector, there is a need for the development of research projects that w o u l d both facilitate a better understanding of the uti l ization process, i n the C a n a d i a n hospital environment , and permit the development of a clearer methodological approach to the evaluation of uti l ization management programs. In addit ion, there is a need for the development of alternatives to length of stay as a measure of the effectiveness of uti l ization management programs. M o r e informative outcome measures might include the number of inappropriate days of care, the n u m b e r of inappropriate admissions, reasons for inappropriate use, and changes over time in hospital case mix severity. In s u m m a r y , because there has been so little research done on uti l ization in the C a n a d i a n hospital sector, there are, at present, m a n y issues that need to be addressed. In the op in ion of the present author, it is essential that this research be initiated immediately. Moreover , given the complexity of the issues invo lved , and the need for inter-hospital collaboration, the author w o u l d concur wi th a suggestion made recently by Suttie (1988) that a national, or a series of interconnected regional, research centres be established with a mandate to perform hospital uti l ization research a n d to oversee p r o g r a m development and evaluation. T h r o u g h the establishment of such centres, it w o u l d be possible to develop the knowledge , expertise a n d coordination required to improve the state of C a n a d i a n hospital util ization research. In addit ion, by coordinating p r o g r a m development and evaluation, it w o u l d be possible to avo id some 2 0 4 of the problems encountered in the present study, by m a k i n g it more feasible for evaluation plans to be incorporated into the development of p r o g r a m initiatives, right from the start. VI.5 Pol icy Implications W h i l e it is too early to make specific recommendations for the development of uti l ization management policies, based o n the S W I T C H Index System, certain general recommendations can be made. First, an impetus for the development and refinement of uti l ization management procedures could be p r o v i d e d by provincial governments if, as was suggested by Suttie (1988), they were to begin requiring that hospitals prov id e evidence of hav ing performed appropriate uti l ization analyses prior to their receiving approval for requests for capital expansion or deficit funding . Second, in order to ensure that standards of appropriate util ization are deve loped, implemented , and proper ly evaluated, regional committees (along the lines of the A m e r i c a n PRO's) , comprised of physicians, health care administrators, and other relevant health service professionals, cou ld be established with a specific mandate to monitor hospital uti l ization and quality of care activities. T h i r d , to ensure that pol icy development is based o n adequate knowledge , continuing support and promot ion of research and development, particularly in the areas of hospital uti l ization, quality of care assessment, and p r o g r a m development, is essential. F inal ly , since the single, most significant determinant of hospital uti l ization is the medical condit ion of the patient (Berki et al., 1984; Boaz, 1979; C l e n d e n n i n g et al. , 1976; Goldfarb , Hornbrook , & Rafferty, 1980; Goldfarb , H o r n b r o o k , & Higg ins , 1983; L a v e & Leinhardt, 1976a, 1976b; M u g h r a b i , 1976; Ro, 1969), there is a need for the development of policies a imed at encouraging 2 0 5 ongo ing development of effective medica l technologies, whi le promot ing more intensive evaluation of technologies that are of little k n o w n value (Anderson & L o m a s , 1988). VI.6 C o n c l u s i o n In conclusion, al though this study was unable to conf irm the expectation that the S W I T C H Index System w o u l d be an effective mechani sm for reducing hospital length of stay at the P A D H , further research is needed before conclusions about its potential effectiveness in other hospital settings can be drawn. However , before any attempt is made to promote the use of the S W I T C H , even for pilot testing purposes, it is essential that the S W I T C H Index criteria be. evaluated for reliability and validity. If possible, once it has been demonstrated that a reliable and va l id "appropriateness of care" instrument has been developed, a pilot project i n v o l v i n g a new set of hospitals should be initiated wi th an appropriate evaluation p lan inc luded as a condit ion of its implementat ion. Furthermore , the proposed evaluation, if undertaken, should be expanded to inc lude a broader spectrum of util ization measures, such as length of stay, number of inappropriate days of care, a n d reasons for inappropriate use, as wel l as quality of care indicators such as changes in illness severity from admiss ion to discharge and readmission and mortality rates. In this way, more useful, and less ambiguous , information could be obtained about the effectiveness of the S W I T C H approach to uti l ization management. 206 References A d l e r , N . E . , & Mils te in , A . Evaluat ing the Impact of Physic ian Peer Review: Factors Associated with Successful PSROs . A m e r i c a n Tournal of Publ ic  Health. 1983; 73 (10): 1182-1185. A n d e r s o n , G . F . , & Steinberg, E . P . Hospi ta l Readmissions in the Medicare Populat ion. T h e N e w E n g l a n d Tournal of Medic ine . 1984: 311 (21): 1349-1353. A n d e r s o n , G . M . , & Lomas , J. Determinants of the Increasing Cesarean Birth Rates - Ontario Data 1979 to 1982. N e w E n g l a n d Tournal of Medic ine . 1984; 311: 887-892. A n d e r s o n , G . M . , & Lomas , J. Expla in ing Variations in Cesarean Section Rates: Patients, Facilities or Policies? C a n a d i a n M e d i c a l Associat ion  Tournal. 1985; 132: 253-259. A n d e r s o n , G . M . , & Lomas , J. T h e Deve lopment of Ut i l i zat ion Analys i s :  H o w , W h y and Where It's G o i n g . Paper prepared for the First A n n u a l Hea l th Pol icy Conference conducted by the Centre for Hea l th Economics a n d Pol icy Analys i s , Hami l ton , Ontario. M a y 27, 1988. A v e r i l l , R . F . , & M c M a h o n , L . F . A Cost Benefit Analys i s of C o n t i n u e d Stay Certification. M e d i c a l Care. 1977; 15 (2): 158-173. Barrable, B. Bed Closures: A n Impact Analys is . Hea l th Management  F o r u m . 