UBC Faculty Research and Publications

Does health care support independence or threaten it : a population based, person specific analysis of… Evans, Robert G., 1942-; Barer, Morris Lionel, 1951-; Hertzman, Clyde, 1953-; Sheps, Samuel Barry; Kazanjian, Arminée, 1947- Jun 30, 1998

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

Item Metadata


52383-Evans_RG_et_al_Support_independence.pdf [ 15.18MB ]
JSON: 52383-1.0048538.json
JSON-LD: 52383-1.0048538-ld.json
RDF/XML (Pretty): 52383-1.0048538-rdf.xml
RDF/JSON: 52383-1.0048538-rdf.json
Turtle: 52383-1.0048538-turtle.txt
N-Triples: 52383-1.0048538-rdf-ntriples.txt
Original Record: 52383-1.0048538-source.json
Full Text

Full Text

SENIOR'S INDEPENDEN~~~SEARCHPROGR_AM (SIRP) AND CANADA'SDRUG STRATEGY (CDS):PROGRAMS OF APPLIED RESEARCHDoes Health Care Support Independence or Threaten It:A Population Based, Person Specific Analysis ofPatterns of Use by SeniorsNHRDP project No. 6610-2117-602Robert G. Evans'>'Morris L. Barer-vClyde Hertzman-vSamuel B. Sheps?Arminee Kazanjian>[Department of Economics, University of British Columbia2Centre for Health Services and Policy Research, University of British Columbia3Department of Health Care and Epidemiology, University of British Columbia"Population Health Program, Canadian Institute for Advanced ResearchFor Further Information Please Contactthe Investigators at:Centre for Health Services and Policy Research429-2194 Health Sciences MallVancouver, BC V6T 1Z3phone: 604-822-4810fax: 604-822-5690Release Date: 30 June 1998Funded by: Health CanadaNational Health Research and Development Program (NHRDP)through a special competition co-sponsored by theSenior's Independence Research Program (SIRP) andthe research arm of Canada's Drugs Strategy (CDS)Telephone: (613)-954-7943fax: (613)-954-7363TABLE OF CONTENTSPage No.Executive Summary 3Overview of the social and theoretical importance of the work 4Seniors ' Use of Health Care: The Long Term Trends 4The Era of "De-Institutionalization" 5The Linked Health Data Base in BC 6Overview of the research methodology 7Objectives 7Methodology 7Phase I: A cross-sectional analysis of trends in the use ofhospital and physician services 8Hospitals 8Physicians 8Phase II: An analysis of the use of health care services prior to death 11Phase III: A longitudinal cohort analysis 12Statement of major findings 15Phase I: A cross-sectional analysis of trends in the use ofhospital and physician services :.. 15Hospitals.. .... .... ...... .................... .................... ... .................. ....... .......... ............. 15Day Surgery 16Physicians................. ................... ................. ................ ....... ..... 16Phase II: An analysis of the use of health care services prior to death 18Phase III: A longitudinal cohort analysis 19Potential practical and theoretical uses of the research fmdings and methods 21Phase I: A cross-sectional analysis of trends in the use ofhospital and physician services 21Phase II: An analysis of the use of health care services prior to death 22Phase ill: A longitudinal cohort analysis 22References 24APPENDICESAppendix 1 27Appendix 2 28Appendix 3 29(e'I (""I'OI Health ,\eriv;('(" and Policy Research429·2/94 Health S, ien rrs Mallvuucnnve) , Be V611 7.i.lunc /9982EXECUTIVE SUMMARYMajor Findings1 During the 1980s, decreasing rates inhospital use occurred even though rates forthe elderly were continuing to increase.During the 1990s, downsizing has affectedall age groups, the elderly included. Daysurgery has increased dramatically, butthere is no clear pattern of a transfer of carefrom inpatient to outpatient services.2 Over this same period, age-specific percapita expenditures on physicians continuedto increase, mostly because of increasedspending on specialists. These increasesoccurred because a slightly largerproportion of Be's population was referredto specialists, they were more likely to see alarger number of specialists than in thepast, and they were being provided withmore expensive services (after adjusting forfee increases).3 People who died in hospital accounted foran increasing proportion of total hospitalcare, but only because their use wasdeclining at a slower rate than non- .decedents. In the mid-1990s, people whodied were as likely to use facility-based careas their counterparts in the 1980s, but hadshorter lengths of stay. They were far morelikely to use community-based services suchas homemaker care and doubled their use ofpharmaceuticals. Despite claims to thecontrary, only a small fraction of theincrease in pharmaceutical costs can beexplained by hospital downsizing.4 Those who lost independence were alsoaffected by downsizing, with those in themid-1990s using fewer facility-basedservices. More importantly, patterns of careprior to loss of independence suggest arelatively rapid transition, which indicateseither that there may have been unmetneeds prior to the loss of independence, orthat many people went through thistransition because of a 'traumatic' event.Cell INJ(Jr Healtli ,>erivin'\ and {'oli,'." Re searrh429-2/94 Health S, ienrrs MallV(lfU:OUV"J : sc \16r 1/3Iuu« 199/:1PolicyfProgram Implications1 The complex outcomes that have occurred as a result ofhospital downsizing suggest that there should be a moreexplicit coordinating (and perhaps budgetary) link betweenfacility-based and community-based care. Some mechanismsfor this have been put into place with 'regionalisation'. Theseshould be strengthened to ensure continuity of care, and tominimise 'off-loading' of clinical care from the formal healthsystem to informal caregivers.2 Since the years on which these analyses are based, BC hasmoved to a 'hard cap' system for physician reimbursement.The hard cap has been relatively successful in controllingphysician expenditures, but these analyses are quite revealingin what expenditure shifts have taken place 'under the globe',and should be useful to policy-makers in identifying areas thatmay cause pressure in the future. For example, if specialistsare 'better' at keeping their portion of the global budget up,there may be interest in moving to split caps for GPs andspecialists, or perhaps in the creation of regional/specialtycaps.3 The doubling in adjusted pharmaceutical costs among theelderly -- for all age groups, and all 'groups' of patients,whether they used hospital services or not -- suggests thatpharmaceutical expenditures must become a high priority areafor policy attention. The BC program of reference-basedpricing is one such form of intervention; calls for a nationalPharmacare program are another. Any policy approach inthis area would benefit from being both top-down (e.g. globalbudget) and bottom up initiatives to improve theappropriateness and cost-effectiveness of prescribingpractices. Policy in this area would also benefit fromcoordinated action, at least at the provincial level, if not thenational, or there is little hope of having control in thisrapidly increasing sector.4 A decreased use of facility-based services in the 1990smeans that the BC government policy of 'closer to home' mayhave been achieved. What should remain of concern iswhether this was accomplished through the deployment ofmore suitable health care services, or at the expense ofinformal caregivers.3OVERVIEW OF THESOCIAL ANDTHEORETICAL IMPORTANCE OFTHEWORKSeniors' Use ofHealth Care: The Long Term TrendsThe increasing use of health care by seniors is widely documented in the literature, particularlyin Canada. In popular discussions this is usually attributed to the aging of the population. But theattribution is false , and the resilience of this error in the face of repeated refutation is itself worthyof some exploration [Barer et aI., 1994a; Barer et al., 1998] Two decades ago Boulet andGrenier [1978] demonstrated the very limited impact of demographic forces per se on health careuse. More recent studies have consistently found large increases over time in age-specific userates by seniors; these include: hospital use (acute and extended care) , physicians' services, anddrugs in BC [Evans et aI., 1989; Hertzman et aI., 1990; Barer et aI., 1989; Anderson et aI., 1990;Anderson et aI., 1993]; hospital use and physicians' services in Manitoba [Roch et aI., 1985; Bareret aI., 1994b]; hospital use, physicians' services and drugs in Saskatchewan [Gormley et aI., 1990];and Canada-wide [Nair, 1991]. Less detailed but consistent data are found in the US [Barer etaI., 1992] and Sweden [Evans, 1991].All studies show substantial increases in the share of health care services taken up by seniors,increases which are largely unexplained by changes in population structure. Barer et ai. [1994a]provide an updated survey. The increases in time spent in hospital-based long term care by the"oldest-old" are particularly marked [Evans et aI., 1989]. These patients are predominantlysuffering from psychogeriatric disorders or related chronic conditions which would seem topreclude independent living, yet curiously there is no evidence in the wider epidemiologicalliterature of an "epidemic" of Alzheimer's disease or other psychogeriatric problems sufficient toexplain the large increases in institutional use. If the population is not "sicker", why is there somuch more care?Common explanations offered by those "in the field" are that the resources that oncesupported such people in the community are no longer available. A darker view is sometimesexpressed, that for the very elderly the experience of institutional care itself destroys the capabilityfor independent living. Our impression is that this view is quite widely held, though rarelyexpressed publicly; remarkably, recent findings in neuro-physiology appear to be providingsupportive evidence [Sapolsky, 1993].These explanations are not in conflict, however, and both support the general opinion as tothe desirability of greater independence for seniors -- better quality of life, and (less clearly)perhaps some reduction in costs of health care. Such "independence" seems to be represented byliving at home or in a more home-like atmosphere -- the hospital is not the only model forcongregate or communal living! But there is a more subtle question that is more rarely addressed.Does not the increased intensity of health care servicing of all types represent a form ofdependence which extends beyond the institution?If the home becomes a "therapeutic milieu" , the "hospital without walls" has in effect extendeditself into the home. The patient is dependent upon drugs, regular diagnostic and otherinterventions, home visits and other support services. The overwhelming opinion seems to be thathomes are still better than hospitals. But a broad reading of the concept of "independence" shouldlead to some concern for the implications of "community-based care" in the context of ever­increasing intensities (per capita rates) of servicing. The theme of "Closer to Home" was intendedto mean instead of in hospital; if