UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Recent trends in the nursing labour market in Canada Vujicic, Marko 2003

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

Item Metadata

Download

Media
831-ubc_2003-854159.pdf [ 8.99MB ]
Metadata
JSON: 831-1.0091303.json
JSON-LD: 831-1.0091303-ld.json
RDF/XML (Pretty): 831-1.0091303-rdf.xml
RDF/JSON: 831-1.0091303-rdf.json
Turtle: 831-1.0091303-turtle.txt
N-Triples: 831-1.0091303-rdf-ntriples.txt
Original Record: 831-1.0091303-source.json
Full Text
831-1.0091303-fulltext.txt
Citation
831-1.0091303.ris

Full Text

RECENT TRENDS IN T H E NURSING LABOUR M A R K E T IN C A N A D A by Marko Vujicic B.Comm., McGill University, 1997 M.A., McGill University, 1998 A thesis submitted in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Economics) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 2003 © Marko Vujicic, 2003 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 The University of British Columbia Vancouver, Canada DE-6 (2/88) A B S T R A C T There is alleged to be a severe nursing shortage in Canada. While the shortage is attributed in large part to fiscal restraint in the hospital sector in the early 1990s, there are competing claims addressing why nursing employment levels changed over this period. Supply-side proponents argue that deteriorating working conditions and stagnant wages led nurses to voluntarily leave the profession, province, or country for better employment prospects. Demand-side proponents argue that hospitals reduced staff levels in response to a decline in inpatient utilization. There is also considerable disagreement on what impact, if any, reduced nursing employment levels had on access to hospital care. However, while there is no shortage of anecdotal evidence and plausible rhetoric, the debate is being carried out in a largely data-free environment. This thesis attempts partially to fill this void. Part I of this thesis examines trends in the nursing employment level in Canada over the hospital restructuring period. Results indicate that the number of nurses employed in hospitals decreased significantly during the cut-backs period and that the decrease was particularly severe among young nurses. The employment level is decomposed into three separate components for each age group: the change in the potential supply of nurses, the change in the employment rate of this group and the change in the likelihood that an individual will work in the nursing profession conditional on being employed. Results indicate that the third factor is most important. To determine whether the observed shift toward non-nursing employment was voluntary, an occupational sector choice model is developed and the pattern of nursing wages, non-nursing wages, and hospital expenditure (a proxy for demand) is examined. The evidence strongly suggests that the reduction in the nursing employment level in hospitals during the downsizing period was a result of a decrease in the demand for nursing labour and did not represent voluntary movement out of the nursing sector. That the decrease in demand primarily affected young nurses appears to reflect the influence of seniority in the highly unionized nursing sector. 11 Part II examines the relationship between the availability of nursing resources in hospitals and access to surgical care in British Columbia. The surgical care focus is motivated by persistent claims that important surgeries were cancelled or unreasonably delayed due to a shortage of nurses. After modelling the pathways through which the availability of nursing resources might affect surgery rates, data from the BC Linked Health Database is used to examine rates for twelve common surgical procedures. Results indicate that a combination of shorter lengths of stay and a shift to day surgery resulted in stable or increasing surgery rates for several procedures during the downsizing era. Furthermore, in most cases where the surgery rate declined there are more plausible reasons for the decline than a reduction in surgical capacity resulting from fewer nurses in hospitals. It appears the cut-backs had little effect on access to beneficial surgical care. in TABLE OF CONTENTS Abstract ii Table of Contents iv List of Tables v List of Figures vi Acknowledgements vii CHAPTER 1 Introduction and Motivation 1 P A R T I 11 CHAPTER 2 Framework and Existing Evidence 12 2.1 Definition of Terms and a Framework for Part I 12 2.2 Existing Evidence Related to Part I 18 CHAPTER 3 Trends in Nursing Employment Levels in Canada 31 3.1 Description of the Data 31 3.2 Trends in R N Employment Levels in Canada 31 3.3 Factors Explaining the Pattern of Nursing 38 Employment Levels in Canada 3.4 Discussion 90 PART II 91 CHAPTER 4 Nursing Resources Available to Patients 93 4.1 The Nursing Employment Level and Hours of 93 Nursing Care 4.2 The Changing Pattern of Acute Care in 99 Canadian Hospitals 4.3 Trends in Acute Care and Patient Health 101 4.4 Trends in Acute Care and Total Nursing 111 Requirements in Hospitals 4.5 Trends in Acute Care and Average Nursing 113 Requirements in Hospitals CHAPTER 5 The Availability of Nursing Resources and the Pattern 119 of Surgical Care in British Columbia 5.1 Defining Terms 120 5.2 The Aggregate Surgery Rate 120 5.3 Surgery Rates for Specific Procedures 128 5.4 Discussion 143 CHAPTER 6 Conclusion 146 References 150 Appendix 158 iv LIST OF TABLES Table 1. R N Employment Level in Canada Table 2. R N Employment Level in Canada by Province Table 3. Age Breakdown of R N Workforce by Province Table 4. Health Care and Non-Health Care Wages of Individuals With an Education in Nursing Table 5. Breakdown of Non-Health Care Occupations for Individuals Under 35 Years of Age With an Education in Nursing Table 6. Variable Definitions and Means, 1991 Sample Table 7. Variable Definitions and Means, 1996 Sample Table 8. OLS Wage Equation Estimates for Health Care Sector Table 9. OLS Wage Equation Estimates for Non-Health Care Sector Table 10. Wage Equation Estimates for Health Care Sector Using Two-Stage Heckman Procedure Table 11. Wage Equation Estimates for Non-Health Care Sector Using Two-Stage Heckman Procedure Table 12. Analysis of Census PUMF Major Field of Education Categories 20 33 37 67 68 74 75 77 78 83 84 161 V LIST O F FIGURES Figure 1. Health Expenditure in Canada 2 Figure 2. Graphical Representation of the Nursing Labour Market 4 Figure 3. Relationship Between the Nursing Labour Market and 7 Patient Care Levels Figure 4. The Nursing Education and Employment Sectors 13 Figure 5. Venn Diagram Distinguishing Potential Supply of 15 RNs from Employment Level of RNs in British Columbia Figure 6. The Nursing Labour Market and the Hospital 17 Production Process Figure 7. R N Employment in Canada 34 Figure 8. R N Employment in Canada by Province and Place of Work 35 Figure 9. R N Workforce in Canada by Age 39 Figure 10. Change in Size of Potential Supply, Labour Force, 41 Employment Rate and Health Care Labour Force Participation Rate, 1991 - 1996 in B C Figure 11. Change in Size of Potential Supply, Labour Force, 42 Employment Rate and Health Care Labour Force Participation Rate, 1991 - 1996 in Alberta Figure 12. Change in Size of Potential Supply, Labour Force, 43 Employment Rate and Health Care Labour Force Participation Rate, 1991 - 1996 in Ontario Figure 13. Change in Size of Potential Supply, Labour Force, 44 Employment Rate and Health Care Labour Force Participation Rate, 1991 - 1996 in Quebec Figure 14. Decomposition of Change in Nursing Labour 46 Force, 1991 - 1996 Figure 15. Age Structure of Potential Supply of Nurses 49 and the Female Population VI Figure 16. Health Expenditure by Province 59 Figure 17. Hospital Expenditure Trends 60 Figure 18. Regression Adjusted Health Care Wage Premium by Age 79 Figure 19. Nursing Wages in Alberta and British Columbia 87 Figure 20. Full-Time Employment Among RNs 96 Figure 21. Production Functions for Hospital Care and Health 103 Figure 22. Paid Nursing Hours per Acute Care Separation 115 by Province Figure 23. Paid Nursing Hours per Day Surgery and Outpatient 116 Visit by Province Figure 24. Paid Nursing Hours per Inpatient Day 117 by Province Figure 25. Aggregate Surgery Rates in British Columbia 122 Figure 26. Proportion of Cases Performed on a Day Basis and 124 Cataract Surgery Rate, BC Figure 27. Length of Stay (in Days), B C 127 Figure 28. Surgery Rates by Procedure, B C 129 Figure 29. A Model of the Determinants of the Surgery Rate 131 for Various Procedures Figure 30. Supply of Specialty Physicians in British Columbia 135 vii ACKNOWLEDGEMENTS I would like to thank Robert G. Evans, my thesis supervisor, who has taught me so much during the writing of this thesis. Bob, your wisdom, support, patience and desire to see me succeed were matched only by your unwavering enthusiasm toward my research. You have made me a better thinker and I end this journey having been trained by the best. I would also like to thank Thomas Lemieux for guiding much of the research in Part I of this thesis and for always making time to answer questions. Finally, none of this would have been possible without the support of my family, especially my sister Vera. You have instilled in me, among other things, a passion for learning that I wil l take on all journeys to come. God Bless. Vll l C H A P T E R 1 - INTRODUCTION AND M O T I V A T I O N Over the past several years there has been considerable unrest in the nursing labour market in Canada. In almost every province over this period nurses held public rallies, adopted work-to-rule campaigns and in many provinces went on strike. The situation in BC in the spring and summer of 2001 was particularly severe. When an agreement could not be reached through labour negotiations, legislation was introduced imposing a 'cooling off period of two months in order to avert an all-out nursing strike. When nurses walked-out on the job subsequent legislation was passed ordering nurses back to work at which point nurses threatened to resign en masse if their demands were not met. Eventually an agreement was reached yet protests continue. Nursing representatives argue that there is a severe nursing shortage in Canada and that strikes and protests are necessary in order to pressure provincial governments to increase nursing wages. The claim is that increased wages are needed in order to retain the current workforce, to discourage emigration (mainly to the United States), and to make nursing more attractive as a profession for young individuals deciding on a career1. Increased wages would thereby alleviate the nursing shortage. Although there are numerous explanations offered as to why there is a current nursing shortage in Canada, the alleged shortage is quite often attributed to a single event: the period of fiscal restraint in the hospital sector in Canada during the early to mid 1990s [Nursing BC (1999), CNW (1999a,b), Steffenhagen (1998), Ryten (1997)]. Beginning in 1992 nominal expenditure fell in the hospital sector for the first time in Canadian history (Figure 1). In other sectors of health care, by comparison, there was simply a decrease in the growth rate. The claim is that falling hospital budgets led to mass layoffs, wage stagnation, and deteriorating working conditions. Since the majority of nurses work in hospitals this made nursing, as an occupation, less attractive. As a result, many nurses left the profession (or the country) and the number of graduates from nursing programs 1 See Lee (2000b), Sibbald (1999), Nursing B C (1999) and Morton (2001) for examples of these arguments. 1 2 decreased over this period. By 2001 these effects cumulated into a severe nursing shortage. It is important at this point to discuss the term 'nursing shortage'. According to basic economic theory, a nursing shortage occurs when at the current wage, the demand for nursing labour exceeds the supply. Furthermore, with perfectly flexible wages, any excess demand for labour is eliminated eventually as wages increase to their equilibrium level. Thus, a nursing shortage, as an economist would define it, is a result of non-equilibrium pricing and its presence depends crucially on inflexible wages. However, protagonists in the current debate do not discuss nursing shortages in these terms but instead in the context of clinical need. A nursing shortage, in the clinical sense, occurs when the amount of nursing care being provided is insufficient to meet the needs of the patient population2. Simply put, a clinical shortage implies that there are unmet needs and, unlike the economic concept of a shortage, it is independent of price flexibility. The key difference between the economic and clinical concepts of a nursing shortage rests with the different interpretations of the demand for nursing labour associated with each concept. Figure 2 illustrates this difference within the familiar supply and demand framework. The clinical demand for nursing labour depends solely on patient needs, which are independent of nursing wages. This is represented by a vertical demand curve. The economic demand for nursing labour reflects the idea that factors other than clinical needs influence the demand for nursing labour. Hospitals in Canada - the main employers of nurses - receive funding from provincial governments in the form of global budgets. These global budgets are used to pay for various hospital expenses including salaried staff, the majority of which are nurses. However, the hospital sector is Actual patient need is an ill-defined term depending on the criteria used as well as the particular interest group assessing the need. However, in discussing the two concepts of shortages, it is assumed that there is some method of evaluating the extent to which patient needs are unmet. Exactly what this method involves is not crucial to this section and is, therefore, not discussed. 3 Figure 2. Graphical Representation of the Nursing Labour Market n 0 ni N(Nurses) (a) The Nursing Labour Market n 3 n 2 n 0 ni N(Nurses) (b) The Nursing Labour Market After a Decrease in Hospital Budget Levels 4 incompletely vertically integrated (Evans 1981, 1984). As a result, many parties, each with their own often-divergent objectives interact simultaneously in deciding how to allocate the hospital's global budget. Related to nurse staffing levels, the clinical needs of the patient population are certainly one factor influencing hiring decisions but when one observes the fluctuations in hospital staff levels in Canada - especially over the past decade - one is forced to consider that other factors play an important role as well. Barer, Stark and Kinnis (1984) provide insight into factors other than patient needs that affect the demand for nursing labour: There is general, although not universal, agreement that the supply and financing sides of the health care market in Canada are at least as important in determining demand as are the consumers of the services, the patients. In this type of environment, the demand for labour is still a derived demand, but the factors that determine numbers employed at various wage levels are not the traditional equilibrating forces ofprimary markets (p.254-5). The authors examine the pattern of nursing position vacancies in BC and find strong evidence that in periods of expanding health care budgets, the demand for nursing personnel increases while the opposite is true when budgets decrease. This indicates that hospital budgets are an important determinant of the demand for nursing labour. Since hospital budgets depend on many factors other than the health care needs of the population (e.g. political priorities of provincial governments) it is obvious that the economic rather than the clinical demand for nursing labour more accurately represents hospital behaviour. Returning to Figure 2, the economic demand for nursing labour is sensitive to nursing wages and is represented by a downward sloping demand curve. The demand curve is represented as unit elastic so that the wage bill is constant at all points. This indicates that hospitals are willing to hire as many nurses as their global budget permits, which seems consistent with the Barer, Stark and Kinnis findings. 5 With the two distinct concepts of demand now well defined, one may use Figure 2 to see how nursing labour market dynamics and inferences concerning clinical shortages interrelate. In panel (a) the curves are drawn such that at wage wo and employment level no the labour market is in equilibrium. However, there is a clinical shortage of («/ -no). This produces an important result: equilibrium in the nursing labour market carries no special significance in assessing whether patient needs are being met. If the economic demand for nursing labour then decreases (as a result of a decrease in global budgets, for example) employment falls to n2 if wages are flexible or to n3 if wages are fixed [panel (b)]. The latter is the more likely scenario given that nursing wages are set through collective bargaining. Although employment falls, without knowing whether patient needs changed one can say nothing about whether the reduction in demand led to a larger clinical shortage. This produces a second important result which is illustrated in Figure 3: in order to evaluate the impact of any change in the supply of or demand for nursing labour on patient care one must examine this change relative to changes in patient activity levels. This last statement provides the backdrop to this thesis research. There are several competing claims relating to the reasons why nursing employment levels changed over the hospital downsizing era and what impact these changes had on patient care. In terms of factors driving employment levels, on one side of the debate are the supply-side proponents, most notably nursing representatives. They argue that recent changes in nursing employment levels in Canada were driven primarily by changes in nursing labour supply3. Deteriorating working conditions and stagnant wages, it is alleged, have led to an exodus of nurses from the profession (and the country), and greater difficulty in recruiting new students into nursing. In considering the validity of these claims, it is 3 See Ryten (1997), Cohen (1989), Growe (1991), Sibbald (1999) and Nursing B C (1999) for a summary of the views of nursing representatives. 6 GO a G O o p r1 H a 1-1 GO H a H O O tt °> tr q pa ft ( i >< GO tr o d erg tr. a> a Go o' •a ~ p ct>' > o 2 H t-1 8-<^ P S' o GO tr „ ° J5 3- W IB O CJQ O ST t/3 5 3 (ft B* a •I" to (ft a Si « S3 (ft ft a Si. (ft (ft (ft -1 P. cro o P 7 important to understand that nurses (and their representatives) are not third-party observers in this debate. They have a direct incentive to promote policies that benefit the current nursing workforce. As a result, the claims (and suggested solutions) put forth by nursing representatives in this debate can not be impartial. On the other side of the debate are the demand-side proponents who argue that changes in nursing labour demand drove employment over this period [e.g. Rachlis et al. (2001)]. It is well documented that the rate of use of inpatient services has been declining in Canada for several decades and especially during the 1990s. This trend is thought to have reduced the need for nursing resources and hospitals are thought to have responded by eliminating nursing positions. The Barer Stark and Kinnis result also suggests that the demand for nursing labour decreased during the early 1990s as global budgets were significantly reduced. More recently there is also disagreement on the extent to which funding increases have restored demand. Over the past five years funding levels have grown substantially in Canadian hospitals (Figure 1). This suggests that the demand for nursing labour has increased. Concerning the impact changes in nursing employment levels have had on patient care there is also considerable debate and rhetoric. Some argue - quite vociferously - that patient care significantly deteriorated during and subsequent to the hospital cut-backs4. It is alleged that even though inpatient utilization decreased steadily over this period, population aging and technological progress increased average acuity so much that the same amount of (or perhaps more) nursing resources were required in Canadian hospitals (O'Brien-Pallas, 1999b). As will be shown, this line of reasoning is flawed and, if anything, increasing average acuity suggests that fewer total nursing resources are required in hospitals. Perhaps most relevant to motivating this thesis research is that the current debate seems to be carried on in a largely data-free environment. While there is no shortage of anecdotal evidence and plausible rhetoric, there is very little research that examines what 4 See Nichols (2001) for an excellent summary 8 actually happened to employment levels and the use of nursing resources in hospitals before, during, and after the cut-backs period. This thesis attempts partially to fill this void by asking three questions: 1. What actually happened to nursing employment levels before (1980 - 1991), during (1992 - 1996), and after (1997 - 2000) the hospital restructuring period? 2. Were the observed changes in employment levels driven mainly by changes in the supply of nursing labour or the demand for nursing labour? These two questions are addressed in Part I of the thesis. 3. From a patient perspective did changes in nursing employment levels matter? The third question is addressed in Part II of the thesis and is sub-divided into two questions: 3a. Have changes in the pattern of acute care increased or decreased nursing resource requirements in hospitals? 3b. Specific to BC, have changes in the amount of nursing resources available in hospitals affected access to surgical care? The last question is motivated by the highly publicized claims in BC that during the hospital downsizing period operating rooms were being closed and important surgeries were not being performed due to insufficient nursing resources in hospitals [Lee (2000a), Fong (2000), Fayerman (2000), Wigod (1999), Steffenhagen (1998)]. It is thought that a lack of nurses to staff operating rooms and, more significantly, a lack of nurses to staff recovery unit beds (especially ICU beds) led to a reduction in the volume of surgeries performed in hospitals. 9 The remainder of this thesis is divided into four chapters. Chapter two presents the framework and discusses the existing evidence as it relates to Part I. Chapter three first identifies important trends in nursing employment levels in Canada since 1980 and then assesses whether the observed changes were primarily demand or supply driven. These two chapters make up Part I. Chapter four discusses the effect of changing patterns of acute care on nursing resource requirements in Canadian hospitals. Chapter five examines the relationship between nursing resource availability and surgical rates in British Columbia. These two chapters make up Part II. Chapter six concludes. 10 P A R T I 11 C H A P T E R 2 - F R A M E W O R K AND EXISTING EVIDENCE 2.1 Definition of Terms and a Framework Some important terminology must first be clarified before proceeding. Throughout this thesis, the term 'nurse' is used in the broadest sense. It refers to RNs (registered nurses), LPNs (licensed practical nurses5), RPNs (registered psychiatric nurses) and nursing aides. When discussing specific types of nurses, the full title will always be written (registered nurse or RN, LPN etc.). The term 'potential supply of nurses' is a measure of all individuals who could be employed as nurses. Nursing is a highly regulated profession and every individual working as a nurse in Canada must meet three criteria. First, she must have the appropriate educational background in nursing. Second, she must be registered with the appropriate self-governing body6. Registration requirements are set at the provincial level and include an education requirement and an hours-worked requirement7. The educational requirement for RNs, for example, is either a bachelor's degree in nursing from a university (BScN) or a diploma in nursing from a college8. Third, she must have been offered a nursing job and must have accepted it. The potential supply of nurses is a head count of all individuals who meet the first criteria and is represented by the box outlined in blue in Figure 4. Since educational attainment is the sole criterion for inclusion, the potential supply of nurses at time t does not necessarily represent the number of individuals who could work as nurses at time t. There will always be a sub-set of these individuals who do not meet registration criteria at time t. In defining potential supply this way, one can then interpret registration criteria as a potential barrier or 'friction' to flows from the potential supply to the nursing workforce. Potential supply 5 In Ontario L P N s are referred to as Registered Nursing Assistants (RNAs) 6 RNs, LPNs , and RPNs all have separate self-governing bodies. 7 See the appendix for the registration requirements of the Registered Nurses Association of British Columbia. 12 00 o X era' o o •z Po H o -5 O >-> •a 5' 2; 5' <? era P 'a _ S O re e. o • TI o >-t o re re fa Sd ^ o S £5 S re a. 3 r a" 5' g •a' 53 I H g 3 ° C/> 3—' era ffl O d o > H o 2; 00 w o H o r 1 o H m o H O s a. a a 5 rs 3 13 was defined in this way so as to provide the broadest measure of the stock of individuals in Canada who, after meeting other non-education registration requirements, could join the nursing workforce9. The terms 'nursing employment', 'nursing employment level', 'nursing workforce' or 'nursing labour force' all denote the same thing: the sub-set of the potential supply who are currently employed and are working as nurses. This stock is represented by the box outlined in red in Figure 4. Figure 5 illustrates the distinction between these terms as they relate to RNs in British Columbia. The term 'total nursing hours' is a measure that incorporates part-time versus full-time status, number of weeks worked during the year, and average hours worked into the measure of nursing employment. It is a measure of all of the hours of nursing care supplied by the nursing workforce. It is important to be very clear in defining these terms since the existing literature has no consistent lexicon. Almost all of the literature that addresses trends in nursing employment levels labels itself as research into the 'supply' of nurses. To the economic audience this tendency is considered a serious flaw in methodology since, according to economic theory, market quantities reflect interaction between supply and demand. Only in the case of a persistent economic shortage do alterations in supply always drive employment. While persistent economic shortages may be implicit assumptions underlying the existing body of research. For this thesis, however they most certainly will not be. Some provinces are modifying the registration criteria such that for new registrants a bachelor's degree in nursing is required (CIHI, 2001c). Some older nurses may hold diplomas from hospital nursing education programs and not colleges. Most of these hospital education programs were closed during the early 1980s. 9 A n immediate implication of defining the potential supply in this way is that policy implications that may emerge from an analysis of the potential supply are most likely medium term rather than short term. 14 Figure 5. Venn Diagram Distinguishing Potential Supply of RNs from Employment Level of RNs in British Columbia 15 Figure 6 illustrates the various pathways linking the nursing labour market to the hospital production process and provides a convenient framework within which the questions addressed in this thesis may be cast. In Part I, the pattern of hospital budget levels, nursing wages, and the potential supply of nurses is examined to determine whether the observed changes in employment during the downsizing period were driven mainly by changes in the supply of or demand for nursing labour. The relationship between nursing labour market conditions (i.e. whether there is an economic shortage or surplus), net migration, nursing school capacity and the demand for nursing education programs is left for further research. The relative contribution of net migration and graduation levels to the change in the potential supply of nurses over time is also left for further research. In Part II, the relationship between nursing employment levels and the volume of hospital services is discussed. A model of the hospital production process as it relates to surgery is developed and data from BC are examined to test several hypotheses. Specific to the objectives of Part I, Figure 4 illustrates that nursing employment levels may change as a result of decision making in two sectors: the education sector and the employment sector. The education sector is the main source of the potential supply of nurses. The most recent published data indicate that in 1997 less than 7% of all working RNs completed their nursing education outside Canada (ACHHR, 2000). The employment sector is simply the nursing labour market where supply and demand interact to determine nursing employment levels. The education decision for the individual has two dimensions. First, it refers to the decision whether to pursue an education in nursing or in some other area such as business or history. Second, for an individual intent on pursuing nursing as a career, it refers to the decision whether to complete a diploma program at a college or a bachelor's program at a university. Since governments underwrite the majority of post-secondary education costs in Canada, the important decision in the education sector from a government perspective is how many educational places to fund. This decision has a great impact on the enrolment capacity of nursing schools. As noted above, this thesis does not examine decision making in the education sector. 16 Figure 6. The Nursing Labour Market and the Hospital Production Process State of the Economy Fiscal Policy (e.g. initiative to fight deficit) Transfers from Federal Government Provincial Government Expenditure Hospital Budget Level "~l Nursing Wages ~~l Nursing School Budget Applicants to Nursing Programs Graduates from Nursing Programs Demand for Nurses I— Net migration I t t Is There an Economic Shortage or Surplus? Health Care Needs of the Population Potential Supply of Nurses Nursing Employment Level in Hospitals Level of Other Inputs to Hospital Services Need for Hospital Services Is There a Clinical Shortage or Surplus? i 1 Volume of Hospital Services Hospital Services Capacity Nursing Labour Market P A R T I Analysis Hospital Production ' Process P A R T II Analysis 17 The employment sector is where the potential supply of nurses is transformed into the nursing employment level and, in turn, total nursing hours. There are two key measures of interest in this sector. The employment rate is the fraction of the potential supply that is employed. The health care labour force participation rate is the fraction of the potential supply that is employed in nursing conditional on being employed. Changes in either of these rates will affect the nursing employment level. Part I contains a detailed analysis of trends in these two measures. Figure 4 also illustrates the various avenues via which nursing employment levels may be altered. This makes it a convenient framework within which one may discuss nursing human resource policy. For example, for a given potential supply of nurses, increasing the nursing labour force participation rate will, ceteris paribus, bring about an increase in nursing employment. Alternatively, a long-run policy may be to increase the potential supply of nurses through increased educational capacity in order to achieve the same result. But Figure 4 also demonstrates that the relationship between nursing employment levels, total nursing hours, and the potential supply of nurses is complex. A complete analysis of all of the relationships in the figure is an enormous undertaking. Instead, one turns to the existing literature to gain an insight into which factors have been identified as important in influencing nursing employment levels and total nursing hours and where further research is required. 2.2 Existing Evidence Related to Part I Trends in Nursing Employment Levels There is a fair amount of information available on RN employment levels in Canada. By contrast, few studies examine trends in LPN or nursing aide employment levels10. This is not surprising since the Canadian Institute for Health Information (CIHI) has for many 1 0 The sources for the literature search were Medline, HealthStar, PubMed, EconLit, N B E R , W o P E C , C H E P A , C H S P R , M C H P E , CIHI, Statistics Canada. 18 years compiled information from RN registration forms into a national database while no such database exists for LPNs". Nursing aides are unregulated workers and there is no annual registration process. The lack of data related to LPNs and nursing aides is a drawback. There has always been debate on the extent to which these different types of nursing resources are substitutable and recent studies have provided some empirical results [Needleman et al. (2002); McGillis Hall et al. (2001)]. It would be interesting to examine whether hospital restructuring in the early 1990s had a differential effect on LPN and RN employment levels. Investigating this important issue of substitutability would involve a significant data-collection initiative above and beyond that required for RNs and is considered beyond the scope of this thesis. Instead, this thesis focuses on trends in the employment level of RNs - who make up over 75% of the nursing workforce - in answering question one of Part I. The most recent data available on the RN workforce in Canada are summarized in CIHI (2000a,b). These two studies provide information on the employment level of RNs from 1994 to 2000, the final years of the hospital cut-back period and the subsequent period of financial recovery. Table 1 summarizes the important results from these studies. The RN employment level in Canada decreased by 0.8% from 1994 to 2000. On a per capita basis this translated into a 6.1% decrease. There was also considerable inter-provincial variation in per capita RN employment in any given year as well as significant differences in the pattern over time. The per capita measure is often reported as a rough estimate of the amount of nursing resources available to the population. In Part II of this thesis a more precise measure is utilized that takes into account changes in the way the population uses hospital services. 