1987; 8 (3): 38-44. Bean, E . Latest Survey Shows Hospi ta l Charges Increasing Far M o r e Q u i c k l y the CPI . T h e W a l l Street Tournal. January 6, 1988. Becker, S .W., Shortell , S . M . , & Neuhauser , D . Management Practices and Hospi ta l Length of Stay. Inquiry. 1980; 17: 318-330. Bennett, J . E . , & K r a s n y , J. Heal th-Care in Canada: A Series o n the Nation's Heal th . Reprinted from the series appearing in The Financia l Post. M a r c h 26 - M a y 7,1977. Berki , S .E. G r a d i n g D R G s . M e d i c a l Care. 1984: 22 (12): 1065-1066. 207 Berki , S .E. , Ashcraft , M . L . F . , & Newbrander , W . D . Length-of-Stay Variations W i t h i n I C D A - 8 Diagnosis-Related Groups . M e d i c a l Care. 1984; 22 (2): 126-142. Berry, N .J . P S R O Impact on Util ization: Forecast in a C l o u d y Crystal Ball. Hospitals . 1977; 51: 57-60. Boaz, R . F . Ut i l i zat ion Rev iew a n d Conta inment of Hosp i ta l Ut i l izat ion. M e d i c a l Care. 1979; 17 (4): 315-330. Bolley, H . Determination of Ac tua l Costs in Accordance wi th M I S Guidel ines: P a r t i . Dimensions in Heal th Service. 1987; 64 (8): 55-57. Boone, C . R . , C o u l t o n , C .J . , & Keller, S . M . The Impact of Ear ly and Comprehens ive Social W o r k Services o n Length of Stay. Social W o r k in  Heal th Care. 1981/82; 7: 1-9. Botz, C k . , & Singh, C . Calculat ing the Cost of C M G s in a C a n a d i a n Hospi ta l . Hea l th Care. 1985: February: 53-56. Brewster, A . C . , Jacobs, C M . , & Bradbury, R . C . Classifying Severity of Illness by U s i n g Cl in ica l Findings . Hea l th Care F inanc ing Rev iew ( A n n u a l  Supplement) . 1984; N o v e m b e r : 107-108. Brewster, A . C . , K a r l i n , B . G . , H y d e , L . A . , Jacobs, C M . , Bradbury, R . C , & Chae, Y . M . M E D I S G R P S : A Cl in ica l ly Based A p p r o a c h to Classi fying Hospi ta l Patients at A d m i s s i o n . Inquiry. 1985; 22: 377-387. Brian , E . W . G o v e r n m e n t C o n t r o l of Hosp i ta l Uti l izat ion: A Cal i fornia Experiment. The N e w E n g l a n d Tournal of Medic ine . 1972; 286 (25): 1340-1344. Brian , E . W . Foundat ion for M e d i c a l Care Contro l of Hosp i ta l Uti l izat ion: C H A P - a P S R O Prototype. T h e N e w Eng land Tournal of Medic ine . 1973; 288 (17): 878-882. British C o l u m b i a Hea l th Association. Ut i l i za t ion Management . A W o r k s h o p presented b y the British C o l u m b i a Heal th Associat ion. June 4,1987. British C o l u m b i a Min i s t ry of Heal th . Hosp i ta l Programs D i v i s i o n ,  Institutional Services: Operat ional Descript ion. Province of British C o l u m b i a : Min i s t ry of Health. October, 1985. 208 Brook, R. H . , & Wi l l i ams , K . N . Evaluat ion of the N e w M e x i c o Peer Review System 1971 to 1973. M e d i c a l Care (Supplement). 1976; 14 (12): 1-122. Brook, R . H . , Wi l l iams , K . N . , & R o l p h , J .E. Use , Costs, and Qual i ty of M e d i c a l Services: Impact of the N e w Mexico Peer Review System: A 1971-1975 Study. Anna l s of Internal Medic ine . 1978; 89: 256-263. B r o w n , D . E . , & L e v y , J .D. Uti l izat ion Review Al tered to Advantage . Hospitals . 1980; 54 (5): 83-86. B r o w n , L . D . Introduction to a Decade of Transit ion. Tournal of Heal th  Politics, Policy and L a w . 1986; 11 (4): 569-583. Bur ford , R. , & A v e r i l l , R .F . The Relationship between Diagnostic Information Ava i lab le at A d m i s s i o n and Discharge for Patients in O n e P S R O Setting: Implications for Concurrent Review. M e d i c a l Care . 1979; 17 (4): 369-381. C a m p b e l l , D . T . , & Stanley, J .C. Experimental and Quasi -experimental  Designs for Research. Chicago , Dlinois: R a n d M c N a l l y & C o m p a n y . 1969. C a n a d i a n Comprehens ive A u d i t i n g Foundat ion . C a n a d i a n Hospi ta l s  Accountabi l i ty and Information for Cost-Effectiveness: A n A g e n d a for  A c t i o n . Ottawa, Ontario: C a n a d i a n Comprehens ive A u d i t i n g Foundat ion . 1987. C a n a d i a n Hospi ta l Association. C a n a d i a n Hospi ta l Directory. Ottawa, Ontario . 1982; 1983; 1984; 1985. C a n a d i a n Hosp i ta l Association. Exp lor ing the Future of Hospitals in  Canada: A Definition Study. Ottawa, Ontario . 1984. C lemenhagen , C , C h a m p a g n e , F . , Contandr iopoulos , A . , & Pineault, R. Hospita l -Based Qual i ty Assurance: A Starting Point for Physicians and  Managers . Ottawa, Ontario: C a n a d i a n Hospi ta l Association. 1985. C lendenning , M . K . , Wol fe , H . , Shuman, L . J . , & H u b e r , G . A . T h e Effect of a Target Date Based Ut i l izat ion Review Program on Length of stay. M e d i c a l Care. 1976; 14 (9): 751-764. C o n k l i n , J . E . , L ieberman, J .V. , Barnes, C . A . , & Louis , D . Z . Disease Staging: Implications for Hosp i ta l Reimbursement and Management . H e a l t h  C a r e F inanc ing Review ( A n n u a l Supplement) . 1984; N o v e m b e r : 13-22. 209 C o n n o r , M . D . , M a c k , G . A . , & H a n d l e m a n , E . V . Dynamics of Ut i l izat ion  M a n a g e m e n t . Chicago: A m e r i c a n Hospi ta l Associat ion Publ ishing C o m p a n y . 1983. C o o m b s , R. & Richter, K . Alberta's Tria l F u n d i n g System Raises Physicians' Cost Consciousness. Dimensions in Hea l th Service. 1986; 63 (2): 21-23. Dans, P . E . , Weiner , J.P., & Otter, S.E. Peer Review Organizations: Promises a n d Potential Pitfalls. The N e w E n g l a n d Tournal of Medic ine . 1985; 313  (18): 1131-1137. D a v i d s o n , S . M . , Wacker , R . C , & K l e i n , D . H . Professional Standards Review Organizat ions - A Crit ique. Tournal of the A m e r i c a n M e d i c a l  Associat ion. 1973; 226: 1106- 1108. Davis , K . , A n d e r s o n , G . , & Steinberg, E . Diagnosis Related G r o u p Prospective Payment: Implications for Hea l th Care and M e d i c a l Technology. Heal th Policy. 1984; 4: ,139-147. DesHarnais , S., K o b r i n s k i , E . , Chesney, J. , L o n g , M . , A m e n t , R., & F leming , S. T h e Ear ly Effects of the Prospective Payment System on Inpatient Ut i l izat ion and the Qual i ty of Care. Inquiry. 1987; 24: 7-16. Dobson , A . , Greer, J . G . , Car l son , R . H . , Davis , F . A . , K u c k e n , L . E . , Steinhardt, B.J., Ferry , T . P . , & A d l e r , G.S. PSROs: Their Current Status and Their Impact to Date. Inquiry. 1978; 15: 113-128. D o d d s , J.J. Surveil lance Reduces Admiss ions . Hospitals . 