interpreted as support for a further overall expansion ofservicing of the elderly its implications for independence are much more ambiguous.Centre tor tlcalt}: ~'el'iI'i('I'S and Policv Rescarrli429-2194 Healt], Sriern n Mu]!V(lI/(:OIl I-er. lie vsr 1i'3June 1(1984We did not include, in the present program, any specific research proposals for addressingthese deeper questions. But the linked data set now available will permit us to identify the wholerange of services being received by each individual, and so at a later stage to draw inferencesabout the actual conditions of those in "independent" versus "institutional" living.The Era of "De-Institutionalization"On the other hand, since the mid-1980s a process of "de-institutionalization" has beengathering momentum in Canada as in many other countries. There has been a slow decline inrates of inpatient use in Canada, starting in the early 1970s. Per capita use of General and AlliedSpecial hospitals rose from 1632.9 patient days per thousand population in 1961 to 1907.6 in1971, but fell to 1805.7 in 1981/2 and to 1529.6 in 1991/2, or 6.33% below the 1961 level [Nair,1991; unpublished data, 1994]. The most rapid decline was in the latter part of the 1980s.In BC the number of patient-days per capita in acute and rehab. beds fell by 22% between1987/8 and 1992/3. Thus the provincial government's "New Directions" program was launched atthe end (February, 1993) of a five-year period which had already seen the largest and mostprolonged contraction in inpatient use in our history. Nor has there been any offsetting increase inother forms of institutional care .The first order of business for this research program was to identify the distribution of theserecent reductions by age-group. Even before starting, we knew that there had been a markedincrease in the use of day care surgical facilities by elderly people, which might suggest acorresponding decrease in inpatient use. But previous studies in paediatrics showed thatcoincident increases in day surgery and decreases in inpatient use were in fact very weakly linkedcausally. Most of the decrease in patient days was among non-surgical diagnoses, or surgicaldiagnoses (T&A) which were not at that time considered eligible for day care [Evans et al., 1988].In any case the observation of a recent and large drop in use, combined with the introductionof policies which include the expectation of substantial future cuts, raised obvious questions notonly about which groups in the population are experiencing the reductions, but also about theeffects upon them and upon the rest of the health care delivery system. By the beginning of 1994,some had begun to describe inpatient bed capacity and use as being in "free fall", in response to avariety of recent changes in provincial funding and labour relations policies. They feared that"over-shooting" might be placing patients at risk . On the other hand, the more recent data raisethe possibility that the Seaton Commission's recommended targets were too conservative,underestimating the potential for "downsizing" inpatient care, and moving health care services"Closer to Home". If institutional use by the elderly was being substantially reduced, with moreto come, was it possible that seniors were being "squeezed" rather than "supported" intoindependence?Conventional wisdom among care-givers is that greater community support -- i.e. resources -­is an essential precursor to reduced use of institutions and greater independence. But both thelong term increase, and the recent decrease, in institutional use occurred without identified ormeasured offsetting changes in community support. The decline in informal care-giving andsupport -- the increase in "needs" -- which is typically asserted as the "cause" of the previousexpansion is not independently documented, only inferred from the expansion itself.It may have happened, but one must recall that study after study, at different times in differentjurisdictions, has found that the principal determinant of institutional use is institutional capacity -­"if you build it, they will come" . And if you shut it, they won't. An independent role forCent re IIJr Healtn Scrivices and Policv Rcscn rcl:429-21 94liealth Scie nce »MallVancouver. Be HiT 113June I CJ985population needs in influencing institutional use is difficult or impossible to find. And yet theremust be some connection, at some level. Hence the importance of developing a morecomprehensive picture of what is happening to the patterns of care (and to the extent possible,outcomes) for seniors as the acute care system or at least bed use, shrinks.The Linked Health Data Base in BeThe work of the BC Linked Health Data set (BCLIID) Project has created a powerful newcapability for addressing such questions. This Project, undertaken and housed at the UBC Centrefor Health Services and Policy Research and supported by the BC Ministry of Health, hassuccessfully attached a (coded) common patient identifier to each of the service records fromHospital Programs, the Medical Services Plan, Pharmacare Plan A (seniors) the Continuing Caresystem, and births and deaths, over the period from 1°85/6 to 1995/6 [Chamberlayne et al.,1998]. It is now possible to define the trajectory of health care use by a particular individual,from 1985/6 to 1995/6, as reflected in these records. As far as we know, the only other researchgroup with such a capability at the present time is the Manitoba Centre for Health Policy andEvaluation, which pioneered the development of this sort of data base (MCHPE, 1993), althoughlinkage within a single year has recently been used to study drug use in New Brunswick[Davidson et aI., 1995].Historical data linkage is largely a probabilistic process ; the BCLIID Project has succeeded inlinking more than 95% of each of the program areas in each year. These are high success rates,comparable to those being achieved by Manitoba researchers [Roos and Wajda, 1990].The development of the BCLIID Project proceeded under the direction of a jointManagement Committee representing both the Ministry and the UBC Centre. This Data AccessCommittee continues to oversee the routine updating and utilization of the data base, dealing withthe 'ethical challenge' of proposed projects and the use and dissemination of results, underguidelines set out in the Access Policy for Research Uses ofLinked Health Data , which has beenin effect since April 1996. These guidelines were designed to comply with BC's Freedom ofInformation and Protection ofPrivacy Act. Thus the link between research and public policy hasbeen in place from the beginning, providing a direct channel for dissemination of results inaddition to the usual presentation and publication of research papers.Since this research program involves analysis of retrospective data, rather than interventions,the only ethical issues which arise are those of confidentiality and privacy. The records used forthis project include no personal identifiers ; they are coded by scrambled versions of identifyingnumbers, which allow analysis at the individual level without the ability to identify a specificperson.Centre [or ttcalth Scriviccs and Policv Rescarcl:429-2/94 Health ~·l.'i~I/I:1'I MallV(JIIC()/(I'£'r, BC V6T 1:1.:1June 19'186OVERVIEW OF THERESEARCH METHODOLOGYObjectives1) to extend previous research on the patterns of health care use by seniors in BC, to documentthe impact of health reforms (e.g. hospital bed closures) on the patterns of service use amongthe elderly (for instance, is there evidence that utilization constraints have increased theproportion of the frail elderly in the community?)2) to develop a data-based indicator of the point in time at which ' loss of independence' may bedeemed to have occurred, more or less permanently, which can then be used in the analysis ofpatient care trajectories among the elderly3) to explore the precursors, correlates, and consequences of loss of independence includingsocioeconomic and geographic factors, and to explore the impact of various policies andprograms which may have an effect on the care trajectories of seniors before and after loss ofindependenceIn order to meet these objectives, the project has been divided into three phases; a 'cross­sectional' phase, to update previous research on trends in hospital and physician serviceutilization; a 'deaths trajectory' phase to look at the impact of time and hospital downsizing onthe care of a (retrospectively) 'at risk' population in the 24 months prior to death; and a 'cohort'phase, to look at longitudinal use, and changes in patterns of use, of health services by the elderly.MethodologyThis was a retrospective study, utilising routinely collected administrative data on vitalstatistics and health services utilisation in the province of British Columbia. Probabilistic linkagemethods were used to identify an individual to whom each utilisation record could be attributed(Chamberlayne et al., 1998). The data files included were: fee-for-service payments to physiciansand alternative providers; acute, rehabilitation, extended, and surgical day care hospitalseparations; assessments and records of service use from the Continuing Care system; paymentsfor pharmaceuticals for community-dwelling elderly aged 65+; and vital statistics records of death.These health care services are universal in BC, and linkage of greater than 95% of records fromeach program file and year was achieved , making this work population-based and individual­specific.All analyses were done using a combination of custom-written perl and C programs, SAS, andExcel. A combination of approaches was used because of the large data sets involved. Thespecific approach taken was determined by the head programmer/analyst in consultation with theprincipal investigators and project manager.The methodological approach for each phase, including the data used and cohort selection,where appropriate, is below . Further details for all phases can be found in publications relating tothose analyses (Meddings et al., 1997; Meddings et al., 1998a; Meddings et al., 1998b; Morgan,et al.. 1997; Barer et al., 1998; McGrail et al., 1998).Ccntr« tor Ilealt}: Serivices antt Policv Rcscarctt429-2!9411ea/r/r S , iaue-, YlallYancouver. Be Y'f>T!Z3June !9lJ87Phase I: A cross-sectional analysis oftrends in the use ofhospital and physician servicesHOSPITALSThis analysis was intended to provide a detailed understanding of how 'hospital downsizing'or 'deinstitutionalization' actually 'played out' in the data for Be over the past decade or so, andto provide an update to previous work by the principal investigators in this area. It is the onepiece of analysis for this project that does not involve linked data.Included are all acute, extended, rehabilitation and surgical day care separations from BChospitals in 1985/86, 1993/94 and 1995/96. The latter year was added well into the project inorder to keep the analysis current, and