1 1 CIHI is planning to collect standardized data on LPNs by 2003 (CIHI, 2000b). 19 o cu =3 < > CO cu ST cn CD o CD N J CD CO o c a CD O I N J O O O m 0) N Z3 JU 5>' 3 N J O O O cr cn cn c o ex. C O C O 03 C D CO C O -o CD O O O T J O " D c_ 5' Z3 o CD CD o o o o £1 cu c CO "D| Q> H 3 CD CD o CD 3 > < CD CD 4 CD 70 |CQ cn (—«-CD - i CD O . cn CD CO TD CD C D C O "CD O o w 00 N J o cn CD o cn CD cn ST C cn O O O 3" •a cn CD cn 3 l"2L 'I c l l l t o N J C O I C D C D m 3 " D O << 3 CD zs CD < CD O —h. 70 w 5' O CD =3 0 ) C L 0 ) CD CD O I i ° § C D C D O N J Vi C D C D N J C O I*-C O C D C O C D C D N J C O N J 00 C D 00 C D C D c n "M N J o CO N J N J I C D | C D N J C D C D CD =0 I I r -o 0) N J N J C D 00 C O N J N J -v| f c D c n N J N J l o o c n o N J I - J - J N J N J I C O | N J N J N J O O o 20 Most of the decrease in employment during this period occurred in the hospital sector where 155,521 RNs were employed in 1994 compared to 148,366 in 2000. The lowest level of hospital employment occurred in 1996 - the final year of financial cut-backs -when 140,709 RNs were employed in hospitals. On a per capita basis there has been little change in the number of RNs employed in hospitals since 1995. The proportion of RNs working full time decreased steadily from 1994 to 1998 and has increased since then to 57% in 2000. Ryten (1997) and CNA (2000) confirm this trend, adding that it was well established prior to 1995 and that during the downsizing period it accelerated among younger RNs. The average age of RNs increased by two years between 1994 and 2000. The two CIHI studies are valuable in that they provide up-to-date information on the employment level of RNs in Canada. However, they are less useful in answering the first question posed in this thesis since they do not provide any information on employment levels prior to the hospital cut-backs period. Kazanjian et al. (2000b) partially fill this information gap by describing demographic characteristics, education levels and employment variables for all RNs in Canada for 1990 and again for 1997. This study complements the CIHI analysis by providing data from the pre-cut-backs period (1990 data) and providing a breakdown of employment by age. In 1990 there were 223,964 RNs in the nursing labour force in Canada. This translated into approximately 8.1 RNs per 1,000 population. In Table 1 the remaining cells under the '1990' column were deliberately left empty since the 1990 data from Kazanjian et al. are not comparable to the 1994 - 2000 data. This is because the full-time/part-time split and the hospital/LTC/other split was not available for RNs in Quebec in 1990. 21 Kazanjian et al. (2000b) remove Quebec from the 1997 data in order to make consistent comparisons across time. They find that the number of RNs working full time decreased from 93,234 in 1990 to 85,315 in 1997. They also find that there was a substantial decrease in the number of RNs working in hospitals (120,087 in 1990 versus 105,211 in 1997). There were 74,134 RNs under 35 in 1990 compared to 54,842 in 1997 and there were 69,240 RNs over 45 in 1990 compared to 97,625 in 1997, confirming the workforce aging trend12. The CIHI (2000a,b) and Kazanjian et al. (2000b) results can be combined into five important trends in the RN workforce over the period 1990 to 2000: 1. An increase (1994 versus 1990), a decrease (1995 - 1998) and a subsequent increase (1999 - 2000) in the RN employment level. In 2000 there were approximately 8,500 more RNs working than in 1990 2. A steady decrease in per capita employment 3. A sharp decrease and subsequent increase in hospital employment 4. A decrease and subsequent increase in the proportion of RNs working full time 5. Workforce aging due to fewer RNs under 35 years of age and more RNs over 45 years of age Trends 1 - 4 are important since they indicate that significant changes in the employment level of RNs occurred during and subsequent to the hospital downsizing period. Refining some of these trends and determining whether they are mainly a result of supply-side or demand-side effects are the two objectives of Part I. Workforce aging is of particular interest due to its implications for future RN capacity in Canada. As Ryten (1997) and CNA (2000) note, fewer young RNs in the workforce today implies that there will be fewer nurses available to replace the large number who are expected to retire over the next five to ten years. Although this is a valid claim, from a policy perspective what is important is the extent to which the nursing workforce is 1 2 These last figures include Quebec. 22 aging more rapidly than the workforce in general in Canada. If there are fewer young people in the Canadian labour force as a whole, then aging is a population phenomenon and policy makers may have very few options to alter the aging trend in the nursing workforce. If, however, the nursing workforce is aging more rapidly than the workforce in general, this indicates that behaviour in the nursing labour market is unique and there may be more policy options available. Gunderson and Riddell (1999) show that the Canadian workforce is in fact aging as the demographic structure of the Canadian population changes but no study to date has explored the relative aging of the nursing workforce. This is addressed in chapter three. The Determinants of Nursing Employment Levels There are far fewer studies examining the factors driving nursing employment levels in Canada compared to the number that document the trends. As noted, this may be due to an implicit assumption of a shortage of labour in which case labour supply always determines employment. On the other hand there are numerous analyses of the nursing labour market in the US but most of these studies examine the extent to which the nursing labour market is monopsonistic [e.g. Staiger et al. (1999); Hirsch et al. (1995)]. Furthermore, they are almost always based on a model where employers act to maximize profits. Profit-maximizing behaviour on the part of hospitals is inconsistent with incentives in the Canadian health care system since hospitals operate as strictly not-for-profit institutions [Evans (1984), Evans et al. (2000)]. This renders these studies less relevant to the Canadian context and, in fact, since the majority of US hospitals (when weighted by size) also operate on a not-for-profit basis, less relevant to the American context as well. Ryten (1997) indicates that any combination of three factors could lead to the observed decrease in the number of young RNs in Canada. First, the average age at which individuals graduate from nursing programs may be increasing. Second, there may be fewer young individuals entering (and exiting) nursing education programs. This might be because there are fewer university-aged individuals in Canada, or because the 23 likelihood of pursuing an education in nursing conditional on pursuing an education is falling, or both. Third, the health care labour force participation rate may be falling for younger individuals. It is important from a policy perspective to ascertain which of these three factors is most responsible for the aging of the RN labour force. If it is the first factor, then policy makers ought not to be so concerned about the future availability of nursing resources since the rate of inflow to the potential supply is not expected to decrease. The second factor, however, implies a decreased inflow to the potential supply. Moreover, if the decreased inflow is a result of decreased nursing education take up, then there may be a significant role for policy makers should they wish to increase the inflow to the potential supply. If Ryten's third factor is driving workforce aging, then policy makers ought to shift their focus away from the education sector and focus on the employment sector in effecting any changes to RN employment levels. In chapter three the relative importance of these factors is quantified. Ryten (1997) finds evidence that graduates from nursing programs are getting older. One factor likely contributing to this is the increased emphasis on the BScN as an entry-to-practice requirement which is expected to increase the average length of education and, ceteris paribus, produce older graduates. Addressing Ryten's second factor, Sabuhoro (2001) finds that total admissions to university, college, and hospital nursing programs were relatively stable from 1980 to 1989, increased slightly to a peak of about 14,000 in 1991, then fell sharply to just over 8,000 in 1998. This downward trend in admissions is reflected in both enrollment and graduation levels in the expected direction. Graduation from and enrollment in nursing programs decreased in Canada over the period 1995 - 199813. Enrollment fell by 14% 1 3 Personal correspondence, Canadian Health Services Research Foundation. 2 4 while graduations fell by 24%14. Clearly, some of the workforce aging trend can be attributed to decreased nursing school enrollment and graduation. Going one step further and asking why nursing school enrollment decreased, Baumgart (1997) argues that most of the current nursing workforce in Canada was educated during the 1960s and 1970s when nursing was seen as a very rewarding career - an 'elite' profession for women. Also during this period, training large numbers of nurses was one of the top priorities of provincial governments (Baumgart and Larsen, 1988). However, because of broad social changes in gender roles over the past forty years, she argues, nursing is no longer an elite profession for women. As a former president of the Ontario Nurses' Association observed, "[t]oday, women have lots of choices. They don't have to pick between nursing and teaching" [qtd. in Cohen (1989)]. Baumgart (1997, 1988) also argues that low nursing wages have contributed to the decrease in uptake of nursing education. She concedes, however, that no comprehensive nurse wage data are available and simply asserts that nursing wages have shown little or no growth since the 1980s. It is clear from the existing research that the factors driving the decrease in enrollment in nursing programs have not been adequately identified. However, as noted earlier, this thesis focuses on the employment sector and does not attempt to model decisions in the education sector. As a final note to the education debate, there are two hypotheses that have not been considered in the literature that could potentially explain the decrease in nursing school enrollment. Both are related to the recent emphasis on the BScN as the preferred educational requirement to practice. Since Canadian governments underwrite most of the cost of a post-secondary education, the number of enrollees in universities and colleges is sensitive to government spending These rates are disproportionate, in part, due to the fact that all of the decrease occurred at the college level. Enrollment and graduation levels in university programs increased steadily between 1995 and 1998. With the increased up take of four-year university programs during this period, enrollees per graduate (all programs) increased from 3.8 in 1995 to 4.0 in 1998. 25 on education (Figure 6). Since it costs more to produce a graduate from a four-year university program compared to a two-year or three-year college program, any shift away from colleges toward universities with fixed education budgets will, ceteris paribus, lead to a decrease in nursing school capacity. If a bachelor's degree becomes the entry-to-practice requirement and nursing school budgets do not increase this 'BScN-Fixed Budget' effect may become significant. The increased emphasis on the BScN also affects the career decisions of university-aged individuals. A switch from diploma programs to bachelor's programs entails a significant increase in human capital investment. An individual must then decide if the return on this increased human capital investment is highest in the nursing profession or whether it is more profitable to pursue a university degree in some other field such as commerce. The BScN requirement forces nursing to compete with other professions that are accessed through a university education. Unless nursing wages increase relative to other professions, this 'BScN-BComm' effect will, ceteris paribus, result in a decreased demand for a nursing education on the part of university-aged individuals. CNA (2000) addresses Ryten's third factor, the health care labour force participation rate. Using data from the National Graduate Survey the authors examine labour market outcomes for three cohorts of recent graduates from nursing programs in Canada (1986, 1990, and 1995 cohort). Five years after graduation, 29% of the 1990 university-educated cohort was working in a non-health care field compared to only 10% of the 1986 cohort15. For college graduates, there was a similar but less pronounced change with 10% of the 1990 cohort working in a non-health care field compared to almost none of the 1986 cohort. This evidence suggests that a decrease in the health care labour force participation rate may be important in explaining the decrease in employment among the youngest RNs over the hospital downsizing period. The decrease in the health care labour force participation rate among the youngest age groups - particularly for BScN graduates - ought to be of particular concern to policy 1 5 Five-year follow up data for the 1995 cohort was not available at the time the study was written. 26 makers. It takes more resources to produce a university-trained nursing graduate compared to one that is college trained. If these university-educated graduates are more likely to take up employment in non-nursing fields, governments could view this as a smaller return on investment than if these individuals applied their skills in the health care system. It seems important, therefore, to ascertain why young graduates were less likely to work in nursing during the downsizing period. More specifically, was the decrease in the health care labour force participation rate mainly demand driven or supply driven? Sabuhoro (2001) most recently explored the factors affecting the labour force participation decision for individuals educated in the nursing field using 1996 Census data. His findings appear consistent with general female labour supply research: wages have a positive effect, age has a positive but decreasing effect, and the presence of young children has a negative impact on participation. The magnitudes of the estimated elasticities are also within the range normally found in the female labour supply literature (Chaykowski and Powell, 1999). However, Sabuhoro does not distinguish employment in health care from employment in other sectors. The study is useful in that it points to the factors one might examine in order to explain changes in the health care labour force participation rate over time (i.e. wages, demographics) but it does not examine the occupational decision conditional on employment. CNA (2000) investigates annual wages as one possible factor driving the decrease in the health care labour force participation rate of recent graduates over the downsizing era. They examine income levels for recent graduates between 1988 and 1997 and find that annual salaries for RNs grew at a higher rate than salaries for all other post-secondary graduates. The difference-in-difference estimate (i.e. the increase in the nursing premium) was approximately $2,000 with almost all of the relative salary gains at the bachelor's level. This evidence is consistent with the hypothesis that, at least for BScN graduates, the decrease in the health care labour force participation rate was demand 27 driven. Nursing wages increased relative to other occupations but there simply were not enough nursing jobs for all graduates16. It is very important to note that no study examines trends in the health care labour force participation rate for older cohorts of the potential supply and that the only study examining the trend for new graduates is based on limited survey data. Chapter three of this thesis contains an extensive analysis of the health care labour force participation rate for all ages using a much larger data set with a particular emphasis on whether any observed changes were driven by demand or by supply. This analysis is viewed as one of the major contributions of this thesis. Moving beyond the three factors identified by Ryten, there is also the possibility - often forwarded in media reports - that the employment level of young RNs is decreasing because new nursing graduates are leaving Canada for better employment opportunities in the United States. CNA (2000) provides data indicating that by 1997, 9.3% of the 1995 cohort of nursing graduates were living in the United States compared to 1.3% of all graduates with 89% indicating the reason was work related. A separate survey by the Registered Nurses Association of Ontario found that of the RNs who left Ontario between 1991 and 2000, 70% indicated that 'job opportunities' were the main reason while 25% indicated 'family or personal reasons' (CIHI, 2001c). However, neither of these studies track these trends over time17. There is a tendency in many studies of international labour flows to focus only on outflows from Canada and nursing is no exception. As Barer and Webber (1999) show, ignoring immigration flows into Canada in the analysis of the potential supply of health care providers often leads to inaccurate inferences. Neither CNA (2000) nor Ryten (1997) take these flows into account. As a result, these studies do not provide insight into 1 6 This would also imply mainly a quantity adjustment to a decrease in demand rather than a wage adjustment. 1 7 Ryten (1997) shows that the number of requests for verification of credentials - a proxy measure for intent to leave Canada to work as a nurse abroad - increased significantly between 1992 and 1996. Schumacher (2001) provides evidence from the United States of significant wage stagnation in nursing 28 the extent to which net migration has affected the age distribution of the RN workforce in Canada. At this point little can be said regarding the extent to which net migration has contributed to RN workforce aging in Canada. Advancing this area of research is important but such an undertaking is left for further research. The trend toward part-time status is of interest as it has direct implications for the total amount of nursing hours provided over the downsizing period. Since average hours per nurse decreased, total nursing hours changed disproportionately to employment. Very few studies explicitly address the factors driving the increase and subsequent decrease in part-time status. CNA (2000) argues that part-time status among young nurses was increasingly involuntary during the early 1990s. This seems reasonable given that young individuals are expected to have a strong preference for full-time work and implies the trend was demand driven. However, a recent study also found that the majority of RNs working in hospitals on a casual basis were doing so voluntarily (CIHI, 2001c). Canada in the International Context Finally, some results from abroad are briefly summarized to place the Canadian situation in an international context. Nursing shortages appear to be a North American as well as European phenomenon [Scott (2001); Buerhaus, Staiger and Auerbach (2000a); Shields and Ward (2001); Antonazzo et al. (2001)]. It is interesting to find countries with vastly different health care financing and provision infrastructures apparently experiencing the same situation in the nursing labour market. The underlying factors driving the current nursing labour market trends in Canada may be the same as those in other countries and, therefore, independent of the setup of the Canadian health care system. However, it may also be the beginning in the 1990s after a period of sizable growth during the 1980s. This would suggests that i f net migration of RNs from Canada to the U S increased, factors other than higher U S wages were the reason. 29 case that the expansion of managed care in the US, NHS reforms in Britain (namely general practitioner fundholding and the establishment of Hospital Trusts) and reductions in global budgets in Canada - three vastly different policies - all had a similar effect on the nursing labour market. Such a comparative analysis is well beyond the scope of this thesis. The workforce aging phenomenon appears to be occurring in the US as well. The number of RNs under 30 years of age fell by 41% over the period 1983 - 1998 and this is thought to have led to shortages in areas of nursing that were traditionally dominated by younger RNs (Buerhaus, Staiger and Auerbach, 2000a,b). Moreover, there is evidence to suggest that the decrease in the number of young RNs was driven mainly by a decrease in the probability that a high school graduate pursues a nursing education (Auerbach, Buerhaus and Staiger, 2000). This decrease in the age-specific education take-up rate was driven, in turn, by a decreased interest in nursing as a career as opposed to a reduction in nursing education capacity (Staiger, Auerbach and Buerhaus, 2000). The decreased employment opportunities for nurses in the hospital sector for the period 1990 - 1997 observed by Kazanjian et al. (2000b) also occurred throughout the US but to a lesser extent. In Canada the employment level actually fell among RNs working in hospitals while increasing for RNs as a whole. In the US it merely increased at a much lower rate than that among RNs as a whole. Buerhaus and Staiger (1999) show that in the US there were 1.04 million FTE RNs working in the hospital sector in 1990 compared to 1.18 million in 1997- Using their method of calculating FTEs and using data from Kazanjian et al. (2000b) there were 93,667 FTE RNs working in hospitals in Canada in 1990 compared to 78,908 in 199718. Buerhaus and Staiger (1999) calculate FTEs as follows: F T E RNs = #RNs working full time + 0.5*#RNs working part time. Since Kazanjian et al (2000b) do not provide a sector specific part-time/full-time split, the ratio for RNs as a whole was applied to the hospital sector. Casual employees were counted as part-time employees in the calculations for Canada. Data from Quebec are excluded from the calculations. 30 C H A P T E R 3 - TRENDS IN NURSING E M P L O Y M E N T L E V E L S IN CANADA 3.1 Description of the Data Primary data used in this chapter are drawn from the Registered Nurses Database administered by the Canadian Institute for Health Information and the Census of Canada administered by Statistics Canada. Descriptions of the databases, details concerning particular shortcomings, and methods used to overcome these shortcomings are discussed in the appendix. 3.2 Trends in RN Employment in Canada Table 2, Figure 7 and Figure 8 describe the evolution of RN employment levels in Canada, BC, Alberta, Ontario and Quebec from 1976 to 2000. RNs account for three-quarters of the entire nursing workforce. At the national level, RN employment increased steadily from the early 1980s until 1993 (Figure 7). From a peak of 235,625 in 1993 it then decreased steadily until 1998 when there were 227,651 RNs employed in nursing in Canada. This represented an average annual decrease of 0.7% over this five-year period compared to an average annual increase of 4.0% between 1980 and 1993. Employment increased substantially from 1998 onwards so that by 2000 there were 232,412 RNs working in nursing in Canada, roughly the same number as just prior to the cut-backs period. The decrease in RN employment in Canada during the restructuring era was solely a result of substantial reductions in hospital employment (Figure 7). From 1993 to 1998 the number of RNs working in hospitals fell at a rate of 2.2% per year while non-hospital employment grew steadily since about 1980. Examining data prior to 1990 leads one to conclude that the early 1990s brought about an end to a twenty-year trend of steady growth in RN employment in Canadian hospitals. 31 CO o c —1 o CD 70 CD CO co' i—i-CD —i CD Q. CO CD CO D 01 5T CX _> CO CD z Canada Alberta Ontario Quebec mber of RNs employed nursing per 1,000 pop. cn A (J) CJl Ol to CD 4- s i bo bo --4 cn Ol 4^ CD cn cn to CD cn sj b> 4*. s i s i P> C) S X U l CO s i CD M CO -.1 CO CD cn cn cn cn CO k i O CD CO N -si CO cn o i N o i cn CO CO KJ CD Ko co CO o O) cn s i cn cn to cn cn A b w CO cn s i a> cn cn _J> CO cn b b t . ^ CO K> s i s i s i si cn co o IJ . i o -»• b i 00 CO s i ^ CO ^ O) 1 CO CO 4_ co cn b i CO 4> SJ -si CO N N CO b i •--1 N OJ b CO Ol s i CD CO si s| to CO •-J- O sj I—. 00 cn BC Alberta Ontario Quebec Canada Number of RNs employed in hospita s 1976 1977 1978 1979 1980 1981 1982 1983 1984 145,159 1985 1986 hO —st CO CD CD —1 CO ^ - s | O M Ol J CD Ol M U 4- "-1^  bo V i co oo cn oo co o i w CD j s . c n _ > . _ i j - o -t- -P-"o "cn oo K> S s i o " 4- Ol 4=. co cn - _ co -si INJ cn "cn "j- js. "i^ j S 4^ _ CD cn cn 4- -si co cn ->• CD CD co cn "cn "-- "CD "--s i o cn 4-NJ CD O Nl 4- cn - i M CO Ol 4- O CO CO Ol Ol CO O CD O 4a. Ol -v 4- cn cn s i "-si 4- "->• O) o cn o NJ CD IM 03 4- CD _>• cn cn -si cn "cn "-_ "o CD O CD CO o o 4- si cn 4- CD _* si CD 0 0 CD 0 CO O CO 01 - 05 CD s i CD 4^  4-M O Ol CO W U t> Ol o cn 4- o O) Ol Ol 4. cn co 4- o Ol 4- 4- 4-Z 5' c 3 c tx •n CD on - i 5' 0 CQ —h 73 Z (A a 3 •D O >< a a 32 CO CO CO 00 --J co b u ^ b c O CO Nl N I 00 CO 00 --1 CD CO •T^  CO fo 0 } CO CO CD CO CO s ^ J £ cn co co b ) 00 CO 00 CO CO N N I CO b bo cn in CO o CO 00 Nl CO cn b) N I CO 00 ( D Q CO M _ l CD CO <= co ^ V CO K> CO to Q Co s CO ro 4^ Q fo Vl CO CO 00 CO CD 00 Nl CO cn Vi cn CO 4a. Nl (D CO S N| co CO Vi 4=. bo 4*. CO Ol —J CD CO Nl S CD Vi b cn cn co CO CO N | CD CO N | Nl CO K> CO CT> CO CO Nl N I 00 CD Nl Nl CD cn CO 4^ 05 O CD CO N I CO CD Nl CD CO cn ^ CO CD CO CD N I CD CD Nl o> ro o cn N) Vl 4^ 03 o o 1987 155,300 1988 1989 14,678 57,337 1990 165,298 1991 1992 160,038 1993 17,856 14,281 • 47,921 41,719 155,521 1994 18,057 12,931 46,794 34,207 145,809 1995 18,853 12,187 48,393 30,759 140,709 1996 18,855 12,609 45,648 38,104 145,467 1997 18,199 13,110 44,422 36,132 142,043 1998 18,112 13,400 43,925 36,884 142,752 1999 18,007 13,983 48,686 37,279 148,366 2000 CO o ro ro co co O ) 4=* co co co ro CO ro "oo o> oo ro ro CO CD co ro ro N | ro ro 00 "4^  ro ~4^  co N I cn ro o co co o Co 4^  CO CO - J . ro cn cn 03 ro cn ro CO 45» b CO 03 CO ro CO CO CO N| cn 4=* cn cn 00 - j - ro Nl ro CO 4=. cn CO *-v "CD o cn CT> 4^ 4^ CD -» • - i co ro ro N I -cn co ro ro NI cn - i - cn "co 4^ V co —* cn CD o co — -OO CO 03 CO CD 4^ . CO 00 o co cn 4^ cn co ro ro "ro "co oo "cn ->• O O) Nl oo - i o cn cn N I ro ro J O CD —^ N J "o V "ro "oo Ol - i CD O) 00 O N | CO cn oo ro ro NJ o o oo "ro "-^  V i co co cn 4^  00 - i oo cn N I ro ro CD CO —^  03 "->• O -P> CO Ol O) M Nl O N | 03 4^ oi N I ro ro C7> CO —^  ~ bo bo "co ~o ro ro oo o Ol Ol 03 4^  cn Ni ro ro N I 0 3 ro N I CD "-^ b CD CO CO A O N | 4>. oi oo to ro oo ro N( N) N| co co ro o 33 Figure 7. RN Employment in Canada Canada - • — Number of RNs Employed in Nursing -&— Number of RNs Employed in Hospitals RNs Employment per 1,000 Population 12 -, • BC —H— Alberta —A— Ontario -Quebec X Canada Source: Registered Nurses Database 34 3 3 cr cr o CD 73 73 Z Z m m 3 3 o o o o o o o o o o o o o o o o o o o o o o o o o o o cr CD a. 3 3 3 3 z z w en m m 3 3 _L _L o o •a _. 0) CQ CO c o 9. SI CD C O c n ' CD —1 ffi I cn CD cn D s. CD < T 0) cn CD H z z c c: 3 3 C T C T CO CD 73 73 m m 3 3 O 3 —t> 0) o" to Nl CO CO 4^  4- cn cn Ol Ol o Ol _o cn _o cn _o cn o cn o O o "o o o o o o "o o o o o o o o o o o o o o o o o o o o o H z z c c 3 3 a- cr <o to o o —>i —K 73 73 Z Z m m 3 3 o o >< •< <o co D c CD CT CD n 35 In general, the RN employment profile over the downsizing period is quite similar in the individual provinces19, leveling off in the early 1990s after a long growth period (Figure 8). The notable exception is BC where employment levels continued to increase throughout the early to mid-1990s. In the hospital sector, again, BC is the exception. In Alberta, Ontario and Quebec hospital employment decreased over the 1990 - 1997 period while in BC it remained fairly constant. It was noted in chapter two that the per capita measure of RN employment is often reported as a measure of nursing resource availability. According to this measure the pattern of RN employment in BC is very similar to the pattern in the other provinces suggesting that nursing resource capacity also decreased in BC over the downsizing period (Figure 7). However, a more accurate measure of nursing resource capacity is examined in Part II of this thesis. Thus, at this point, all one concludes is that RN employment in hospitals did not fall in BC while it did in the other provinces but based on this data nothing can be said concerning the pattern of the availability of nursing resources to patients in hospitals. Table 3 shows the age breakdown of RNs employed in nursing for each of the four provinces. In every province, there was a significant decrease in the number of RNs 25-34 years old and, in particular, a significant decrease in the number of RNs less than 25 years old. These figures demonstrate clearly that the hospital downsizing period was associated with a significant reduction in the number of young RNs. Data are not available at the provincial level on the age structure of the RN workforce prior to 1990. However, national data provide some insight into the extent to which this workforce aging phenomenon pre-dated the restructuring era. 1 9 Only the four largest provinces were selected in order to reduce the workload of analysts at Statistics Canada and CIHI who provided the data and because of the substantial variation in the pattern of hospital expenditure within these provinces. 36 cn -t- co ho A cn cn cn cn ro cn 4-4^ 4* _- NJ _-o o cn NJ N O 4 > 0 0 CO NJ CO 4-co cn 4- cn cn _- NJ __ cn CD NJ cn NJ NJ 4- NJ ->• CO -si O CO -k NJ CO O CO CO cn co co NJ CO CO -k CO -si co NJ cn co cn CO CO —4—• cn oo co NJ —^  cn co to co co NJ o co o cn O CO -s! CO CO 4- 00 CO NJ ->• 4- CO 4— CO CD —^  cn 4— co NJ ->. -k 4- CO 4k CO CO NJ NJ cn -si o NJ 4^ 4^ - NJ CO CO NJ cn O 4k CD cn 4k 00 4k CD CO NJ _v _v O 00 - i . NJ CD Ol 00 NJ " CO 00 o -sJ OJ -k D c CD tx CD o CO CO CO CO 4k CD CO cn CO CO CD CO CO -J CD CO CO CO CO CO NJ o © cn 4^ - CO NJ A cn cn cn cn M + cn 4k co 4» 4k 4k —^  NJ NJ CO CD CO —>• NJ cn NJ co co —1 -sJ 00 CD CO —I —J co cn cn co NJ NJ —^  CD OJ —I CD " CD CD O o o 00 o -k 00 -si 00 CD NJ o cn NJ NJ -x CO CO CD CO -si co CD cn a> co 4^ - o cn co -si oo cn OJ -4 —^  NJ NJ —^  O 4» CO -s| NJ —J 4^ CO CO CO NJ cn —k 4^ -si OJ -J cn 4*. —^  NJ NJ —^  o cn cn CD -k 4^ - 4*. CO CD CO -si O NJ CD -si NJ OJ NJ —^  NJ NJ o CD cn CD cn - i -si -k CO CO CD NJ cn CO -si -4 CD NJ -k —^  NJ NJ —^  O ) 4> 4> 4^ 00 Ol 4k 00 CO Ol -si CD -v O) O 4k 4^ -x NJ NJ CO CO 4> NJ co cn -x CO CO 00 CO 4k NJ CO O 00 4k -s| o 3 DJ —I o CO CO o CO CO CO CO O l CO CO CD CO CO -s| CO CO 0 0 CO CO CO NJ © o o cn 4^ - co NJ A cn cn cn cn NJ + cn 4k co 4k 4k 4k -k co CD cn 4k CO -4 00 00 NJ —k CD NJ cn -Si —J —J —k CD NJ co —J cn M O O) CO A O 4k CO CD si 4k 4k OO -k -k NJ CD -si cn NJ NJ 4k O NJ CO CO 00 O CO -si 00 NJ —k CD NJ CO -si 4k NJ NJ —k -si CO -k CO CO - J CO oo -k cn cn co NJ CO —J 4^ 4k cn -k - j 4k cn CD CD —si CO NJ oo cn 4k cn NJ CO —I 4k -si 00 -k 00 4k 4k -k NJ 4k 4k CO CO —* -si —J NJ —I CO 4k CO O CD -si CO CD NJ 00 O NJ CO CO O CO CO CO -si CD 4k NJ CO -4 cn NJ OJ OJ cn —j co NJ O0 CD O CO c r <D 3. at CO CO o CO CO 4k CO CO cn CO CO CD CO CO -si CO CO CO CO CO CD NJ o o o cn 4k co NJ A cn cn cn cn NJ + cn 4k co ^ 4k 4^  4k NJ CD 00 CD cn -k cn o -si CD CO OJ O CO O 00 OJ 00 CD CD CO —k CO O 4k -k -k cn co NJ CD NJ -si CD NJ co co CD cn 4k cn co oo co CD cn —k 4k co o cn oo 4k CO 00 CD Ol —4 CD CO —J CO NJ 4k o o cn co oi cn -k NJ 4k CD CD cn o co cn co 4k OJ CO 4k O O CO NJ CD —k CD 4k CD CD Ol O CO O O 4k CO CO OJ CD CO CD NJ O CO O 4k CO 00 4k CO Ol -si -J 4k NJ 00 -si CO 4k NJ O —k —Si —k 4k CO 00 4k cn oo OJ cn 4k 4k CD OO NJ -k NJ 00 O CO —*-00 o CO CO o CO CD 4k CO CO O l CO CO cn CO CO -si CD CO CO CD CO CO NJ o o o 73 z t/J CO 3 •a o (D Q . 3 c 3 CQ c 3 cr s1 cr Co *>. CD 03 Q. o I o I CD I o CD NJ CO i NJ NJ —si NJ —s| CO CO cn 4k —k co co 4k -si 4k CO 4» NJ CO NJ CD CD -si O —k CD 1^ CD NJ cn CD CO CO o I to o o o o 3 " 0 J 3 ca CO 37 The number of RNs less than 35 years old has been steadily decreasing in Canada since the mid-1980s (Figure 9). Clearly, the trend of decreasing numbers of younger RNs was present prior to the expenditure reduction period. This suggests that demographic factors might partially explain the decrease in the number of young RNs in the nursing workforce during the downsizing period. This subject is explored in the next section. 