1974; 48 (23): 52-55. Donabedian , A . A G u i d e to M e d i c a l Care Admini s tra t ion , V o l u m e II: M e d i c a l Care A p p r a i s a l - Qual i ty and Uti l izat ion. N e w York: A m e r i c a n Publ ic Hea l th Association. 1969. Doremus , H . D . D R G s may be Raising False Expectations. Hospitals . 1980; 54  (15): 47-51. D y c k , F.J . , M u r p h y , F . A . , M u r p h y , J .K. , Road , D . A . , B o y d , M . S . , Osborne, E . , de Vheger , D . , Korchinsk i , B. , Rip ley , C , Bromley, A . T . , & Innes, P.B. Effect of Surveil lance o n the N u m b e r of Hysterectomies in the Province of Saskatchewan. T h e N e w E n g l a n d Tournal of Medic ine . 1977: 296 (23): 1326-1328. 210 Eastaugh, S.R., & Eastaugh, J .A . Prospective Payment Systems: Steps to Enhance Qual i ty , Efficiency, and Regionalization. Hea l th Care  M a n a g e m e n t Review. 1986; 11 (4): 37-52. El iasoph, H . & Hassen, P. V I P Streamlines Care and Reduces Length of Stay. Dimens ions in Heal th Service. 1986; 63 (7): 42,44-45. El iasoph, H . , & Hassen, P. Hospi ta l Eyes Cataract Surgery. Hea l th  Management F o r u m . 1987; 8 (3): 45-53. Enthoven , A . C . Shatruck Lecture - Cut t ing Cost Without Cut t ing the Qual i ty of Care. The N e w E n g l a n d Tournal of Medic ine . 1978; 298 (22): 1229-1238. Enthoven , A . C . Reforming the U.S . Hea l th Care E c o n o m y T h r o u g h the  E c o n o m i c Compet i t ion of Alternative Heal th Care F inanc ing a n d  De l ivery Systems. Reading, M A . : A d d i s o n Wesley Publ i sh ing C o . 1983; 1-13. Evans , R . G . Hea l th Care in Canada: Patterns of F u n d i n g and Regulation. Tournal of Politics, Policy and L a w . 1983: 8 (1): 1-43. Evans , R . G . Strained Mercy: The Economics of C a n a d i a n Heal th Care . Toronto: Butterworth & C o . (Canada) L t d . 1984. Fetter, R .B . Diagnosis Related Groups : Th e Product of the Hospi ta l . C l i n i c a l  Research. 1985: 32 (3): 336-340. Fetter, R .B . , Shin, Y . , Freeman, J . L . , A v e r i l l , R . F . , & T h o m p s o n , J .D. Case M i x Definit ion b y Diagnosis-Related G r o u p s . M e d i c a l Care  (Supplement). 1980; 18 (2): 1-53. Flashner, B . A . , Reed, S., C o b u r n , R . W . , & Fine, P.R. Professional Standards Rev iew Organizations: Analys i s of Their Deve lopment and Implementat ion Based o n a Pre l iminary Rev iew of the Hosp i ta l A d m i s s i o n a n d Surveil lance P r o g r a m in Illinois. Tournal of the  A m e r i c a n M e d i c a l Association. 1973; 223 (13): 1473-1484. Flashner, B . A . , Reed, S., White , R. , & Norr i s , J. T h e Hosp i ta l A d m i s s i o n a n d Surveil lance P r o g r a m i n Illinois. Tournal of the A m e r i c a n M e d i c a l  Associat ion. 1972: 221 (10): 1153-1158. Foster, J.T. Ut i l izat ion Review is a Great Idea, But Does it Affect Util ization? M o d e r n Hospi ta l . 1968: February: 95-96,101. 211 Frederick, L . D . Peer Review N o w : P S R O at W o r k . M o d e r n Healthcare. 1975; 3_(1): 49-55. Fulchiero , A . , M i l l e r , S., Foley, C . R . , Ballantine, H . T . , Jr., & A m o r o s i n o , C.S . , Jr. C a n the PSRO's be Cost Effective? The N e w E n g l a n d Tournal of  Medic ine . 1978; 299: 574-580. Ger tman , P . M . , & Lowenstein, S. A Research Paradigm for Severity of Illness: Issues for the Diagnosis-Related G r o u p System. Hea l th Care  F inanc ing Rev iew ( A n n u a l Supplement) . 1984; N o v e m b e r : 79-90. G e r t m a n , P . M . , & Restuccia, J .D. T h e Appropriateness Evaluat ion Protocol: A Technique for Assessing Unnecessary Days of Hospi ta l Care. M e d i c a l  Care. 1981; 19 (8): 855-871. Gi l l ings , D . , M a k u c , D . , & Siegel, E . Analys i s of Interrupted T i m e Series Morta l i ty Trends: A n Example to Evaluate Regional ized Perinatal Care. A m e r i c a n Tournal of Publ ic Health. 1981; 71 (1): 38-46. Ginzberg , E . The Restructuring of U.S . Hea l th Care. Inquiry. 1985; 22: 272-281. G i n z b e r g , E . Sound ing Board: A H a r d L o o k at Cost Containment . The N e w  E n g l a n d Tournal of Medic ine . 1987; 316 (18): 1151-1154. G o l d b e r g , G . A . , & H o l l o w a y , D . C . E m p h a s i z i n g 'Level of Care' over 'Length of Stay' in Hospi ta l Uti l izat ion Review. M e d i c a l Care. 1975; 13 (6): 474-485. Goldfarb , M . G . , H o r n b r o o k , M . C . , & Higg ins , C S . Determinants of Hosp i ta l Use: A Cross-Diagnostic Analys is . M e d i c a l Care. 1983; 21 (1): 48-66. Go ld farb , M . , H o r n b r o o k , M . , & Rafferty, J. Behavior of the M u l t i p r o d u c t F i r m : A M o d e l of the Nonprof i t Hosp i ta l System. M e d i c a l Care. 1980; 18 12): 185-201. G o r a n , M . J . T h e Evo lut ion of the P S R O Hospi ta l Review System. M e d i c a l  Care (Supplement). 1979:17 (5): 1-46. G o r a n , M . J . , Roberts, J.S., Ke l logg , M . A . , F ie ld ing , J . , & Jessee, W . The P S R O Hosp i ta l Review System. Medica l Care. 1975; 131(4): 1-33. G r i m a l d i , P . L . , & Michelett i , J .A . Ut i l izat ion and Qual i ty Rev iew U n d e r the Prospective Rate System. Qua l i ty Review Bulletin. 1984; February: 30-37. 212 H a l p e r i n , W . , & Neuhauser , D . M E U : M e a s u r i n g Efficient Ut i l izat ion of Hosp i ta l Services. Hea l th Care Management Review. 1976; 1 (2): 63-70. Hea l th Insurance for the A g e d : Condit ions of Participation. U .S . Department of Hea l th , Educat ion and Welfare, Social Security Admin i s t ra t ion . Washington , D . C : U.S . Government Print ing Office. 1966. H i r s c h , E . O . , R i m m , A . A . , & Welsch, R . G . Evaluat ion of a N e w Strategy for C o n t i n u e d Stay Review in Hospitals . M e d i c a l Care. 1977; 15 (11): 906-914. H M R I : Th e Hospi ta l M e d i c a l Records Institute. Inpatient Program. 500 -1985 West Broadway, Vancouver , B . C . September, 1983. H M R I : T h e Hospi ta l M e d i c a l Records Institute. C o m p a r i s o n of Hosp i ta l  Activit ies Program Part 1 ( C H A P 1). 500 - 1985 West Broadway, Vancouver , B . C . June, 1984. H o d g s o n , M . , & O r m e r o d , C . Case M i x Management - Management T o o l of the Future? T h e C a n a d i a n Nurse . 1985; N o v e m b e r : 17-20. H o r n , S.D. , Bulkley, G . , Sharkey, P . D . , Chambers , A . F . , H o r n , R . A . , & S c h r a m m , C.J . Interhospital Differences in Severity of Illness: Problems for Prospective Payment Based o n Diagnosis-Related G r o u p s (DRGs) . T h e N e w E n g l a n d Tournal of Medic ine . 