because we discovered that significant downsizingcontinued to occur in these last two years. Tables were generated outlining separations(discharges from or deaths in hospital) and days of separations (i.e. accumulated lengths of stayfor patients separated), by five-year age group to 90+, and 'bed type'. For purposes ofcomparison to previous work including data from 1969 and 1978, all stays recorded as 'acute'were counted as such, while all other inpatient stays were counted as 'extended' . This means thesmall number of rehabilitation stays were counted as extended, as were stays recorded as 'longterm care in acute', which were present in the 1985/86 and 1993/94, but not the 1995/96 data.Surgical day care events were kept separate, adding to the total number of separations, but not tothe days of separations.Age was calculated at the date of admission, after determining that choosing age at admissionvs. age at discharge made virtually no difference in the results. Days of separations werecalculated by subtracting the date of admission from the date of separation; i.e. admission on Jan 2and separation on Jan 4 of the same year yields a length of stay of 2 days.All numbers are reported as rates per thousand population, using population figures providedby BC Stats' p.E.a.p.L.E. projection model, version 21.PHYSICIANSThis phase of analysis is based on fee-for-service payments to physicians in 1985/96 and1993/94, contained in the Payment Information Masterfile of the Medical Services Plan (MSP) inBe. Fee-for-service payments account for more than 90% of all payments to physicians in bothyears under study, and the linkage rate was greater than 99% in both years. The working data setrepresented payment records totalling approximately $780 million in 1985/86 and $1.27 billion in1993/94. In order to compare utilization patterns, we first needed to remove the effects of feechanges. This was done by valuing each fee item in each of the two years using the same (April 1,1988) fee levels. Details of the methods are described elsewhere [Pascali, 1995]. Removing theeffects of fee changes on expenditures results in fee-adjusted expenditures of $832 million in1985/86 and $1.19 billion in 1993/94.An individual's utilization of physician services in any given period may be described in waysthat reveal considerably more about their patterns of use than summary statistics such asexpenditures or utilization per capita. Specifically, for each person who contacts at least onephysician during a given period, it may be of interest to know how many different physicians theysaw and, for each such physician, how many times the patient was seen, and the patterns ofservicing per visit. Using data from the BCLHD, we were able to disaggregate patterns of use foreach person in the province who received physician services, as follows :Ccnt rr f or Hcnlth. Scrivices rind Polirv Rcscarcli42 <)-2194 Health ~rie" " l'\ ;VlallVrJ/lCf){(I'I'r. He vsr 1/.3June ICNS8of the period). This is not a problem when counting services, contacts, or expenditures, all ofwhich are attributed based on the age of the person at the date of service . But it is a potentiallylarge problem when attempting to create an additive system of MDPPs and DPs . In the case ofDPs, each person who sees a physician during the fiscal year must contribute exactly '1' to theDiscrete Patient count. But in the course of that year, an individual will spend some proportion oftime at age 'x' and 1 minus that proportion at age 'x+ 1'. We have thus 'split' each DP accordingto the proportion of the year spent in each age. The total, rather than a cross-sectional count,represents person-years, by age, of the subset of the Be population who visit a physician at leastonce during that year.MDPPs are a bit trickier, but only because there is no necessary correspondence between thenumber of MDPPs and the number of people who make physician visits. If all visits to aparticular physician occur while an individual is age 'x' or age 'x+ 1', then no additionalcomputation is required; 1 MDPP is attributed to the appropriate age for each physician-patientcombination. If, however, a patient visits a particular doctor both before and after his/herbirthday, into which age category should that MDPP be placed? We considered assigning theseMDPPs based on age at a particular point in time - the beginning of the period, the middle of theperiod, or the end of the period - but in all cases this caused large biases for the under 1 agecategory. Instead, we decided to calculate age at service date for all services, and to assign aproportion to each MDPP based on the proportion of the fiscal year the patient spent in each agecategory. For example, a person born in the sixth month of the fiscal year who received servicesfrom a single MD both before and after the sixth month, this physician-patient pairing wouldcount as 0.5 MDPP at age 'x ' and 0.5 MDPP at age 'x+1', where 'x' is the age of the patient atthe beginning of the fiscal year. (The 1985/86 data contain only service month and year, so forease of computation we have assumed all patients are born on the last day of the month.)But this is really only half the story. The data available are in the form of a payment file,which means records included are for services which were paid during the fiscal year rather thanservices that were received during the fiscal year. This means that services paid during a fiscalyear starting in April and ending the following March actually represent services rendered fromapproximately the beginning of March through the following February. To most accuratelycalculate proportion of time spent in each age, then, a frame of a March to February 'service year'is used rather than an April to March fiscal year. For example, a person born in May would count0.25 towards a discrete patient in age 'x' and 0.75 of a discrete patient in age 'x+l'.Initial Discrete Patient/Population results showed several older age groups with ratios above1.00. We hypothesized that part of what we were seeing were the effects of increased mortalityat these older ages. In our discrete patient calculation, a patient's proportion of time spent in anage was not adjusted for mortality, so we were effectively counting patients in the numerator whowere not in the denominator at the time the population estimates were developed. We consideredmaking mortality adjustments to the discrete patient numbers using linkages between MSP anddeaths information, but were not able to do so because of poor death linkage rates in the first partof 1985/86. As an alternative, mortality adjustments were added to the denominator (population)rather than subtracted from the numerator (discrete patients). One person was added to thepopulation count for each person who died between March 1 (the start of our 'service year') andJuly 1 (the date of the population count). These 'population' figures, then, overstate the numberof people alive at July 1, but accurately reflect the number of people who could have receivedservices at a given age.( 'cntr« tor Health. Scriviccs (Jl ld l 'olicv !?(' .\'f'<II'CIt429-2/94/ha/rh S', ;<'IIU'I MallVW/CO/lI'l'r, Be VIST /"1.3Jun e tvvs10As an example, a person who turned 80 on May 31, and died on June 30 would count as 0.25of a discrete patient in age 79, and 0.75 at age 80 (because in the discrete patient count, mortalityis not considered). In all likelihood, this person would have received services at age 80, butwould not be counted in the population at age 80, because she died before the population count atJuly 1. Therefore, since we are unable to adjust her proportion of 'exposure ' in the numerator,we include her in the denominator, by adding' l' to the population count for age 80 (age atdeath). This may seem like over-counting, but in the example above, without adjusting, theperson contributes' l ' to the numerator (though divided into two age groups), but does notcontribute to the denominator, because they die before the July 1 cross-sectional populationestimate. With the adjustment, the person is included, as '1', in both the numerator anddenominator. 1 These adjustments increased the denominator count by less than 1% for all agegroups except the two eldest; the 75-84 year old group increased by about 2% in both years, and85+ by about 5%. This methodology also had the expected effect of moving all the DiscretePatientIPopulation ratios down to or below 1.00.Phase II: An analysis ofthe use ofhealth care services prior to deathThis phase of the analysis utilised data from all five program areas of the linked data set, asoutlined above. We identified all BC decedents in 1988 and 1993, and using the linked data,extracted all health care utilisation information for the 24 months (730 or 731 days) precedingdeath. Subsequent analyses were limited to those aged 65 and over at the time of death, whocomposed approximately three-quarters of total decedents in both years.Using these data, records of each decedent's level of the following health care utilizationmeasures in the two years preceding death were constructed: (1) the number of days spent inextended or long-term care hospital beds; (2) the number of days spent in acute care hospitalbeds; (3) the number of day surgeries; (4) fee-adjusted payments to physicians and other healthcare providers; (5) price-adjusted pharmaceutical payments (Pharmacare Plan A); (6) the numberof hours of home maker services used; and (7) the number of days spent in adult day care.Two of these measures of health care utilization are fee- or price-adjusted expenditures; therest are utilization counts. Utilization proxies were created from payments made by the MedicalServices Plan (MSP) and Pharmacare Plan A (Pharmacare) on behalf of each decedent, by'adjusting out ' the effects of changes in fees/prices.The adjustment of MSP payments employed the methodology referred to above (under 'Phase1'). Pharmacare payments were all converted to what they would have been in 1993 prices.'Prices' for drugs in 1993 were assigned based on the per-unit reimbursement for each uniquedrug in the 1993 pharmaceutical file, identified by the Canadian Drug Identity Code (i.e., brandedand generic drug costs were calculated separately). For the (very few) drugs that were prescribedin earlier years and not in 1993 , costs were inflated using an overall Laspeyres index ofpharmaceutical prices, calculated at the two-digit therapeutic code level.I The population adjustments were not 'split' between ages as the DP counts were for the sake of simplicity. Sucha split would have required a good deal of extra programming, and would have affected the resulting numbers verylittle.Centre tor Heaith ~'(,I"I\'IC(,S and Polin- Research429-219411ealrh Stienre-; ;\'/al!VOI/COIIIW, Be V6T J:/.3JUIIC 19'1$11Previous analyses of the utilisation and cost of care for the dying have often focused on theelderly decedent population as an aggregate