3.3 Factors Explaining the Pattern of Nursing Employment Levels in Canada 3.3.1 Quantifying the Importance of Various Factors The following identity describes the relationship between the nursing labour force and the potential supply of nurses: WF = PS x ER x HCLFPR where WF = Nursing Workforce (Employment Level) PS = Potential Supply ER = Employment Rate [E/(E+U+OLF)] HCLFPR = Health Care Labour Force Participation Rate The health care labour force participation rate is the probability that an individual is employed in the health care sector conditional on being employed. In discrete time the change in the size of the nursing workforce can be decomposed into three components to yield the following relationship: Ln(WFl+))-Ln(WFJ = fLn(PSl+]) - Ln(PS,)J + [Ln(ERt+1) - Ln(ERt)] + [Ln(HCLFPRl+i) - Ln(HCLFPR,)J It is then possible to quantify the relative importance of each of the three factors in explaining the change in the size of the nursing workforce over the hospital downsizing 38 m A cn m to CO 4^ 0 CO I O l £ l 11 c!n B cn cn + cn o o o o 1982 1985 1988 1990 1991 1992 1993 1994 1995 1996 hi 1997 H 1998 1999 2000 •to -23-— ^ of] , - cn -—6H -8-..CO o o o o o cn o o o o o p o o o cn o o o o - o f 4* LraJ -3- -8-a • ' a-4^  -H-- N F CD a: i_. C O cn -1 C O i C D t C O r-o ro . O J o> N) r M - J - c , -C O i -' 4^ - • U1 C O C J Nl s --a-CD , <p~-O D 4^ 2 r CO : •o — S r -- Nl C 3 cr o -h 73 Z 5" O 0) 3 0) a. u cr > (Q CD Q o c 3 I 5-S CD o 3 a> Cr «t> itD -s-N) CO T5h . Nl' -a-K) -5t-: T n CD 'N| N to_ co co D tu 5T cr 03 co CD 39 period. For example, if the reduction in the number of young RNs is driven mainly by decreased output from educational institutions and young RNs leaving Canada, then [Ln(PS,.,) - Ln(PSt)] ~ [Ln(WF,+,) - Ln(WFd] and [Ln(ERt+l) - Ln(ERt)] + [Ln(HCLFPR,+,) - Ln(HCLFPRt)] ~ 0. Ideally, one would want to carry out such a decomposition for the RN workforce separately so that the analysis can be matched to the RN employment trends identified in section 3.2. However, there is no way of distinguishing the potential supply of RNs from the potential supply of LPNs, nursing aides etc. in the Census of Canada Master File data 20 that is used in the decomposition exercise . Therefore, the decomposition analysis in this section is for the nursing workforce as a whole and not exclusively for RNs. Of course, one keeps in mind that RNs dominate the nursing workforce and are expected to drive the results. Figures 10, 11, 12 and 13 illustrate graphically the change from 1991 to 1996 in the size of the nursing workforce, the potential supply of nurses, the employment rate and the health care labour force participation rate for each age group in BC, Alberta, Ontario and Quebec. Again, it is emphasized that these charts are based on a combined sample of RNs, LPNs, Aides, Orderlies etc. The potential supply sample was constructed by selecting all individuals in the Census data who indicated that their major field of education was nursing. The employment rate is the fraction of these individuals who are employed. The health care labour force participation rate is the fraction of these employed individuals who are working in the nursing field. The timing of the Census in Canada is ideal for the purposes of this thesis: 1996 Census data provide a picture of the nursing labour market in a period of deep hospital expenditure reductions while 1991 data provide a picture of the labour market prior to the cut-backs. For an explanation see the appendix. 40 41 42 43 44 In each of the four provinces there were fewer nurses under 40 and more nurses over 40 in the workforce in 1996. This is consistent with the results from the Registered Nurses Database summarized in section 3.2. At least some of the 'tilt' in the age structure of the nursing workforce was accounted for by the aging of the potential supply. There were fewer individuals under 40 and more over 40 in Canada whose major field of education was nursing. The employment rate did not change significantly between 1991 and 1996. The exception is the youngest and oldest age groups in Quebec where the employment rate fell by ten percentage points. In other provinces any changes were negligible. While the health care labour force participation rate decreased by only three or four percentage points for all age groups combined, it decreased significantly for those under 30 and decreased quite dramatically for those under 25 in every province. For the less than 25 age-group it decreased from 78% to 52% in BC, 83% to 60% in Alberta, 78% to 55% in Ontario and 81% to 57% in Quebec. Clearly, over the hospital downsizing period, young individuals with an education in nursing who were employed were much less likely to be employed in the health care sector. In 1996 there were 2,684 individuals under 30 in BC who had an education in nursing but were employed in a non-health care occupation. In Alberta there were 1,372 while in Ontario and Quebec there were 7,066 and 4,211 respectively. Figure 14 conveniently summarizes the relative importance of each of the three factors in explaining the change in nursing employment levels over the downsizing period (i.e. between 1991 and 1996). The total height of the stacked bars represents the change (in logs) in the employment level for each age group. It is composed of the change (in logs) in the potential supply of nurses, the employment rate and the health care labour force participation rate. 45 V) o c —I o CD O CD CO c CO O CD W CL 0) • o rn, o Tl 73 m -a ml 3 CO 51 I O Tl T l Ln(1996) - Ln(1991) o o b o o o o o o o o o o o o Voicn^UNJ^b^kJCo^cna jV [SJ 3 " CD cn oi S 1~ 73 CD CO CO 3 w — N" O <° 3. ° s° 5 cn < CD -3 73 O 3. tu 3 O CD o T l fi) f) |-*-o 1 CO m x •a 0) T l O 3 CO O 3" 0) 3 ca CD Ln(1996)-Ln(1991) o o o o o o o o o o o o o o p Vbcn^wi\3^b^i\3sj-f>-sncn-ssi L L T 7J CD CO CO CO CO a> 5" co — N" 0 «> £ o CD - h cr z CB (fl 3" ca |— cr o < CD 3 T3 O 3 a> 3 O CD O 0) o o m x TJ_ B) 5" ca o 3" n> 3 ca CD Ln(1996) - Ln(1991) o p o o o o o o o o o o o o o ^ b 5 c n ^ c o K j ^ b ^ t s ) W ^ c n b ) s ro CD Oi OJ CO I I '—s 7J CO CD CO DJ ^ 1 <-\ CD CO — CO 3 -0 0 3 3 CO an N' 0 CD CD o O —•* —ts z T l c 0) O cn 0 3 - ™c ca en r~ m DJ X cr 73 0 DJ c - * 3 T l 3 ' O ca -1 O O CD 3-. 01 00 3 0 <o ' 1 CD Ln(1996) -Ln(1991) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Vajcn-r^cofsj^b^isou^bicnN) ^ CD - n «? 0 ro » o CO c 3 o o 3 73 O to 46 Displaying the data in this manner illustrates the fact that the relative importance of each of the three factors differs by age. For those under 25 the main factor by far is the decrease in the health care labour force participation rate. The only exception is Alberta where the decrease in the potential supply is equally important. For those 25-29 the decrease in the health care labour force participation rate remains significant, but the decrease in the potential supply begins to dominate. For those 30 and older, most of the change in employment is accounted for by changes in the potential supply. The employment rate in all cases accounts for very little of the change in the size of the nursing workforce. As noted in the previous chapter, the reduction in the number of young individuals in the nursing workforce is of particular interest. These young individuals are new to the nursing profession and, unlike their older counterparts, have most of their working lives ahead of them. Since the number of these individuals appears to be decreasing - quite dramatically in most provinces - a significant portion of the future nursing capacity in Canada seems to be disappearing. Understanding why the potential supply and the health care labour force participation rate of this age group decreased is extremely important as this knowledge will provide policy makers with an understanding of the potential interventions that may reverse this trend. The potential supply of nurses is affected by graduation levels from nursing education programs and net migration (Figure 4). Graduation volumes, in turn, depend on the number of individuals in Canada who are of college/university age and the nursing education take-up rate of these individuals21. If the nursing education take-up rate and net migration are both stable, this implies that a change in the number of college/university aged individuals in Canada is the main factor driving the decrease in the potential supply of young nurses. More significantly from a policy perspective, it also implies that a shift in population demographics - which is beyond the control of policy makers - is driving the decrease in the potential supply. 47 Figure 15 compares the change in the size of the potential supply of nurses to that of the female population in general. The female population was selected as a comparison group since over 95% of nurses in Canada are female. In Alberta and BC the potential supply of young nurses decreased at a faster rate than the female population in general. This indicates that either the nursing education take-up rate decreased, or net migration of young nurses between 1991 and 1996 decreased relative to that of young females in general, or both. In Quebec and Ontario, population demographics account for most of the change in the potential supply of young nurses. An in-depth analysis of the relative contributions of net migration, educational take-up rates, and changes in population demographics to the potential supply of nurses is an interesting research topic but is left for further research. One notes, however, that although this thesis does not explore any of these individual relationships directly, their net effect is captured by examining trends in the potential supply of nurses. This thesis now turns to explaining why the health care labour force participation rate of young individuals decreased so dramatically over the hospital cut-backs period. Of principal interest is whether the observed decrease was demand driven or supply driven. This is the topic of the next section. 3.3.2 A Model of the Occupational Sector Decision One of the earliest analyses of sectoral choice is found in Roy (1951). According to the Roy model individuals choose to work in the sector where they earn the highest income. An individual's income differs by sector due to the assumption that individuals are endowed with various levels of sector-specific skills. The model assumes that skills are exogenously given and that income is the only variable affecting the choice of sector. 2 1 In 1999 5,729 students were admitted into nursing programs in Canada while 12,000 applications were received (CIHI, 2001c). This suggests that any recent changes in the take-up rate may be driven primarily 48 Ln(1996) - Ln(1991) > FJ TJ O CD Ln(1996) -Ln(1991) o o o Co Ko —k o o o o o o o o Ko co jk cn cn o 3 " &> 3 CQ 0 ° 1 3 I s o ? o 5 E ET CD — =*. CO o c 3 TS — "O 03 t< 5 o GO (D 3! -A • 3 3" Ln(1996)-Ln(1991) Ln(1996) - Ln(1991) p o o p o o o o o o O 3 " 0 ) 3 CQ CD 5' co N' CO ~ 9. TJ O CO 3 0) Q> r-t- _ 10 ' c TJ •D < O z e —I cn CD GO 49 The model further assumes that every individual desiring a job in a given sector is able to fmd a job in that sector. Heckman and Sedlacek (1985, 1990) make several modifications to the Roy model. One key modification is to base the sectoral decision on utility maximization rather than income maximization and to allow for sector-specific non-wage job attributes. According to their model, an individual chooses to work in the industry in which his utility is maximized. Similar to Roy (1951), Heckman and Sedlacek assume that there are no constraints on the availability of jobs. These two models provide the background to the sectoral choice (health care sector versus non-health care sector) model developed in this section. Following Heckman and Sedlacek (1985, 1990), utility rather than income is the basis upon which individuals evaluate jobs. However, the model used in this section does not assume that there is an unlimited supply of jobs in each sector. The idea of constraints on the supply of jobs is taken from the 'union-queuing' literature of the 1980s. For example, Farber (1983) and Abowd and Farber (1982) describe models where there is some portion of individuals who prefer union sector jobs but there is only a limited supply of these jobs. Applying this idea to the nursing labour market, there may be some individuals who prefer jobs in the health care sector but employers (e.g. hospitals) are not willing to hire these individuals. These individuals then take up a job in the non-health care sector even though it entails a lower level of utility. One recalls that price rigidity is a necessary condition of such models. If wages were flexible, then any excess demand for jobs in one sector would be eliminated and there would be no queues. The model in this section is developed to help answer the question: Which empirical observations are consistent with the hypothesis that during the cut-backs period the by restrictions on the number of spaces available. 50 decrease in the health care labour force participation rate was demand driven, and which are consistent with the hypothesis that it was supply driven? (i) The Model There are two sectors in the economy: the health care (nursing) sector (HC) and the non-health care (non-nursing) sector (NHC). Each job is summarized by two characteristics: a wage (W) and working conditions (WC). The working conditions variable in the health care sector is meant to capture job aspects such as patient workload, length and type of shift and stress levels in hospitals. Preferences over wages and working conditions are characterized by a direct utility function U(W, WC) where U\ > 0 and U2 < 0 2 2 . Individual * has the choice of working in either the health care sector or the non-health care sector. It is assumed that individuals first choose whether to work, then choose the sector they prefer conditional on working23. For now, it is assumed that there are no 'switching' costs associated with inter-sectoral flows. Individual i will prefer to work in the health care sector if lf(Whc,WChc) > U(Wnhc,WChc). Define a function as /, = lf(Whc, WChc) - U(W"hc, WChc) so that HC >' NHC => i , > 0. The partial derivative is arbitrarily set negative so that working conditions are a 'bad'. 2 3 The model is developed to potentially explain the increased likelihood of individuals with an educational background in nursing working in the non-health care sector observed in the Census data. Therefore, the model does not allow for the case when individuals leave the labour force in response to constraints on the availability of nursing jobs. Since the age-specific employment rates, in general, did not change significantly between 1991 and 1996, this simplification in the model is not considered restrictive. 51 Turning to the supply of jobs, at any given time employers of nurses - mainly hospitals -may not wish to hire all individuals who desire a job in the nursing field. If an arbitrary hiring index Ht is defined such that Hi > 0 when employers are willing to hire individual i and Hi < 0 when employers are willing to lay off or not hire individual i, what factors might affect Ht ? In the union membership literature, the relative marginal product of labour determined whether an employer was willing to hire a unionized rather than non-unionized worker given the union to non-union wage differential. That is, ceteris paribus, firms are more likely to hire an additional worker when the worker's marginal product of labour increases. The assumption that the demand for nursing labour increases as the marginal product of labour increases seems inconsistent with hiring behaviour in the nursing sector, both theoretically and empirically. The theoretical relationship between the marginal product of labour and the demand for labour is a result of profit maximizing behaviour on the part of firms which, a priori, is inapplicable to the health care environment in Canada (Evans, 1984). As noted in the introduction, Barer, Stark and Kinnis (1984) provide empirical evidence that the demand for nursing labour is most sensitive to global budget levels, which are expected to be uncorrelated with labour productivity. Taking into consideration these findings, the following labour demand function is consistent with observed employer behaviour: = f(B, W10), where B is the global budget level, W10 is the nursing wage and f\ > 0,/2 < 0. Whenever the current level of employment is below M1, employers are willing to hire more nurses and when the current level is above or at A ^ they are willing to lay off or not hire nurses. 52 Furthermore, in situations when ^ 7V^  it is argued that changes in employer hiring policies are expected to affect new graduates and young nurses the most. This is for two reasons. First, the nursing profession is highly unionized and there is a high degree of job security among senior nurses (i.e. those with many years experience). It is more difficult for employers to lay off experienced nurses compared to recently hired nurses. This is discussed in greater detail in the next section. Second, when employers are willing to hire more nurses, the largest pool from which to draw is recent graduates. Thus, any fluctuations in the demand for nursing labour are expected to affect Ht for younger, less experienced individuals more than older, more experienced nurses. One other factor that may affect the probability of being offered a nursing position is education level. As noted in chapter two, many nursing associations considered making the BScN the minimum education requirement for entry to practice during the period in question. Since registration is a prerequisite to being offered a nursing position, whether an individual holds a BScN might also be an important factor (indirectly) affecting the likelihood of being offered a job in the health care sector. Taking into consideration these nursing labour market characteristics, the H function can be written as Hi = g(B, W^^Edud, Expi) where gu g 3 , g 4 > 0 and g2 < 0. Immediately one notes that the only individual-level variables that affect H are education and experience. That is, for a given level of B and W110 - which vary only by province -the probability that employers are willing to hire an individual is the same for all individuals in a given experience/education group24. One other caveat concerning the H function is the possibility that experience and education enter asymmetrically. That is, g 3 and gA may be positive when dBldt < 0 but close to zero when dBldt > 0. In 53 Note that up to this point the process that generates W1'0, W"hc, WChc, WCnhc and B has not been discussed. For now, these variables are assumed to be exogenous to the model. The probability that individual i works in the health care sector - Pr(HG = 1) - is the joint probability that an individual prefers a health care job and that employers are willing to hire the individual and is written as Pr(HG = 1) = Pr(I, > 0) x Pr(Ht > OJ. The probability that individual i works in the non-health care sector is written as Pr(NHG = 1)= Pr(h > 0) x Pr(Ht < 0) + Pr(h < 0). The first term is the joint probability that an individual prefers a health care job but is not offered one. These individuals are involuntarily employed in the non-health care sector. The second term is the probability that an individual prefers a non-health care job. (It is assumed that there are no constraints on the availability of non-health care jobs). These individuals are voluntarily employed in the non-health care sector. In the special case where there are no constraints on the availability of health care jobs individuals sort into sectors based only on utility maximization. It is easy to see that a change in the likelihood of observing an individual in the health care sector results from a change in either the willingness of the individuals to accept a job in the health care sector, or the willingness of employers to hire that individual, or both. other words, in periods of budgetary reductions, experience may affect the probability of being laid off, but in periods of budgetary expansion experience plays less of a role in determining the probability of being offered a job. 54 (ii) The Canadian Experience During the hospital downsizing period the health care labour force participation rate decreased for individuals under 30 and decreased dramatically for individuals under 25 while remaining constant for older age groups. It must be the case, therefore, that either Pr(Ii > 0) decreased for the young relative to all other age groups, or Pr(H, > 0) decreased, or both. According to preference specifications in the model, Pr(It > 0) can change only as a result of three effects: changes in tastes, changes in relative working conditions, and changes in relative wages. Each of these effects is examined in turn. It is easiest to illustrate the effect of a change in tastes by using a particular utility specification. Suppose that lf(W, WC) = dlnW-(l- d)WC, d e [0,1 ] so that /, = dln(Whc/Wlhc) -(1- d)(WChc - WChc). AW low a •A / high a 55 This specification is a modification on Abowd and Farber (1982) i . For a given set of Wxc, W"hc, WChc and WCnhc, individual variation in a will result in some individuals preferring health care jobs and some preferring non-health care jobs. The a parameter measures the relative importance of wages versus working conditions to an individual. Indifference curves for this particular functional form are depicted in the above diagram. The curves correspond to /, = 0 for a high-a and low-oc individual with AW = Whc - W"hc and AWC = WChc - WChc. If everyone faces the same wages and working conditions (represented, for example, by point A), individuals with a low a will prefer to work in the health care sector while individuals with a high a will prefer to work in the non-health care sector. One hypothesis potentially explaining the decrease in the health care labour force participation rate for the younger age groups is that these individuals experienced a change in preferences in the form of a decrease, on average, in a. It is difficult to imagine why preferences would change among this group and not for individuals over 30. This 'preference-based hypothesis', therefore, seems implausible. Another factor affecting Pr(Ij > 0) is the nursing wage premium (AW). If individuals face the same set of working conditions in both sectors and have identical preferences, then individuals with a higher nursing wage premium will be more likely to prefer jobs in the health care sector. A second hypothesis potentially explaining the decrease in the health care labour force participation rate for the younger age groups is that these individuals experienced a decrease in the nursing wage premium relative to all other age groups. This is referred to as the 'wage-premium' hypothesis. Finally, the working conditions premium also affects Pr(7, > 0). The 'working conditions-premium' hypothesis is analogous to the wage-premium hypothesis. If the 2 5 Alternatively one may use an additive utility function of the following form lf(W, WC) = InW'- WC + ci in which case the indifference curves would be parallel. 56 youngest age groups experienced an increase in AWC relative to older age groups, then a higher proportion of these individuals would prefer non-health care jobs, ceteris paribus. Up to this point it was assumed that there were no costs associated with switching occupational sectors. If the model is extended to include switching costs, this allows for a fourth hypothesis that explains why a higher proportion of young individuals may prefer non-health care jobs. Suppose the switching cost (in utils) associated with an inter-sectoral move for individual i is Ttj. Individuals employed in the health care sector will prefer to switch to the non-health care sector if and only if + a, < 0. If switching costs increase with age, then even if wages and working conditions change in the same way for everyone, only the young might prefer non-health care jobs. This would imply that the health care labour force participation decision is much more elastic for younger individuals. Whether switching costs are expected to increase with age depends on how they are defined. If they include such things as loss of accrued pension benefits, vacation time and, specific to the nursing profession, benefit hours26, then ni will increase with age as these benefits tend to be low for nurses with little or no seniority. Individuals who have spent significant time out of nursing must typically complete re-certification exams, pay registration fees and pass a probationary period before being allowed to practice. These barriers to entry to the nursing sector lead to another potential switching cost. If an individual is contemplating leaving the health care sector but is considering returning at some time in the future, then this individual might discount the future cost of re-registering into a present-day cost associated with leaving the health care sector today. It is unclear, however, how this component of the switching cost varies by age. As part o f their benefits RNs are entitled to a certain number of hours each year for which they receive pay but do not actually work. 57 Each of these four hypotheses attribute the decrease in the health care labour force participation rate to a decreased preference for health care jobs. The discussion now turns to a demand-side explanation - the 'hiring freeze' hypothesis. According to the model of nursing labour demand, one expects the large-scale expenditure reductions in the hospital sector of the early 1990s to have led to a sharp decrease in hospitals' willingness to hire nurses. Furthermore, nurses with seniority are less likely to face a reduction in Pr(Hi > 0) so that during the cut-backs period, any change in hospitals' willingness to hire is expected to affect most significantly those individuals with little or no nursing experience. The 'hiring-freeze' hypothesis states that the reduction in the health care labour force participation rate for the younger age groups during the downsizing period was driven by a reduction in labour demand and was, therefore, involuntary. (iii) Exogenous Variables and Equilibrium in the Model It is useful to refer to Figure 6 in the following discussion. In a world of perfect price flexibility, whenever there is involuntary non-health care employment (i.e. an economic surplus of nurses) wages in the health care sector would fall, bringing about three effects. First, employers would be able to hire more nurses. Second, some individuals currently employed in the health care sector would switch to a non-health care job. Third, some individuals working in the non-health care sector but preferring the health care sector would be able to find a health care job. In aggregate, employment in the health sector would increase and employment in the non-health care sector would decrease until all involuntary non-nursing employment is removed and equilibrium in the nursing labour market is restored. However, nursing wages are extremely rigid. They are set through collective bargaining between provincial health employers and nursing unions. The collective bargaining agreements typically last two to four years during which nominal wages are fixed. 58 (000,000s) cn o c —I o p. O 01 3 01 Q . 0)' 3 5~ CO c CD cT I CD 01 3 —ti O ro CM CO o ro b b b b b b b o o o o o o o o o o o o o o o o 3 CD 01 m x 73 CD 3 a. i-+ c -i CD i o z (000,000s) io ro co co 3 SL X CD 01 m x 73 CD 3 Q. ,-*• c CD • cn o ro co (000,000s) cn c co co b b O-O 71 | O 0 o 3 3' 01 CD 0) m x 73 CD 3 Cl r i ' c CD • TJ D (000,000s) _->• ro ro b "cn "0 "cn 0 0 0 0 0 0 0 0 Z o 3 3' 5L I CD 01 m x 73 CD 3 Q. p+ C CD > 00 3 p i o 3 59 c •s HE CD CD 3 —H o —I 3 CD o' &> if) c « « 9 —k to CO 4k cn CD oo CO o o o o o o o o o o o o o o o o o o o o o o 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 73 CD cu_ "0 CD —t O 0) •o I - * CO m x T3 CD 3 CL e —c CD o 3 X o (A •g SL cn a: o co •fc ST I CD 3 8: 5 60 Moreover, factors completely exogenous to the nursing labour market have an extremely important impact on the negotiated nursing wages. The relative strength of government and nursing negotiators, the economic environment in the province at the time of negotiation, and even the timing of elections27 all have an important effect on the government's willingness to cede wage increases. For these reasons it is unlikely that nursing wages are flexible to a degree that they equilibrate the nursing labour market. Non-health care wages are assumed to be exogenous to the nursing labour market. This assumption is reasonable given that flows between the health care and non-health care sectors are small compared to the size of the latter28. This (and the fact that nursing wages are rigid) implies that the health-care wage premium is exogenous; a decrease in the premium could induce movement out of the health care sector, but movement out of the nursing sector can not affect the premium. If global budgets are endogenous they could potentially adjust until the nursing labour market is in equilibrium. However, the financing structure of the Canadian health care system suggests that hospital budgets are, in large part, exogenous to the nursing labour market. Hospital budgets are set by provincial governments and depend greatly on total expenditure levels which, in rum, vary with economic conditions, the willingness of governments to run deficits, and the level of transfer payments from the federal government. The fiscal priorities of provincial governments - i.e. the relative priority of education, social assistance, health care, infrastructure, tax reductions etc. - also determine the size of hospital budgets. Al l of these factors are exogenous to the nursing labour market. Of course, the health care needs of the population affect budgetary decisions as well. In addition, campaigns by nursing associations to increase hospital funding levels during a perceived economic shortage of nurses have, on occasion, been successful. But during the period of study, reductions in hospital budgets reflected broader deficit reduction initiatives and were not motivated, for example, by reduced The typically negative publicity associated with nursing strikes is more important to governments in election years versus non-election years. 2 8 In addition, one must assume that labour demand in the non-health care sector is not so inelastic that even negligible shifts in labour supply cause significant wage changes. 61 health care needs of the population . As a result, changes in global budget levels during the period of study can be viewed as completely exogenous to the health care sector. Working conditions are, in part, endogenous. If, for example, total patient volume does not change, then a reduction in nursing staff levels, ceteris paribus, implies an increase in per-nurse workload - a worsening of working conditions. This, in turn, leads more individuals to leave the nursing sector thereby decreasing nursing staff levels further. Unless patient volume adjusts, workload in the non-health care sector changes, relative wages change, or a combination of these factors occurs, the positive feedback loop will continue until employment in the health care sector falls to zero. Only by asserting that factors in addition to per-nurse patient workload affect working conditions is this feedback loop potentially avoided. Therefore, the assumption that Whc, W"hc, WChc, WChc and B are exogenous to the nursing labour market, particularly during the period of study, seems very reasonable. Figure 6 illustrates that equilibrium could potentially be restored if graduation levels from nursing programs and net migration are endogenous and respond to nursing labour market conditions. However, nursing school budgets, the demand for diplomas/degrees in nursing and immigration policies are determined in part by factors exogenous to the nursing labour market. Education budgets are set by provincial governments and depend on economic conditions, fiscal policy, and federal transfer payment levels. The number of applicants to nursing programs also depends on exogenous factors such as population demographics and aggregate educational take-up rates. Of course, in the event of an economic shortage of nurses, for example, university-aged individual's might be more likely to choose nursing as their educational field and governments might increase funding to nursing education programs. But given that it takes two to four years to attain an education in nursing, such responses are long term and are unlikely to restore equilibrium in the nursing labour market. It is impossible for the dramatic variation across provinces in expenditure patterns over the period of study (Figures 16 and 17) to be explained by differences in the pattern of health care needs. 62 Immigration policy is set at the federal government level. While based on several factors exogenous to the nursing labour market, in principle, the point system can be altered to make it easier for nurses from abroad to immigrate to Canada if, for example, there is an economic shortage of nurses. However, past experience suggests that in practice immigration policy is not particularly sensitive to nursing labour market conditions (Mclnnes, 2000). Furthermore, registrations requirements set by provincial nursing associations sometimes make it difficult for individuals trained abroad to register as nurses (Steinhauer, 2001). Thus, immigration flows are unlikely to restore equilibrium in the nursing labour market. Therefore, since the factors affecting the demand for and the supply of nursing labour are largely exogenous to the nursing labour market, one does not expect the nursing labour market to be in equilibrium. 3.3.3 Empirical Evidence — Testing the Hypotheses In the previous section, several hypotheses were identified as potentially explaining the decrease in the health care labour force participation for the youngest age groups over the hospital downsizing period. In this section, data are examined to determine which hypotheses - if any - are consistent with the empirical evidence. Working Conditions Premium Hypothesis In the occupational sector choice model, the working conditions variable is meant to capture non-wage job attributes such as patient volume per.nurse, length and type of shift work, stress levels associated with particular nursing positions, availability of support services, workplace autonomy, and the performance of non-nursing tasks30. Other than patient volume per nurse, most of these factors tend to be subjective. As a result, data measuring these variables is almost always based on surveys of working nurses. For 3 0 Aiken et al. (2001), for example, provides a comprehensive list of such workplace factors. 63 example, a recent and comprehensive study by Aiken et al. (2001) shows that 36% of a sample of nurses surveyed in Canadian hospitals between 1998 and 1999 showed signs of 'emotional exhaustion' at work and 17% planned to leave their current job within a year. Based on results from other questions, the authors conclude that, in general, nurses reported increased dissatisfaction with working conditions in hospitals in Canada. However, nurses were asked to compare current working conditions to those of a year ago. This short time frame provides little insight into the trend in working conditions over the period of interest for this thesis. Regarding patient volume per nurse (a measure of workload), in addition to subjective assessments (Aiken et al, 2001), some administrative data are available. O'Brien-Pallas et al. (2001) find that in Ontario, nursing hours per patient day remained relatively stable from 1994 to 1998 while the average Resource Intensity Weight - a measure of case severity - of hospital separations increased. This suggests that acuity-adjusted workload per nurse increased, contributing to the worsening of working conditions. Other than the O'Brien-Pallas et al. study, no time series data are available on patient volume per nurse. Even if they were available, they would comprise only one component of working conditions. For these reasons, very little can be said concerning the evolution of working conditions over the hospital downsizing period. This data void, however, is not such a crucial drawback. In testing the working conditions hypothesis, it is not so much the change in working conditions that is of concern, but the change in working conditions for young nurses relative to older nurses. According to the model, working conditions in the health care sector would have to worsen for the young relative to the old in order to explain the decrease in the health care labour force participation rate of the young. This scenario can be ruled out by assuming that working conditions in the health care sector did not change differentially by age. Is such an assumption reasonable? 