1985; 313 (1): 20-24. H o r n , S .D. , & H o r n , R . A . Severity Data aid in Hosp i ta l Management . Hosp i ta l F o r u m . 1984: 27 (4): 55-60. H o r n , S.D. , H o r n , R . A . , & Sharkey, P . D . The Severity of Illness Index as a Severity Adjustment to Diagnosis-Related G r o u p s . Hea l th Care  F i n a n c i n g R e v i e w ( A n n u a l Supplement) . 1984; N o v e m b e r : 33-45. H o r n , S .D. , H o r n , R . A . , Sharkey, P . D . , Beall, R.J. , Hoff , J.S., & Rosenstein, B.J. Misclassif ication Problems in Diagnosis-Related Groups : Cyst ic Fibrosis as and Example . T h e N e w E n g l a n d Tournal of Medic ine . 1986; 314 (8): 484-487. H o r n , S .D. , Sharkey, P . D . , Chambers , A . F . , & H o r n , R . A . Severity of Illness wi th in D R G s : Impact o n Prospective Payment. A m e r i c a n Tournal of  Publ ic Health. 1985; 75 (10): 1195-1199. H o r n b r o o k , M . C . Techniques for Assessing Hospi ta l Case M i x . A n n u a l  Reviews in Publ ic Heal th . 1985; 6: 295-324. 213 Hs iao , W . C . , & D u n n , D . L . The Impact of D R G Payment o n N e w Jersey Hospitals . Inquiry. 1987; 24 (3): 212-220. H u b e r , G . A . , Wolfe , H . , & H a r d w i c k , C P . Evaluat ing Computer i zed Screening as an A i d to Uti l izat ion Review. Inquiry. 1974; 11 (3): 188-195. H u g h e s , R . A . , G e r t m a n , P . M . , A n d e r s o n , J.J., Fr iedman, N . L . , Rosen, M . R . , W a r d , A C , & Kreger, B . E . T h e Anc i l l ary Services Rev iew Program in Massachusetts. Tournal of the A m e r i c a n M e d i c a l Association. 1984; 252  (13): 1727-1732. Iglehart, J.K. T h e N e w E r a of Prospective Payment for Hospitals . Th e N e w  E n g l a n d Tournal of Medic ine . 1982a: 307 (20): 1288-1292. Iglehart, J .K. N e w Jersey's Experiment with D R G - B a s e d Hospi ta l Reimbursement. T h e N e w E n g l a n d Tournal of Medic ine . 1982b; 307 (26): 1655-1660. Iglehart, J .K. Medicare Begins Prospective Payment of Hospitals . T h e N e w  E n g l a n d Tournal of Medic ine . 1983; 308 (23): 1428-1432. Iglehart, J.K: Canada's Heal th Care System (First of Three Parts). The N e w  E n g l a n d Tournal of Medic ine . 1986a; 315 (3): 202-208. Iglehart, J .K. Canada's Heal th Care System (Second of Three Parts). T h e  N e w E n g l a n d Tournal of Medic ine . 1986b: 315 (12): 778-784. Jencks, S.F., & Dobson , A . Refining C a s e - M i x Adjustment: Th e Research Evidence. T h e N e w E n g l a n d Tournal of Medic ine . 1987; 317 (11): 679-686. Jencks, S.F., Dobson , A . , Wi l l i s , P. , & Feinstein, P . H . Evaluat ing and I m p r o v i n g the Measurement of Hospi ta l Case M i x . Hea l th C a r e  F i n a n c i n g Rev iew ( A n n u a l Supplement) . 1984; N o v e m b e r : 1-11. Jenkins, L . , & Cole , J. Measure for M e a s u r i n g Hea l th Service Output . Hea l th a n d Social Service Tournal. 1984; 94: 884-885. Johnson, A . N . , & A p p e l , G . L . D R G s and Hospi ta l Case Records: Implications for Medicare Case M i x Accuracy. Inquiry. 1984; 21: 128-134. Kel ley , D C , W e n g , J. , & Watson, A . T h e Effect of Consultat ion o n Hospi ta l Length of Stay. Inquiry. 1979; 16: 158-161. 214 Ke l ly , J., & Bankhead, C D . D R G s : H o w A r e They Stacking U p ? M e d i c a l  W o r l d News . 1985: 26 (5): 80-82, 84, 87,91-92,94,100,102-103. K e m p e r , K . Inappropriate Hospi ta l Days for Pediatric Patients ( Abstracts). A m e r i c a n Tournal of Diseases of C h i l d r e n . 1987a: 141 (4): 393. K e m p e r , K . Pediatr ic -Modif ied Appropriateness Evaluat ion Protocol. N e w H a v e n , Connecticut: Department of Pediatrics, Yale Univers i ty School of Medic ine . 1987b. Kl ic ius , R., & Sawyer, E . Balancing Qual i ty and Cost: A Public Decision-M a k i n g Process. Dimensions in Hea l th Service. 1986; Tune: 11-13. K n a p p , D . A . , Speedie, M . K . , Yaeger, D . M . , & K n a p p , D . A . D r u g Prescribing a n d Its Relation to Length of Hospi ta l Stay. Inquiry. 1980; 17 (3): 254-259. K n a u s , W . A . , Draper , E . A . , Wagner , D .P . , & Z i m m e r m a n , J .E. A P A C H E II: A Severity of Disease Classification System. Crit ical Care Medic ine . 1985; 13 (10): 818-829. K n o w l t o n , H . C . A Retrospective and Concurrent M e d i c a l A u d i t . H o s p i t a l  Progress. 1975; 56 (6): 68,73,76,78. K o m e s , A . Peace A r c h / B C H A . Hea l th Care C o m p u t i n g & C o m m u n i c a t i o n  Canada . 1988; 2 (2): 26. L a m p r e y , J. , & Jacobs, C . The I S D - A Review System with A d u l t Criteria. Westborough , M A : InterQual, Incorporated. 1987. Lantos , R . L . Part One: H o w D o Hospi ta l Directors Plan to C o p e with the Constraints of D R G s ? Pharmacy Times. 1984; 50 (4): 66,68-69. Lave , J.R., & Leinhardt , S. A n Evaluat ion of a Hosp i ta l Stay Regulatory M e c h a n i s m . A m e r i c a n Tournal of Publ ic Heal th . 1976a: 66 (10): 959-967. Lave , J.R., & Leinhardt , S. The Cost and Length of a Hosp i ta l Stay. Inquiry. • 1976b; 13: 327-343. Lazare , E . H . F o r u m : Challenges of the '80s and '90s. Hea l th Management  F o r u m . 1987; 8 (4): 57-74. L e C l a i r , M . T h e C a n a d i a n Heal th Care System. In S. A n d r e o p o u l o s (Ed.), Nat iona l Hea l th Insurance: C a n W e Learn F r o m Canada? (pp. 11-93). N e w York: John Wi ley . 1975. 215 L i p p , C.S. T he Effect of the Prospective Payment System on Hospi ta l Q A / U R Systems. Qual i ty Review Bulletin. 1984; September: 283-287. L o w r y , B. M.I.S. Project - O v e r v i e w and Framework. Dimensions in Heal th  Service. 1983; 60 (3): 40-47. Luft , H . S . H o w D o Health-Maintenance Organizat ions Ach ieve Their 'Savings'?: Rhetoric and Evidence. T h e N e w E n g l a n d Tournal of  Medic ine . 1978; 298 (24): 1336-1343. L u k e , R . D . Dimensions in Hospi ta l Case M i x Measurement. Inquiry . 1979; 16: 38-49. Macies , J.S. Strategic Information: P lanning for the 1990's. Hea l th Care. 1984; February: 22-23. M a c K e n z i e , T . A . , M a r k l e , F . , & Croke , M . C M G s : Variations on a Theme. Hea l th Management F o r u m . 1987; 8 (1): 21-24. M a y , J.J. Th e Greatest Benefit is the Disc ip l ined Use of Resources. Hea l th  M a n a g e m e n t F o r u m . 1983; 4 (4): 75-77. M a y , J.J. Impact of Diagnosis Related G r o u p s on M e d i c a l Practice. T h e  A m e r i c a n Tournal of Cardio logy . 1985; 56: 16c-26c. M a y , J.J., & Wasserman, J. Selected Results from an Evaluat ion of the N e w Jersey Diagnosis-Related G r o u p System. Hea l th Services Research. 