cohort. Comparisons are often made between"decedents'" and "survivors'" use of health care services in a given year. The approach takenhere is different. To better describe treatment patterns, the population of elderly BritishColumbian decedents in 1988 and 1993 was divided into categories according to the nature oftheir exposure to different types of institutional care during the last 24 months of life. Theprincipal criterion for distinguishing decedents was their length of stay (if any) in acute carehospitals and extended or long-term care (nursing home) facilities during their last two years oflife.The linked database is central to this analysis because it allows one to categorize individualsbased on their use of selected services, acute care for example, and then measure their use ofother services conditional on the level of the first. This would not be possible with data files thatdid not permit the analyst to identify services used by a specific individual across different healthcare programs (e.g., the Pharmacare, MSP, and Hospital insurance programs). Categorizingdecedents based on the form of institutional exposure (the conditioning variable) provides themeans to ascertain whether there were changes in (1) the relative magnitudes of populationswithin different institutional settings over time, and (2) the intensity of services provided topatients in similar institutional settings during the two periods of analysis.The division of categories was as follows: Decedents who spent their last 24 months withinhealth care institutions were broken into two groups: those who were in extended care or longterm care facilities exclusively (the 'All Facility, no AC' group), and those who spent at least oneday in acute care (the 'All Facility, with AC' group). This distinction was made because of thedifferences in estimated average costs of caring for patients in these types of beds ($650 per dayin acute care versus $150 per day in extended care)-reflecting differences in the acuity/needs ofpatients.Those decedents who had some form of institutional care, but were not continuously withininstitutions in their last two years of life, were disaggregated into three groups: 'Some Facility' ,'Died in Hospital', and 'Some Hospital'. The 'Some Facility' category are those who spent atlease one night in an extended-care facility or a nursing home in the two years prior to death. Asmall minority of individuals in the 'Some Facility' group did not spend any time in acute care.The 'Died in Hospital' cohort are those who had no contact with either extended care or nursinghomes, but died in an acute care hospital. Those in the 'Some Hospital' group spent at least onenight in acute care during their last two years of life, but did not die in hospital. The distinctionbetween the latter two groups is made because of perceived differences between those who die inhospital and those who die elsewhere---either with formal supports and preparation, or suddenly.Finally, the last of the six categories, 'No Hospital ,' is made up of people who had no in-patienthospital care in the two years preceding their death.Phase III: A longitudinal cohort analysisThe study population in this phase was people aged 65, 75-76, 85-87 or 90-93 on the first dayof January in 1986 and 1993. We limited analysis to specific ages so that comparisons ofutilisation could be made over time without the necessity of adjusting for changes in agecomposition. Increasing the 'band widths' with age was done to ensure that the numbers ofCentre tor Il cnlt}: Seriviccs and Policv Resonrch429-2194 lieu/Iii ~'ne"r:('\ Mot!V(J/J(;f)lI\W, He HST 1/3June /9lJS12people involved would remain relatively high. There were just under 80,000 people included inthe earlier time period, and just over 92,000 in the later, with even the smallest groups (90-93)containing 7,820 patients in the 1986 cohort and 9,158 in the 1993 cohort.Health services utilisation data for these cohorts was extracted for the subsequent three years,or until death, whichever came first. The first part of this analysis involved repeating, with someminor modifications, the methodology developed for the deaths cohorts, as described above. Theonly changes made were to the 'institutional use' groups, removing the 'died in hospital' group,and adding a 'Some home care' group, which included people who did not stay overnight in eithera hospital or long-term care facility, but who did receive home-based (formal) care servicesthrough the continuing care system. The former category was dropped, because it was felt that itwould be of limited use/interest since this analysis is not based on a death outcome, as was theprevious. The latter group was added partly because this phase has a larger focus on community­based care, but more importantly because our knowledge of the data, how to organise them, andwhat measures could reasonably be created, was more robust at this stage of the project.The second part of this phase involved attempting to define the point at which people lost'independence' . Loss of independence (LOI) can clearly be defined in many ways, and it was notour intent, nor do we think it possible, to arrive at the 'definitive' parameters for this construct.Because this project is based on administrative data, we developed the following definition: 'Lossofindependence' occurs at the point in time at which a person becomes permanently dependentupon institutional care (extended, nursing home, or acute), or upon a significant level ofcommunity-based continuing care services.For our purposes, permanent refers to the receipt of services until death. Because our dataare 'right censored' by the end of our study period (except for those who die, but that is arelatively small proportion of the total), we had to estimate our notion of permanent. To do this,we looked at patterns of care during the third year of study for those people who were in facilitiesat the end of the second year. A survival analysis suggested that after about 60 days, individualswere very unlikely to return to the community. This sixty day cut-off was then used at the end ofour three-year study period; i.e. only those people who met our criteria for loss of independencefor a minimum of 60 days before the study period ended were included in the LOI group.We then had to decide how 'a significant level of community-based continuing care services'could be defined within the administrative data. The Be continuing care system (which is auniversal program) assigns 'levels of care' based on an assessment by a case manager.' Theseassessments consider both clinical factors -- how sick the person is, and how much 'professional'care is required -- and social factors -- whether the person lives at home, or has informalcaregiving available. The assigned levels of care are associated with both the type and intensityof services that can be received by an individual through this system. For example, homemakerservices are available at any level of care, but more hours per month are possible for peopleassessed at higher levels of care. Assuming that all care in the community implies a level ofindependence greater than care in a facility, we ranked the levels of care as follows:o=no care -- may be receiving physician services or pharmaceuticals, but nofacility-based services, and no community-based services through Continuing Care2 Definitions for these care levels have not changed over time, though in practice, of course, their applicationprobably has (but this is difficult. if not impossible. to quantify) .Centre [or Health Scriviccs and Policv Rcsearct:429-219411ealch Srienre, A'lallVancouver. Be vsr JZ3June leNS13I =home nursing care, no LTC assessment -- home nursing care is the only carereceived, and is anticipated to be of short duration2 =personal care (community)3 = ICI (community) -- Intermediate care level I in the community4 =IC2 (community) -- Intermediate care level IT in the community5 =IC3 (community) -- Intermediate care level ill in the community6 =EC (community) -- Extended care in the community7 =personal care (facility)8 =ICI (facility) -- Intermediate care level I in a facility9 =IC2 (facility) -- Intermediate care level II in a facilityA= level IC3 (facility) -- Intermediate care level ill in a facilityB =EC (facility) -- Extended care in a facilityC =acute careD =day surgeryThe administrative data often contain 'errors' in the sense that an individual may appear to bereceiving two types of services at one time. For example, he or she may be reported as being inextended care and in acute care on the same day, or, as is perhaps more common, may bereceiving community-based services, but on one particular day also have a 'day surgery' eventrecorded. In these cases, we used the hierarchical structure above to decide which service typewould override. We chose day surgery as the 'highest', not because it is necessarily the mostintense in terms of service, but because it will appear for only one day , and, because of the natureof the data, will often conflict with the other service types.We then had to decide where to draw the line, above which a person would be considered'dependent' and below which 'independent'. We conferred with several people who work in theContinuing Care system for the Vancouver/Richmond Health Board. Based on that meeting, wedefined 'loss of independence' as the point in time at which an individual begins (permanent)receipt of services at coded level '5' and above (IC3 - community). This cut-point was thoughtrealistic by our consultants from continuing care, because they believe it reflects a level of carewithout which an individual would be forced to enter an institution. In other words, at level IC2,there is potential that without those services the individual in question could stay at home (thoughthere may be hardship to informal caregivers, and eventual decline). At IC3 - community,however, if those services were removed, the person would have to be moved to a long-term carefacility. The services are being provided precisely to keep a person at home. (The consultantsalso noted that many people at this level of care are those with diagnoses of advanced dementia oralzheimer's disease.)For comparison purposes, a more conservative definition, continuous (permanent) carestarting at coded level '7' (the lowest level of facility care) was also used. In both cases, in orderto allow for data inaccuracies and temporary absences for various reasons, a period(s) 'out ofcare' for up to 31 consecutive days was allowed. A period of 7 days was also examined, but therewas little difference in 'assignment' based on 7 vs 31 days.With these definitions in place, we proceeded to building 'arrays' for each patient thatrecorded their level of care (as above) on each day of the study period (for those who died duringthe study period, the date of death , and all subsequent days', were recorded as 'X'). The result isa string