64 If it were the case, for example, that in 1996 younger nurses were more likely to work in stressful, understaffed areas of the hospital compared to 1991, then the answer is no. If in 1996 younger nurses were more likely to work inconvenient or excessively busy shifts compared to 1991, the answer would also be no. However, there is very little information available on whether working conditions changed disproportionately for younger nurses over the downsizing period. It is assumed that when working conditions change they change in the same way for everyone. As a result, the working conditions hypothesis can not explain the decrease in the labour force participation rate for the youngest age groups. Wage Premium Hypothesis The wage premium hypothesis states that an increase in non-health care wages relative to health care wages induced young nurses to leave the health care sector. According to the model, there are three possible wage scenarios that support this hypothesis: 1. If relative working conditions do not change, then AW falls for the young while remaining constant or increasing for all others. 2. If relative working conditions deteriorate (AWC increases), then AW increases for everyone, but it increases least for the young. 3. A W declines in the same way for everyone, but switching costs for the young are lower than for all others. The strategy is to compare health care and non-health wages over time to determine the extent to which the available data support this hypothesis. 65 Table 4 contains information on weekly wages for the potential supply of nurses in Canada who are employed full time and earn at least the minimum wage. The sample is divided into those who are employed in the health care sector and those who are not. The average weekly wage for nurses under 25 was $490 in 1991 and represented a 31.3% premium over the average wage of an individual with a nursing education working in the non-health care sector. The health care wage premium falls rapidly with age. For those 25-29, 30-34, 35-39, 40-44 the premium is 21%, 13%, 9% and 8% respectively. For those under 35 and especially those under 30, the large premium is clearly a reflection of low non-health care wages as opposed to high health care wages. The reasons behind these low non-health care wages are explained in Table 5 which gives an occupational breakdown of the non-health care sector for the under 35 age-group. It is clear that the majority of these jobs are low skilled clerical and sales or service occupations which typically pay low wages. It is important at this point to state a crucial assumption that is implicit in the wage premium hypothesis. According to this hypothesis, non-health care employment is always voluntary. This implies that either working conditions are so bad in the health care sector (i.e. AWC is very large) or wages are of so little importance (i.e. a' is very small) that a 31% health care wage premium is insufficient to induce some young individuals working in the non-health care sector to switch to nursing. Should the reader find this assumption unreasonable, then non-health care employment is by definition involuntary and, a priori, the wage premium hypothesis can be dismissed. To assess the validity of the wage premium hypothesis, one must examine the change in the age-specific wage premium between 1991 and 1996. Table 4 shows that the wage premium increased quite significantly for those 25 and older. For those under 25, however, it decreased from 31% in 1991 to 24% in 1996. These data, therefore, could support version #1 of the wage premium hypothesis. 66 CO < ( 5T CD 3 CD Q . 77 3- S CD CD =! 1 0 O <1> o + CO ST Q . 0) CD CD 7£ •< ^ g CD 03 O CD 4> C J C J ro o cn o cn 03 CD J > 4> => CO <; 3" co w CO ST CD CO J > 3 CO g S" CD ro CO 3 CO g M CD 3 CD 3 CD 3 CD 3 CD 3 CD Q . Q . 7 T Q . 7 T Q . 7 T Q . 7T 03 C?" 03 0?" - l 03 v< 03 — i 03 .<" ° - s ° - N 3- S ° - s 3- S CD 03 CD 03 CD 03 CD 03 CD 03 3 CO 3 CO 3 CO - i CO - i CO o w -1 o CD — i o n> —t O CD — i o CD A cn > CO CD Cr CD" a: CD 03 92 ro o C D • —j 00 - « l • Zl 00 CJ) C » cn cn 4> ro - cn 2 M C O S3 C D P cn U 0 1 4 v FO V N J CO •<> • •ij 00 00 00 03 C D 00 C D ~ , Cn 22 ro c» 23 a, cn \ ) N 3 0 - N | £ 2 cn 4^ ro N J 4 ^ p fe 0 0 C O — J 00 cn eg ro N J « s S — i . O N | cn cn w cn ro o ^ ^ C D O 4> C D 00 Ko 0 3 C D N ) ~ cn oo ro cn o, - g 4^. ro cn CD S C D P CO 4^ cn ro cn cn CT> _•. cn N J O -± cn a> 0 1 ° o o C D ro o CO co CD ^ 23 S C D 0° O o cn 4 i . CO N l CO C D y ho CD O ZL CD 03 ° 03 CD 3 o CD o o o c ra CD 5' 3 o 03 3 Q. =» O I CD o 03 s IQ CD </> O 3" Q. 5r CO o c -1 o CD o CD 3 cn c C/3 O O 03 3 0) CL 03 _k " N l _ L N 3 ->• S s ^ CO N l N J J 9? cn ro £ -M co cn cn 0 0 C D CO k, CO oo ro co O -M Cg w a , 2 8 bo w C D co ro 0 3 cn co oo CO 4 ^ N j CD CO N J 00 ° 0 1 w CO C D cn w cn co oo • o cn cn . N J C D cn CO 0 3 ° ro C D 'ro « U O ) N l o ro ro cn cn co CD cn § i i £ c, - g CD O bo N> CD cn N | ~, cn oo co cn § ro N CD CD O cn W -NI J > U ^ o cn ^ N | 00 00 . CO CO o ro o CD CT3 cn co o 8 2 £ CD * ° 4> -^1 N l 00 j> cn ro 00 M cn .-^  ro oo ro CD 4^  g U1 N O ~ J ^ o co ro 4 ^ CO 4> N l . oo cn C D N | CO O CO j> cn C D N | to CO X CD 03 O 01 O DJ 3-03 3 I O &> s* o 3 3" c 3* cf — i CD 3 O CD CD 03 T l CD" 67 o 01 o 3 I X C D M _ p + 3 " o o fil N J .—. O ) t o -° s CO o CO o CO o c o CD O CD 3 Cn c cn O O 03 3 01 C L 0 ) CO co 5; e n CD -P. t o o So H § | [1 -a ^ CD 0) K. cn o o CD cn cn 3 ' CQ 3 3 c 5T o I f s= cy) 3 T J CD m 3 O C L " O CD 3. X I 3 £. 0) T J -2 3 _ . CD 3 3 C L - » <= O cn ~o (—t-o —t cn CD CD CD CO CO O 9L 3-0) 0) 3 — CL CD CD CD CD N O) 01 A CD 3 § CD 5. q o o o o o o c c T J T J S- 21 5' 5' 3 3 C/l cn o o o 3-O CD T J 0) 0) CD CD O J N J O TJ 0) CD cn 3-CD ^ " 9>. ° CO — . = : CD CD ->• o •> CO CD. CD cn CD "n rn CO o o OJ o o ( D e c CO T J T J m CU 0) =! a-. ~ cn cn cn CD -O 3 3 Q-O O O o S 3 o cn ai T J T J 0) T J S3". O 3 CO 0) 3 Q. CD cn 0) 3 C L 2L 0) CD C L o o o cr. T J CD-S' 3 cn cn 5. " ° O o CD =r T J § 2-CD CD oT a 3 < Q. CD 0) cn O CD cn o o 3 3 o Q. CD 5' 3 5' O 3 CQ CD 3 Q. 0 ) 3 i . 9 cn Q3_ CD cn TJ CD —1 cn O 3 cn g CD cn 0) 01 CQ CD cn CD i o CD CO CD CD cn o_ CD 0) 3 C L —x CD sf. CQ 3 3' CT C >< CD 3 CL cn CD 2 O CD C/l C7. T J CD i . cn O 3 0) CO 8 O 0)_ CD CD 3 3 CO CD CD O -CD JD 0) g-O 0 J o CD CD CD cn_ ET CD CO pj_ CD CO cn TJ CD g 0)' o 3 OJ 3 Q . -6 o a ? 3 3 CD 3 co CD i o CD OJ 3 C L —i CD_ eg' o' 3 NJ CO NJ cn N J co 0 0 0 0 CO - k N J 4> N J o -ft- N J cn cn o cn cn o cn N J C D o cn cn C D 0 0 0 0 CO -Pk 0 0 cn cn cn © CO o> cn N J CO -Pk CD © cn o 0 0 C D "-Pk V l CO o o N J N J cn cn co ~vj CD 0 0 CD -vl CJl o cn o cn cn cn 4^ - k 0 0 CD o cn cn CD C D N J 0 0 cn cn o cn CD © o cn co CD - J -cn o CD o o 0 c T J DJ r-t-o ' 3 CO c n c o c o N J cn O z 03 0 J S kD C L cn OJ N T J L L T J CD — - 1 CD O' Q_ 03 03 > 03 cn o O CD o 3 c O T J CD 03 cn a-. 03 3 C L —I cp_ 03 .—»-CD C L O O O tz T J 03 ,—. o' 3 cn CD 03 -Pk O J O > CD % 3 . 3 0 3 ' 03 =!. — C/3 cn c T J CD 2. cn O < CD 01 3 Q. C" - . CD CQ g_ 03_ o •3 o o o s= T J CD 5' 3 c/l CD CD NJ - k CO TI £ 5' q a, s- 8 5' e> " I 3 ' cn cz —1 03 3 O CD 03 Q. CD <P O 03 3 03 CQ CD 3 CD 3 TJ 3 3, 0) cn CD cn O O O c T J Q) 03 CD cn < CD O O O S3 T J 03 t-o' 3 cn 8 3 T § = G) (/) § s cn ? 3 CD cr O c o cn o cn 0) cn c * 03 O 3 3 Q. CO 3' 03 3 O CD -pk o N J -Pk ^Pk cn —k 0 0 - k C D cn co 4>-o o o cn o cn co o CO N J cn N J © cn CD —k cn co cn —* •pk NJ "vi co cn CD cn o o cn cn o co cn cn CD - k © OJ © cn 01 o s CO CO cn 5 5' «5 CT CD Ol 00 0) >r a o 1 o o 3 I C D 0) O CD o o o c "O CD £«. o" 3 CO o* t 3" 9. Q: c CD Co" 8-CO Ol OJ 2 CQ CD CD 3 S c o CD jj. O" 3 68 However, when one examines the source of the decrease in the wage premium one is forced to re-examine such a conclusion. The wage premium for the under 25 group decreased, in part, as a result of a fall in health care wages. Given that when nursing wages change they change in the same direction for all experience levels - a characteristic of the collective bargaining process - it is very puzzling that health care wages decreased for the youngest age group but increased for all other age groups. Since LPN wages are lower than those of RNs one explanation might be that the ratio of RNs to LPNs decreased for the under 25 group relative to all other ages between 1991 and 1996. (Recall that the sample of health care workers in Table 4 includes RNs, LPNs and nursing aides). To test whether this is the case Table 4 was replicated restricting the sample of health care workers to RNs only. The result was unchanged: RN wages decreased for the under 25 age-group ($526 versus $507) while increasing for all other ages (results not shown). Nurses working outside the hospital setting are not covered by collective bargaining agreements and, as a result, tend to have lower wages than their hospital counterparts31. It might be the case that the proportion of nurses working in hospitals in the health care sector decreased for the under 25 group relative to all other ages between 1991 and 1996. Unfortunately, it is not possible to test this hypothesis with the data from the Registered Nurses Database that was provided for the purposes of this thesis. If such a change did occur, one could think of two interpretations. The movement toward nursing jobs outside hospitals could have been voluntarily. That is, working conditions in hospitals deteriorated and individuals were willing to accept lower paying nursing jobs outside hospitals. According to this interpretation, the observed wage pattern supports the working conditions premium hypothesis. Such an interpretation, however, can not explain why older individuals - presumably facing similar working conditions - were not also willing to accept lower wages to work in nursing positions outside the hospital setting. The alternative interpretation is that the movement toward nursing jobs outside hospitals on the part of the youngest individuals was involuntary and arose because there 3 1 Personal communication, Canadian Nurses Association. 69 were constraints on the availability of nursing positions within hospitals. Not able to secure hospital employment, younger individuals accepted lower paying nursing jobs outside hospitals. Finally, the classification of employment into part-time and full-time is not very precise in the Census data. It could be the case that part-time employment increased disproportionately for the under 25 group between 1991 and 1996 and that this change was not captured in the Census. The decrease in the health care wage premium among the youngest individuals during the downsizing period observed in the Census data is difficult to accept and, at this point, remains a puzzle. Nevertheless, these data can be interpreted as supporting the wage premium hypothesis. But once again, implicit in this line of reasoning is the assumption that despite a 24% health care wage premium, non-health care employment is voluntary. Accounting for Selection Bias The data in Table 4 must be interpreted carefully. What is of interest in assessing the validity of the wage premium hypothesis is the change in the average health care wage premium for a given age group. This measure - AWj* - is defined as AWj* = 1/nj It (Wihc- W,nhc) V i e age group j,j = {<25, 25-29...} where rij is the number of individuals in age group j. To calculate AWj* directly from the Census data one would need to know the health care wage of every individual working in the non-health care sector and the non-health care wage of every individual working in the health care sector. But one observes either the health care wage or the non-health care wage for every individual but never both. The measure of the health care wage premium that appears in Table 4, therefore, is written as AWj = l/n,j % (Wihc) - l/n2J 2J (WPhc) V / e age group j,j = {<25, 25-29...} 70 where «/ individuals are observed in the health care sector, n2 are observed in the non-health care sector and «/ = Zpij and h2 = Zfliy. If - a s implicitly assumed in the wage premium hypothesis - selection into occupational sectors is not random, AW, is a biased measure of AWj*. To show this formally one can rewrite the wage premium measure used in Table 4 as AWj = l/nu It (W-c I /,• > Oj - l/n2J 2J (W"hc, \ I, < 0) Vie age group j, j={<25, 25-29...} For a given a' and AWQ this reduces to AWj = l/n,j Ei (Wthc I Wihc > Winhc) - l/n2j 2} (W^ \ W?c < W,nhc) or AWj = 1/njj 2J (W?0 \ AW(* > 0) - l/n2J 2J (W"hc \ AWt* < 0). Now suppose the data generating process for wages is represented by Whc = Po + pXi + e W"hc = S0 + 5X2 + T] where Xi and X2 are vectors of variables affecting health care wages and non-health care wages, respectively. Then, W* = (Po- 80) + PXH -5X2i + (Si - n) which implies A Wi* > 0 <=> - [(Po - 80) + PXn -SX2i\ < £i - Vi 71 AW* < 0 <=> - [(B0 - 50) + BX,,-dX2i] > £t - n, Two sources of selection bias associated with the AWj measure now become clear. Individuals observed working in the health care sector will, on average, have higher values of variables in X; that have a positive B associated with them compared to individuals observed in the non-health care sector32. In this case, the average health care wage for a given age group in Table 4 does not represent the average health care wage available to individuals in the same age group who are working in the non-health care sector. This type of bias is based on variation in observable characteristics across occupational sectors. Individuals with a high (s, - 77,) are also more likely to be observed in the health care sector. Again, this implies that the average health care wage for a given age group in Table 4 does not represent the health care wage available to individuals in the same age group who are working in the non-health care sector. This type of bias results from variation in unobservable characteristics across occupational sectors. Only when there are no differences across sectors in the returns to these unobservable characteristics or these unobservable characteristics are distributed equally across sectors is this type of selection bias not a concern. To correct for selection bias based on observables the strategy is to estimate the following wage equations for 1991 and 1996 and then to construct age-specific wage premia using the regression coefficients: (1) In W!IC = B0 + frAGEi + B2EDUG + fiPROVi + st, for all health care workers If A^andX2 contain the same variables then the statement should read "...a positive (/} - 5) associated with them..." 72 (2) InWi =50 + 8iAGEi + 52EDUG + S3PROVi + 84EDNRSNG + 77,, for all non-health care workers?'5 Equation (1) is estimated using the Census Master File sample of all individuals working in nursing occupations. In equation (1), age and education were included mainly because collective bargaining agreements specify different wages for different experience levels34. These two variables also capture the fact that nurses with higher levels of education are also more likely to hold supervisory positions which, as specified in collective agreements, carry higher wages. Equation (2) is estimated using the Census PUMF sample of all non-health care workers. It is not restricted to only non-health care workers with an educational background in nursing since this sample is too small to produce robust estimates. However, an EDNRSNG variable (which equals 1 for individuals with an educational background in nursing) was included to account for the fact that individuals with an educational background in nursing may systematically earn different wages than individuals whose educational background is in some other field3 5. Age and education were included in equation (2) to capture life-cycle patterns in earnings. The province variable was included in both equations to allow for regional variation in wage levels. Collective agreements are set at the provincial level in the health care sector and provincial fixed effects in wage levels in Canada are well documented. A l l variable definitions and means are given in Tables 6 and 7. Since over 95% of nurses are female, the non-health care workforce was restricted to only females in the equation estimation. Furthermore, the Census P U M F data were used to estimate the non-health care wage equation since the sample size was sufficiently large. 3 4 The alternative approach is to construct an experience variable that embeds education and age but the data did not contain accurate information on years of schooling. 3 5 One might argue that a nursing education produces sector-specific human capital that is valued less in the non-health care sector compared to an education in business or history, for example. Others argue that an education in nursing produces mainly 'general' human capital and, accordingly, non-health care wages should not differ between those educated in nursing and those educated in some other field. 73 I I CD m m o 30 CO CD O O O O 33 O • N CD O O 5 O m D CD o m D c 5 m o o o i— i— m o O Oi O O O D C 4^ 4x co co > g cn p rj 4 \ CO CO 4> Co 4i. 33 Co O — co' 2:1 ro 11 11 5" N oi S CQ CD CD CD Cn CO CO CD s CD CD cn o - i pr CD O -CD Q . c o CU 0' 3 II CD . 3 3 ^ 5 ' S 3 §_ a a c CD_ CD_ _ Q- Q- ~ cn CD II CD O _ < o S < CD o N < CD Q . CD 5' _. O g" 0) 3 3 O CO CD Q . — CO cn > 11 cr ->• CD —. O < 3' o CD CD C/J Q-' CD 3 O CD ^ O o < O 11 3 _>. 5T =^  ~" T3 3 O CD CD cn CL CD 3 O CD O 11 C _ i CD _ . CT - * CD t 3 ° 3 < 3' o CD CD cn CL CD 3 O CD c r II CO _ . O 3" =^  5- 3" ° CQ" cn 3" o CD CD £ <£; Q . CD CD CD CO —1 CD CD 3" CD Q . CZ 3 <" CD II II =; CD =v CO ? CD Z CQ n tQ 3" ~ 3" CD 32. CD co o cn d 3 CD CO =r 11 11 11 11 11 11 11 CO . . _ 1 . _ 1 . _ 1 . _ 1 . _ i _ > . CL CD = K =^ =h = ; ^ a - * ^ u c o i o _ —• CO 4^ CO 4^ CD CQ A A A A O . CD CQ — CD CD CD. Q . CD cn Q . CD CQ —1 CD CD 3" cp_ Q_ DO 01 01 CO CQ CO CO CO CD CD CD CD A CO CO CD A CO cn CD A 00 O CD O 3 4^ Jso "cn 00 4^ a> 00 0 0 O 0 O ho 03 1_ _ v 1_ '__ O cn 00 4^ 0 cn —V CD 00 N l ro 4^ 0 CO cn cn 0 to 0 L_ 0 cn M O 00 CD ro cn 0 O O O 0 0 0 0 0 O 2 ho O 4^ ro . L_ 1_ !_. 1_ CO CD N j N l CO ro —>. O 4^ cn N j —1. CO O 4^ CD — cn cn CD CO O - — 3 SL CD c/i GO o C CD S o CD ro • • " N l O CD 3 cn c cn O 01 3 01 Q_ 01 Ol ai co co P N J CO ro o CO O 4^ O O O CO cn co o o o N l co N j 0 0 O O 0 0 0 0 O 0 0 O O O <: 0 4^ 1. _ v 0 cn '__ L_ 1_ L_ ^ CD ro CO O CO ro CO CO 00 00 cn N j CD N l 01 cn 4^ N J w N J 00 CD CO CD CO ro 3 101 __ | M T i CD |__. 3 IQ _L CD T i cn ° - § o S c 1 co cn CD — T l -»'_? Ocn CD cn 3 N | cn c cn O CO 3 01 Q . 0) TJ C CD CJI 00 0 0 0 0 0 O O O 0 O O O O O O O O 2 O 0 0 L_ 0 CO fo ro 0 4^ ro L_ O 1- Li- '__ CO CD cn 0 0 4^ . CD 00 CD CD N j 0 ro O CD 4^ en N | CO 01 0 CD 0 0 cn —^ cn CD — v 4^ O CD N | 4^ CD — v —^ cn CO 3 o ro b J> o o ro co O 3" CD —1 T l CD CD cn 74 CD CO m m I CD rn m D 70 co CD O O O O TJ 70 O CD o J> CD o TJ o 8 Oj o m m m o D O c o ~H § o 5" i— j> r— O m D D O D C •*\ -t\ Co co > (J O Ul " s -Ck -C* Co CO -ti CO -tk To co O •P* co co CO -Pk II 3 =£ 3 < 1 0 7n CD CD —• CD CO 3 3" II CD II CD . 3 , CD S CD C Q SE. CQ - 7 T CD C L S O o CD C L cn CD C D 00 P - s i o CD O CO s 1= i i CD. CD Q . :/: CD 3 7 3 CD Q . C o CD 5' 3 cn „ 3" CD O CO CO o •< CD CL T J —\ O < 3' o CD O O O O CD II o ->. T J O < 3' o CD O > 1 1 CT -k CD _ . CD TJ O 11 3 _k CD =Z H O o < 3 ' o CD O ;• TJ —1 o < 3' o CD O 11 C _ i CD _ . CT CD TJ 0 3 < CT II CD _k O 3" = 2. 3-° C Q ' - cn 3-5' o CD o CD  Q . CD CD SO-< ° Q . C Q -I —1 —1 —1 —1 —1 CD CD CD CD CD CD CO cn cn CO CO CD C L CL C L C L C L 3" CD CD CD CD CD CD 3 3 3 3 3 O O 0 O O Q . CD CD CD CD CD cn' cn C/l cn CO cn D J O O 0 0 O O !_k !_k 0 C O N J C O -pk N J C O C O C O 0 C O Co C D —k C D N J n 11 q- 11 11 11 11 11 11 11 g__kCD _ k _ k _ k _ k _ k _ k _ k c f - S - - - - - - - -m - T a - T 4 - J k U « M 0 ) CQ" " O ' C D ' CO -pk CO -pk CO CQ ~ 3" ^ =T A A A A A 0 1 5 fS CD CD CD CD CD ' - f CQ CQ CQ CQ CQ Q_ CD CD CD CD CD CD CQ CD "3 CD CO o cn C"- CD-CD CD fl) C Q C Q CD A O J O CD_ CL O CD CD zr CD C L Co' DJ O *k CD O cn o A co cn C Q CD CD 3" cp_ CL cn' CD O O O O O O O CD -Pk - J O co cn cn —1 co co cn 00 co cn o co o - J N J ~ - J CO o C -Pk 3 - 0 CD N J CD 3 cn c cn O ~*\ O CD 3 CD CD-CD 00 C D O C D P° -k N J cn co O CD O CO o 00 CD - J -k O O O O -si cn cn oo NJ O 00 N J •pk O 00 00 -si CO cn o O O O O O O O CD b i o J CO O ^ (fl CD CO -k J N J o c o o - s j c n o o o I0J £ I CO CD IS. 3 IQ. SL CD CO CD °- § 1 o 92 c 1 ^ cE O w CD -si 3 CO cn c o CD 3 CD Q -CD TJ c CD 0 0 cn co cn -pv N J NJ O cn o o N J C O 0 0 0 O O 0 0 O 0 0 0 0 0 p 0 O 0 2 0 b l_k Lk b C O N J C O b 4> Ik Lk Li Lk Lk Lk N J CD 0 0 -Pk N J C O Co •Pk —k C D C O 00 N J N J cn C D 00 -Pk CD 0 0 C D - J -s| C O —k C O -si -si -pk C O -s| 0 —k cn 00 3 TI 75 Tables 8 and 9 present the OLS estimates of equations (1) and (2). The coefficient signs on the age and education variables are all as expected; higher education brings about higher earnings as does increased age. Individuals with an education in nursing working in the non-health care field earned, on average, 3.8% less than those with an education in some other field in 1991. In 1996, however, this difference disappeared. The 3.8% wage difference in 1991 is accounted for entirely by the fact that individuals with an education in nursing who worked in the non-health care field were more likely to be employed in low-paying occupations. When equation (2) is re-estimated adding a series of occupation dummies, the coefficient on the EDNRSNG variable is insignificant (results not shown). To test the wage premium hypothesis, one may use the coefficients from the four regressions to construct age-earnings profiles in the health care and non-health care sectors for 1991 and 1996. Health care wages may be calculated by adding the coefficient on the appropriate age dummy variable to the constant. Non-health care wages may be calculated in a similar way but the coefficient on the EDNRSNG variable must be added in 1991. One may then calculate the change in the health care wage premium over time. This analysis is summarized in Figure 18. The regression-adjusted health care wage premium in 1991 was 31% for those under 25, 26% for those 25-29 and between 10% and 17% for all others. In 1996, the regression-adjusted premium dropped to 22% for those under 25 but changed only negligibly for all others although the change is statistically significant36/ Since the change in the regression-adjusted wage premium through time is very similar to the change in the unadjusted premium, the conclusions based on Table 4 remain valid. As a final caveat, one notes that the coefficient estimates in Tables 8 and 9 may themselves be biased. Recall that according to the wage-premium hypothesis for every 3 6 The fitted wages in Figure 18 are for an individual with a university below bachelor's education living in a province other than Alberta, B C , Ontario and Quebec. For other levels of education or other provinces, the absolute value of the premium would change but it would change by the same amount for all age groups. In other words the difference-in-difference estimate in Figure 18 is independent of province and education level. 76 Table 8. OLS Wage Equation Estimates for Health Care Sector 1991 1996 Coefficient Std. Error Coefficient Std. Error Constant 6.072 * 0.008 5.987 * 0.010 Aqe Dummies a D2529 0.153 * 0.000 0.263 * 0.011 D3034 0.127 * 0.008 0.336 * 0.011 D3539 0.154 * 0.008 0.357 * 0.010 D4044 0.190 * 0.008 0.403 * 0.010 D4549 0.206 * 0.009 0.427 * 0.010 D50+ 0.220 * 0.009 0.436 * 0.010 Education Dummies b EDTRADE -0.153 * 0.005 -0.216 * 0.005 ED UNIV 0.108 * 0.006 0.102 * 0.007 EDBACH 0.164 * 0.006 0.169 * 0.005 Province Dummies c PQ 0.037 * 0.006 0.052 * 0.005 ON 0.026 * 0.005 0.088 * 0.005 AB 0.046 * 0.007 -0.017 * 0.007 BC 0.043 * 0.007 0.125 * 0.006 Dependent Variable: Ln(W) Sample size d : 53,584 55,101 R-squared: 0.060 0.106 Notes: * significant at 5% level a ommitted category is "under 25" b ommitted category is "college degree" c ommitted category is "all other provinces" d sample is all employed individuals working full time, earning above minimum wage, working in the health care sector, whose major field of study is nursing 77 Table 9. OLS Wage Equation Estimates for Non-Health Care Sector 1991 1996 Coefficient Std. Error Coefficient Std. Error Constant 5.926 * 0.007 5.927 * 0.007 Aqe Dummies 3 D3034 0.186 * 0.007 0.236 * 0.007 D3539 0.240 * 0.007 0.321 * 0.007 D4044 0.269 * 0.007 0.329 * 0.007 D4549 0.274 * 0.008 0.359 * 0.008 D50+ 0.232 * 0.008 0.340 * 0.008 Education Dummies b EDTRADE -0.113 * 0.006 -0.122 * 0.006 EDUNIV 0.113 * 0.009 0.112 * 0.009 EDBACH 0.264 * 0.005 0.269 * 0.005 Province Dummies c PQ 0.072 * 0.007 0.098 * 0.007 ON 0.160 * 0.007 0.184 * 0.007 AB 0.078 * 0.009 0.080 * 0.009 BC 0.074 * 0.009 0.143 * 0.008 Field of Education Dummy (Nursing = 1) EDNRSNG -0.038 * 0.011 0.002 0.341 Dependent Variable: Ln(W) Sample size d : 38,143 40,357 R-squared: 0.233 0.183 Notes: * significant at 5% level ** significant at 10% level a ommitted category is "under 30" b ommitted category is "college degree" c ommitted category is "all other provinces" d sample is all employed individuals working full time, earning above minimum wage 78 Figure 18. Regression Adjusted Health Care Wage Premium by Age 1991 Ln(W) 1996 Ln(W) Age Group Health Care Non-Health Care Difference Health Care Non-Health Care Difference Difference-in-Difference Repression Adjusted (From Tables 9 and 10) <25 6.07 5.76 0.31 5.99 5.77 0 22 -0.09 25-29 6.23 5.96 0.26 6.25 6.01 0 24 -0.02 30-34 6.20 6.08 0.12 6.32 6.17 0 16 0.03 35-39 6.23 6.13 0.10 6.34 6.25 0 09 0.00 40-44 6.26 6.16 0,10 6.39 6.26 0 13 0.03 45-49 6.28 6.16 0.12 6.41 6.29 0 12 0.01 50+ 6.29 6.12 0.17 6.42 6.27 0 15 -0.02 Unadjusted (From Table 5) <25 0.31 0 23 -0.08 25-29 0.21 0 26 0.05 30-34 0.13 0 21 0.07 35-39 0.09 0 18 0.09 40-44 0.08 0 16 0.08 45-49 0.09 0 14 0.05 50+ 0.11 0 14 0.03 Regression Adjusted Wages by Age 6.50 -i : 5.70 4— 1 1 : 1 1 1 1 <25 25-29 30-34 35-39 40-44 45-49 50+ •a-Health Care 1991 - » - N o n - H e a l t h Care 1991 - * - Health Care 1996 -A-Non-Hea l th Care 1996 Notes: Ln(wage) and differences are those of a university below bachelor's graduate, in a province other than Alberta, BC, Ontario and Quebec. 79 individual working in the health care sector AWj* > 0. This means that the expected value of the dependent variable in equation (1) is E(lnWt \AWi*>0) = Po + pXu + Efa \AWi*>0) where Xn = [AGEi ,EDUQ .PROV, ]. This can be rewritten as E(lnWt | AWt* > 0) = p0 + pXn + E{si | - [(po - S0) + pXn -SX2i] < £i-T?i}. From expression (3) it is clear that the error term and the regressors are correlated implying OLS estimates of equation (1) are biased. In equation (2), for all non-health care workers whose educational background is in a non-nursing field E(rj) = 0. However, for those whose educational background is in nursing, the expected value of the error term can be written as (4) E(TJ) =E(n\ AW*<0)=E{TJ\- [(p0 - S0) + pX, -SX2] > s-77} * 0 and once again OLS estimates will be biased. Assuming that s ~N(0, CTI2), n ~N(0, cr/), Cov(_-, 7) = ai2 and letting Z = - [(p0 - S0) + PX, -SX2], the expected value of the error terms in (3) and (4) can be written as E{rj\Z >-(?•/-£)} = - ( C T 22 - C T 1 2 ) J(Z) a\ + G\ - 2CT1 2 O(Z) E{s Z < s - 7} = x <f>{Z) 12 1 - O ( Z ) 80 where (j> and O represent the pdf and the cdf of the normal distribution, respectively (Maddala, 1983). The terms <Kf)l$>{^) A N D ^ ' V O ^ O ) ] a r e known as Inverse Mills Ratios. Heckman (1979) provides a two-step method to obtain unbiased OLS estimates of equations (1) and (2). The first step is to maximize the following likelihood function where the sample is the potential supply of nurses who are employed: The indicator variable p equals 1 if the individual works in the health care sector and 0 otherwise. Parameter estimates from equation (5) can then be used to fit for every individual in the potential supply of nurses. The second step is to re-estimate equations (1) and (2) but to add the Inverse Mills ratio as an explanatory variable. That is, one estimates the following two equations by OLS: (2a) lnWi"hc = S0 + 8,AGEi + 82EDUCt + 83PROVi + 84EDNRSNG + 85Xt + vt ( 5 ) L = [l-cD(Z)] p[cD(Z)] i-p Z,=-[(J30-S0) + /3XU-SX2I] (la) InWt = Po + Pi AGE j + p2EDUG + p3PROVt + p4X{ + % for all health care workers for all non-health care workers where <Kzt) for all health care workers 1-0(Z,) for all non-health care workers whose educational background is in nursing 0 for all other non-health care workers 81 In such a case, E(Q = E(v) = 0 so that OLS estimates of (la) and (2a) are unbiased. Specific to the data used in this thesis, there is an important identification issue that arises when using the two-stage Heckman approach. Since Z is a combination of the other regressors in equations (la) and (2a), and S; are identified only to the extent that 2.,- is a non-linear function of Z. A much more desirable estimation strategy is to find an instrument variable that affects occupational sector choice but does not affect wages thereby identifying p4 through an exclusion restriction. In the nursing case, the costs of taking a nursing refresher course in order to re-register with a nursing association, for example, might be a suitable instrument. The Census data do not contain any variables that might be deemed suitable instruments. As a result of this identification issue, ex ante, the two-stage Heckman estimation may produce parameter estimates that are unreliable since one of the regressors is a function of all the others. Tables 10 and 11 present the estimates of equations (la) and (2a). Comparing OLS estimates and Heckman-procedure estimates shows that correcting for selection bias has no effect on the wage equation estimates in the non-health care sector (Table 9 compared to Table 11). However, in the health care sector - especially in 1991 - the coefficient estimates from the Heckman procedure are drastically different from those using OLS (Table 8 compared to Table 10). The coefficients on the age variables in Table 10 are all of the 'wrong' sign in 1991. This renders the Heckman results highly suspect. As noted earlier, A, and the regressors are highly correlated in equations (la) and (2a) which makes it difficult to identify the other coefficients separate from the coefficient on X. In fact, the Inverse Mills Ratio was found to be highly negatively correlated with age (results not shown) which might explain the negative coefficients on the age variables. Since the two-step Heckman procedure produces highly suspect results from the health care wage equation estimation, this procedure is viewed as much less reliable than OLS. 82 Table 10. Wage Equation Estimates for Health Care Sector Using Two-Stage Heckman Procedure 1991 1996 Coefficient Std. Error Coefficient Std. Error Constant 7.013 * 0.161 6.182 * 0.177 Aqe Dummies a D3034 D3539 D4044 D4549 D50+ -0.133 * -0.174 * -0.201 * -0.225 * -0.249 * 0.046 0.061 0.075 0.079 0.085 0.143 * 0.186 * 0.248 * 0.293 * 0.299 * 0.027 0.026 0.043 0.065 0.068 Education Dummies b EDTRADE EDUNIV EDBACH -0.305 * 0.088 * 0.109 * 0.033 0.019 0.021 0.085 * 0.146 * 0.027 0.027 Province Dummies c PQ ON AB BC 0.128 * -0.073 * -0.100 * -0.052 ** 0.023 0.032 0.042 0.028 0.066 * 0.178 * 0.069 0.176 * 0.020 0.053 0.061 0.058 Inverse Mills Ratio -3.223 * 0.720 0.850 0.734 Dependent Variable: Ln(W) Sample size d : R-squared: 53,584 0.089 55,101 0.081 Notes: * significant at 5% level ** significant at 10% level a ommitted category is "under 30" b ommitted category is "college degree" c ommitted category is "all other provinces" d sample is all employed individuals working full time, earning above minimum wage, working in the health care sector, whose major field of study is nursing 83 Table i 1. Wage Equation Estimates for Non-Health Care Sector Using Two-Stage Heckman Procedure 1991 . 1996 Coefficient Std. Error Coefficient Std. Error . Constant 5.926 * 0.007 5.927 * 0.007 Age Dummies a D3034 D3539 D4044 D4549 D50+ 0.186 * 0.240 * 0.269 * 0.273 * 0.232 * 0.007 0.007 0.007 0.008 0.008 0.236 * 0.321 * 0.329 * 0.359 * 0.341 * 0.007 0.007 0.007 0.008 0.008 Education Dummies b EDTRADE EDUNIV EDBACH -0.113 * 0.113* 0.264 * 0.006 0.009 0.005 -0.122 * 0.112 * 0.269 * 0.006 0.009 0.005 Province Dummies 0 PQ ON AB BC 0.072 * 0.160 * 0.079 * 0.074 * 0.007 0.007 0.009 0.009 0.098 * 0.184 * 0.080 * 0.143 * 0.007 0.007 0.009 0.008 Field of Education Dummy (Nursing EDNRSNG = 1) -0.056 0.125 0.160 0.147 Inverse Mills Ratio -0.008 0.059 0.O77 0.289 Dependent Variable: Ln(W) Sample size d : R-squared: 38,143 0.233 40,357 0.183 Notes: * significant at 5% level ** significant at 10% level a ommitted category is "under 30" b ommitted category is "college degree" c ommitted category is "all other provinces" d sample is all employed individuals working full time, earning above minimum wage 84 Hiring Freeze Hypothesis The hiring freeze hypothesis is best summarized in the following passage from Aiken et al. (2001): New graduates [in Canada] were for a time unable to find work in hospital settings, and the seniority rights negotiated by nurses' unions caused a high proportion of younger and relatively inexperienced nurses to lose their jobs when hospital staffs were cut (p.46). For this hypothesis to be supported empirically one must observe a positive relationship between the demand for nursing labour and the health care labour force participation rate for younger individuals. That reductions in labour demand primarily affect younger individuals reflects the important role seniority plays in a unionized sector such as nursing. In fact, seniority clauses are directly written into collective bargaining agreements between nursing unions and employers. For example, Article 19.01 of the BC collective bargaining agreement stipulates that In the event of a reduction in the work force, regular employees shall be laid off in reverse order of seniority, provided that there are available employees with greater seniority who are qualified and willing to do the work of the employees laid off. Agreements in other provinces contain similar articles. To test this hypothesis the strategy is to exploit variation across provinces in the size of negative labour demand shocks (i.e. hospital expenditure reductions). Figures 16 and 17 show that not all provinces reduced hospital budgets to the same extent during the downsizing period. Alberta, for example, reduced hospital expenditure by an astonishing 30% between 1992 and 1996. In BC, however, although the growth rate decreased, expenditure actually increased by 4% over the same period. 