1984; 19 (5): 547-559. M c D o n o u g h , C , & V a z , A . P . Hospi ta l Ut i l izat ion Management . H e a l t h  M a n a g e m e n t F o r u m . 1987; 8 (1): 40-52. M c G e o r g e , K . , de M o r a , J .A . , & Legros, G . L . A Case M i x Management System. Hea l th Management F o r u m . 1985: 6 (2): 53-64. M c L u r e , W . T o w a r d Development a n d A p p l i c a t i o n of a Qual i tat ive Theory of Hosp i ta l Uti l izat ion. Inquiry. 1982; 19: 117-135. M c M a h o n , L . F . , Jr. Diagnosis-Related G r o u p Prospective Payment . Eva luat ion & T h e Heal th Professions. 1984; 7 (1): 25-41. M e n d e n h a l l , S. D R G s M u s t be C h a n g e d to take Patient's Illness Severity into Account . M o d e r n Healthcare. 1984:14 (15): 86,88. 216 Mercer , K . Responding to Fiscal Restraint: B . C . Closes Beds and Cuts Staff. Dimens ions in Hea l th Service. 1986; 63: 38-39. M I S Project Sets Framework of Standards. Heal th Care. 1986; A p r i l : 27. M u g h r a b i , M . A . T h e Effects of Selected Demographic a n d Cl in ica l Variables o n the Length of Hospi ta l Stay. Hosp i ta l Admin i s tra t ion in C a n a d a . 1976; 18: 82-84,86,88. M u l l i n , R . L . Ut i l izat ion Review Based on Practitioner Profiles. Tournal of  M e d i c a l Systems. 1983; 7 (5): 409-412. M u l l i n , R. L . Diagnosis-Related G r o u p s and Severity: I C D - 9 - C M , the Real Problem. Tournal of the A m e r i c a n M e d i c a l Association. 1985; 254 (9): 1208-1210. Nathanson , M . 'Double D i p p i n g ' can H i k e Profits, but Advantages m a y be Short-l ived. M o d e r n Healthcare. 1984a; 14 (3): 154,156. Nathanson , M . Report ing, Care Di lemmas L u r k in 'Gray' Changes Spurred by D R G s . M o d e r n Healthcare. 1984b; 14 (4): 104,108. Nathanson , M . M o r e Hospitals T u r n to SOIs, but Experts Quest ion their Usefulness. M o d e r n Healthcare. 1985; 15 (4): 63-64,69-75. Nat iona l Hea l th Expenditures in C a n a d a 1975-1985. Ottawa, Ontario: Minis ter of S u p p l y and Services C a n a d a . N o v e m b e r , 1987. Neter, J . , & Wasserman, W . A p p l i e d Linear Statistical M o d e l s . H o m e w o o d , Illinois: R i chard D . Irwin, Inc. 1974. N u s b a u m , M . H . Cross ing the Thresho ld to Strategic Management . H e a l t h  C a r e C o m p u t i n g & C o m m u n i c a t i o n C a n a d a . 1988; 2 (3): 24, 26. O T A , Office of Technology Assessment. O T A Examines Variations in Hospi ta l Length of Stay. O T A N e w s A d v i s o r y . 1983; August : 1-4. O m e n n , G.S . , & C o n r a d , D . A . S o u n d i n g Board: Implications of D R G s for Cl inic ians . T h e N e w E n g l a n d Tournal of Medic ine . 1984; 311 (20): 1314-1317. O w e n s , T . R . , & A v e r i l l , R . F . T h e Role of Ut i l izat ion Management U n d e r PPS. Healthcare Financial Management . 1984; October: 60-62, 64, 66. 217 Payne, S . M . C . Identifying and M a n a g i n g Inappropriate Hosp i ta l Uti l ization: A Pol icy Synthesis. Heal th Services Research. 1987: 22 (5): 709-769. Pettingill , B .F . , & Westel l , H . C . Unanswered Questions Facing D R G ' s . Hospi ta l Topics. 1984: 62 (4): 3-4. Pointer, D . D . , & Ross, M . B . D R G s K i c k O f f a W h o l e N e w Game; Hospitals N e e d N e w Tactics to W i n . M o d e r n Healthcare. 1984:14 (3): 109-110,112. - Pokras, R., & Kubishke , K . K . Diagnosis-Related G r o u p s U s i n g Data F r o m the Nat iona l Hosp i ta l Discharge Survey: U n i t e d States, 1982. Advancedata . 1985; 105: 1-8. Q u e r i d o , A . T h e Efficiency of M e d i c a l Care: A Crit ical Discussion of  M e a s u r i n g Procedures. Leiden: H . E . Stenfert Kroese N . V . 1963. Rachlis , M . & Fooks, C . Ut i l izat ion Analys is : Current Initiatives Across  C a n a d a . Paper prepared for the First A n n u a l Heal th Policy Conference conducted by the Centre for Hea l th Economics and Pol icy Analys i s , H a m i l t o n , Ontario . M a y 27,1988. Restuccia, J .D. The Effect of Concurrent Feedback in Reduc ing Inappropriate Hosp i ta l Uti l izat ion. M e d i c a l Care. 1982: 20 (1): 46-62. Restuccia, J .D. , & H o l l o w a y , D . C . Barriers to Appropr ia te Ut i l izat ion of an Acute Facility. M e d i c a l Care. 1976:14 (7): 559-573. R i c h , M . W . , & Freedland, K . E . Effect of D R G s on T h r e e - M o n t h Readmiss ion Rate of Geriatric Patients wi th Congest ive Heart Fai lure. A m e r i c a n Tournal of Publ ic Heal th . 1988: 78 (6): 680-682. Ro , K . Patient Characteristics, Hospi ta l Characteristics, and Hospi ta l Use. M e d i c a l Care. 1969: 7 (4): 295-312. Roberts, R. F . , Re id , A . , & Irwin, A . L . Case-mix G r o u p i n g and D R G s : M a k i n g the Princ ipal Diagnosis. The M e d i c a l Tournal of Austral ia . 1985; 143: 243-245. Roos, L . L . , Jr. Alternative Designs to Study Outcomes: T h e Tonsi l lectomy Case. M e d i c a l Care. 1979; 17: 1069-1087. Roos, L . L . , Jr. Supply , W o r k l o a d a n d Uti l izat ion: A Population-Based Ana lys i s of Surgery in Rura l Mani toba . A m e r i c a n Tournal of Publ ic  Heal th . 1983; 73 (4): 414-421. 218 Roos, L . L . , Jr., & Roos, N . P . Ut i l izat ion Analys is : W h a t H a v e W e Learned?  W h e r e Shou ld W e Go? Paper prepared for the First A n n u a l Heal th Pol icy Conference conducted by the Centre for Heal th Economics and Pol icy Analys i s , H a m i l t o n , Ontario. M a y 27,1988. Roos, N . P . , Henteleff, P . D . , & Roos, L . L . , Jr. A N e w A u d i t Procedure A p p l i e d to an O l d Question: Is the Frequency of T & A Justified? M e d i c a l Care. 1977; 15: 1-18. Rosser, R . M . T h e Reliabil ity and Appl i ca t ion of Cl in ica l Judgement in Evaluat ing the Use of Hospi ta l Beds. M e d i c a l Care . 1976:14 (1): 39-47. Rothberg, D . L . , & Gertman, P . M . The U s e of Att i tudinal Data for Public Policy: The Case of Unnecessary Hospi ta l Use. M e d i c a l Care. 1981; 19: 47-54. Sanazaro, P.J. The P S R O Program: Start of a N e w Chapter? . T h e N e w  E n g l a n d Tournal of Medic ine . 1977; 296 (16): 936-938. Sanazaro, P.J., Goldste in , R . L . , Roberts, J.S., Maglott , D . B . , & McAl l i s t er , J .W. Research a n d Deve lopment in Qua l i ty Assurance: T h e Experimental M e d i c a l C a r e Review Organizat ion Program. T h e N e w E n g l a n d Tournal  of Medic ine . 