of alphanumeric characters, 1095 or 1096 bytes long, for each member of the cohort.These arrays allowed us to: 1) define the point of LOI, according to the definitions above; and 2)compare patterns of utilisation (excluding pharmacare and MSP) both before and after LOr.Ccntr« /,)1' Health Serivices and Polin' [(('S('(II'C/I429-2194 llealll, Srienrc» uon\'(//1<:0111'£'1', BC VIS) 173June 19<;814STATEMENT OF MAJOR FINDINGSSpace limitations for this report precluded us from including tables/charts that we know wouldbe helpful complements to these findings. All such material is available upon request from theauthors. A more detailed analysis and discussion for each phase can also be found in publicationseither currently available or in draft ((Meddings et aI., 1997; Meddings et aI., 1998a; Meddings etal., 1998b; Morgan, et al., 1997; Barer et aI., 1998; McGrail et aI., 1998).Phase I: A cross-sectional analysis oftrends in the use ofhospital and physician servicesHOSPITALSThe use of acute and extended hospitals in BC has changed dramatically over the past quartercentury. Acute days per thousand pop ilation fell from over 1,800 to 764 between 1969 and1995/96. At the same time, the use of extended care increased from 150 days/thousand to over750/thousand. The overall conclusions from our analysis of changes in the hospital sector were:1. The proportion of hospital inpatient days in British Columbia used by the elderly hascontinued to rise during the past decade. Downsizing has had its principal impact on youngerpatients, lowering both case rates and lengths of stay .2. Declining use of acute care by non-elderly patients is in fact a long-established trend, goingback into the 1960s. This decline has , however, accelerated over the last decade, and evenover the last couple of years, raising questions as to where the bottom might be? This periodof steeper decline may well reflect the effects of policies intended to reduce inpatient use.3. It is notable that acute care use has, over the last decade, fallen significantly among the elderlypopulation as well. This rate had been declining prior to 1985/86, but appeared to bebottoming out. Again this may be a consequence of policies to discourage inpatient use.4. The decline in acute care use among the over sixty five population is , however, less rapid thanamong younger patients. Lengths of stay are falling at the same rate in both populations, butthe separation rate is falling less rapidly among the elderly. This suggests that there may befewer alternative sites for elder-care, or less discretion as to hospital use.5. The rapid increase in use of surgical day care by elderly patients does not appear to explainmuch of the drop in their use of inpatient days, though it may account for a significant share ­but by no means all - of the reduction in separation rate.6. While acute care use by the elderly is falling, extended care use is not. The extended caresector has, however, largely ceased to expand (per capita) over the last decade. Thecontinuing increase in the proportion of all hospital care in the province that is extended careis therefore the result of stability in rates of extended care use, while acute care use hasdeclined. This contrasts with the experience prior to 1985/86, when extended care use rateswere increasing to offset the declines in acute care.7. There was a decline in rates of use of extended care between 1985/86 and 1993/94, but incontrast to acute care, where declines accelerated after 1993/94, the rate of use of extendedcare flattened out in the more recent period. This suggests that efforts to reduce extended careuse may have run into increasing resistance. Alternatives may not have been developed formeeting unavoidable needs.Centre [or ilcalth Seriviccs and Polu:v Resoarcli429-2194 Health ~·";ell{·e\ Mot!VaI/COIII 'c r. I1C vsr 17.3.IUftC lY'IS15Centre I(JI" Health Serivircs and Policv Research429-2194 J-lea/!h S'I'I<,II'l',\ !YlallVm/C()III'a, Be For 1:1.3JUrIe 1CjY88. Among long stay patients, it is only those whose episodes end in death for whom patient dayuse rates have not fallen. Long stay patients discharged alive have accounted for a more orless stable share of total days since 1978, implying that since 1985/86 their use has declined inparallel with the rest of the hospital system. Care of the dying does not appear to be placingincreasing pressure on hospital capacity -- use by those over seventy five has been relativelystable over the last decade, though use by (the much smaller group of) younger long-staydecedents appears to be rising. But care of the long-stay dying is taking up a rapidlyincreasing share of the total as the overall system shrinks.9. It would appear that the overuse of acute care beds, which has been alleged to be a failing ofthe Canadian hospital system for nearly half a century, has been largely solved. No one, inCanada or elsewhere, knows what "the" appropriate rate may be, but whatever it is, we arelikely to be much closer now.10. As acute care use declines, however, the care of the very elderly, and particularly of the dying,stands out in much sharper relief. The policies of the last decade, at least in British Columbia,have had limited effect on this component of use. Perhaps they have obviated the need to buildstill more extended care beds - these data do not tell us. But the prospect of an ever-largerextended care sector, most of whose patients will depart only through death, seemsunsatisfactory on a number of grounds (of which cost is not necessarily the most important).DAY SURGERYThe use of surgical day care services grew dramatically between 1985/86 and 1995/96,particularly among the elderly. Cataract surgery alone accounted for over one-third of the growthin day surgery for the elderly between 1985/86 and 1993/94, which prompted further investigationinto this service. A survival analysis comparing people who underwent cataract surgery in1985/86 and 1989/90 to controls who did not, found that the subsequent mortality of people whohave undergone cataract surgery at younger ages (50-80) was higher than for controls(controlling for age, sex, and presence of diabetes). This suggests that the development ofcataract may be indicative of generalised tissue aging and thus a marker for increased probabilityof 'premature' death. We also found that this association did not diminish between the two timeperiods, during which time, as noted above, rates of cataract surgery increased dramatically.While we do not have information on pre-operative lenticular changes, it seems safe to assumethat at least part of the increase in utilisation is associated with operating at earlier stages ofcataract development -- i.e. more minor lenticular changes. If this is true, then the fact thatassociation did not diminish between the two time periods suggests that even minor lenticularchanges may serve as a marker for early mortality (Meddings et al., 1997a, 1997b).The results of this paper prompted us to look more closely at the relationship betweencataract and early mortality at younger ages (50-65). A socioeconomic analysis showed that earlycataract is inversely associated with SES, as well as with subsequent mortality independent ofSES. This suggests that generalised tissue aging may in part explain the gradient in health statusseen across levels of SES (Meddings et al., 1998).PHYSICIANSFee-for-service expenditures on physician services grew by almost two-thirds between1985/86 and 1993/94. After taking out the effect of fee increases, the growth was 43%, whichwas almost equally split between the effects of a growing provincial population and increases in16age-specific patterns of utilisation. Separating the effects of changes in the age structure fromoverall population growth showed that 'aging' explains increases in physician expenditures ofapproximately 0.2% per year, and is by far the least important component of change. Most of thegrowth in expenditures is attributable to increases in specialist services, but patterns of careprovided by OPs were changing as well. The analytical approach described above helps toidentify the sources of those changes.Despite little (for the non-elderly) to virtually no (for the elderly) growth in per capitaexpenditures on OP services , we found significant changes in the patterns of care that patientsreceive from OPs. A slightly larger portion of the population visited a OP at least once in1993/94 compared to 1985/86, and those who did make a OP visit were (on average) seeing alarger number of different GPs. At the same time, each patient made fewer visits to each OP. Adecrease in services per visit and expenditure per service, along with the decrease in visits perunique OP (almost completely) offset the increase in number of different OPs seen. In short,people of all ages in BC are (on average) making more trips to a wider variety of OPs, but arehaving fewer (and slightly less costly) things done to/for them once there. There may beimplications in these changes for quality of care, but the resulting impact on expenditures perperson in BC has been relatively small.Per capita expenditures on care provided by specialists, in contrast, increased for all agegroups, ranging from 13% for the youngest age group to 60% for ages 75-84 (all three elderlyage groups grew between 50 and 60%). As with OPs, a larger proportion of each age group(except the youngest) saw a specialist in 1993/94 compared to 1985/86. The increases were, infact, larger for specialists, though the proportion of any age group that sees a specialist remainsconsiderably smaller than the proportion that sees OPs. Also similar to GPs, each patient (onaverage) saw a larger number of specialists in 1993/94 compared to 1985/86, though theseincreases were larger for the elderly than they were for the non-elderly. In contrast to OPs,however, there was no compensating decrease, in fact there was virtually no change at all, ineither visits per unique physician or the number of services provided per visit. And there has beena shift toward the provision of more expensive services, again with increases more pronouncedamong the elderly. Overall, then, a larger proportion ofBC's population was seeing specialists bythe end of the period, they were more likely to see a larger number of specialists than in the past,they were being provided with more expensive services, and these trends were not offset by a'group practice' phenomenon or by decreases in service intensity (Barer et al., 1998).The increase in the number of MDPPs for GPs was large enough to generate curiosity aboutwhy this might have occurred. One explanation might be that 'doctor shopping' caused this --­that a small number of people are seeing a very large number of OPs, and are thus pushing theaverages up. This explanation does, of course, assume that not only are 'doctor shoppers' aproblem, but that they became a far larger problem in 1993/94 compared to 1985/86. Anotherexplanation might be the phenomenon of group practices and/or after-hour clinics, which mayresult in more people seeing a larger number of GPs.A cumulative frequency distribution of the number of GPs seen by each patient suggests thatthe latter is a better explanation; the vast majority of patients still saw relatively few GPs -- 1 to 3- at the later point in our analysis , but there has been a definite shift in this lower end of thedistribution, e.g. with more seeing three in 1993/94 compared to 1985/86. In addition, however,there are, indeed, those who would appear to fall in the 'doctor shopping' category, seeing morethan 50 GPs in a single year, but their numbers are very small- they cannot account for theCentre [or Health Serivices run! Pol icy {((',\'(,(/,.