85 Reductions in health care spending during this period were largely exogenous to the nursing labour market. Fiscal reform in the health care sector reflected broader deficit reduction initiatives and was not motivated by, for example, reduced health care needs of the population. Furthermore, the variation across provinces in hospital expenditure reductions was mainly a result of divergent priorities on the part of provincial governments. For example, a right-wing majority governed in Alberta during this time while in BC a left-wing government was in power. Thus, the difference in health care expenditure trends between Alberta and BC provides a natural experiment that may be used to test the hiring freeze hypothesis. According to the labour demand function frf1 = f(B, W10) the large reduction in health care expenditure in Alberta led to a decrease in demand for nursing labour while the decrease in demand in BC was very small. The hiring freeze hypothesis predicts that the health care labour force participation of the young should decrease most in Alberta and least (if at all) in BC. As Figure 14 shows, this is not the case. In fact, the decrease in the health care labour force participation rate is fairly uniform across provinces and is virtually identical in Alberta and BC. But one can not dismiss the hiring-freeze hypothesis due to two important confounding factors. First, nursing labour demand is a function of nursing wages as well as budget levels and nursing wages in Alberta decreased during the hospital restructuring era while in BC they did not (Figure 19). The health care wage equation estimates in Table 8 confirm this result. The coefficients on the province variables indicate that, out of the four provinces, Alberta had the highest health care wages in 1991 and BC the lowest, while in 1996 Alberta had the lowest and BC (and Ontario) the highest. Since f2 < 0 the wage cut in 86 CO" 3 co CD CO > CO CO o o DJ" o 3 5° c S5 3* I <D Co 3" 5= & &) Q. Co O 87 Alberta may have, either fully but certainly partially, offset the decrease in demand resulting from reductions in hospital budgets. Second, and more significantly, between 1991 and 1996 the potential supply of nurses under 30 decreased substantially in Alberta but only slightly in BC. If the decrease in potential supply in Alberta represents individuals leaving the province because they are unable to find jobs in the health care sector, then the health care labour force participation rate in Alberta ought to be adjusted downward in 1996 to be comparable to the BC rate. This would result in a much larger decrease in the health care labour force participation rate in Alberta compared to British Columbia. An alternative interpretation, however, is that individuals with an education in nursing left Alberta to seek nursing employment in other provinces or countries in response to the sharp decrease in nursing wages imposed 37 by the provincial government in Alberta . According to this interpretation, the potential supply of nurses is endogenous to such a degree that the large reduction in the nursing employment level in Alberta was accompanied by a less-than-expected increase in the level of involuntary non-health care employment. Taking these two factors into consideration, the data are consistent with the predictions of the hiring freeze hypothesis. One might alternatively test the hiring freeze hypothesis by examining nursing vacancy data. According to the hypothesis, at the national level the number of nursing vacancies in hospitals should have decreased over the downsizing period as hospital budgets were reduced. Moreover, one would expect the number of vacancies to have fallen dramatically in Alberta but only slightly, if at all, in British Columbia. In recent years, one expects the number of nursing vacancies to have increased as hospital budgets were restored to their pre-downsizing levels. Unfortunately, data on the number of nursing In principle Census data can be used to track inter-provincial movements between Census years. However, at the time of the data request this type of analysis was unforeseen and the appropriate variables were not requested. Of course, one must still determine where the individuals who left Alberta went. Most provinces were reducing nursing employment levels during this period. 88 vacancies in hospitals are not available at the national level or for BC for the period of study38. As a result, this strategy was not pursued. As a final note, the hiring freeze hypothesis predicts that employment reductions were least severe among nurses with high levels of seniority. If there is a correlation between seniority and specialization - e.g. if you must be a senior nurse in order to work in the ICU or OR - then the hypothesis predicts that hospital layoffs should not be distributed evenly across nursing specialties. Unfortunately, data on nursing employment levels within particular areas of specialization in hospitals that is consistent across the period of study is not available. Therefore, this avenue of research was not pursued. 3 8 Knowledgeable experts at the Canadian Nurses Association indicated that nursing vacancy data are not routinely collected at the national level. In B C , the Health Manpower Research Unit at the University of British Columbia compiled monthly nursing vacancy data on a monthly basis but only up until the late 1980s. 89 3.4 Discussion The hospital cut-backs were associated with significant changes in the nursing labour market in Canada. Nursing employment levels in hospitals decreased during this period and the decrease was accounted for entirely by reduced employment among nurses under 30. There is little evidence to support the claim that the decrease in nursing employment among the youngest age groups was a result of individuals leaving the health care sector for better job opportunities in non-nursing occupations. For young individuals with a nursing education, working as a nurse was associated with a very large wage premium relative to the non-health care sector and it is difficult to imagine a scenario where someone with a nursing education would voluntarily forego such a large wage increase. The alternative claim that reductions in employment over this period reflected a sharp decrease in the demand for nurses - a result of severe hospital budget cuts - is much more consistent with the evidence. The correlation between reductions in hospital expenditure and employment suggests very strongly that the nursing employment level over this period was determined by labour demand. This, in turn, implies that the hospital downsizing period in Canada was associated with a large economic surplus in the labour market for nurses. That the reductions in labour demand primarily affected young nurses can be explained by the important role unionization plays in the nursing labour market. 90 PART II 91 The analysis of the nursing labour market in Canada in Part I produced several important results. Nursing employment levels in hospitals decreased significantly in Ontario, Alberta and Quebec during the hospital restructuring period (1992-1997). In BC the nursing employment level in hospitals remained stable. Nevertheless, this period brought an end to a fifteen-year trend of steady increases in the number of nurses working in hospitals in British Columbia. The reductions in employment during the downsizing period were focussed entirely on young nurses. Employment levels for nurses over the age of 45 actually increased during this period. An analysis of nursing wages relative to wages in other occupations strongly suggests that the employment reductions during the downsizing period were involuntary. Reductions in the demand for nursing labour on the part of hospitals - stemming from sharp reductions in global budgets - resulted in layoffs and reductions in new hiring. Since the nursing profession is highly unionized, less senior staff were much more likely to be affected by hiring freezes and layoffs. Part II of this thesis examines the effect on patients of the reduction in the nursing employment level in Canadian hospitals associated with hospital downsizing. Two avenues of research are pursued. Chapter 4 discusses the impact - both expected and observed - of reduced nursing employment levels in hospitals on the number of hours of nursing care patients receive. Chapter 5 investigates the relationship between the availability of nursing resources in hospitals and access to surgical care in British Columbia. The motivation behind examining surgical care and focussing only on BC is provided at the beginning of chapter five. 92 CHAPTER 4 - NURSING RESOURCES A V A I L A B L E TO PATIENTS To understand the relationship between the nursing employment level and the amount of nursing care patients in hospitals receive one must consider two important points. First, the nursing employment level is a somewhat crude measure of the total hours of care provided by the nursing labour force. Second, during the downsizing period, treatment patterns in hospitals changed in a way that reduced the nursing resource requirements in hospitals. 4.1 The Nursing Employment Level and Hours of Nursing Care The following identity summarizes the relationship between the nursing employment level and total hours of care provided in hospitals: CAREt = N, x (PdHrs/N), x Pt x C, where CAREt = Total hours of nursing care provided in hospitals in period t Nt = Nursing employment level in hospitals in period t (PdHrs/N)i = Average number of hours for which a nurse is paid in period t Pt = Fraction of paid hours that are worked in period t Ct = Fraction of hours worked that are spent providing care in period t In principle, existing administrative data can be used to provide a reasonably accurate measure of CAREt. The Annual Hospital Survey (AHS), established in 1994/5 and administered by CIHI, contains a field where hospitals report the total number of hours worked by nurses. This field corresponds to the quantity N, x (PdHrs/N), x P, in the identity and provides an extremely accurate approximation of CARE,. The predecessor survey - the Annual Return of Health Care Facilities: Part 1/Part 2 (HS1/2) - contains a 93 field where hospitals report the total number of hours for which nurses are paid. The HS1/2 survey was administered by Statistics Canada and was discontinued in 1992/3. The paid hours field corresponds to the quantity Nt x (PdHrs/N), in the identity. Although not as accurate a proxy for hours of care as the AHS measure, paid hours represents a considerable improvement over the employment level. Unfortunately, after examining these data with Statistics Canada and CIHI analysts, it became clear that it is not possible to constmct a time series of either total paid hours or total worked hours for nurses in hospitals that is consistent both across provinces and through time. This is a result of three major flaws in the data. First, the reporting rate for the HS1/2 survey decreased gradually from 1990 to 1993 as the survey was phased out. This is problematic as the period corresponds precisely to the beginning of hospital cut-backs in Canada. There is no way of knowing whether any observed changes in variable levels are due to hospital cut-backs or to a decreasing reporting rate. Second, facilities that submit surveys do not always complete all the required fields. As a result of this under-reporting total nursing hours summed across all reporting facilities can only be interpreted as a lower bound. Attempts to correct for non-reporting and under-reporting, which are summarized in the appendix, do not appear to have succeeded in producing a reliable time series. This suggests that even among facilities that complete all required fields the accuracy of the data are questionable. Third, survey responsibility shifted from Statistics Canada to CIHI. This in itself would not have been a concern had it not been for the timing of the shift. One might question the rationale in discontinuing an existing hospital survey and implementing its replacement precisely at a time of major hospital restmcturing when accurate data are required to study the effects of such a 39 restructuring . As a result of data shortcomings, one must infer the pattern of total hours of nursing care from the pattern of each of the variables on the right-hand-side of the identity. 3 9 Officials at Statistics Canada and CIHI put forth considerable effort trying to improve the data for the purposes of this thesis. This ongoing process occurred over a one-year period and included a week-long visit to the Health Statistics Division of Statistics Canada and several meetings and conference calls with 94 In Part I it was shown that the nursing employment level in hospitals decreased significantly during the downsizing period in Ontario, Alberta and Quebec. It remained stable in British Columbia. The proportion of RNs working M l time has decreased gradually over the past 25 years in Alberta, BC, Ontario and Quebec (Figure 20). However, during the downsizing period the trend appears to have accelerated rapidly in BC while in Alberta and Ontario the full-time rate stabilized40. From 1989 through 1991 about two-thirds of RNs in BC worked full time but by 1994 the proportion was down to half. In Alberta and Ontario by contrast, the full-time rate remained stable at around 55% over this period. Although it would be rash to draw any strong conclusions from data that display such large year-to-year fluctuations, these observations are consistent with a large decrease in full-time status among RNs in BC during the downsizing period that did not occur in the other provinces. Another indication that {PdHrs/N), decreased in BC during the downsizing era emerges from an examination of the Health Labour Adjustment Accord, an agreement between the Health Employers Association of BC and, among other parties, the BC Nurses' Union. Pursuant to the recommendations of the Royal Commission on Health Care and Costs, the BC Ministry of Health in 1993 declared that hospital employment should be reduced by 4,800 FTEs over three years. The purpose of the Health Labour Adjustment Accord is summarized in the following passage: The government and the unions and employers directly affected by [the PTE] reductions believe it is in the public interest to proceed with restructuring in a cooperative way. [The Health Labour Adjustment Accord aims to] reduce the impact of the FTE reductions on the employees in the system (HLAA, 1993; p. 1). CIHI representatives. Unfortunately, the end result simply demonstrated that the current hospital data are insufficiently reliable for use in certain health care services research. 95 -z. c co CD CO o £D CD CT CD CO CD CO CO •£> cn cn CD CD ->l o c n O cn o cn O cn O cn s p - o s P s p P s p v O , c s p 0 S 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 TJ CD - i O CD 3 m 3 T3 O ><• CD Q . O 5' CQ 3 CD 3! C O 3" 3 CD o 1 CD 3 o 3 (Q =0 ST 96 One of the provisions of the Accord was to reduce the work week for nurses working in hospitals from 37.5 to 36 hours beginning July 1, 1993. The Accord also stipulates that reduction in hours and other voluntary arrangements (e.g. early retirement, relocation, etc.) will be part of the FTE reduction. Failing voluntary resolution, positions to be reduced will be identified by the employer. Employees identified by this process will be laid off only if they chose layoff instead of options for continued employment with the same employer in another position, including a temporary position, or they refuse placement into a generally comparable position with another employer in the region (HLAA, 1993; p. 7). This suggests that layoffs were to be a last resort in BC and that reducing hours worked per nurse, possibly from full-time to part-time status, was emphasized as a preferable alternative41. Nurses are paid benefits in the form of hours for which they receive pay but for which they do not actually work. These benefits include paid vacation and paid sick leave. As nurses gain seniority, they are entitled to more of these types of benefits. For example, in 1998 in BC an RN with 1 year of experience was entitled to 20 days of paid vacation, an RN with 10 years of experience was entitled to 26 days vacation while an RN with 29 years of experience was entitled to 45 days (the maximum level). Since the nursing workforce has aged steadily over the past twenty years P, at the aggregate level is expected to decrease during the hospital downsizing period42. 4 0 Data from Quebec are not available between 1983 and 1993. 4 1 Knowledgeable experts at the B C Nurses Union have confirmed that the reduced work week was implemented province wide in 1993. Annual operational review reports from Vancouver General Hospital - B C ' s largest hospital - indicate that the reduced work week was implemented for nurses in 1993 at the hospital. 4 2 The A H S records both paid hours and hours worked for nurses working in hospitals. In Alberta and B C the fraction of paid hours for which nurses worked decreased from 85.2% to 84.2% and from 83.1% to 81.9% respectively during this period. In Ontario the rate remained constant at 86.5%. No data are available for Quebec. However, given the inconsistencies observed in the A H S it is doubtful that these figures are accurate. 97 Aside from caring for patients nurses also perform various administrative and supervisory duties. It is difficult to find accurate data on the fraction of time nurses spend taking care of patients during a typical shift. There are some studies that address this issue. For example, Aiken et al. (2001) show that more than one-third of RNs in Canada reported doing some form of non-nursing work - e.g. housekeeping, transporting patients, ordering ancillary services, delivering food trays - on their last shift. However, no information is available on the proportion of total hours worked spent doing such tasks or whether this proportion has increased through time. More fundamentally, the definition of non-nursing work used by Aiken et al. is different from the definition used in the identity. Transporting patients, performing housekeeping duties and ordering ancillary services are tasks that ought to be considered patient care. Such tasks contribute to patient well-being independent of the fact that perhaps nurses are somewhat overqualified to perform these duties. As a result little can be said concerning trends in C,. Taken together, these factors indicate that total hours of nursing care available to patients in hospitals decreased during the downsizing period but the source of decrease varied by province. In BC the main factor was a large reduction in average hours worked per employed nurse. The employment level remained fairly constant. This contrasts with the experience in Alberta, Ontario and Quebec where average paid hours per employed nurse might have decreased slightly but the employment level decreased dramatically. Although data are limited, the fraction of paid hours for which nurses actually work is likely to have decreased over the downsizing period in all provinces as a result of workforce aging. Such a conclusion implies that in Alberta, Ontario and Quebec the total hours of nursing care available to patients in hospitals decreased at least as fast as the employment level during the hospital downsizing period. In BC, even though the employment level did not change the total hours of nursing care available to patients in hospitals decreased. 98 4.2 The Changing Pattern of Acute Care in Canadian Hospitals The way in which Canadians use acute care services has changed dramatically in the past few decades. The number of inpatient separations per capita has decreased while the number of day surgeries has increased. These effects have roughly balanced each other out so that total separations per capita has remained stable. Lengths of stay have decreased considerably. Taken together, these trends have led to a large reduction in the number of inpatient days per capita. Although clearly established prior to the early 1990s, the downsizing period brought about an acceleration of these trends. For example, at the national level the inpatient separation rate was 146 per 1,000 population in 1982, 144 in 1985, 135 in 1990 and then fell dramatically to 97 in 1999 [Randhawa and Riley (1995); CIHI (2002)]. At the national level, data on the day surgery rate are unreliable. This is due in large part to the inconsistent labeling across provinces of the various types of alternatives to inpatient care. At one end of the spectrum is a service such as a visit to a walk-in clinic located within a hospital. The patient is awake during the entire visit which can be as short as a few minutes. At the other end of the spectrum is a day surgery procedure where the patient walks into the hospital in the morning, is examined by a nurse or physician, has local or general anesthetic administered, is operated on by a surgeon in an operating room, is examined by a physician or nurse, waits in the post-anesthesia recovery area and finally goes home in the evening. Over the past few decades the spectrum of procedures performed in hospitals where the patient does not remain overnight has increased dramatically. The way in which these types of episodes are recorded in administrative databases is not consistent. As a result, each province - and sometimes each hospital - has its own method of recording the volume of these various inpatient alternatives. Therefore, it is useful to examine data from individual provinces. 99 Brownell et al. (1999) show that between 1992 and 1996 the decrease in the inpatient separation rate in Manitoba was offset by the increase in the day surgery rate resulting in a fairly constant overall separation rate. Day surgery rates increased particularly fast over this period moving from 29 per 1,000 population to 37. The Manitoba data also show that these trends were well established prior to but accelerated during the cut-backs period. McGrail et al. (1998) present similar data for British Columbia. Between 1985 and 1995 the inpatient separation rate decreased while the day surgery rate increased so that the combined inpatient plus day surgery rate remained fairly constant43. They do not provide data on outpatient visits. Concerning lengths of stay, at the national level for all acute care patients the average fell from 9 days in 1987 to 7 days in 1995 (Tully and St. Pierre, 1997). CIHI (2002) shows that the average continued to decrease between 1995 and 2000 but at a slower rate. In BC the average length of stay for acute care patients was 8.7 days in 1978, 7.7 days in 1986 and 6.2 in 1995 (McGrail et al., 1998). The combination of fewer inpatients and shorter lengths of stay led to a sharp reduction in the number of inpatient days. Nation wide the rate decreased from 1,700 days per 1,000 population in 1985 to 1440 in 1992 (Randhawa and Riley, 1995). In Manitoba, inpatient days per 1,000 population decreased at a rate of 3.6% per year from 1991 to 1996 (Brownell et al, 1999). This represented the continuation of a pre-cut-backs trend. A similar change occurred in BC where inpatient days per 1,000 population decreased at a rate of 3.7% per year between 1986 and 1996 (McGrail et al, 1998). 4 j The day surgery rate in B C increased from 49 per 1,000 population to 69. The large discrepancy in day surgery rates between Manitoba and B C is explained by the fact that Brownell et al. exclude "outpatient contacts which occurred for reasons other than major day surgery procedures" while McGra i l et al. include minor procedures (e.g. scopes). 100 4.3 Trends in Acute Care and Patient Health The implication for patient health of shorter lengths of stay, a reduction in inpatient care and an increase in day surgeries is highly controversial. One school of thought - the 'nursing viewpoint' - contends that nursing care provided during the later days of a hospital stay is necessary and that shorter lengths of stay adversely affect patient health. A second school of thought - the 'pharmaceutical viewpoint' - contends that the nursing care provided at the end of a stay is not required because substitute care is available. The pharmaceutical industry argues that innovations in drug therapies have allowed patients to recover quicker, reducing the need for nursing care44. A third school of thought - the 'ministry (of health) viewpoint' - contends that the nursing care provided at the end of a hospital stay is largely unnecessary and releasing patients from hospitals earlier in their stay does not adversely affect patient health. It is crucial to test the validity of these competing claims since they have very different policy implications. If shorter lengths of stay and the shift to day surgery resulted in beneficial nursing care not being provided, then the hospital downsizing initiative affected patient health. Policy makers ought to consider whether such a result is an acceptable consequence of cost cutting in the hospital sector. If, however, patient health was largely unaffected by the reduction in nursing resources, the cost-cutting initiative resulted in an improvement in hospital efficiency - the same health outcomes were attained using fewer nursing resources - which, in turn, was achieved by employing fewer young nurses. Policy makers ought to then consider the future impact on the nursing labour market of having fewer young nurses in the current workforce. Advocates of home care also argue that patients require less nursing care in hospitals. However, they argue that patient preferences for home care over hospital care rather than better pharmaceuticals is the main reason. Therefore, this school of thought may also be called the 'home care viewpoint'. 101 Before examining the empirical evidence related to the three viewpoints it is useful to review the theoretical relationship between nursing care, hospital services and patient health. 4.3.1 A Production Model for Hospital Services and Health Two common measures of aggregate hospital output are per capita separations and per capita inpatient days (Feldstein, 1968). The separation rate is important as it is a measure of the actual number of episodes of care provided in hospitals. The inpatient day rate is important as it measures the total number of days of care provided in conjunction with those separations. Production functions for each of these hospital outputs can be written as SEP <f(RN;X) (la) IPD < g(RN;X) (lb) where SEP = the aggregate separation rate IPD = the aggregate inpatient day rate RN = the per capita level of nursing resources available in hospitals X = a vector of the per capita level of all other hospital inputs (e.g. physicians) In such models it is typically assumed that BSEP/dRN > 0, dIPD/dRN > 0 and cfSEP/dRN2 < 0, tflPD/dRN2 < 0. While the assumption of strictly positive marginal products is retained, the assumption of diminishing marginal products is not essential to the analysis. The production functions are illustrated in Figure 21 [panels (a) and (b)]. Al l points on the production possibilities frontier are said to be 'technically efficient': there is no way to re-organize nursing resources to increase either the separation rate or inpatient day rate. Technical efficiency implies that (la) and (lb) hold with strict equality. 102 103 Hospital separations and inpatient days are produced in order to improve or maintain the health status of the population. Production functions relating either separations or inpatient days to the 'health' these hospital services produce can be written as HS < u(SEP; Y) (2a) HS < v(IPD; Y) (2b) where HS Y Health status of the population Vector of variables other than hospital care that affect health Although easy to conceptualize, it is somewhat difficult to quantify HS. Common measures include the mortality rate, life expectancy, quality-adjusted life-years (QALYs) and disability-adjusted life-years (DALYs). The production functions for health are illustrated in Figure 21 [panels (c) and (d)]. It is assumed that d2HS/dSEP2 < 0 and cfHSMPD2 < 0. This implies that hospital services are provided in a particular sequence; services with the greatest benefit to the population (e.g. immunizations) are provided first, those with the next highest benefit are provided second, and so on. It is also assumed that beyond a certain level hospital services do not affect or actually decrease population health. The idea behind this assumption is best summarized by Fisher and Welch (1999): Although medical care has obvious benefits, many assume that more medical care must lead to improved health and well-being. There are theoretical reasons, however, to believe that additional growth will be associated with progressively smaller returns. The law of diminishing returns also suggests that at some point additional growth will yield no benefit (the 'flat of the curve'). And while the debate about where we sit on the curve is far from settled the theory suggests that 104 there is some point at which additional growth might actually produce harm (p.446). Health care interventions are often quite damaging to the body. Most surgery, for example, requires some form of incision and anesthetic, both of which harm the body temporarily. In order for a procedure (e.g. appendectomy) to increase patient health the benefit associated with the procedure (e.g. avoiding death, disability) must outweigh the cost (e.g. cutting the abdomen open). When the volume of hospital services is large enough it might signal, as one physician puts it, "we begin operating on healthy people and do more harm than good". A l l points on the production possibilities frontier for health are said to be 'clinically efficient': there is no way of reorganizing hospital services to increase population health. This implies that (2a) and (2b) hold with strict equality. 4.3.2 Casting the Hypotheses Within the Production Model, Examining the Empirical Evidence The nursing, pharmaceutical and ministry viewpoints each correspond with a different set of assumptions concerning technical and clinical efficiency. Nursing representatives argue that the reduction in nursing resources during the cut-backs led to a reduction in separations and inpatient days which, in turn, led to a reduction in population health. In other words, the system prior to the cut-backs was both technically and clinically efficient and was operating to the left of the flat-of-the-curve in Figure 21 4 5. According to this viewpoint the cut-backs corresponded with a move from A to B (Figure 21, all panels). The pharmaceutical viewpoint also implicitly assumes technical and clinical efficiency prior to the cut-backs. However, any reduction in separations and inpatient days did not lead to a reduction in population health because better drug therapies were developed that 105 acted as substitutes for hospital care. According to this viewpoint the cut-backs were associated with an improvement in health technology and corresponded with a move from A to B [panels (a) and (b)] and A to C [panels (c) and (d)]. The ministry viewpoint implicitly assumes that there was technical inefficiency prior to the cut-backs. It also assumes that there was clinical inefficiency46. According to this viewpoint, reductions in hospital budgets brought about an improvement in both technical and clinical efficiency corresponding with a move from D to A [panels (a) and (d)]. With respect to the production of separations the evidence indicates that hospitals were technically inefficient prior to the cut-backs era. The separation rate remained fairly constant in the face of large-scale reductions in the availability of nursing resources in hospitals indicating an improvement in technical efficiency47. With respect to the production of inpatient days, one can not draw the same conclusion since the inpatient day rate decreased dramatically during the cut-backs. There is a well-developed literature addressing the relationship between aggregate hospital utilization and patient health in Canada and the United States. Overwhelmingly, results from the studies suggest that the cut-backs era most likely brought about an improvement in clinical efficiency. One group of studies examines variation in hospital use across geographic regions with similar population characteristics [e.g. Wennberg, Freeman and Culp (1987); Wennberg et al. (1989); Fisher et al. (1994); Wright et al. (1999); Fisher and Welch (1999); Fisher et al. (2000); CIHI (2000c); CIHI (2002);]. A 4 5 Assuming technical and clinical efficiency one can combine equations (1) and (2) and write health status as an implicit function of nursing resources: HS =f(u(RN; X)). S i n c e / > 0 and « ' >0 this implies that 8HS/dRN> 0, consistent with the nursing viewpoint. 4 6 It might also be the case that hospitals were clinically efficient but operating on the flat of the curve. Although theoretically distinct concepts, clinical inefficiency is indistinguishable from flat-of-the-curve medicine empirically. Moreover, even i f it were, the distinction is not that meaningful. In both situations care is being provided that has no benefit to patients. 4 7 Some might argue that hospitals were technically efficient prior to the cut-backs but an improvement in technology occurred during the cut-backs - e.g. less invasive surgery, better drug therapies - that shifted the production function for separations upward. Such an argument, however, ignores a large volume of research showing that the technology required to substitute day surgery for inpatient surgery, although unused, has been available in Canada for decades [Barabas (1977); Evans and Robinson (1973); Robinson and Clark (1980); Evans (1980a, 1980b)]. 