1972; 287 (22): 1125-1131. Sayetta, R.B. Cri t ique of an Earlier Study of the Sacramento M e d i c a l Care Foundation's Certif ied Hospi ta l A d m i s s i o n Program ( C H A P ) . M e d i c a l  Care. 1976:14 (1): 80-90. Schmitz , H . H . , & Schoenhard, W . C . A u t o m a t e d Ut i l i zat ion Rev iew is T i m e l y , Accurate, Efficient. Hospi ta l Progress. 1976; 57 (7): 73-77,98. Schultz, K . Hospitals of the Future - Heal th-Care/Sickness-Care? H e a l t h  Care. 1986; October: 14-15. Shain, M . , & Roemer, M.I . Hosp i ta l Costs Relate to the S u p p l y of Beds. T h e  M o d e r n Hospita l . 1959; 92 (4): 71-73,168. Shortell , S . M . , & Hughes , E . F . X . T h e Effects of Regulation, Compet i t ion , and O w n e r s h i p o n Morta l i ty Rates A m o n g Hospi ta l Inpatients. The N e w E n g l a n d Tournal of Medic ine . 1988: 318 (17): 1100-1107. Sieverts, S. T h e Uses of Uti l izat ion Review. T h e N e w E n g l a n d Tournal of  Medic ine . 1978:299 (11): 601-602. 219 S imborg , D . W . D R G Creep: A N e w Hosp i ta l -Acquired Disease. Th e N e w E n g l a n d Tournal of Medic ine . 1981:304 (26): 1602-1604. S iu , A . L . , Sonnenberg, F . A . , M a n n i n g , W . G . , G o l d b e r g , G . A . , Bloomfield, E.S . , Newhouse , J.P., & Brook, R . H . Inappropriate Use of Hospitals in a R a n d o m i z e d T r i a l of Hea l th Insurance Plans. The N e w E n g l a n d Tournal  of Medic ine . 1986: 315 (20): 1259-66. Smits, H . L . , Fetter, R .B . , & M c M a h o n , L . F . , Jr. Variat ion in Resource Use W i t h i n Diagnosis-Related Groups : T h e Severity Issue. H e a l t h Care  F inanc ing Rev iew ( A n n u a l Supplement) . 1984; N o v e m b e r : 71-78. Snider, M . Ut i l izat ion Review for Medicare . A m e r i c a n Tournal of N u r s i n g . 1977; 77 (1): 107-109. Soderstrom, L . T h e C a n a d i a n Heal th System. L o n d o n : C r o o m H e l m . 1978. Spiegel, A . D . , & Kavaler , F. The Debate over Diagnosis Related Groups . Tournal of C o m m u n i t y Heal th . 1985:10 (2): 81-92. Stern, R.S. , & Epste in , A . M . Institutional Responses to Prospective Payment Based o n Diagnosis-Related Groups : Implications for Cost , Qual i ty , and Access. The N e w E n g l a n d Tournal of Medic ine . 1985: 312 (10): 621-627. Stoughton, V . D R G s w i l l not Replace G l o b a l Budgeting. H e a l t h  M a n a g e m e n t F o r u m . 1983: 4 (4): 78-79. Studnicki , J . , & H o n e m a n n , D . Payment Source as a Variable in Ut i l izat ion Research: A Selected Literature Review. Qua l i ty Review Bullet in. 1983; September: 258-266. Suttie, B. T he Future of Ut i l izat ion Analys is . Paper prepared for the First A n n u a l Hea l th Pol icy Conference conducted by the Centre for Heal th Economics a n d Pol icy Analys i s , H a m i l t o n , Ontario. M a y 27, 1988. T h u r o w , L . C . M e d i c i n e Versus Economics. The N e w E n g l a n d Tournal of  Medic ine . 1985:313 (10): 611-614. T o m , B . C . K . , G r e s h a m , S . C , Goto , U . , Henderson , W . J . , J u d d , C .S . , & M c D o n a l d , B . A . Pay the Peers: Ut i l izat ion Rev iew W i t h a T e a m A p p r o a c h and Payment of Physic ian Reviewers. Tournal of the  A m e r i c a n M e d i c a l Association. 1976: 235 (7): 738-741. Veney , J .D. , & K a l u z n y , A . D . Evaluat ion and Decis ion M a k i n g For Health  Services Programs. Eng lewood Cliffs, N.J . : Prentice-Hall , Inc. 1984. 220 Walter , R.S. The Hottest N e w T h i n g in Uti l izat ion Review. M e d i c a l  Economics . 1983; 60: 177,180-181. W e n n b e r g , J . , & Gitt lesohn, A . Variations in M e d i c a l Care among Smal l Areas. Scientific A m e r i c a n . 1982; 246: 120-134. Wennberg , J . E . , McPherson , K . , & Caper , P. W i l l Payment Based on Diagnosis-Related G r o u p s Contro l Hosp i ta l Costs? Th e N e w E n g l a n d  Tournal of Medic ine . 1984; 311 (5): 295-300. Wi l l i ams , S .V. , K o m i n s k i , G . F . , D o w d , B . E . , & Soper, K . A . Methodolog ica l Limitat ions in Case M i x Hospi ta l Reimbursement , W i t h a Proposal for Change. Inquiry. 1984; 21: 17-31. W i l s o n , C . N . Financial Management Issues: Peer Review Organizat ion (PRO) to Moni tor D R G s . Hospi ta l Pharmacy. 1984; 19: 306, 312. Winche l l , J.R. Reduced Costs can Improve Care. Hea l th Care . 1985; October: 15-16,18. W y l i e , C M . , & Flashner, B . A . Reassessing the Ear ly Effect of Concurrent Review on Length of Stay: Illinois, 1971-72. M e d i c a l Care . 1979; 17 (6): 607-617. Wyszewiansk i ,L . , Thomas , J .W. , & F r i e d m a n , B . A . Case-Based Payment and the Contro l of Qual i ty and Efficiency in Hospitals. Inquiry. 1987; 24: 17-25. Y o u n g , W . Incorporating Severity of Illness and C o m o r b i d i t y in C a s e - M i x Measurement . Hea l th C a r e F inanc ing Review ( A n n u a l Supplement) . 1984: N o v e m b e r : 23-31. Y o u n g , W . W . , Swinkola , R .B . , & H u t t o n , M . A . Assessment of the A U T O G R P Patient Classification System. M e d i c a l Care . 1980; 18 (2): 228-244. Z i m m e r , J . G . Length of Stay and Hospi ta l Bed Misuti l izat ion. M e d i c a l Care. 1974; 12: 453-462. Z u c k e r m a n , A . M . Diagnosis Related Groups : Appl icat ions for C a n a d a . Hea l th Management F o r u m . 1983; 4 (4): 66-74. 221 A P P E N D I X A : Medicare Levels of Care Criteria for M e d i c a l / S u r g i c a l  Patients Medica l -Surgica l F l o w D i a g r a m I. Avai lab i l i ty of Ski l led N u r s i n g Services at A l l Times Definition: A skil led service is one w h i c h cannot safely and adequately be performed by the average, rational, nonmedica l person without direct supervis ion of trained medical or paramedica l personnel ( including observation a n d instruction). I . A . I . B . Observation. Does the unstabilized condition of the patient require the skills of a nurse to detect and evaluate (i.e., observe) the patient's need for possible modification of treatment or institution of medical procedures? N O Direct Service. Does the patient require direct skilled nursing services (excluding observation) every day? N O Y E S Patient's Condi t ion. In v iew of the patient's condition, are the range (number of different skilled services) and intensity (frequency of duration) of al l ski l led services (e.g., extensive diagnostic tests) furnished such that they cannot be performed outside an institution. Go to II.A I.D. Is the patient terminal? N O patient's condition unstable? N O Y E S Acute Hospi ta l N o t Acute Hospi ta l or S N F (Appendix A cont...) 222 A P P E N D I X A (cont). LT. Acute Hosp i ta l M e d i c a l Services Il.a Having established that the patient requires the availability of skilled nursing services at all times (or requires broad and intense skilled services in an institution), does the patient today require the constant availability of medical services provided by a Skilled Nursing Facility (SNF)? Skilled Nursing Facility Acute Hospital Source: Qted in Restuccia and Holloway (1976). 