('II429-2194 Health S, ·i,;//{:n Mallvancouver. Be V6T 173June J<;'IS17overall trend. Nevertheless, they are a subject of some interest in their own right, and we arehoping to pursue further research on "high users" in a subsequent project being considered byNHRDP.Phase II: An analysis ofthe use ofhealth care services prior to deathThere was little change, comparing the 1988 and 1993 cohorts, in the proportion of eachcohort that fell into each of the 'institutional use' categories, but there have been decreases in theamount of facility care that they receive. In other words, the same proportion of people are goinginto hospitals and long term care facilities, but they appear to remain there for shorter lengths oftime. People who die, who are known to be heavy U1':f.'1"'; of health services, have clearly not beenentirely shielded from the effects of the downsizing in the acute care hospital sector. They appear,however, to be affected to a lesser degree than the rest of the population -- this group continuesto increase its share of total inpatient days (as noted above).The average number of days spent in acute care by elderly decedents fell 17 percent between1988 and 1993, whereas use of all other services increased markedly. Declining use of acute carecould have a number of influences on the demand for other health care services by the elderly.Not only could demand for outpatient and community based health care rise, but a trend towardlong-term care may also be noted as some chronically ill patients are 'moved' from acute carefacilities to extended or long-term care facilities. The former appears to have had a largerinfluence than the latter, as we see large increases in the community-based services, such as adultday care, homemaker services and pharmacare, and only small changes in the distribution betweengroups.Some of the observed increase in adult day care use and homemaker services can be explainedby a change in priorities for BC' s Continuing Care Systems, in accordance with the provincialgovernment's New Directions policy. This policy, implemented in 1993, was designed to provideservices 'closer to home', in effect, to de-emphasize the use of institutional services in favour ofthose based in the community. This shift in priorities away from facility care runs counter to somepressures created by downsizing of acute care (from which one might expect a transfer to afacility, albeit a less clinically-intensive one), which may further exacerbate pressures oncommunity based services.More dramatic changes occurred in pharmaceutical use and day surgery rates amongdecedents. It may be conjectured that the rise in day surgery rates and pharmaceutical use is dueto the shift away from acute care for some patients. For example, some people are nowundergoing surgical day procedures for conditions that would previously have required aninpatient stay. If pharmaceuticals are a necessary part of follow-up care for these procedures (e.g.antibiotics), then as day surgery use goes up, we would expect pharmaceutical costs to rise tosome degree as well, as what would have been covered as part of an inpatient stay is provided inthe community instead (and the pharmaceutical expenses that we have are for people who areliving in the community). But we already know that the increase in day surgery is not fullyexplained by the decrease in the provision of acute inpatient services, and both pharmaceuticalexpenditures and day surgery rates are increasing at about the same rate across all groups -­including those who had no acute stays .Centre [or Hralth Scrivices and Policv Research429-2]94/leah" \"';,,1/1:('\ MallVrJ/1co!I\"{'r.8C \/61' 17.3June I'J'JS18Centre tor Health Scrivices and Polin' {(esear,,11429-2194 Ileal I}, )"'1,,1/1:(,\' ,HallVancouver. Be Vt5T173June 1CJ'.i8The utilization changes in the 'no hospital' group do, in fact, provide the strongest evidencethat other factors are affecting outpatient services and pharmaceutical use among the decedentpopulation. If the contraction of hospital services were the only cause of the increased 'need' forday surgeries and pharmaceuticals, then we would expect that the relative size of this groupwould grow -- as fewer people use acute care services, more would be found in the 'no hospital'group . If this group did grow, then increased utilisation could be explained by suggesting thatsome proportion of this group were 'kept out' of acute care hospitals, but in order to accomplishthat, increased spending on community-based services (pharmaceuticals, day surgery) wasrequired. In fact, however, the relative proportion of the 'no hospital' group was remarkablystable between the two cohorts; the proportion was actually slightly lower in 1993 than 1988. Atthe same time, utilisation of day surgery and expenditures on pharmaceuticals per capita doubled.This is not to say that acute care downsizing had no part in these increases. The complexity ofchanges that occurred in the acute care system would make this statement difficult to supportwithout a far more detailed analysis. But it is clear that this is not the only, and certainly not eventhe most important, explanation for tl e changes.Phase III: A longitudinal cohort analysisThe first part of this phase, in which we analyzed care use by all individuals who met the agecriteria, showed there were some small changes in the composition of the 'institutional use'groups. Though small, they were somewhat larger than those seen in the deaths cohort analysis,and suggested a trend away from institutional care, with increases in the proportion of people inthe 'home care' and 'no home care ' (meaning no facility and no home-based continuing care)groups. It should be noted that nearly half of all people in these cohorts fell into the 'no homecare' group -- 45 .2% in 1986 and 47.4% in 1993.The use of services within groups followed the same general pattern seen for the deathscohorts; acute days and long-term facility days declined, while the use of homemaker services,adult day care, and pharmaceutical use increased. Also similar to the deaths cohorts, the increasein the use of these services occurred at the same rate among the 'no home care' group as amongthe others, suggesting that there are multiple factors explaining the growth. Here , however, thereis stronger support for the argument that a move away from care in acute hospitals (at leastpartially) explains these increases.An argument common among community caregivers is that one effect of changes in servicedelivery over the past decade has been a transfer of responsibility for care from the formal systemto informal caregivers. One interpretation of the trend we identified, of a (slightly) greaterproportion of people in the 'no home care' group in the later cohort, and a general push'downward' from the more facility-based groups to the less, is consistent with this argument.Decreasing sizes of the 'higher intensity' groups suggests that an individual must be sicker in 1993before being admitted to facility-based care. It would follow, then, that the general state of healthof those in the community would be lower if the underlying health status of the population had notchanged in the intervening years. An alternative hypothesis, of course, would be that the health ofthe average resident had improved so that there was less need for higher intensity levels of care.This might be the case, for example, if morbidity was becoming more compressed over time, thatis, a greater proportion of each age cohort was living to older ages, and people are healthy for a19larger proportion of their lives. In any case, these are all small numbers, and while they suggest atrend, they should not be considered conclusive until it is demonstrated that this pattern continuesover time.The second part of this phase was the development and implementation of a definition for lossof independence (LOI) . We identified people who lost independence in the latter two years ofeach of the study periods, to allow analysis of the twelve months of health service utilisationpreceding that loss. The two definitions that were used (as outlined above) identified 6,926 (outof 66,011 'eligible':') individuals who lost independence based on the more liberal definitionduring the earlier time period, and 3,168 for the more conservative definition, compared to 7,837(out of 75,637) and 3,427 for the later cohort. The proportion of the cohort who lostindependence was quite stable over time, while the ratio of the second definition to the first (theproportion of the total LOI group that also lost independence according to the more conservativedefinition) decreased slightly in the later cohort. This finding is consistent with the decrease ininstitutional services seen in the initial analysis, and with a greater emphasis on community-basedcare.A logistic regression looking at loss of independence showed that the use of facility care(acute or long-term), and the use of homemaker services and adult day care, are all strongpredictors of the transition from independence to dependence. While this may at first appear to bean expected finding , since our definitions for LOI are based on the use of health services, it shouldbe kept in mind that LOI only occurs once there is a pennanent transition to continuous use ofthese services . The fact that less permanent use of these is a good predictor for LOI actuallysuggests people who use these services at all are at greater risk of loss of independence, thatthese services mark the initiation of a downward, and permanent, functional trend which could beused to plan needs for services.What is more interesting, and perhaps a bit confusing given the above findings, is that >50%of people who DID lose independence in each time period did not use any facility or continuingcare services in the 12 months prior to their loss of independence, and >90% used only acute care,homemaker and home nursing care services. They did have higher than average use ofpharmaceuticals and physicians, but there does not seem to be a generalisable 'pattern' to the useof health care services in the year preceding loss of independence; except that for most people it isgenerally quite low. So, while the use of homemaker services and adult day care may be goodpredictors of LOI, they will actually predict LOI for only a fraction of people who experience it.These findings suggest that for many people, the loss of independence may be caused by atraumatic event; a health-related accident of some sort, such as a hip fracture, or perhaps thedeath of a spouse or other informal caregiver. Again, this would be consistent with the'compression of morbidity' hypothesis . Alternatively, what we may be seeing is a great deal ofunderlying need that is only addressed, or addressed adequately, following an acute stay -­caregiver burden again . It should be a priority to 'dig a bit deeper' with these data to see if one orthe other of these alternatives can be supported.3 The number 'eligible' to lose independence is based on the number of people who had not died by the beginningof the second year of study, and who had at least 32 recorded 'out' of care (i.e. were known to still be 'independent'by our definition).Ccntr« torIlcalth. Srriviccs and Policv UC'.