106 second group examines variation in hospital use across time [e.g. Brownell et al. (1999)]. A third group examines variation in hospital use across payment system [e.g. Brook et al. (1983); Siu et al. (1986); Pipel et al. (1992); Bernstein et al. (1993)]48. The consistency of study results is remarkable: among populations with similar health care needs, increased aggregate use of hospital services (both separations and length of stay) is not associated with any improvement in mortality. One study is particularly relevant to the Canadian context. Brownell et al. (1999) examine the effect of hospital cut-backs on health outcomes in the Winnipeg area. Between 1991 and 1996 the inpatient day rate decreased from 1,473 per 1,000 population to 1,208. Lengths of stay decreased by 14% for surgical inpatients but did not change significantly for medical inpatients. The inpatient separation rate decreased from 99 per 1,000 population to 86 and the day surgery rate increased from 35 per 1,000 population to 45 so that the overall separation rate remained fairly constant. The authors examine mortality rates within 30, 60 and 90 days of discharge from hospital from 1989 to 1996 for patients admitted for acute myocardial infarction, hip fracture, and cancer surgery. For all methods and conditions, the authors find no significant increase in mortality after the hospital cut-backs. In fact, for both hip fracture and cancer surgery, significant decreases in mortality rates were observed49. The authors also examine readmission rates for patients treated for 13 medical, surgical and obstetric conditions to determine if patients with longer stays were less likely to be readmitted to hospital. No relationship between length of stay and readmission was evident. 4 8 A fourth group might include studies that review hospital charts to judge what proportion of patients in hospital could have been treated in some alternative setting based on their medical needs. Two examples of such studies are Wright and Cardiff (1998) and Cardiff, Wright andKilshaw (1997). 4 9 One must keep in mind that statistically significant differences in mortality do not always translate into clinically significant differences. A statistically significant difference of a few days life expectancy is not likely to be considered very meaningful to an individual. 107 The authors go on to conclude that results from their study - the only one of its kind in Canada - appear to contradict reports by both health care professionals and patients about the stress on the hospital system;[...]Our findings indicate that there are apparently few, if any, long-term health effects or access to services effects of hospital bed closures, despite what may be real short-term tensions for those adjusting to fewer beds in the system (p. 49). Taken together, the available evidence clearly indicates that when health status is measured by mortality and hospital re-admission rates, hospitals were not clinically efficient prior to the cut-backs. Moreover, results from Manitoba suggest that fiscal restraint in the hospital sector brought about an improvement in clinical efficiency. This evidence, therefore, supports the ministry viewpoint. However, when one moves beyond mortality and hospital readmission rates and examines other measures of health status the evidence is mixed. Needleman et al. (2002) examine the relationship between the hours of nursing care provided and the incidence of 14 in-hospital adverse health outcomes believed to be sensitive to nursing care in a sample of US hospitals. Although they find no association between hours of nursing care and mortality - consistent with the previous discussion - they find a significant relationship among 6 other outcomes: length of stay50, the rates of urinary tract infections, upper gastrointestinal bleeding, hospital-acquired pneumonia, shock or cardiac arrest, and failure to rescue (the death of a patient with one of five life threatening complications - pneumonia, shock or cardiac arrest, upper gastrointestinal bleeding, sepsis, or deep venous thrombosis). Hospitals with higher levels of RN hours per inpatient day (and a higher proportion of total nursing hours provided by RNs) had lower rates of adverse outcomes. Interestingly, among surgical patients only 2 of the 14 5 0 The positive relationship between length of stay and R N hours per inpatient day must be interpreted carefully. Shorter lengths of stay are expected to increase nursing hours per inpatient day (see Section 2.5) so that the direction of causality between these two variables is difficult to determine. 108 adverse outcomes were sensitive to the availability of nursing care . Nevertheless, this extensive study demonstrates that fewer nursing resources in hospitals can have a negative impact on patient health when certain outcomes beyond mortality are examined. McGillis Hall et al. (2001) carry out a similar analysis for a sample of Canadian hospitals. However, they focus on the relationship between the ratio of RNs and LPNs to 52 other (unregulated) nursing staff and patient outcomes and do not examine the effect of total nursing hours. This study - although widely cited in the current nursing shortage debate - is not very relevant to the discussion of clinical efficiency53. One aspect of the hospital experience that has received little attention in the literature is the 'psychic' dimension. If patients value not only the health outcomes associated with hospital care but also value the time a nurse spends talking to them or sitting by their side, then the health outcomes examined in the literature may be too narrowly defined. As a result, if one were to include the psychic dimension of hospital care as a health outcome, one might find that hospitals were clinically efficient prior to the downsizing era. The fact that the existing evidence focuses on mortality and morbidity is not a severe drawback for two reasons. First, very little is known on the extent to which patients value the psychic dimensions of hospital care. One might argue that it is important for patients to feel 'cared for' and not to feel that their hospital care is rushed. However, one could just as easily argue that a hospital is a very unpleasant environment and patients would much rather be sent home than spend an extra day or two under the care of a nurse. In fact, recent evidence suggests that patient satisfaction levels in regions with high rates of utilization of hospital care are not above those in regions with low rates (Fisher et al., 5 1 The authors do not appear to have an explanation for the different result for surgical patients. They suggest that either surgical patients are healthier or that the smaller number of surgical patients rendered the relationship undetectable. 5 2 They also examine the effect on nurse outcomes. 5 3 Although a higher proportion of regulated staff resulted in better patient outcomes (e.g. ability to perform daily functions, pain, general health, vitality) at the time of discharge, there was no effect at 6 weeks past discharge. This indicates that having a higher proportion of regulated nursing staff is not associated with any long term benefits to patients. 109 2003). Second, the debate surrounding the health effects of hospital downsizing actually focussed on mortality and morbidity, not on the psychic aspects of hospital care. Returning to the available evidence, although certain to alarm the general reader the fact that hospital care is sometimes provided that is of little or no apparent benefit to patients ought to come as no surprise to an economist. The forces that drive the production processes of firms to the efficient frontier are conspicuously absent in hospitals and, more generally, in the health care sector. A hospital is not a single decision-making entity whose only objective is to maximize patient health subject to some set of constraints. Rather, it is a setting where various groups with differing objectives interact with each other in a mixture of co-operation and competition. The characteristic 'two lines of authority' in hospitals - administration and medical staff — and the problems created by conflicts between them can be represented as two separate firms-within-a-firm, in which the administration assembles inputs and produces services which are then supplied to physicians, who demand and direct such services on behalf ofpatients (Evans, 1984; p. 173). In turn, there is no single objective associated with each of the firms-within-a-firm. Neither physicians nor administrators can be characterized as simple profit or income maximizers [Rice (1998); Evans (1984); Reinhardt (1972, 1973, 1975)]. Several empirical studies have shown that physicians do not simply maximize patient health but rather some combination of patient health, personal income, leisure, prestige etc. 5 4 As a result, the apparent utility function of a hospital could include a variety of interests, including high-quality care, fancy machinery for diagnosis, subsidized obstetrical service, pleasant working environment and high wages for the nurses, total For examples of some of these studies see Evans (1984), Rice (1998). 110 numbers of patients served, the quality of the meals, or any endless set of things (Phelps, 1997: p.272). 4.4 Trends in Acute Care and Total Nursing Requirements in Hospitals Section 4.3 described three trends in the hospital sector during the downsizing period. The following diagrams are useful in discussing the implications of these trends for total nursing requirements in hospitals. Days Since Admission They illustrate the relationship between the hours of nursing care a patient actually receives and time since admission. The upper graph depicts care profiles for two types of 111 inpatient episodes (A and B) as well as a profile that is the weighted average of the two (C) 5 5. The lower graph depicts the average care profile taken across all day surgeries (D). The inpatient profiles are drawn skewed to capture the fact that care peaks shortly after admission and then, as the patient recovers, less and less care is provided56. The day surgery profile is represented by a vertical column since all care is provided in a single day. The area under curve C represents the average hours of nursing care provided during a typical inpatient episode. The area under curve D represents the average hours of nursing care provided during a typical day surgery. Of course, every patient will have his own care profile depending on the type of episode and, for a given episode, his age, sex, comorbidity level etc. The profiles in the graphs represent an average taken across all patients. Total nursing requirements in hospitals can be calculated as Area Under Curve C x Number of Inpatient Separations + Area Under Curve D x Number of Day Surgeries With fewer inpatient separations and more day surgeries total nursing requirements are expected to decrease. Day surgeries, on average, use far fewer nursing resources than inpatient procedures. For example, Evans and Robinson (1977) find that day surgery cases in BC require, on average, one third the nursing resources of comparable inpatient cases. Such a large discrepancy implies that the reduction in nursing requirements resulting from this trend ought to be extremely large. The weight for each procedure would correspond to the proportion of all inpatient separations that are accounted for by that procedure. 5 6 The profdes represent what actually happens during a hospital stay. They do not necessarily represent the 'medically necessary' level of nursing care at various points in a hospital stay. 112 When lengths of stay are reduced the nursing care normally provided at the end of a stay is eliminated. This is represented by truncating the right tails of profiles A and B and also implies a reduction in total hours of nursing care required in hospitals. As a counter argument, it is often claimed that technological advances are allowing sicker patients (with very high nursing requirements) to be admitted to hospitals. Furthermore this 'technology effect' is allegedly so large that it outweighs the effect of fewer inpatients and shorter lengths of stay and acts to increase the total hours of nursing care that ought to be provided in hospitals. In the current framework, the technology effect is represented by an upward shift in profiles A and B. However, such a claim amounts to no more than an assertion since the anecdotal evidence indicates that new technologies can both decrease and increase the nursing requirements in hospitals57. Thus, fewer inpatient separations, more day surgeries and shorter lengths of stay act to decrease total hours of nursing care required in hospitals58. 4.5 Trends in Acute Care and Average Nursing Requirements in Hospitals The dramatic shift to day surgery, ceteris paribus, acts to increase average hours per inpatient separation. This reflects the fact that as more and more relatively simple surgeries (i.e. type A) move out of the inpatient setting and are performed on a day basis, 5 / For example, self-administered analgesic reduces the need for nurses to monitor patients. Pulse-oxymetry equipment is operated by nurses and increases nursing resources requirements. 5 8 Nursing advocates tend not to employ this line of reasoning. Instead they argue that total nursing requirements in hospitals have increased because hospital inpatients, on average, require more nursing care. Undoubtedly, this line of reasoning reflects the fact that nurses on hospital wards no longer see 'easy' patients since these patients are treated in day care units. Nevertheless such logic is fundamentally flawed. While it is true that the average inpatient requires more nursing care, this is because there are fewer inpatients with low requirements and not because there are more inpatients with high requirements. Nursing requirements o f the average inpatient have increased but total nursing requirements have decreased. In the diagram, profde C shifts upward because there are fewer cases of type A not because there are more of type B . 113 the average inpatient surgery becomes more complex. In the current framework, this is represented by an upward shift in profile C but no change in profiles A and B 5 9 . Shorter lengths of stay, ceteris paribus, decrease average hours per inpatient separation as the care provided in the later days of the stay is eliminated. The effect of the shift to day surgery is expected to outweigh greatly the effect of shorter lengths of stay. As a result, one expects average hours per inpatient separation to increase. Shorter lengths of stay and the shift to day surgery both act to increase average hours per inpatient day. Figures 22, 23 and 24 show average nursing paid hours per inpatient separation, outpatient visit, day surgery, and inpatient day for BC, Alberta, Ontario and Quebec. The data are from the HS1/2 survey administered by Statistics Canada (1976 - 1994) and the AHS survey (1994 - 2000) administered by CIHI. As discussed earlier, it is not possible to construct a time series for these variables that is consistent across the period of study. For example, the reduction in paid hours per acute care separation observed in all four provinces from 1992 to 1996 (Figure 22) is consistent with the preceding discussion but the magnitude of the decrease in Ontario and BC is quite inconceivable. The increase in paid hours per inpatient day within each of the surveys in BC, Alberta and Ontario is consistent with the preceding discussion. However, the drop in the level across the two surveys in BC and Ontario is not plausible. Even within each survey the year-to-year fluctuations in many of these variables are extremely large and cast doubt on the accuracy of the data. The average profile (C) shifts upward as a result of an increase in the relative weight of Profile B in the calculation of the average. 114 115 GO o CO D | CO •< O l CO o i CO <—*-o •5 II T| CQ' C c3 NJ Co i ? c 3 •-«„ tQ a: o Tj CD o 0) Co c <a CO 0) 3 Q. O c •5-CD" 3 cn" ? 3 ' o CD — CO 116 C a :> c 3 5" CQ 5 117 In some provinces for some periods these data provide some information on the reduction in total hours of nursing care relative to patient volumes. However, the data are comparable neither across province nor across time. At this point little can be said concerning the extent to which reductions in nursing employment levels affected the average amount of nursing care available to patients in hospitals in Canada. 118 C H A P T E R 5 - THE A V A I L A B I L I T Y OF NURSING RESOURCES AND T H E PATTERN OF SURGICAL CARE IN BRITISH C O L U M B I A The late 1990s in BC saw highly publicized, persistent claims that important surgeries were being cancelled or unreasonably delayed due to a shortage of nurses in hospitals [Lee (2000a), Fong (2000), Fayerman (2000), Wigod (1999), Steffenhagen (1998)]. It is alleged that constraints on nursing resources led to a reduction in surgical volumes through two mechanisms. First, a lack of operating room nurses forced the closure of several operating rooms and greatly diminished surgical capacity in BC hospitals. Second, a lack of hospital ward nurses forced the closure of beds (particularly ICU beds) forcing the cancellation of surgeries that require an inpatient stay. The goal of this chapter is to test the validity of these claims. Data from the BC Linked Health Database are used to determine what effect, if any, the reduction in the availability of nursing resources in BC hospitals had on the surgical case mix. It is important to distinguish between these two hypotheses as they have very different policy implications. If constraints on operating room capacity led to a reduction in the volume of surgeries then an appropriate policy response is to increase the number of operating room nurses. However, if recovery bed capacity is the binding constraint then increasing the number of staffed hospital beds - by increasing the number of ICU or ward nurses - is an appropriate policy response. In light of the results in Part I one might question the motivation behind choosing BC to study the relationship between the availability of nursing resources and surgical volumes. BC was the only province where nursing employment levels did not decline during the cut-backs era. One might instead be inclined to study a province like Alberta where nursing employment levels dropped quite considerably. There are two main reasons to justify studying British Columbia. First, as discussed in the previous section, although the employment level in BC hospitals remained fairly 119 constant total hours of nursing care decreased during the downsizing period. Second, although there is no shortage of rhetoric, no research has directly examined changing patterns of a wide variety of surgical procedures over the past decade in BC and how these patterns might have been affected by nursing resource constraints. A third reason is purely practical. Through a carefully worked out access protocol, the BC Linked Health Database provides graduate students with free access to aggregate tabulations based on individual hospital records60. 5.1 Defining Terms Before continuing with the analysis it is important to clarify some terminology. For the purposes of this thesis, the term 'procedure' denotes a specific type of surgery. Gall bladder surgery, total hip replacement and C-section are examples of procedures. The term 'case' denotes a single instance of a particular procedure. The term 'day surgery' refers to any case where the patient does not stay overnight while 'inpatient surgery' refers to any case where he does. The term 'surgery' represents all types of surgery combined. For example, 'surgery rate' represents the fraction of the population that receives any type of surgery. 5.2 The Aggregate Surgery Rate Recall the following identity describing the 'production' of nursing care available to patients: CARE, = N, x (PdHrs/N), x Pt x C, The individual records themselves remain in a secured computer database and can not be accessed due to privacy/confidentiality concerns. 120 Focussing only on surgical cases the following identity describes the 'distribution' of this care: SCARE, = DS, x NCjs,, + INP, x NQnpJ where SCARE, = Total hours of nursing care per capita available to surgery patients in year t DS, = Number of day surgeries per capita in year t NCds,, = Average hours of nursing care per day surgery in year t INP, = Number of inpatient surgeries per capita in year t NCinpj = Average hours of nursing care per inpatient surgery in year t In the previous chapter it was argued that total hours of nursing care in hospitals decreased during the downsizing period. When measured on a per capita basis - as required in the identify - this is certainly true61. Therefore, during the downsizing era either the inpatient surgery rate decreased, or the day surgery rate decreased, or average hours of nursing care per inpatient surgery decreased, or average hours of nursing care per day surgery decreased or any combination of these effects occurred. Figure 25 [panel (a)] shows the aggregate surgery rate in BC from 1985/6 to 1998/962. The inpatient surgery rate decreased during the downsizing era from 15.6 per 1,000 population in 1992/3 to 11.9 in 1998/963. This rate of decrease (4% per annum) was 6 1 The per capita nursing employment level was discussed in Part I. 6 2 The aggregate measure is composed of twelve 'marker' procedures that account for most of the total surgical volume in British Columbia. The only notable omission is cardiac surgery. At the time of the data request, only procedures for which major technological advancements have not occurred in the past 15 years and for which the inclusion criteria are clear were included. The plan for Part II was slightly different at that time and cardiac surgery was excluded as it violates both of these conditions. 6 3 A l l surgery rates have been adjusted to the age structure of the 1998/9 population. The surgery rate for each procedure in each year for each age group (age groups are divided into 10 year intervals) is first calculated. The age-specific surgery rate for each procedure in a given year is then multiplied by the age-specific population in the reference year. Summing across all age groups yields the age-adjusted surgery rate for each procedure in that particular year. Since the proportions of the B C population in the older age 121 GO o c 3 CD 03 O _ 7 T CD Q . I CD CO • CO (—t-co CT C0 cn CD C D ( D ( D C O C D ( D C D C D C O C O C D 0 0 C X ) C » C 0 S 0 ) 0 1 ^ C 0 M - » O ( D C B N 0 ) 0 1 C D ( D ( 0 C D C D _ C D _ C D C D O 3 C O C O C O C D C D N I O I O I A O O M - ' O C D C O N O ) —>• r o r o CO 4^ cn c n CD CD CD -Nl 00 CO 00 CD c n CO CO __ r o cn CO CO 00 CD CD CD NJ c n w r o o CD r o Nl cn CO 4^ CO —k 00 00 00 N | CD CD c n c n 4^ CO r o NJ _>. __ CD 4^ NO b c n CO r o bo CD 00 '"Nl NO co o cn CD N I r o ^Nj CD NJ NJ 4^ CO ho 4*. NJ O r o NJ NJ r o o r o NJ NJ r o r o o r o o r o o r o o CD cn 4^ b co b r o cn CD oo CO oo r o oo CD r o c n CO cn N| CO CD bo c n c n CD bo cn C O C O C D C D C D C O oo -N| C D cn co CO CD CD CD CD CD c o c o N I c n c n j i CD CD CD 00 00 M - i O CD CO CO CO CD CD O0 c o r o —* o co r o N J C J G 0 4 ^ 4 ^ c n c n c n c n -NI 4 - -NJ CD CD c n CO C J CO —»• CD CO ro N J - i _k -k b b bo c n cn 00 -NI N J Nl -k C D ->• 4^ - ->• 4^ O l N J J i O N l N | ( J ) CD 00 CD CO c o 4 i . 4 ^ c n c n c n c n c D C D c n oo cn cn —k ro N I o cn oo o N | CD N J N J CO —i - J - cn N J cn o oo CQ' c 122 larger than in the preceding period (2.5% per annum) although the difference is not statistically significant. The day surgery rate increased from 5.7 per 1,000 population in 1992/3 to 8.6 in 1998/9. The rate of increase during the downsizing era (8% per annum) was considerably lower than in the previous period (31% per annum). The decrease in the number of inpatient surgeries was exactly offset by the increase in the number of day surgeries so that the overall surgery rate remained quite stable over the downsizing era. There were 21.2 surgeries per 1,000 population in 1992/3 compared to 20.5 in 1998/9. This appears to indicate that access to surgical care was not jeopardized during the cut-backs period and, given the reduction in the availability of nursing resources, is consistent with an improvement in technical efficiency with respect to the production of surgical separations. However, excluding cataract surgery from the total produces quite a different result. During the downsizing period the increase in the day surgery rate did not completely offset the decrease in the inpatient surgery rate so that the total surgery rate declined significantly after remaining stable in the preceding period [Figure 25, panel (b)]64. The surgery rate excluding cataracts decreased from 16.0 per 1,000 population in 1992/3 to 13.9 in 1998/9, a decline of 2.2% per annum. Thus, access to non-cataract surgical care appears to have decreased during the downsizing period. There are several reasons one may wish to exclude cataract surgery from the total. First, the cataract surgery rate increased at a dramatic annual rate of 6.4% during this period (Figure 26). This is far greater than the growth rate for any other procedure. This increase - which accounted for 29% of the increase in the total volume of day surgeries between 1985 and 1993 - was a result of a sharp lowering of inclusion criteria for cataract surgery [Meddings et al. (1996); Wright and Robens-Paradise (2001)]. By 1998/9 cataract surgery accounted for one-third of total surgical volume (inpatient and day surgery) and, as a result, including cataracts in the total completely masks the trends groups increased over this period, the unadjusted rate for most procedures would have increased more rapidly than the adjusted rate. 123 Figure 26. Proportion of Cases Performed on a Day Basis and Cataract Surgery Rate, BC Proportion of Cases Performed on a Day Basis Mjtr Gall Hyst Csec Mntr APP Absc Ex.brst THR T'KR Hern Cat 8586 0% 0% 0% 0% 27% 0% 17% 73% 0% 0% 14% 28% 8687 0% 0% 0% 0% 28% 0% 17% 74% 0% 0% 16% 30% 8788 0% 0% 0% 0% 29% 0% 18% 75% 0% 0% 20% 44% 8889 0% 0% 0% 0% 30% 0% 19% 77% 0% 0% 22% 53% 8990 0% 0% 0% 0% 32% 0% 19% 80% 0% 0% 28% 66% 9091 0% 0% 0% 0% 32% 0% 19% 82% 0% 0% 29% 74% 9192 0% 0% 0% 0% 34% 0% 20% 85% 0% 0% 32% 81% 9293 0% 9% 0% 0% 35% 0% 22% 86% 0% 0% 35% 84% 9394 0% 3% 0% 0% 38% 0% 22% 86% 0% 0% 39% 90% 9495 0% 4% 0% 0% 41% 0% 23% 88% 0% 0% 45% 94% 9596 0% 7% 0% 0% 42% 1% 22% 89% 0% 0% 52% 96% 9697 0% 11% 0% 0% 44% 1% 24% 90% 0% 0% 60% 98% 9798 0% 14% 0% 0% 44% 1% 26% 90% 0% 0% 62% 98% 9899 0% 20% 0% 0% 46% 1% 25% 89% 0% 0% 67% 98% Source: BC Linked Health Database 124 in the surgery rate of other important procedures. Second, cataract surgery is performed to some extent in specialty clinics and operates outside the hospital system. Thus, the cataract surgery rate is constrained by a different set of factors than all other procedures and should not be affected by reductions in the availability of nursing resources to the same extent as other procedures. Average hours of nursing care per inpatient surgery decreased during the downsizing period to the extent that length of stay is a suitable proxy for hours of nursing care65. Figure 27 shows that length of stay averaged across all inpatient cases decreased from 7.5 days in 1992/3 to 6.3 days in 1996/7. This represents an acceleration of the pre-downsizing downward trend (1.9% per annum versus 1.3%). It is not possible to determine whether average hours of nursing care per day surgery decreased during the downsizing period with the available data. The available evidence clearly indicates that during the downsizing period in BC the reduction in total hours of nursing care available to surgery patients was accounted for entirely by a reduction in the inpatient surgery rate and the average hours of nursing care per inpatient surgery. If the availability of operating room nurses was the binding constraint affecting surgical volumes, one might not expect to observe such a divergent pattern between inpatient and day surgery rates since both types of surgery are performed exclusively in operating rooms66. Rather, the fact that the inpatient surgery rate decreased but the day surgery rate increased seems to suggest that the availability of nurses to staff recovery beds affected surgical volumes during the downsizing period. A test of a break in the trend in 1992/3 confirms this. 6 5 Once again, the pattern of nursing hours per inpatient surgery is, in principle, traceable using data from the HS1/2 and A H S surveys but since these data are not reliability this remains infeasible in practice. 6 6 Cataract surgery is the exception. 125 However, the data are inconclusive and can support both claims. If hospital administrators are most concerned with surgical throughput (i.e. number of surgeries performed), then in the face of constraints on OR capacity they may decide to increase the surgery rate for procedures that require little OR - or 'skin-to-skin' - time. Given that inpatient procedures typically have a longer skin-to-skin time than day surgeries, this type of policy response would produce the pattern observed in the data. As discussed earlier, there is no single maximand associated with hospital behavior and there is no way of knowing how important throughput is to hospital administrators and physicians. Moreover, knowledgeable experts in BC all assert that it is the availability of nurses to staff hospital beds and not ORs that could potentially affect surgical volumes in hospitals. The next step is to determine why the inpatient surgery rate declined sharply. Specifically, the following two claims are tested: The surgery rate for procedures that require an inpatient stay decreased while the surgery rates for procedures typically performed on a day basis increased The proportion of cases performed on a day basis increased for most procedures Identifying which of these two claims is supported by the data is extremely important. The first claim - the 'access hypothesis' - implies reduced access to surgical care during the downsizing period and provides motivation for further research into the potential health effects (positive or negative) associated with reduced access to inpatient surgical care. The second claim - the 'efficiency hypothesis' - is consistent with no change in access to surgical care and an improvement in technical efficiency. Such a finding might motivate research similar to Brownell et al. (1999) into the health effects of performing more procedures on a day basis. 126 Figure 27. Length of Stay (in Days), BC Average Length of Stay, BC Mjtr Gall Hyst Csec Mntr App Absc THR TKR Hern AN 8586 24.2 9.7 7.2 6.8 4.2 5.3 9.0 18.3 18.7 4.1 7.9 8687 25.3 9.6 7.1 6.7 4.8 5.3 8.5 19.1 20.2 4.0 8.2 8788 24.2 9.0 7.0 6.6 4.7 5.2 8.8 18.8 18.3 3.9 8.0 8889 23.7 9.0 6.9 6.6 4.8 5.2 8.2 17.2 18.9 3.8 8.0 8990 22.5 8.7 6.8 6.3 4.4 5.1 8.6 17.5 16.8 3.7 7.8 9091 22.4 8.1 6.6 6.2 4.6 4.9 8.0 18.0 18.2 3.5 7.9 9192 21.3 6.5 6.3 6.0 4.1 4.8 8.2 16.7 16.3 3.4 7.5 9293 20.6 5.2 6.0 5.7 4.2 4.6 7.4 15.6 14.8 3.1 7.2 9394 19.9 4.8 5.6 5.5 4.3 4.4 8.1 14.3 13.3 2.9 6.9 9495 19.2 4.1 5.0 5.2 4.2 4.3 7.4 12.6 11.7 2.6 6.6 9596 19.3 •3.9 4.4 5.0 4.0 4.2 8.3 10.7 9.5 2.5 6.4 9697 18.4 3.8 4.2 4.8 4.2 4.2 7.3 10.1 8.8 2.4 6.3 9798 18.3 3.9 4.2 4.7 4.0 4.0 7.1 9.7 8.7 2.4 6.2 9899 18.4 4.0 4.1 4.5 4.2 3.9 7.0 9.7 8.3 2.6 6.3 127 Source: BC Linked Health Database To test these two claims it is necessary to move beyond the aggregate data of Figure 25 and examine surgical rates for specific procedures. 5.3 Surgery Rates for Specific Procedures The top half of Figure 26 shows the proportion of cases performed on a day basis for twelve common procedures that account for the majority of all surgeries in British Columbia. The procedures are major trauma (mjtr), gall bladder surgery (gall), hysterectomy (hyst), C-section (csec), minor trauma (mntr), appendectomy (app), drainage of abscess (absc), excision of breast lump (ex.brst), total hip replacement (THR), total knee replacements (TKR), hernia repair (hern) and cataract surgery (cat). The only major class of procedures not included is cardiac surgery. Major trauma, hysterectomy, C-section, appendectomy, total hip replacement and total knee replacement are performed only on an inpatient basis. Did the rate for these procedures decline during the cut-backs era as predicted by the access hypothesis? The surgery rate for C-section and hysterectomy decreased, remained fairly stable for major trauma, appendectomy, and total hip replacement, and actually increased for total knee replacement (Figure 28). Moreover, the surgery rate for procedures performed primarily on a day basis did not necessarily increase. The rate for hernia repair clearly decreased during the downsizing period while the rate for excision of breast lump remained stable. Thus, the data do not seem to support either the access hypothesis or the efficiency hypothesis. The story is more complex. To gain further insight into how constraints on nursing resources affect the pattern of surgical care in BC it is useful to develop a model that identifies the determinants of surgery rates. One may then use the model to describe the various avenues through which constraints on nursing resources are expected to affect the pattern of surgical care. 128 Figure 28. Surgery Rates by Procedure, BC Age Adjusted Surgery Rates per 1,000 Population, BC 3.0 2.5 \ 2.0 1.5 -1.0 0.5 A 0.0 oo cn CD CD N- OO CD CD —•— Mjtr -©-Gal l - A Hyst Csec -ft-Mntr - e - A p p -©—Absc Ex.brst ~~m— THR m TKR —x— Hern Age-adjusted Surgery Rate per 1,000 Population, BC Mjtr Gall Hyst Csec Mntr App Absc Ex.brst THR TKR Hern 8586 0.75 2.06 1.86 2.52 1.35 1.18 0.84 1.44 0.54 0.20 2.50 8687 0.81 2.14 1.84 2.53 1.40 1.24 0.85 1.47 0.66 0.28 2.46 8788 0.78 2.11 1.86 2.58 1.48 1.20 0.81 1.48 0.69 0.30 2.49 8889 0.73 2.04 1.69 2.65 1.41 1.13 0.83 1.45 0.74 0.31 2.43 8990 0.73 1.98 1.57 2.86 1.44 1.15 0.79 1.44 0.57 0.31 2.42 9091 0.73 2.11 1.65 2.81 1.41 1.17 0.76 1.32 0.66 0.40 2.51 9192 0.72 2.16 1.62 2.77 1.40 1.13 0.71 1.45 0.67 0.42 2.51 9293 0.73 2.31 1.49 2.67 1.45 1.08 0.70 1.44 0.65 0.44 2.37 9394 0.73 2.34 1.43 2.57 1.37 1.06 0.67 1.38 0.70 0.51 2.42 9495 0.70 2.24 1.33 2.47 1.41 1.02 0.67 1.38 0.64 0.51 2.34 9596 0.73 2.31 1.40 2.42 1.40 0.96 0.60 1.29 0.71 0.61 2.24 9697 0.69 2.13 1.32 2.33 1.30 0.97 0.62 1.23 0.69 0.62 2.20 9798 0.67 2.16 1.29 2.28 1.31 0.99 0.62 1.24 0.67 0.62 2.21 9899 0.69 2.04 1.20 2.26 1.33 0.98 0.55 1.26 0.62 0.60 1.94 129 Source: BC Linked Health Database 5.3.1 A Model of the Determinants of the Surgery Rate for Particular Procedures Figure 29 illustrates the flow of patients through the health care system. Four decisions jointly determine the surgery rate for various procedures in any given time period. 