223 A P P E N D I X B. Pediatr ic -Modif ied Appropriateness Evaluat ion Protocol  M e d i c a l Services 1. Procedure in OR 2. Pre-op eval. or post-op evaluation (1 day) 3. Cardiac catheterization 4. Angiography 5. Biopsy of internal organ, N O T bone marrow 6. Thoracentesis/paracentesis/cysternal/ventricular tap 7. Test requiring (a or b) a. Strict dietary control b. collection of timed sample; > 8 hours 8. N e w or experimental treatment requiring frequent dose adjustments under direct  M P supervision 9. Documented M D exam >= 3 times within 24 hours N u r s i n g / L i f e Support Services 10. Vent i la tor / IPPB/Ches t PT >= TID 11. Mis t Tent 12. IV Therapy >= 8/24 hours 13. >= I M / s q injections/24 hours, except insulin 14 Strict I's&O's, ordered & done 15. Major surgical wound & Drainage care (chest tubes, hemovacs, Penrose, traction) 16. C R monitor & 24 hours after monitor D C ' d 17 >= 3 Vi ta l signs ordered and done. Patient Condi t ions 18. A n y two conditions: a. RT, PT, OT, or ST b. > 5 yo, non-ambulatory, and requires help with transfers c. defined patient education program, documented teaching d. Rehabilitation, Psychological, or Social services documented Within 24 hours before day reviewed: 19. Unable to void, not due to neurologic problem 20. Transfusion: acute bleed 21. M D ' s suspect suicide attempt 22. M D ' s suspect child abuse/neglect Within 48 hours before day reviewed: 23. Temperature >= 101 if patient admitted for other reason 24. Coma 25. Acute confusional state 26. Acute hematologic problem, eg, ANC<500, H C T <20,000 27. Progressive, acute neurological difficulties Source: Kemper (1987b) 224 A P P E N D I X C . List of Potential Barriers to Appropr ia te Uti l izat ion Barriers: Phys ic ian Responsibi l i ty 1. Patient is admitted for problem outside the area of competency of attending physician (e.g., should have referred patient to physician in different specialty). 2. Some or all of work-up could have been done on an outpatient rather than inpatient basis; e.g., preadmission work-up prior to surgery is inadequate. (Preoperative day, however, is to be regarded as appropriate unless clearly inappropriate). 3. Physician admits patient to hold bed (for possible future patient). 4. "Political admission," as with VIP (Please comment). 5. Physician admits patient for acute condition, then decides to work-up patient for chronic condition, causing delay. 6. Physician delays scheduling of test or procedure. 7. Unavailability of physician causes delay in performing a procedure, (e.g., surgery which could have been performed on a Wednesday is delayed until Thursday because the physician is unavailable on Wednesday). 8. Test sequencing is inadequate, i.e., ordering test in the wrong sequence causes delay i n diagnosis and/or treatment. (Please comment). 9. Physician requests consultation—delay in carrying out consultation or in receiving report of consultation. 10. Failure of attending physician to provide patient (or family) with necessary information for informed consent causes delay. 11. Failure of other physician (e.g., anesthesiologist) to provide patient (or family) with necessary information for informed consent causes delay. 12. Physician is not visiting the patient. 13. UR review physician is not forceful enough in supporting UR function. 14. Physician has legal problem with patient, i.e., malpractice threat or involvement with an attorney. 15. Physician's medical management of patient is conservative. 16. Physician delays decision regarding further treatment of the patient because of the complex medical nature of the case, yet does not request a consultation. 17. Patient is kept in hospital because it is easier for his/her physician to have "all" patients in one facility. (Please comment). 18. Physician refuses alternative facility because "its too far away," for him/her to visit patient there. 19. Physician delays transfer or discharge of patient because, "I want to watch him/her for a few more days myself." 20. Physician has no confidence in quality of services available at SNF. 21. 'Interesting case" is kept in hospital for teaching purposes. 22. Patient's transfer is delayed because of late date at which physician writes order. 23. Physician does not want to use SNF on prolonged case because "Then Blue Cross will expect this treatment on all cases like this," (e.g., orthopedic problem which could be treated in several different ways). 24. Other. (Please comment). (Appendix C cont...) 225 A P P E N D I X C . (cont.) Barriers: Hosp i ta l Responsibil i ty 25. Problem in hospital scheduling for test or procedure causes delay. 26. Failure of nursing service to obtain consent for a procedure causes delay. 27. Return of test results is delayed, causing delay in diagnosis or treatment. (Please specify type of test). 28. Scheduling or transfer of patient to another institution to perform a special procedure causes delay. 29. Communication failure, e.g., missed doctor's order. (Please comment). 30. Hospital has legal problem with patient, i.e., malpractice threat or involvement with an attorney. 31. Lack of administrative support contributes to inappropriate utilization, (e.g., inadequate clerical support, excessively heavy workload for coordinator). (Please comment). 32. Patient's transfer or discharge is delayed because of inadequate discharge planning on the part of hospital personnel. 33. Other. (Please comment). Barriers: Patient or Fami ly 34. Patient or family insists on admission to hospital. 35. Patient is admitted because he/she is uncooperative with therapeutic or diagnostic program outside the hospital. 36. Indecisiveness of patient (or family) regarding a procedure (despite provision of adequate information by physician and hospital personnel) causes delay. 37. Patient is uncooperative with therapeutic program in the hospital, causing delay. 38. Patient or family refuses alternative facility because it is too far away. 39. Patient or family insists on patient remaining in hospital. 