\'C',lITII429-2194 lJealch ~'I 'i<'II( 'I'\ }/a/lvancouver. BC 1"6'/ I!.:?JU!1e 19'JS20Centre [or Hcaltn Scriviccs and Polirv Research429-2/94 l-lealrh SI'i~I1(:C,\ ,HallVancouver. Be V67' lZ3June /9'18POTENTIAL PRACTICAL AND THEORETICAL USES OF THE RESEARCHFINDINGS AND METHODSPhase I: A cross-sectional analysis oftrends in the use ofhospital and physician servicesDuring the 1980s, rates of hospital use in B.C. declined, although rates for the elderly werecontinuing to increase. During the 1990s, downsizing has affected all age groups, the elderlyincluded. In the latter period, these decreases have occurred in both short «60 days) and long­term (>=60 days) stays. At the same time , the use of day surgery has increased dramatically,particularly for the elderly, but there is no evidence to suggest that this increase is entirely (oreven mostly) a result of a transfer of care frominpatient to outpatient services.This project has identified the macro-level trends in the use of hospitals, but there needs to bea clearer understanding of the types of patients now cared for by the hospital system -- inpatientand outpatient -- and the ways in which the 'mix ' of patients has changed over time. Specifically,are there significant 'sub-populations' of people who are being excluded from necessary hospitalcare, because of age, medical condition, socioeconomic status, location of residence?The changes in the mix of clinical care provided, and in the types of patients who are receivingthis care, as a result of 'downsizing' in the acute care sector are clearly very complex. Thissuggests that there should be a more explicit coordinating (and perhaps budgetary) link betweenfacility-based and community-based care . Some mechanisms for this have been put into placewith 'regionalisation'. These should be strengthened to ensure continuity of care, and to minimise'off-loading' of clinical care from the formal health care system onto informal caregivers.While the acute care sector was shrinking between the mid-1980s and the mid-1990s, age­specific per capita expenditures on physicians continued to increase, mostly because of increasedspending on specialists. These increases occurred because a slightly larger proportion of BC ' spopulation was referred to specialists, they were more likely to see a larger number of specialiststhan in the past, and they were being provided with more expensive services (after adjusting forfee increases). There was also a trend toward patients seeing a greater number of different GPs,but this did not have the same effect on expenditures, because it was offset by a decrease in thenumber of visits per GP, the number of services provided during each visit, and the provision of(slightly) less expensive services.The specific reasons for continued increases in expenditures on physician expenditures (e.g.changing technologies, changing patterns of practice) , and their explicit connection to changes inother components of health services (e.g. pharmaceuticals) require further investigation.Physician associations across the country continue to claim that "patient needs" are outstrippingthe meagre increases provided by Ministries of Health under global budget arrangements. Yetthese budgets, including that in BC, are growing at rates sufficient to cover population growthand aging. Our work here has offered some preliminary indications of what else is going onwithin the underlying utilization growth dynamics, but much more work remains.Meanwhile, changes in policy have been enacted across the country to try to limit the growthin this area of health expenditure. Since the years on which these analyses are based, BC hasmoved to a 'hard cap' system for physician reimbursement. The hard cap has been relativelysuccessful in controlling physician expenditures, but these analyses are quite revealing in whatexpenditure shifts have taken place 'under the globe', and should be useful to policy-makers in21identifying areas that may cause pressure in the future. For example, if specialists are 'better' atkeeping their portion of the global budget up, there may be interest in moving to split caps forGPs and specialists, or perhaps in the creation of regional/specialty caps.Phase II: An analysis ofthe use ofhealth care services prior to deathBy the mid-1990s, people who died in hospital accounted for an increasing proportion of totalhospital care, but only because their use was declining at a slower rate than that of non-decedents.At that time, people who died were as likely to use facility-based care as their counterparts in themid-1980s, but had shorter lengths of stay. They were far more likely to use community-basedservices such as homemaker care and doubled their use of pharmaceuticals, Despite claims to thecontrary, only a small fraction of the increase in pharmaceutical costs can be explained by hospitaldownsizing.For all types of patients, there needs to be a better understanding of the direct connectionbetween decreased use of acute care and increased use of community-based services. Aresufficient community services available? How much of the increase in pharmaceutical costs is dueto ' shifting' from hospitals to communities, and if that amount is small, what is the majorcontributor to the large increases in the ambulatory use of pharmaceuticals? What has been theincreased demand on informal caregivers? And are there qualitative data available that might tellus something about whether people are 'better off?The doubling in adjusted pharmaceutical costs among the elderly -- for all age groups, and all'groups' of patients, whether they used hospital services or not -- suggests that pharmaceuticalexpenditures must become a high priority area for policy attention. The BC program ofreference-based pricing is one form of intervention; calls for a national Phannacare program areanother. Any policy approach in this area would benefit from combining top-down (e.g. globalbudget) and bottom up initiatives to improve the appropriateness and cost-effectiveness ofprescribing practices. Policy in this area would also benefit from coordinated action, at least atthe provincial level, if not the national, or there is little hope of having control in this rapidlyincreasing sector.Phase III: A longitudinal cohort analysisA decreased use of facility-based services in the 1990s means that the BC government policyof providing services 'closer to home' may have been achieved. What should remain of concern iswhether this was accomplished through the deployment of more suitable health care services, orat the expense of informal caregivers.Comparing people who lose independence in the mid-1980s to those who go through thistransition in the mid-1990s suggests two things; first, that even this vulnerable group of patientshas been affected by acute care downsizing, and second, that many of these people appear tomake this transition quite rapidly, perhaps because of a specific medical (or social) event. Analternative view may be that this 'rapid' transition (based on analysis of the use of health careservices) actually signals a level of unmet need prior to the diagnosis and response that leads toCentre f or Health Scrivices and Pol in ' Resc.u cl:429 -2194lJealrlt Srien re» A'fallYancon vcr. Be V6T 173JUli e 19'1822the loss of independence. Clearly, either of these alternatives should be of some concern topolicy-makers.The transition to 'dependence' needs to be better understood. Are there particular diagnosesthat predict this transition, or particular patterns of care from physicians, or certain categories ofpharmaceuticals? How often are there truly 'traumatic' events associated with the transition, andcan the incidence of those, or the treatment of them, in some way be changed? Can 'predictive'information such as that generated from this project be used to project resource needs for thatsegment of the population that will lose independence in future years?Overall, on the basis of this project, there is no compelling information to suggest that acutecare downsizing has had major deleterious effects on the elderly, although we have no way ofassessing the impact on care-givers, which is an important issue. The mix of services received bythe elderly has clearly changed, with the use of community-based continuing care servicesincreasing rapidly as facility-based car ~ has declined, but these changes are consistent with statedpolicy objectives of the BC Ministry of Health, and with the wishes of the public, as summarisedby the BC Royal Commission on Health Care and Costs (B.c., 1991). There has, in addition,been a huge increase in the use of community-based pharmaceuticals. These changes suggest thatthe National Forum on Health, and now the federal government, have identified appropriatepolicy targets for the coming term.Centrefor Heahh Serivices and Policv Research429-2194 Health Scien, ('\ MallVal/COIII 'U, Be V6T 1'13June 199823REFERENCESAnderson, G.M., LR Pulcins, M.L. Barer, RG. Evans and C. Hertzman (1990) "Acute CareHospital Utilization under Canadian National Health Insurance: The British Columbia Experiencefrom 1969 to 1988" Inquiry 27:4 (Winter) pp. 352-8.Anderson, G.M., K. 1. Kerluke , LR Pulcins, C. Hertzman, and M.L. Barer (1993) "Trends andDeterminants of Prescription Drug Expenditures in the Elderly: Data from the British ColumbiaPhannacare Program", Inquiry 30:199-207.Barer, M.L. LR Pulcins, RG. Evans, C. Hertzman, J. Lomas, and G.M. Anderson (1989)"Trends in use of medical services by the elderly in British Columbia" Canadian MedicalAssociation Journal 141:1 (July 1) pp. 39-45.Barer, M.L., C. Hertzman. R Miller, and M.V. Pascali (1992) "On Being Old and Sick: TheBurden of Health Care for the Elderly in Canada and the United States" Journal of Health Policy.Politics and Law 17:4 (Winter) pp. 763-82.Barer, M.L., RG. Evans , C. Hertzman, and M. Johri (1998), "Lies, Damned Lies, and HealthCare Zombies: Discredited Ideas that will not Die", HPI Discussion Paper #10, Health PolicyInstitute, University of Texas-Houston Health Science Center.Barer, M.L. , RG. Evans and C. Hertzman (1995), "Avalanche or Glacier: Health Care and theDemographic Rhetoric", Canadian Journal on Aging 14(2): 193-224Barer, M.L., M. Brownell, and S. Sheps (1994b), Adult Surgical Utilization in Manitoba: 1981­1991, report prepared for Capital Planning, Manitoba Health .Barer M.L., K.M. McGrail, B. Green, RG. Evans, C. Hertzman, S.B. Sheps, and A. Kazanjian(1998) "Utilization of Physician Services in BC 1985/86 and 1993/94: An analysis using linkeddata" (in draft).Berk, M.L. , and A.c. Monheit (1992) "The Concentration of Health Expenditures: An Update"Health Affairs. 