1. The Diagnostic Decision A portion of the population in any given time period will be diagnosed by a physician as having some specific condition (e.g. arthritis, cataract, pregnancy, a broken arm). For the purposes of this thesis this 'incidence rate' for a particular condition is assumed to be determined wholly by the age and sex structure of the population. This assumption is quite reasonable in many cases - e.g. pregnancy, arthritis of the hip - but is, in general, a gross simplification. In reality, the definition of disease or illness is far more subjective and varies across time, place and culture and is subject to modification in response to economic as well as other interests of particular provider groups. Payer (1988) describes dramatic differences between France, Germany, the US and Great Britain in the way doctors view various medical conditions. For example, French doctors are much more likely to attribute medical symptoms to the liver compared with doctors from the other countries. German doctors will diagnose a heart condition based on symptoms that US and British doctors would almost never attribute to this organ. Although each of these countries has access to the same set of medical knowledge, cultural difference result in drastic variations in diagnostic methods. Wright (2002) cites examples of how pharmaceutical companies - motivated by a desire to increase profits - are increasingly promoting conditions such as shyness, periodic sadness, and impotence as illnesses that can be treated with drug therapies. For the purposes of this thesis, however, the diagnostic decision is assumed to be made independently of all other factors in the model so that the incidence rate for each condition is exogenous. 130 131 2. The Inclusion Decision Of the people diagnosed with a certain condition some will qualify for surgical intervention while others will not. This is often a joint decision between physician and patient but given the asymmetry of (clinical) information it is mostly a physician decision. The inclusion decision is greatly influenced by the views of the medical profession on appropriate interventions. In turn, this 'medical convention' is, in theory, based primarily on scientific evidence on the effectiveness of a particular procedure compared to its non-surgical substitutes. For example, if a broken femur (or other major trauma) is not repaired the patient faces a life of severe disability and sometimes a serious chance of death. There are few alternatives to surgery. Repair of major trauma, thus, has very clear inclusion criteria. For some ailments there is no clear scientific evidence on when surgery is appropriate. For example, the inclusion criteria for C-section are very unclear. A pregnant woman who has undergone a C-section in many cases could have given birth vaginally without any adverse health effects (CIHI, 2000c). In the absence of - even sometimes despite -clinical evidence on appropriateness, the inclusion decision is strongly influenced by patient and physician preferences (Wright et al., 1999), the particular culture of the hospital or country (Payer, 1988), hospital capacity67 [Fisher et al. (2000); Fisher et al. (1994); Wennberg, Freeman and Culp (1987)] and even the legal environment (Bassett et al., 2000). Even when scientific evidence becomes available that clearly indicates when a particular treatment is appropriate, there is a considerable lag between evidence and practice in the health care field. Procedures that have clear inclusion criteria are described as 'necessary' while those that do not are described as 'discretionary'. Al l surgical procedures can then be classified along the Necessary-Discretionary spectrum. Of course, practitioners in different 6 7 This is reflected by the lower-most arrow in Figure 6. 132 hospitals, cities, countries will disagree on where each procedure ought to fall on this spectrum. For the purposes of this thesis only the dichotomous classification is used. The incidence rate and the inclusion rate together determine the pool of surgical candidates in any given time period for a particular procedure. 3. The Delayability Decision The delayability decision entails judging the extent to which a particular procedure can be delayed without harming the patient. For example, procedures such as major trauma need to be performed very soon after diagnosis in order to avoid adverse health effects, the most notable of which is death. C-section is another example of a procedure that can not be delayed by more than a few hours. Conversely, a patient with some loss of visual acuity resulting from cataracts can often wait months before undergoing eye surgery without any negative health effects. Procedures can be ranked along a Delayable-Undelayable spectrum. Of course, in practice the extent to which a procedure is delayable is determined on a patient-by-patient basis. Gall bladder surgery may be delayable for one patient but quite undelayable for another. For the purposes of this thesis, rankings along the Delayable-Undelayable spectrum are based on the average case. For example, the statement "the average hysterectomy candidate can wait longer between the decision to have the surgery and having the surgery than the average minor trauma patient without detrimental effects to her health" implies that hysterectomy is more delayable than minor trauma. Just as with the inclusion decision, in the absence of - or sometimes in spite of - clinical evidence on the health effects of waiting for surgery, factors such as physician preferences play an important role in determining which procedures can be delayed. As a result, there may be no clear consensus among providers on where a particular procedure ought to be placed on the Delayable-Undelayable spectrum. 133 4. The Decision to Perform the Surgery The total volume of surgeries performed in any period t depends on the surgical capacity of the health care system. Surgical capacity has several components. For procedures that must be performed in hospitals, operating room availability is an important determinant of surgical capacity. This includes the availability of surgeons, the availability of nurses to assist in the operating room and post-anesthesia recovery (PAR) areas68, the availability of technicians, pathologists, and any other personnel participating in the surgery. For cases where the patient must remain in hospital for at least one night the availability of a bed and the necessary nursing resources to staff the bed are also important components of surgical capacity. Where reliable data are available they show that the pattern in the availability of important non-nursing inputs to surgical capacity — most notably physicians - did not change during the downsizing period (Figure 30). To close the model, it is assumed that all individuals who do not receive surgery within period t are returned to the pool of surgical candidates at the beginning of period t+1. As the time horizon increases, this backflow decreases and if the time horizon is extended sufficiently - to about one year - this backflow will be close to zero. 5.3.2 Behavioral Predictions Implied by the Model With the various components of surgical capacity outlined, one may now return to the discussion of how the availability of nursing resources might affect the pattern of surgical care. One may rewrite the previous identity as: Post-anesthesia recovery (PAR) is a separate area near the operating room where patients may rest for a few hours before either going home or being transferred to a hospital bed. 134 2! CQ C 3 CO © CO ^" o o 0> 5 CD to" -J" O o 5" 3 fy 5' 135 SCARE, I, {PROC,, x CARE,,} where SCARE, total hours of nursing care available to surgery patients in BC hospitals in year t total number of cases of procedure i in BC hospitals in PROQ, year t CAREu average hours of nursing care per case for procedure i in year t A decrease in SCARE, implies at least one of the following responses in hospitals for at least one procedure: Decrease length of stay. This implies fewer nursing resources per case and if lengths of stay decrease enough, surgery rates need not change in response to constraints on the availability of nursing resources in hospitals. Perform more cases on a day basis. This is a special case of decreasing the length of stay to zero so that no nursing resources are required beyond the OR and PAR. Perform more cases in specialty clinics outside hospitals. This policy response attempts to remove cases from the identity. Current medical convention renders this policy response inappropriate for most surgical procedures. Decrease the surgery rate. This, in turn, implies either an increase in the size of (i.e. more people on) waiting lists or stricter inclusion criteria. 136 According to the model, the feasibility of each policy option varies from procedure to procedure. For example, medical convention may make it difficult to decrease lengths of stay for some procedures but not for others. Medical convention also affects whether it is possible to perform more cases on a day basis for a particular procedure. One might divide procedures into three categories: those for which there is a consensus that an inpatient stay is appropriate, those for which there is a consensus that an inpatient stay is inappropriate and those for which there is no consensus or for which the consensus has changed over the past 15 years. For the first category of procedures the policy option of performing more cases on a day basis is not feasible. For the third category it is. Once again, it is emphasized that the consensus may or may not be based on clinical evidence. Reducing the surgery rate is also subject to medical convention. For procedures that are undelayable and have very clear inclusion criteria reducing the surgery rate is completely inconsistent with basic medical ethics. These procedures must be performed in order to avoid serious harm to the patient. For procedures with unclear inclusion criteria, however, the surgery rate can be lowered by lowering the inclusion rate. Similarly, for procedures that are delayable the surgery rate can be lowered by increasing the size of waiting lists for these procedures. Eventually, growing waiting lists may lead to stricter inclusion criteria. Returning to the individual procedures, the surgery rate for C-section, hernia repair, gall bladder surgery, drainage of abscess and hysterectomy decreased during the downsizing period. The surgery rate increased steadily for total knee replacement, dramatically for cataract surgery (Figure 26), and remained fairly constant for the remaining procedures69. Regressing the surgery rate on a constant and a time trend confirms these results. 137 These results motivate two questions: For the procedures where the surgery rate remained stable or increased, which of the first three policy responses were invoked and for which procedures? For the procedures where the surgery rate declined, is there any evidence of stricter inclusion criteria or increased waiting times? Before addressing these questions it is important to discuss what the model does not do. In situations where more than one policy response is available, the model does not attempt to explain why the observed policy response was chosen. Such an analysis would involve, among other things, modeling the process through which administrators and medical staff interact in the hospital setting - an endeavor with a questionable likelihood of producing any useful insight. For as Evans (1984) notes, [the relationship between administrators and medical staff in hospitals] is a very complex, non-zero-sum game, not a sequence of self-contained spot contracts at explicit or implicit prices. Realistic description and analysis of such complex processes can lead to very useful generalizations...but they may never be expressed in a formal analytic framework which is either realistic or useful (p. 173). 5.3.3 Procedures with a Stable or Increasing Surgery Rate As noted earlier, the dramatic increase in the rate of cataract surgery can be explained by technological advances that led to more lenient inclusion criteria and the fact that some cataract surgery is performed in specialty clinics and is not subject to nursing resource constraints. Health care officials ought to examine whether such a rapid increase in cataract surgery is desirable. A recent study of patients in the Vancouver/Richmond Health Board provides 138 some evidence that the volume of cataract surgery in B C is so high that patients, at the margin, receive no health benefits (Wright and Robins-Paradise, 2001). About one quarter of cataract surgery patients experienced a worsening of visual acuity after undergoing the procedure. It appears that the volume of cataract surgery may have reached beyond the flat-of-the-curve of the health production function. For the remaining procedures with stable or increasing surgery rates, the primary policy response was to shorten lengths of stay. Lengths of stay decreased during the downsizing period for all procedures studied except minor trauma and the average decreased from 7.2 days in 1992/3 to 6.3 in 1998/970 (Figure 26). The decrease in length of stay was particularly drastic for total hip replacement and total knee replacement. In 1992/3 the average for these procedures was just over 16 days but by 1998/9 it dropped to 9.7 for total hip replacement and 8.3 for total knee replacement. Other than cataract surgery, performing more cases on a day basis played a fairly minor role for procedures with a stable or increasing surgery rate. The proportion of cases performed on a day basis increased significantly only in the case of minor trauma where it went from 35% in 1992/3 to 46% in 1998/9 (Figure 27). These empirical observations are consistent with the policy options predicted by the model. Major trauma, total hip replacement and total knee replacement have very clear inclusion criteria and it is highly inappropriate to perform these procedures on a day basis. Therefore, in order to reduce the nursing resource requirements associated with these procedures without reducing the surgery rate, the only option is to decrease length of stay. It is precisely these procedures that had the largest decrease in length of stay. A procedure such as minor trauma has very clear inclusion criteria but the extent to which an inpatient stay is required is unclear. Thus, the proportion of cases performed on a day basis can be increased in response to reductions in the availability of nursing resources and this is precisely what is observed. 139 5.3.4 Procedures with a Decreasing Surgery Rate The decrease in the surgery rate for C-section, drainage of abscess and hysterectomy was well established prior to the cut-backs period suggesting that constraints on nursing resources did not drive the decrease. C-section and hysterectomy are highly discretionary procedures71. Attitudes of the medical profession have changed recently with an emerging consensus that the surgery rate for these procedures has historically been too high in Canada72. Therefore, the decrease in the surgery rate for C-section and hysterectomy most likely reflects a lowering of the inclusion rate due to changing attitudes of the medical profession rather than a response to constraints on surgical capacity. As a result, a lower surgery rate for these procedures should be interpreted as an improvement in clinical efficiency73. The downward trend in the surgery rate for gall bladder surgery and hernia repair, however, ought to be interpreted quite differently. The decrease coincides precisely with the beginning of the hospital downsizing period suggesting that constraints on nursing resources may have had an impact74. The rate for gall bladder surgery decreased from 2.3 per 1,000 population in 1992/3 to 2.0 in 1998/9 after increasing over the previous four years while for hernia repair the rate decreased from 2.4 per 1,000 population to 1.9 after being stable for seven years (Figure 28). The magnitude of the reduction in average length of stay for surgical inpatients in B C is virtually identical to that for surgical inpatients in the Winnipeg area between 1991 and 1996 (Brownell et al, 1999). 7 1 This is strongly suggested by large regional variation across Canada in age-adjusted C-section rates that are documented in CIHI (2002). 7 2 For example, CIHI (2000c) reports that "experts from the Society of Obstetricians and Gynecologists of Canada have weighed the evidence and established guidelines for appropriate care before, during, and after birth. Among other things, they note that most women, even a large proportion of those who have previously had C-sections, can safely deliver vaginally. In fact, they suggest that successful vaginal births after C-sections typically carry lower health risks for mothers and require shorter hospital stays than for those having an optional surgical delivery. There is also recent evidence to suggest that women having C-sections are at higher risk of being re-admitted to hospital than women who have had vaginal births." 7 3 A n initiative was introduced in B C hospitals in 1992 to reduce the number of'inappropriate' surgeries. C-section and hysterectomy may have been two of the targeted procedures. 140 The reduction in the surgery rate for these two procedures is particularly puzzling when one considers that all of the policy options available were employed. The length of stay for gall bladder surgery decreased from 5.2 days in 1992/3 to 4.0 in 1998/9 while for hernia repair it decreased from 3.1 days in 1992/3 to 2.6 days in 1998/9. The proportion of cases performed on a day basis for both of these procedures doubled between 1992/3 and 1998/9, representing the most dramatic increase among all procedures during the downsizing period. Furthermore, since the inclusion criteria for both of these procedures are quite clear it is unlikely that the inclusion rate decreased during this period. One possible explanation for the pattern of gall bladder surgery concerns technological innovation. In 1990 a new, less invasive technology for gall bladder surgery was implemented in BC hospitals. Ceteris paribus, this ought to have increased the inclusion rate for this procedure and, according to evidence from other provinces, it indeed has (Priest, 1997). However, the introduction of a new technology in the health care field often results in overuse in the short term75. After a few years, as more information becomes available on the effectiveness of the new technology, utilization rates start to return to their pre-innovation levels. The gall bladder surgery rate appears consistent with this phenomenon76. If inclusion rates for gall bladder surgery and hernia repair did not decrease during the downsizing period, the model predicts that the size of waiting lists might have increased since both of these procedures are somewhat delayable. As a comparison, the gall bladder surgery rate did not change in the Winnipeg area during the downsizing era (Brownell et al, 1999). 7 5 See Vujicic (1998) for an example of this phenomenon as it relates to angioplasty in the treatment of heart disease. 7 6 Similar to the case of C-section, an alternative explanation is that sometime in the mid-1990s gall bladder surgery may have been targeted as an over-prescribed procedure. Canada has one of the highest rates of gall bladder surgery in the world and many experts believe that many unnecessary surgeries are being performed (Priest, 1997). 141 Unfortunately, current data on waiting list sizes and wait times in BC can not be used to test this hypothesis. Waiting list data are available only on a piece meal basis and the criteria governing when and whether a patient is placed on a waiting list differ from physician to physician. This reflects the fact that in Canada fmjost lists are created and maintained in the offices of individual physicians or hospital surgical or diagnostic departments rather than by a regional authority or other coordinating agency (Sanmartin et al. 2000; p. 1306). Further, waiting lists are not audited in Canada (Lewis et al., 2000). As a result, they tend to be an inaccurate representation of the number of people actually waiting for a particular procedure. Studies have found that in regions where waiting lists are subject to periodic audit between 30% and 70% of patients on waiting lists are not (or should not be) waiting to receive a particular treatment [Sanmartin et al. (2000); Lewis et al. (2000); McDonald et al. (1998)]. The reasons behind such large overcounts are numerous: the procedure has already been done or is no longer required; the patient is not aware of being on a list and requests removal when so informed; the patient has died; the procedure is not appropriate for the patient; an alternative treatment is preferable. This suggests that even if waiting list data were available at the provincial level they would significantly overstate the number of people actually waiting for a particular surgery77. In the absence of accurate waiting list information, an alternative strategy was considered to estimate the number of people waiting for gall bladder surgery and hernia repair. Using BC Medical Services Plan (MSP) billings data one might count the number of individuals diagnosed with a particular gall bladder condition or a particular severity of hernia in any given year. These counts would represent the number of people who qualify for gall bladder surgery and hernia repair. The subset of individuals not receiving The Western Canada Waiting List Project - a partnership between the B C , Alberta, Saskatchewan and Manitoba ministries of health and other organizations - has developed a set of tools to help physicians and health care administrators better manage waiting lists for several specific surgical and diagnostic procedures. It is unknown at this time whether results have been implemented. 142 the surgery within the year would be on a 'virtual' waiting list. This method could be used to construct a time series for the waiting list size for gall bladder surgery and hernia repair. After discussing the proposed procedure with health care services researchers at UBC who have expert knowledge of the MSP billings database it became clear that one can not infer a patient's need for a particular surgical procedure from the billings data. Among other things, physicians do not record diagnoses in a consistent manner and even if they did, the ICD-9 classification system used in the database is not detailed enough for the purposes of this thesis. As a result, this strategy was not pursued. Therefore, all that can be said at this point is that the decrease in the surgery rate for hernia repair and gall bladder surgery warrants further investigation as it may represent unmet health care needs resulting from a reduction in the availability of nursing resources. It is unclear why the surgery rate for drainage of abscess has steadily declined over the past fifteen years. This procedure is necessary and non-delayable. It could be the case that the incidence rate has decreased, perhaps as a result of better drug therapies78. In any event, the decrease is clearly uncorrected with the downsizing period. 5.4 Discussion Hospitals in BC appear to have adapted quite well to the reduction in the availability of nursing resources when it comes to access to beneficial surgical care. A combination of shorter lengths of stay and a shift to day surgery resulted in stable or increasing surgery rates during the downsizing era for several procedures. This is consistent with the claim that hospital downsizing brought about an improvement in technical efficiency with respect to the production of surgical separations. 143 In most cases where the surgery rate declined there are more plausible reasons for the decline than a reduction in surgical capacity resulting from fewer nurses in hospitals. Only for two procedures - hernia repair and gall bladder removal - did the reduction in the surgery rate coincide precisely with the period of hospital cut-backs. The fact that both of these procedures are necessary but somewhat delayable suggests that waiting times for these procedures might have increased during the downsizing period. This deserves further study, which should be put off, perhaps, until more accurate waiting list data are available. Part II of this thesis investigates access to surgical care during the cut-backs era in BC and does not examine the health implications of the changing pattern of care. Such research would be a valuable extension of this thesis. The available evidence suggests that shorter lengths of stay and a shift to day surgery within the range observed in BC are not associated with increased mortality or re-admission to hospital. This suggests that, to the extent that mortality and re-admission rates measure health status, the hospital downsizing period brought about an improvement in clinical efficiency, at least with respect to surgical care. As a final note, the empirical analysis of Part II has focussed on the pattern of surgical care in BC, ignoring any potential effect of reductions in the availability of nursing resources on access to non-surgical hospital care. This is an important limitation to this thesis. The available evidence indicates that utilization of non-surgical hospital care (both separations and length of stay) is very sensitive to hospital capacity79 but that the additional care provided is often of little or no benefit to patients [Wennberg et al. (1989); Roemer and Shain (1959); Myers (1954)]. However, Brownell et al. (1999) show that lengths of stay for non-surgical separations did not decrease during the downsizing period in Manitoba and that non-surgical separations decreased at a much lower rate than It might also be the case that more cases are being performed outside of hospitals. 7 9 This phenomenon is known as Roemer's Law, named after Milton Roemer who first posited that almost any additional hospital beds provided to the population wil l be used, up to a ceiling not yet determined. That is, "a bed built is a bed filled". 144 surgical separations. It would be interesting to extend the analysis of this thesis to examine the pattern of non-surgical care in B C during the downsizing period. 145 CHAPTER 6 - CONCLUSION This thesis has examined trends in the nursing labour market in Canada over the past two decades and has produced several important results. The hospital expenditure reduction initiative of the early 1990s was associated with a significant decrease in nursing employment levels in Canadian hospitals. The evidence strongly suggests that this decrease was a result of a fall in the demand for nursing labour in hospitals that, in turn, was driven by reductions in hospital global budgets. The evidence does not support the competing claim that nurses working in hospitals voluntarily quit their jobs and took up employment in the non-health care sector due to stagnant wages and deteriorating working conditions in hospitals. According to these findings the downsizing period brought about an economic surplus of nurses. The decrease in nursing employment levels in hospitals was most severe for nurses under 30 years of age. Employment levels for older nurses actually increased or remained stable during the downsizing era. This appears to reflect the important role seniority plays in an environment of layoffs and hiring freezes in unionized sectors such as nursing. It is still too early to use the methods in this thesis to determine the extent to which recent increases in the demand for nurses - stemming from increases in hospital budgets over the past five years and represented by a rightward shift in the demand curve in Figure 2 -have removed the economic surplus. When 2001 Census data become available, a natural extension of this thesis is to test this hypothesis. If the health care labour force participation rate of the under-25 cohort in 2001 is equal to its pre-downsizing level this would suggest that the excess supply of nurses associated with the downsizing period has been removed. Data from the 2001 Census could also be used to test the extent to which the downsizing era affected the demand for and supply of nursing education programs. If during the cut-backs period the demand for nursing diplomas/degrees or government funding for nursing programs decreased then the size of the under-25 cohort of the potential supply of nurses ought to have shrunk between 1996 and 2001, ceteris paribus. 146 The analysis in this thesis also casts doubt on the widespread claims that fewer nursing resources in hospitals did not have a negative impact on access to beneficial surgical care in British Columbia. A combination of shorter lengths of stay and an increase in the use of day surgery resulted in a stable or increasing surgery rate for most procedures despite fewer nursing hours in hospitals. In cases where the surgery rate declined, it is unlikely that constraints on nursing resources were the principle cause80. Although further research is warranted to measure the health effects of shorter lengths of stay and an increased reliance on day surgery, available evidence from other jurisdictions indicates that while fewer nursing resources in hospitals may increase the likelihood of certain in-hospital adverse events, few if any negative effects on mortality or hospital readmission rates are to be expected. The main impact of hospital downsizing documented in this thesis, therefore, was a reduction in the number of young people in the nursing workforce. This, in turn, stems directly from seniority provisions written into collective bargaining agreements between hospitals and nursing unions. Thus, the findings of this thesis do not apply, in general, to non-collective bargaining settings. There are several important policy implications of having fewer young individuals in the nursing workforce. In the coming years, a large portion of the nursing workforce is expected to retire81. Combined with the fact that recent increases in hospital budgets are expected to increase the demand for nurses in hospitals this is expected to create a significant number of nursing vacancies. One might infer from the results of this thesis that many of these vacancies could be filled by the large number of young individuals who have an education in nursing but were working in non-health care occupations during the cut-backs era. Such a belief, however, may be somewhat optimistic. These 'non-nursing nurses'- over 15,000 under 30 years of age in BC, Alberta, Ontario and 8 0 Even i f the declines may be partially attributed to more limited nursing resources, for most of the procedures this ought to be a matter for praise. See section 5.3.4. 8 1 Sibbald (2002) claims that as much as half of the R N workforce in Canada wi l l retire within the next 15 years. 147 Quebec in 1996 - face barriers to entry into the nursing profession imposed by licensing bodies that may significantly impede re-integration into the nursing workforce. Individuals who have spent significant time out of nursing must typically complete re-certification examinations, pay registration fees and pass a probationary period before being allowed to practice. In other words, due to the presence of barriers to entry in the nursing profession the 'effective' potential supply of nurses may have been reduced during the downsizing era. This would be represented by a leftward shift in the supply curve in Figure 2. For the youngest individuals the large health care wage premium identified in Part I is expected to outweigh the 're-entry costs' so that one expects most of these individuals to be willing to enter the nursing workforce if a job became available. But in a few years all of the non-nursing nurses identified in Part I will be over 30 years of age and the health care wage premium for individuals over 30 is significantly smaller than for younger individuals. As a result, it is quite possible that many non-nursing nurses may voluntarily remain in the non-health care sector if and when future nursing vacancies arise. Again, one could use 2001 and 2006 Census data to test whether the health care labour force participation rate of the 1996 under-25 and 25-29 cohorts ever returns to its pre-downsizing level. If, faced with unacceptable numbers of unfilled nursing vacancies, policy turn their attention on drawing non-nursing nurses back to the nursing profession they might address the barriers to 're-entry' imposed by nursing associations. Some of these might be relaxed or removed, or potential re-entrants might be offered assistance in overcoming them. Otherwise, these non-nursing nurses may be lost forever to the non-health care sector. Policy makers must then either increase nursing school enrolment, or recruit nurses from abroad - both of which shift the supply curve to the right in Figure 2 - or increase nursing wages (i.e. move up along the demand curve in Figure 2) if an economic shortage of nurses is to be avoided in the future. 148 With respect to increasing nursing school enrolment, policy makers must consider the implications of having the BScN as opposed to a diploma in nursing as the entry-to-practice requirement for RNs. Since governments underwrite most education expenses, it costs governments more money to train a nurse through a four-year BScN program than through a two-year diploma program. A BScN also entails two additional years of education over a diploma program - a considerable increase in investment in human capital - and individuals will demand higher nursing wages as compensation, again increasing government costs. Thus, the BScN as the entry-to-practice requirement causes governments to receive a much lower return for each dollar invested in nursing education capacity. As a result, policy makers might find it more cost effective to focus on recruiting individuals who already have the educational credentials to work in the nursing sector but are employed in non-health care occupations rather than increasing nursing school enrolment. However, the results of Part II - although limited to surgical care - show that hospitals adjusted quite well to reductions in nursing capacity and that technical efficiency improved during the downsizing era. In light of such findings, policy makers might wish to re-examine the extent to which more nurses ought to be hired when hospital budgets expand. 