40. Family member (or friend) is unwilling to care for patient after discharge. 41. Other (Please comment). Barriers: E n v i r o n m e n t a l Responsibi l i ty 42. Insurance coverage for diagnostic procedures is more complete on an inpatient than on an outpatient basis. 43. Insurance coverage for therapeutic procedures is more complete on an inpatient than on an outpatient basis. 44. Medicare requires three-day acute hospitalization prior to SNF coverage. 45. Patient must be admitted to remove from environment adverse to health, e.g., unavailability of family or friends to provide care. (Please comment). 46. Terminal patient might die in transit to alternative facility. 47. Patient from unhealthy environment (e.g., home environment) is kept until environment becomes acceptable or alternative facility is found. 48. Patient is convalescing from an illness and it is anticipated that his/her stay in an alternative facility would be less than 72 hours. (Appendix C cont...) 226 A P P E N D I X C . (cont.) E n v i r o n m e n t a l Barriers (cont.) 49. Patient is terminal and stable, yet is kept in the acute hospital for humanitarian reasons. (Please comment). 50. Patient is kept in hospital for short period of time (at the end of hospitalization) because patient or family needs to be taught self care, (as with ileostomy). 51. Patient's insurance coverage is more complete in an acute hospital than in an alternative facility or home health care program. 52. Unavailability of SNF bed. 53. Unavailability of sub-SNF level of care facility (e.g., nursing home, boarding house, etc.). 54. Unavailability of SNF with ability to provide the necessary type of skilled nursing services. (Please specify type of service). 55. Patient is Medicare-MediCal "crossover". No alternative facility available or willing to incur financial risk with patient having both Medicare and MediCal coverage because of MediCal requirement that Medicare be billed first with confirmation of no coverage before assuming coverage for patient itself. 56. Unavailability of SNF with ability to provide the necessary amount of (all types of) nursing services. (Please specify amount of service). 57. Unavailability of SNF with necessary ancillary service(s). (Please specify type of service). 58. Awaiting financial clearance from alternative facility for transfer. 59. Awaiting medical clearance from alternative facility for transfer. 60. Awaiting financial clearance from insurance program for transfer to alternative facility. 61. Family member (or friend) is not available to transport patient from the hospital. 62. Family member (or friend) is not available to care for patient after discharge. 63. Other. (Please comment). Source: Restuccia and Holloway (1976). 227 A P P E N D I X D. Data Elements Included on the H M R I Hospital Discharge  Abstract 1 ITEM NO. COL. COMMENTS SEGMENT 1 -discharge date: day 02 month 02 year 02 institution 04 batch: period 02 number 02 abstract 02 coder 02 chart 02-02-07 register number 05 los 04 SEGMENT 2 province 01 institution 04 batch: period 02 number 02 abstract 02 coder 02 chart 11 register number 05 los 04 health care number 08 postal code 05 residence code 04 sex 01 male, female, undetermined field name not specified 09 field name not specified 01 (Appendix D cont...) 1 An original HMRI Discharge Abstract is not provided because the pale blue colour of the form prohibits mechanical reproduction. 228 A P P E N D I X D (cont). age 01 > 99, unknown birth date: day 02 month 02 year 02 admission: day 02 month 02 year 02 hour 02 discharge: day 02 month 02 year 02 hour 02 institution from entry codes stillbirth institution to date ready for discharge: day month admission category admitted by ambulance 04 01 04 02 02 01 01 direct, emergency, newborn., elective, urgent, emergency main patient service subservice 02 01 exit alive codes death codes: coroner/autopsy < 48 hr after admission in or suite < 48 hr. post op. > 48 hr. post op. trauma pregnancy other complex case abstract incomplete 01 01 01 01 01 01 01 01 01 01 discharged, signed out coroner, autopsy (Appendix D cont...) 229 APPENDIX D (cont). SEGMENT 3 -service transfers: service A: patient service 02 sub service 01 days 02 service B: patient service 02 sub service 01 days 02 SEGMENT 4 most responsibe doctor 05 other doctors #1: type 01 other doctor 03 service 02 other doctors #2 - #7: coded as per 'other doctor #1' SEGMENT 5-most responsible diagnosis 10 other diagnoses (n = 15) 10 per SEGMENT 6-principal prodecure: day 02 month 02 procedure 04 doctor 03 service 02 tissue 01 time 03 OR 01 anaesthesia 03 tech. 01 other procedures: (n=ll) coded as per 'principle procedure' (Appendix D cont...) 2 3 0 A P P E N D I X D (cont). basic options: a - h 01 per i - l 01 per m-n 02 per o 04 SEGMENT 7-therapy 01 other blood given 01 pre-admission workup 01 discharge planning 01 social services 01 infections 01 special care units: unit 1 -unit no. 02 days 02 unit 2-3 coded as per 'unit T weight 04 SEGMENT 8-project codes: project #1-project code 03 1-13 01 per 14 - 15 02 per 16 03 17 04 project #2 - #5 coded as per 'project #1' physio., occup., speech, inhalation, 1,2 prof, social work, other social work pre-adm., non-op., post-op. Source: HMRI (1983) 23 A P P E N D I X E . Release of Data F o r m R E L E A S E O F D A T A F O R M This is to confirm that M s . Sandra Wigg ins has sought and obtained p e r m i s s i o n to i n c l u d e the H o s p i t a l i n the study entitled "Utilization Management of Acute Care Services: A n Eva luat ion of the S W I T C H Index System". Consequently , the Hosp i ta l M e d i c a l Records Institute ( H M R I ) is hereby authorized to release to M s . W i g g i n s the fo l lowing information: F o r each patient d i s c h a r g e d f r o m the Hospi ta l between January 1,1982 and December 31,1985, all available data corresponding to the fol lowing list of data elements (and only the specified elements): Case M i x G r o u p i n g Length of Stay Sex A g e Postal C o d e A d m i s s i o n Category Exit A l i v e C o d e Death C o d e A d m i s s i o n - Institution F r o m Discharge - Institution T o M o s t Responsible Doctor C o d e Special Care Uni t s M a i n Patient Service Sub Service A d m i s s i o n Date Discharge Date It is understood that this data wi l l be used for the sole purpose of conduct ing the study referred to above. Signature: Title: Date: H M R I Institution N u m b e r : 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0097839/manifest

Comment

Related Items