11:4 (Winter) pp. 145-9.Boulet, J.-A. , and G. Grenier (1978), "Health Expenditures in Canada and the Impact ofDemographic Changes on Future Government Health Insurance Program Expenditures",Economic Council of Canada Discussion Paper #123 Ottawa: ECC (October), 98p.British Columbia (1991), Closer to Home: Report of the B.C. Royal Commssion on Health Careand Costs , Victoria: Royal CommissionChamberlayne, R, B. Green, M.L. Barer, C. Hertzman, W. Lawrence and S. Sheps (1998forthcoming), "Creating a Population - Based Linked Health Database: A New Resource forHealth Services Research", Canadian Journal of Public Health.Cen tre [or Healtl: Scriviccs (mel "olin' Resrarch429-2194 l/wilh )', '; el/('('I ,Hall\'(J//(:O/flW, Be vsr l l.3June j'J'J824Davidson, W., D. W. Molloy, M. Bedard (1995), "Physician characteristics and prescribing forelderly people in New Brunswick: relation to patient outcomes" Canadian Medical AssociationJournal, 152(8): 1227-34, Apr 15.Evans, RG., G.C. Robinson and M.L. Barer (1988) "Where Have All the Children Gone?Accounting for the Pediatric Hospital Implosion" in RS. Tonkin and J.R Wright, eds.Redesigning Relationships in Child Health Care, s.c. Children's Hospital: Vancouver, s.c, pp.63-76.Evans, RG., M.L. Barer, C. Hertzman, G.M. Anderson, I.R Pulcins, and J. Lomas (1989) "TheLong Good-Bye: The Great Transformation of the British Columbia Hospital System" HealthServices Research 24:4 (October) pp. 435-59.Evans , RG. (1991) "Reflections on the Revolution in Sweden", in International Review of theSwedish Health Care System [Den Svenska Sjukvarden] (chaired by AJ. Culyer, with l-M. Grafvon der Schulenberg, W.P.M.M. van de Yen, and B.A Weisbrod) Occasional Paper no. 34, SNSSweden, Stockholm: SNS, pp. 118-57.Gormley, M., M. L. Barer, P. Melia and D. Helston (1990) , The Growth in Use of HealthServices 1977/78 to 1985/86, Regina: Saskatchewan HealthHertzrnan, c., I. Pulcins, M.L. Barer, R.G. Evans, G.M. Anderson, and J. Lomas (1990) "Flat onYour Back or Back to Your Flat? Sources of Increased Hospital Services Utilization among theElderly in British Columbia" Social Science & Medicine 30:7, pp. 819-28 .Manitoba Centre for Health Policy and Evaluation (1993) 1992-93 Annual Report Winnipeg:University of Manitoba, MCHPE.McGrail KM" RG. Evans, M. L. Barer, C. Hertzman, S.B. Sheps, and A Kazanjian (1998) ''TheQuick and the Dead: Hospital utilisation in British Columbia, 1969 through 1995/96", HealthPolicy Research Unit Discussion Paper Series, HPRU 98:3D.Meddings DR, Hertzman C, Barer ML, Evans RG, Kazanjian A, McGrail KM, Sheps SB, (1998a)"Socioeconomic Status, Mortality, and the Development of Early Cataract", Social Science andMedicine 46(11):1451-1457.Meddings DR, Marion SA, Barer ML, Evans RG, Green B, Hertzman C, Kazanjian A, McGrailKM, Sheps SB (1998b forthcoming) "Mortality Rates Following Cataract Extraction",Epidemiology.Meddings DR, McGrail KM, Barer ML, Hertzman C, Sheps SB, Evans RG, Kazanjian A (1997)''The Eyes Have it: Cataract Surgery and Changing Patterns of Outpatient Surgery". MedicalCare Research and Review 54(3):286-300 September.Centre !'IY Health Serivices and Policv Research429-2194 Heald, Science» MallYancouver. Be \/6'1' IZ3June IlJ'JS25Morgan S., RG. Evans, M.L. Barer, K.M. McGrail, B. Green, S.B. Sheps, C. Hertzman, A.Kazanjian (1997) "Health Service Utilization in the 24 Months Prior to Death" April 1997 (indraft).Mustard, C. [1994] Presentation to National Population Health Committtee, Health Canada,Toronto, January.Nair, C. (1991) "Trends in Hospital Inpatient Utilization, 1961-1988/89" Health Reports 19913:2 Canadian Centre for Health Information, Statistics Canada Cat.# 82-003 Qly, Ottawa:Ministry of Supply and Services, pp. 189-97, updated with unpublished data provided byStatistics Canada.Pascali, M.V. (1995) Controlling Expenditures for Physician Services: An Evaluation of BritishColumbia's Cost Containment Policies, 1979-1991. University of California at BerkeleyRoch., D.J., RG. Evans and D. Pascoe (1985) Manitoba and Medicare: 1971 to the PresentManitoba Health: Winnipeg, xviii, 249 p.Roos, N.P., E. Shapiro and R Tate (1989) "Does a Small Minority of Elderly Account for aMajority of Health Care Expenditures?: A Sixteen-year Perspective" The Milbank Quarterly 67:3­4, pp. 347-69.Roos, L.L. jr., and A. Wajda (1990) "Record Linkage Strategies: Part I. Estimating Informationand Evaluating Approaches" unpublished ms., Faculty of Medicine, University of Manitoba (June6).Sapolsky, RM. (1993) Stress, the aging brain, and the mechanisms of neuron death CambridgeMass.: the MIT Press.Shapiro, E., and R. Tate (1988) "Who Is Really at Risk of Institutionalization?" TheGerontologist 28:2, pp. 237-245.Shapiro, E. (1993) "The Vulnerable Elderly: Predicting Necessary Resources" Canadian FamilyPhysician 39: pp. 1406-10.Wilkins, R, O. Adams and AM. Brancker (1990), "Changes in Mortality by Income in UrbanCanada from 1971 to 1986", Health Reports 1989 1(2):137-74, Canadian Centre for HealthInformation, Statistics Canada, Cat.# 82-003 Qly, Ottawa: Ministry of Supply and Services,Wilkins, R (1992) "Use of Postal Codes and Addresses in the Analysis of Health Data" HealthReports 1993 5:2 Canadian Centre for Health Information, Statistics Canada Cat.# 82-003 Qly,Ottawa: Ministry of Supply and Services, pp. 157-77.Centre [or E!ealth Serivices find Polin Reseavcl:42?-2 I e!4 llealth ~'('iel/n'\ iI'fallV(IJ/CO//I'£'r, Be V6T I [IJune I'NS26APPENDIX 1: The nature of the collaboration, networking and input of otherresearchers, decision-makers and seniorsAs with all research, the analytical approach taken to this project, and the interpretation offindings, was influenced by contact with research colleagues, policy-makers, seniors, and so on.Faculty and staff at the Centre for Health Services and Policy Research have ongoingrelationships with officials at the British Columbia Ministry of Health. These contacts provedinvaluable in implementing the research methodology. The following people at the Ministry ofHealth provided significant consultation throughout the course of this project: Jim Cruikshank,Bob Hart, Linda Low, Kate Pengelly, Petr Schmidt, Bill Selwood, Ron Strohmaier, JanetVanKlaveren, and Shirley Wong. Dr. Alan Thomson, Dr. Doug Bigelow and Geri Hinton,among others, provided valuable comments and feedback on work-in-progress and drafts ofpapers.Lynn Buhler, Jan Fisher, and Jerry Reichdorf, from the Continuing Care section of theVancouverlRichmond Health Board, were instrumental in helping us modify our data-drivenapproach to the definition of loss of independence.Attendees of the various sessions where we presented preliminary findings from this project,particularly the CAG conferences, provided useful feedback on alternative interpretations, andpotential directions for future research.Attendees at the 3rd Age Spring Lecture Series, where we presented early findings fromphases I and II of this project, also provided useful comments, as well as assurance that there wassupport among the general public for this type of research endeavour.More recently, a potential synthesis/collaboration with a SIRP group from the Centre onAging at the University of Victoria, headed by Dr. Margaret Penning, offers assurance that wewill be able to follow up on some of the more interesting findings fiom this project.Centre ,;,1' llealm Serivices and Pol icv l({'sC'<l rcit429-21 94 Health S','iel/l'('1 Mallvancou ver, Be \/151' //.3Jun e 19'1827Appendix 2: Evidenceof the innovativeness, multi-disciplinarity and complementarity ofthe design approaches and team skillsThis was a multi-disciplinary team from the start, including epidemiologists, economists,physicians, and a sociologist. The team was further enhanced with the clinical knowledge of Dr.David Meddings, at the time a Community Medicine Resident, who headed the work on cataractsurgery.The research approaches used were varied, and were often developed specifically with theintent of exploiting the use of the linked data. Because this data set was quite new when theproject started, many of the analytical techniques had to be developed de novo. The developmentof 'trajectories of care' -- which were essential for phases IT and ITI -- and the 'equation'framework for the analysis of physician expenditures from phase I, were only possible because ofthe linked data . We intend to use both approaches in the future, and the latter has been adoptedfor some work in progress at the BC Medical Association.Beyond the 'data management' techniques required to produce workable data sets, we usedstandard (though modified, in some cases) epidemiological approaches to analysis, such assensitivity analysis, and survival analysis.Centre tor Health Scrtvices and Policv Rescarrh429-2194 Health Srient«, MaltVmICOIII'('T, Be vsr 1/.3June 19<)S28Appendix 3: Plans and prospects for publications and dissemination of results to serviceproviders, program planners, policy-makers, voluntary organizations andseniorsWe have several papers in progress, which will be submitted to journals as appropriate, whencompleted. Abstracts for all papers will be available on our web site (www.chspr.ubc.ca). as willfull text documents when possible. In addition, all papers will be provided to Geri Hinton,Director of the BC Office for Seniors, to other policy-makers and researchers in the BC Ministryof Health, to colleagues, and to the general public when requested.We plan to work with researchers from the Centre on Aging at the University of Victoria, totry to 'synthesize' results from our SIRP projects, specifically to see what overlap we can identifybetween our quantitative approach to loos of independence, and their more qualitative approach.Results from this project will also continue to be incorporated, as appropriate, to the manyformal and informal communications that Centre faculty and staff have with other researchers,interest groups and the general public.Centr e [or Hcalt t: Scrivices and Policv Resea rch429 -2} 94 Health Sri""" C' 1 Mat!Vanconvcr. Be VIS T IZ3June } <;'I 829


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items