149 REFERENCES Abowd, J.M. and H.S. Farber (1982) "Job Queues and the Union Status of Workers," Industrial and Labor Relations Review, (3): 354-367. Advisory Committee on Health Human Resources (2000) "The Nursing Strategy for Canada," Advisory Committee on Health tluman Resources, Working Group on Nursing Resources, October 2000. Aiken, L.H., S.P. Clarke, D.M. Sloane, J.A. Sochalski, R. Busse, Fl. Clarke, P. Giovannetti, J.H. Hunt, A . M . Rafferty and J. Shamian (2001) "Nurses' Reports on Hospital Care in Five Countries," Health Affairs, 20(3): 43-53. Antonazzo, E., D. Skatun, A. Scott and R.F. Elliott (2001) "British Nurses' Labour Supply: An Econometric Investigation," poster presented at the 3 r d International Conference of the International Health Economics Association, July 2001, York, United Kingdom. Auerbach, D.L, P.I. Buerhaus and D.O. Staiger (2000) "Associate Degree Graduates and The Rapidly Aging RN Workforce," Nursing Economics, 18(4): 178-184. Barbaras, M.C. (1977) "Hospital Based Outpatient Surgery: A Viable Alternative," Dipl.Hosp. Admin, thesis, University of Toronto. Barer, M.L. and W.A. Webber (1999) "Immigration and Emigration of Physicians to/from Canada," UBC Health Human Resources Unit Discussion Paper 99:6. Barer, M.L, A.J. Stark and C. Kinnis (1984) "Manpower Planning, Fiscal Restraint, and the "Demand" for Health Care Personnel," Inquiry, 21(Fall 1984): 254-265. Bassett, K.L, N . Iyer and A. Kazanjian (2000) "Defensive Medicine During Hospital Obstetrical Care: A Byproduct of the Technological Age," Social Science and Medicine. 51(2000): 523-37. Baumgart, A.J. (1997) "Hospital Reform and Nursing Labor Market Trends in Canada," Medical Care. 35(10): OS124-131. Baumgart, A.J. (1988) "The Nursing Workforce in Canada," in Baumgart, A.J. and J. Larsen eds. Canadian Nursing Faces the Future: Development and Change, C.V. Mosby Company, Toronto. Baumgart, A.J., and J. Larsen (1988) "Introduction to Nursing in Canada," in Baumgart, A.J. and J. Larsen eds. Canadian Nursing Faces the Future: Development and Change, C.V. Mosby Company, Toronto. 150 Bernstein, S. J., E. A. McGlynn, A. L. Siu, et al. (1993) "The Appropriateness of Hysterectomy: A Comparison of Care in Seven Health Plans," Journal of the American Medical Association, 269: 2398-2402. Bourbonnais, R., A. Vinet, F. Meyer, and M . Goldberg (1992) "Certified Sick Leave and Work Load: A Case Referent Study Among Nurses," Groupe Interdisciplinaire de Recherche sur l'Organisation , la Sante et la Securite du Travail, Laval University. Brook, R.H., Ware, J.E., Rogers, W.H. et al. (1983) , "Does Free Care Improve Adults' Health? Results from a Randomized Controlled Trial," New England Journal of Medicine, 309(24): 1426-1434. Brownell, M.D., N.P. Roos, and C.A. Burchill (1999) "Monitoring the Winnipeg Hospital System: 1990/91 through 1996/97," Manitoba Centre for Health Policy and Evaluation. Bruce, S., S. DeCoster, J. Trumble Waddell, C A . Burchill and S. De Haney (2001) "Acuity of Patients Hospitalized for Medical Conditions at Winnipeg Acute Care Hospitals," Manitoba Centre for Health Policy and Evaluation. Buerhaus, P.I., D.O. Staiger and D.I. Auerbach (2000a) "Why Are Shortages of Hospital RNs Concentrated in Specialty Care Units?" Nursing Economics, 18(3): 1-6. Buerhaus, P.I., D.O. Staiger and D.I. Auerbach (2000b) "Implications of an Aging Registered Nurses Workforce," Journal of the American Medical Association, 283(22): 2948-54. Buerhaus, P.I., D.O. Staiger and D.I. Auerbach (2000c) "Policy Responses to an Aging Registered Nurse Workforce," Nursing Economics, 18(6): 278-303. Buerhaus, P.I. and D.O. Staiger (1999) "Trouble in the Nurse Labor Market? Recent Trends and Future Outlook," Health Affairs, 18(1): 214-222. Canada News Wire (CNW) (1999a) "Devaluation of Nursing Profession Harmful to Ontario's Hospital Health Services, Nurses Say," December 2. Canada NewsWire (CNW) (1999b) "T Quit' Say Ontario Nurses" Canada NewsWire December 8. Cardiff, K., C J . Wright and M . Kilshaw (1998) "Acute Care Beds: How Are They Used in BC," UBC Health Policy Research Unit Discussion Paper 97:07D. Chaykowski, R.P. and L . M . Powell (1999) "Women and the Labour Market: Recent Trends and Policy Issues," Canadian Public Policy, November 1999 Supplement: S l -S25. 151 CHSRF (2001) "Listening for Direction: A National Consultation on Health Services and Policy Issues," Canadian Health Services Research Foundation. CIHI (2002a) "Statistics: Hospitalization Rates," Canadian Institute for Health Information, <http://www.cihi.ca> (Accessed: October 27, 2002). CIHI (2002b) "Health Care in Canada, 2002," Canadian Institute for Flealth Information. CIHI (2001a) "Table 1. Acute Separations and Days, Rates per 100,000 Population and Average Days per Separation by Sex and Age Group, Canada, 1998/99," Canadian Institute for Health Information, <http://www.cihi.ca> (Accessed: September 21, 2002). CIHI (2001b) "Future Development of Information to Support the Management of Nursing Resources: Recommendations," Canadian Institute for Health Information. CIHI (2001c) "Canada's Health Care Providers," Canadian Institute for Health Information. CIHI (2000a) "Supply and Distribution of Registered Nurses in Canada, 2000 Report," Canadian Institute for Health Information. CIHI (2000b) "Supply and Distribution of Registered Nurses in Canada, 2000 Provincial/Territorial Profiles," Canadian Institute for Health Information. CIHI (2000c) "Health Care in Canada, 2000," Canadian Institute for Health Information. CNA (2000) "Labour Market Integration of Graduates in Nursing in Canada 1986 -1997," Canadian Nurses Association. Cohen, L. (1989) "Why are Canada's Nurses Quitting?" Canadian Medical Association Journal 141(12): 1270-72. Dussault, G., M.A. Fournier, M.S. Zanchetta, S. Kerouac, J.L Denis, L. Bojanowski, M . Carpentier and M . Grossman (2001) "The Nursing Labour Market in Canada: Review of the Literature," GRIS, Universite de Montreal. Evans, R.G., M.L Barer, S. Lewis, M . Rachlis and G.L. Stoddart (2000) "Private Highway, One-Way Street: The Deklein and Fall of Canadian Medicare?" UBC Health Policy Research Unit Discussion Paper 2000:3D. Evans, R.G. (1980a) "Alternatives to Traditional Care at Children's Hospital, Vancouver: The Economics Studies of Cost Savings," in Robinson, G.C. and H.F. Clarke eds. The Hospital Care of Children: A Review of Contemporary Issues, Oxford University Press. 152 Evans, R.G. (1980b) "Other Experiences With Alternative Care: Common Problems and Suggested Solutions," in Robinson, G.C. and H.F. Clarke eds. The Hospital Care of Children: A Review of Contemporary Issues, Oxford University Press. Evans, R.G. (1981) "Incomplete Vertical Integration: The Distinctive Structure of the Health-Care Industry," in van der Gaag, J. and M . Perlman (eds), Health, Economics, and Health Economics, North-Holland. Evans, R.G. (1984) Strained Mercy, Butterworth, Toronto. Evans, R.G. and G.C. Robinson (1973) "An Evaluation of the Economics Implications of a Day Surgery Unit," mimeo. Farber, H.S. (1983) "The Determination of the Union Status of Workers," Econometrica, 51(5): 1417-1438. Fayerman, P. (2000) "St. Paul's Doctors to Divert Patients," The Vancouver Sun, December 6: B3. Feldstein, M . (1968) Economic Analysis for Health Service Efficiency: Econometric Studies of the British National Health Service, Chicago, Markham Publishing Company. Fisher E.S., Wennberg J.E., Stukel T.A., Sharp S.M. (1994) "Hospital Readmission Rates for Cohorts of Medicare beneficiaries in Boston and New Haven," New England Journal of Medicine, 331(15):989-95. Fisher E.S., Welch H.G. (1999) "Avoiding the Unintended Consequences of Growth in Medical Care: How Might More be Worse?" Journal of the American Medical Association, 281(5):446-453. Fisher E.S., Wennberg J.E., Stukel T.A., et al. (2000) "Associations Among Hospital Capacity, Utilization, and Mortality of US Medicare Beneficiaries, Controlling for Sociodemographic Factors," Health Services Research, 34(6): 1351-62. Fisher E.S., Wennberg J.E., Stukel T.A., et al. (2003) "The Implications of Regional Variations in Medicare Spending. Part 2: Health Outcomes and Satisfaction with Care," Annals of Internal Medicine, 138: 288-298. Fong, P. (2000) "Hospitals Try to Halt Surgery Shutdown," The Vancouver Sun. September 19: B l . Growe, S.J. (1991) Who Cares? The Crisis in Canadian Nursing, McLelland and Stewart, Toronto. Gunderson, M . and W.C. Riddell (1999) "The Changing Nature of Work: Implications for Public Policy," Institute for Research on Public Policy, Montreal, Canada. 153 Heckman, J. (1979) "Sample Selection Bias as a Specification Error," Econometrica, 47: 153-161. Heckman, J. and G. Sedlacek (1990) "Self-Selection and Hourly Wages," Journal of Labor Economics. 8(1): S329-S363. Heckman, J. and G. Sedlacek (1985) "Heterogeneity, Aggregation and Market Wage Functions: An Empirical Model of Self-Selection in the Labor Market," Journal of Political Economy, 93: 1077-1125. Hirsch, B.T. and J.E. Schumacher (1995) "Monopsony Power and Relative Wages in the Labor Market for Nurses," Journal of Health Economics, 14(4): 443-76. Kazanjian, A. et al (2000a) "Nursing Workforce Study Volume I: Demographic Context and Health System Structure for Nursing Service in Canada," UBC Health Human Resources Unit Discussion Paper 00:3. Kazanjian, A., L. Wood, H. Yip, S. Rahim-Jamal and A. MacDonald (2000b) "Nursing Workforce Study Volume II: The Supply of Nursing Personnel in Canada," UBC Health Human Resources Unit Discussion Paper 00:4. Kazanjian, A., K. Brothers and G. Wong (1986) "Modeling the Supply of Nurse Labor: Life-Cycle Activity Patterns of Registered Nurses in One Canadian Delivery System," Medical Care. 24(12): 1067-1083. Lee, J. (2000a) "Nursing Shortage Delays Surgeries," The Vancouver Sun, February 19: A3c. Lee, J. (2000b) "Show Us the Money, Nurses Tell B.C.," The Vancouver Sun. December 8: A l . Lewis, S., M . Barer, C. Sanmartin, S. Sheps, S. Shortt, P. McDonald (2000) "Ending Waiting List Mismanagement: Principles and Practice," Canadian Medical Association Journal, 162(9): 1297- 1300. Maddala, G.S. (1983) Limited-Dependent and Qualitative Variables in Econometrics, Cambridge University Press, New York. McDonald, P, S. Shortt, C. Sanmartin, M . Barer, S. Lewis and S. Sheps (1998) "Waiting Lists anmd Waiting Times for Health Care in Canada: More Management, More Money," mimeo. McGillis Hall, L., D. Irvine Dorin, G. Baker et al. (2001) The Impact of Nursing Staff Mix Models and Organizational Change on Patients, System, and Nurse Outcomes, Faculty of Nursing, University of Toronto. 154 McGrail, K., R.G. Evans, M.L. Barer, S.B. Sheps, C. Hertzman and A . Kazanjian (1998) "The Quick and the Dead: The Utilisation of Hospital Services in British Columbia, 1969 to 1995/96," UBC Health Policy Research Unit Discussion Paper 98:3D. Mclnnes, C. (2000) "Ottawa's Official Policy: No Nurses Needed," Vancouver Sun. November 21, A l . Meddings, D., K. McGrail, M . Barer, C. Hertzman, S. Sheps, R.G. Evans, A. ICazanjian (1996) "The Eyes Have It: Cataract Surgery and Changing Patterns of Outpatient Surgery," UBC Health Policy Research Unit Discussion Paper 96:7D. Myers, G.W. (1954) "Hospitalization Experience of a Government Hospital Care Insurance Plan, Canadian Journal of Public Health, 45: 372-380. Needleman, J., P. Buerhaus, S. Mattke et al. (2002) "Nurse Staffing Levels and the Quality of Care in Hospitals," New England Journal of Medicine, 346(22): 1715-1722. Nichols, M . (2001) "T Wanted To Help People': Nurses No Longer Have Time to Provide Personal Attention and Emotional Support," Macleans Magazine, January 8. O'Brien-Pallas, L., D. Thomson, C. Alksnis and S. Bruce (2001) "The Economic Impact of Nurse Staffing Decisions: Time to Turn Down Another Road?" Hospital Quarterly, Spring (2001): 42-50. O'Brien-Pallas, L., A. Baumann and J. Dunleavy (1999a) "Fact Sheet: Nurse Supply in Ontario," Nursing Effectiveness Utilization and Outcomes Research Unit, University of Toronto Faculty of Nursing and McMaster University School of Nursing. O'Brien-Pallas, L., A. Baumann and J. Dunleavy (1999b) "Fact Sheet: The Hospital Sector 1992 - 2000: A Changing Workplace for Nurses in Ontario," Nursing Effectiveness Utilization and Outcomes Research Unit, University of Toronto Faculty of Nursing and McMaster University School of Nursing. Payer, L. (1988) Medicine and Culture: Varieties of Treatment in the United States, England, West Germany, and France, New York, Penguin. Phelps, C. (1997) Health Economics, Addison-Wesley Educational Publishers. Pipel, D., G. M . Fraser, J. Kosecoff, et al. (1992) "Regional Differences in Appropriateness of Cholecystectomy in a Prepaid Health Insurance System," Public Health Review, 20: 61-74. Priest, L. (1997) "The Low Scar Surgery With a High Risk," Toronto Star, September 21: A l . 155 Rachlis M . , R.G. Evans, P. Lewis and M.L'. Barer (2001) "Revitalizing Medicare: Shared Problems, Public Solutions," Tommy Douglas Research Institute, Vancouver. Randhawa, J. and R. Riley (1995) "Trends in Hospital Utilization, 1982-83 to 1992-93," Health Reports, 7(1): 41-49. Reinhardt, U. (1973) "Manpower Substitution and Productivity in Medical Practices: Review of Research," Health Services Research, 8(3): 200-227. Reinhardt, U . (1975) Physician Productivity and Demand for Health Manpower, Cambridge, Ballinger Publishing. Rice, T. (1998) The Economics of Health Reconsidered, Health Administration Press. Robinson, G.C. and H.F. Clarke (1980) "Day Care Surgery," in Robinson, G.C. and H.F. Clarke eds. The Hospital Care of Children: A Review of Contemporary Issues, Oxford University Press. Roemer, M.I. and M . Shain (1957) "Hospital Utilization Under Insurance," Hospital Monograph Series 6, American Hospital Association. Roy, A.D. (1951) "Some Thoughts on the Distribution of Earnings," Oxford Economic Papers, 3: 135-46. Ryten, E. (1997) "The Future Supply of Registered Nurses in Canada," Discussion Paper, Canadian Nurses Association. Sabuhoro, J.B (2001) "Determinants of Admissions to Nursing Programs, Labour Force Participation, and Hours Worked," paper presented at the 9 t h Canadian Conference on Health Economics, Toronto, Canada. Sanmartin, C , S. Shortt, M . Barer, S. Sheps, S. Lewis and P. McDonald (2000) "Waiting for Medical Services in Canada: Lots of Heat, but Little Light," Canadian Medical Association Journal, 162(9): 1305-10. Schumacher, E.J. (2001) "The Earnings and Employment of Nurses in an Era of Cost Containment," Industrial and Labor Relations Review, 55(1): 116-132. Scott, W. (2001) "Nurse Workforce: Condition Critical," National Health Policy Forum Issue Brief no.763, The George Washington University: 1-9. Sharon, B., C. DeCoster, J.T. Waddell, C. Burchill, and S. De Haney (2001) "Acuity of Patients Hospitalized for Medical Conditions at Winnipeg Acute Care Hospitals," Manitoba Centre for Health Policy and Evaluation. 156 Shields, M . A . and M . Ward (2001) "Improving Nurse Retention in the National Health Service in England: the Impact of Job Satisfaction on Intentions to Quit," Journal of Health Economics. 20(2001): 677-701. Sibbald, B. (2002) "The Nursing Crisis: Physicians Should Ponder What This Wil l Mean." Canadian Medical Association Journal. 167: 535. Sibbald, S. (1999) " R N = Really Neglected, Angry Nurses Say," Canadian Medical Association Journal 1999(160): 1490-1. Staiger, D.O, Spetz, J. and Phibbs, G. (1999) "Is There Monopsony in the Labor Market? Evidence from a Natural Experiment," N B E R Working Paper 7258. Staiger, D.O, D.I. Auerbach and P.I. Buerhaus (2000) "Expanding Career Opportunities for Women and the Declining Interest in Nursing as a Career," Nursing Economics. 18(5): 230-6. Statistics Canada (1991), "Classification of Occupation, 1991," Statistics Canada, Ottawa, 93-565E. Steffenhagen, J. (1998) "Shortage of Nurses 'Threatens Care'" Vancouver Sun. August 29, B7. Steinhauer, R. (2001) "U.S. Nurse Feels Cold Welcome," Vancouver Sun. January 10, A15. Siu, A . L. , F. A . Sonnenberg, W. G. Manning, et al. (1983), "Inappropriate Use of Hospitals in a Randomized Trial of Health Insurance Plans," New England Journal of Medicine. (315): 1259-1266. Tully, P and E. Saint-Pierre (1997) "Downsizing Canada's Hospitals, 1986/87 to 1994/95," Health Reports. 8(4): 33-39. Vujicic, M . (1998) "Percutaneous Transluminal Coronary Angioplasty: Technological Progress and Use," mimeo. W C W L (2001) "From Chaos to Order: Making Sense of Waiting Lists in Canada, Final Report," The Western Canada Waiting List Project. Wennberg J.E., Freeman J.L., Culp W.J. (1987) "Are Hospital Services Rationed in New Haven or Over-utilized in Boston?" Lancet, 1: 1185-1188. Wennberg J.E., Freeman J.L., Shelton R .M. , Bubolz T.A. (1989) "Hospital Use and Mortality Among Medicare Beneficiaries in Boston and New Haven," New England Journal ofMedicine. 321: 1168-1173. 157 Wigod, R. (1999) "Lack of Nurses Forces V G H to Shut ORs," The Vancouver Sun. September 18: B3. Wright, C J . and Y . Robens-Paradise (2001) Evaluation of Indications and Outcomes in Elective Surgery, RESIO. Wright, C J . and K . Cardiff (1998) "The Utilization of Acute Care Medical Beds in Prince Edward Island," U B C Health Policy Research Unit Discussion Paper 98:14D. Wright, C J . (2002) "Health Care Supply and Demand: Maybe Less is More," presentation to the 10 th Canadian Conference on Health Economics, Halifax, Nova Scotia. Wright J.G., Hawker G.A., Bombardier C , Croxford R., Dittus R.S., Freund D.A., et al. (1999) "Physician Enthusiasm as an Explanation for Area Variation in the Utilization of Knee Replacement Surgery," Medical Care, 37(9): 946-56. Unruh, L . Y . (2000) "The Impact of Hospital Nurse Staffing on the Quality of Patient Care," University of Notre Dame, PhD thesis. 158 APPENDIX Additional Information Pertaining to Data - Part I Registered Nurses Database: 1980 - 2000 (RNDB) This data is based on information collected from the registration form that every R N in Canada must complete once a year. Data from 1980 - 2000 were accessed through the Graduate Student Data Access Program at the Canadian Institute for Health Information. Individual records were requested for Alberta, B C , Quebec and Ontario but the regulating bodies in Quebec and B C would not grant access to these individual records. However, CIHI agreed to provide cross-tabulations based on the individual data. This method was used for the majority of data access. CIHI would not release data on the number of RNs employed in hospitals prior to 1994 as these data were not audited for consistency across time. The data points prior to 1994 cited in this thesis are based on published studies that are, in turn, based on the non-audited RNDB place of employment variable. In 1990 and 1991 Quebec reported only the number of RNs registered. No details were given on employment status. Data prior to 1980 were gathered from secondary sources. Census of Canada: 1991,1996 Census data is used to construct a random sample of all individuals in Canada whose major field of education is nursing. As defined earlier, this is a sample of the potential supply of nurses in Canada. 159 Census analysis is based on data from the Census of Canada Master File located at Statistics Canada . This file contains detailed education and employment information for the one in five Canadians who complete the long form in Census years. There are three main advantages of using this file instead of the public use file (PUMF). First, the master file is much larger. It is a 20% sample of the Canadian population whereas the P U M F is roughly a 3% sample. The larger sample size allows for more powerful analysis and more detailed cross-tabulations. Second, the 'Occupation' field in the PUMF is highly aggregated. There are 25 categories of occupations in the PUMF file that are aggregated from 514 detailed occupations described in Classification of Occupation, 1991 (Statistics Canada, 1991) that are available only in the master file. The detailed occupational analysis of section 3.3 would not be possible using the PUMF. Third, the 'Major Field of Education' field in the P U M F is aggregated from more detailed data that is available only in the master file but the method of aggregation is inconsistent across time. In 1991 the PUMF category 'Nursing and Nursing Assistance' is an aggregation of the detailed educational field categories 399 - 410 from the master file 8 1 . In 1996, however, the PUMF category 'Nursing' is an aggregate of categories 399 - 406 only. Detailed categories 407 - 410 were included in 'Other Health Professions, Sciences and Technologies' along with 58 other non-nurse categories. The inconsistent aggregation method results in several undesirable effects associated with the PUMF data. First, since individuals with educational fields 407 - 410 are expected to have lower wages than those with fields 399 - 406, it is not possible to make wage comparisons across time in a consistent manner. Second, i f any individuals have their major filed of education classified under fields 407 - 410 but actually work as RNs, the 8 1 These categories are: Nursing - General (399), Critical Care Nursing (400), Geriatric Nursing (401), Medical, Surgical, Hospital Nursing (402), Obstetric Nursing (403), Psychiatric Nursing and Mental Health Care (404), Public Health and Community Nursing (405), Nursing - Other (406), Nursing Assistant, 160 1996 PUMF file can not be used as a 3% sample of the potential supply of RNs in Canada. In fact, this is the case. As Table 12 shows, in 1991 22,985 individuals with educational category 407 - 410 were employed as RNs 8 2 . In 1996, the number was 22,710. This demonstrates that roughly 10% of the R N workforce is composed of individual with education category 407 - 410. Furthermore, roughly 10,000 individuals with education category 399 - 406 were employed as LPNs. This represents over 25% of the L P N workforce. Clearly, there is enough classification error with respect to either major field of education or occupation or both so that the 'Major Field of Education' field in the Census data can not be used to construct separate samples of the potential supply of RNs and LPNs. A l l Census data analysis of the potential supply of nurses, therefore, should be interpreted as pertaining to the nursing workforce as a whole. Since Statistics Canada would not release the actual master data to the Regional Data Centre located at U B C , analysts at Statistics Canada performed cross-tabs and regressions through a remote access protocol. Census data is regarded, in general, as extremely accurate because of the large sample size (20% of the population). Using published data and primary data from the Registered Nurses Database one may check the accuracy of the data as it pertains to the nursing sector. According to the Census there were 213,150 individuals working as RNs in Canada in 1991 and 213,875 in 199683. According the RNDB there were approximately 229,046 Assistant Nursing (407), Health Care Aide/Support (408), Long-term Care Aide (409), Nursing Aide, Orderly (410). 8 2 The Census uses the term Registered Nursing Assistant (RNA) which is equivalent to LPN. 8 3 Individuals with occupation codes Dill (HeadNurse/Supervisor) and Dl 12 (RegisteredNurse) were counted as RNs in the Census data . 161 Table 12. Analysis of Census PUMF Major Field of Education Categories 1991 Occupation Within Health Care Field Educational Field RN(D111, D112) LPN (D233) Total 399-406 190,165 10,160 200,325 407-410 22,985 28,745 51,730 Total 213,150 38,905 252,055 RN/LPN Ratio 5.5:1 1996 Occupation Within Health Care Field Educational Field RN(D111, D112) LPN (D233) Total 399-406 191,165 9,040 200,205 407-410 22,710 21,745 44,455 Total 213,875 30,785 244,660 RN/LPN Ratio 6.9:1 Occupation Categories (Within Health Care Field) D111 Head Nurse/Supervisor D112 Registered Nurse D233 Registered Nursing Assistant (Includes Registered Nursing Assistant, Certified Nursing Assisstant, Licensed Nursing Assistant, Licensed Practical Nurse, Nursing Assistant, Operating Room Technician, Registered Nursing Assistant, Surgical Technician; Excludes Nursing Aides, Orderlies, Ward Aides) Major Fields of Education Categories 399 Nursing - General 400 Critical Care Nursing 401 Geriatric Nursing 402 Medical, Surgical, Hospital Nursing 403 Obstetric Nursing 404 Psychiatric Nursing and Mental Health Care 405 Public Health and Community Nursing 406 Nursing - Other 407 Nursing Assistant, Assistant Nursing 408 Health Care Aide/Support 409 Long-term Care Aide 410 Nursing Aide, Orderly 162 individuals working as RNs in 19918 4 and 229,162 in 1996. This represents a Census undercount of working RNs of roughly 7% in each of the years85. According to the Census, the R N / L P N ratio in Canada was 5.5 in 1991 and 6.9 in 1996. Dussault (2001) finds that the ratio was 2.8 in 1992 and 3.0 in 1997. This suggests a considerable undercount of LPNs in the Census data86. This represents an arithmetic average of the 1990 and 1992 figures since data for 1991 is not available. 8 5 Possible reasons for an undercount include the discrepancy between the survey date of the Census (fixed) and the RNDB (any time during the year), the method used by Statistics Canada to weight individual observations in aggregating up to the population file, and to classify self-reported occupation categories written on the Census form into 514 detailed categories. 8 6 See footnote 5. In addition, it may also be the case that a significant portion of LPNS are employed in occupations related to health care that Statistics Canada classifies as non-health care (e.g. Home support worker which is classified under "Sales and Service Occupations"). 163 Additional Information Pertaining to Data - Part II Annual Return of Health Care Facilities: Part 1/Part 2 (HS1/2) The methods for adjusting these data to account for under-reporting and non-reporting are summarized in Figures A and B. Unusual Patterns in HS1/2 Data No provinces recorded number of visits to the emergency room. As a result, nursing hours per outpatient visit was calculated excluding all emergency room activity. This method is consistent because of two characteristics of the data. First, emergency room visits that involve day surgery or an inpatient admission are also recorded as separations under those two categories. The only emergency room visits that are not recorded elsewhere are same-day episodes that do not involve surgery. Including emergency day surgery cases and emergency inpatient admissions would introduce double counting. Second, emergency room nursing hours are recorded separate from day surgery and inpatient nursing hours. Since the only visits excluded from the denominator are outpatient visits that came through the emergency room, it makes sense to exclude emergency room nursing hours from the numerator when calculating nursing hours per outpatient visit. British Columbia Paid nursing hours per patient day for long term care fluctuates considerably from year to year. 164 Paid nursing hours per patient day for acute care spiked sharply in 1977/78. Since paid nursing hours were stable, this suggests that patient days dropped sharply only for that single year, which is unlikely. Paid nursing hours per separation for ambulatory care doubled from 1976/77 to 1980/81. This is quite a large increase for a four-year period. Alberta Paid nursing hours for long term care decreased by 80% from 1989/90 to 1990/91. Almost all of this decrease was accounted for by a decrease in paid nursing hours in the Extended Care Unit (8,792,589 hours in 1989/90 versus 223,213 in 1990/91). It is highly improbable that this reduction represented a real decrease in nursing hours. Nor is it the case that these hours were simply redefined and captured under another category since total paid nursing hours also dropped significantly over this period. Paid nursing hours per day surgery doubled from 2.0 in 1988/89 to 4.0 1991/92 - a remarkable increase over a three-year period. Even more questionable is the quadrupling from 0.5 to 2.0 between 1978/79 and 1991/92. Paid nursing hours per patient day in long term care fluctuated significantly between 1987/88 and 1993/94 - sometimes over 20% from year to year. Quebec Between 1984/85 and 1985/86 a portion of acute care patient days were reclassified as long term care patient days. Long term care patient days increased from 3,601,685 in 1984/85 to 7,413,819 in 1985/86 while acute care patient days decreased by roughly the same amount. However, there seems to have been no corresponding reclassification of inpatient nursing hours. As a result, there is a one time sharp increase in paid nursing 165 hours per patient day in acute care and a corresponding decrease in paid nursing hours per patient day in long term care between 1984/85 and 1985/86. Paid nursing hours per day surgery quadrupled from 0.5 to 2.0 between 1977/78 and 1979/80 and then remained quite stable. This anomaly in the earliest years resulted from a 250% jump in paid nursing hours in surgical day care (203,390 in 1979/80 versus 60,595 in 1977/78) and a decrease in the number of cases (95,040 versus 153,000). It is unlikely that these drastic changes were anything except counting anomalies. There are zero visits to day/night care programs after 1978/79. In that year there were roughly 12,000 visits. 166 Figure A - Adjusting for Under-Reporting Goal: To distribute "Total Nursing Hours"( 191) and "Total Nursing Salaries"(192) into their various components for those facilities that report only the total and do not provide the detailed breakdown. Procedure: 1. 2. 4. Take all facilities that provide the breakdown of the total Group them based on size. There are six categories: 1-24, 25-49, 50-99, 100-199, 200-299, 300+ staffed beds For each category calculate the proportion of "Total Nursing Hours", "Total Nursing Salaries" accounted for by a particular component Multiply "Total Nursing Hours" and "Total Nursing Salaries" by this proportion for all facilities in a given group. There will be six different proportions based on the six bed-size categories A l l Facilities Facilities Providing Details Facilities Providing Only Total ADMIN/TOTAL, X TOTAL, ADMINflttedJ LNGTRM / TOTAL, X TOTAL, = LNGTRMflttedii Calculate Amounts "Adjusted for Under-Reporting": ADMINtotai i = ADMIN, + ADMIN fllted,i i = 1....6 ADMIN total = S ; ADMINtotal.i Implicit Assumption: Distribution of "Total" among facilities not reporting details is identical to the distribution within facilities of a similar size that do report details. 167 Figure B - Adjusting For Non-Reporting Goal: Procedure: To create a time series that accounts for complete non-reporting on the part of some facilities. This becomes especially important in latter years when reporting rates decrease significantly. 1. Take all facilities that completed the survey in a given year 2. Group them based on size. There are six categories: 1-24, 25-49, 50-99, 100-199, 200-299, 300+ staffed beds 3. Calculate the number of staffed beds the survey accounts for in each category in each year 4. Compute this as a percent of all staffed beds in the province 5. To adjust a series for non-reporting, the series is first disaggregated by facility size. Then in each year, the value is divided by the appropriate year-facility size factor calculated in #4. Ot-;/ = Provincial Staffed Bed Capacity Captured by Survey in year t for size group i. - Scaling Factor Zadj.u ~ Zi,t I where Z = Nursing Hours, Patient Days,..., etc. -•adj.t t-i z-'adjI Implicit Assumption: In each year and for facilities of similar sizes, the level of any variable - on a per bed basis - among facilities not completing the survey is identical to the level in facilities that complete the survey. 168 Registration Requirements for Registered Nurses in British Columbia The following is required for registration as a practicing registered nurse with the Registered Nurses Association of British Columbia (RNABC): For First Time Registrants (This also includes new graduates from other provinces/countries seeking employment in BC) • Must be a graduate from a nursing education program • Must have completed the Canadian Registered Nurses Examination • Must undergo criminal records check • Must complete English competency test • If educated outside Canada, must complete Qualifying Program For Members Currently Registered with R N A B C or Other Provincial Nursing Association Practice Hours Requirement • In last 5 years must have 1,125 hours practicing as registered nurse, or • Must have graduated from a nursing program in past year, or • Must have graduated from a refresher course, or • Must have practiced 500 hours in the last 5 years plus completed the reduced hours option (i.e. must have either completed 100 hours of learning activities or must have performance appraisals from three practicing registered nurses) Personal Practice Review Requirement • In last year must have completed a self-assessment of practice based on the Standards for Nursing Practice in British Columbia, and • Must have obtained peer feedback, and • Must have developed and implemented a learning plan based on the self-assessment and peer feedback, and • Must have evaluated the impact of the learning on your practice. Have Not Worked in the Previous Year but Wish to Renew, or Convert From Non-Practicing to Practicing Status • Must complete a self-assessment of your practice based on RNABC's Standards for Nursing Practice in British Columbia, and • Must develop and implementing a learning plan, based on your self-assessment, to address areas you want to enhance in your practice For Those Wishing to Register but Having Practiced Less Than 500 hours in the Past 5 years • Must complete refresher course (ranges from 8 - 2 4 weeks), or • Must complete 400 hours of precepted experience. 169 

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.831.1-0091303/manifest

Comment

Related Items