- Email: [email protected]
The effects of introducing prospective payments to general hospitals on length of stay, quality of care, and hospitals’ income: the early experience of Israel
Social Science & Medicine 55 (2002) 981–989
The effects of introducing prospective payments to general hospitals on length of stay, quality of care, and hospitals’ income: the early experience of Israel Amir Shmuelia,b,*, Orna Intratorc, Avi Israelid a
Department of Health Management, School of Public Health, The Hebrew University-Hadassah, PO Box 12272, 91120 Jerusalem, Israel b Gertner Institute for Health Policy Research, Israel c Department of Statistics, The Hebrew University, Israel d Hadassah Medical Organization, Israel
Abstract A new reimbursement system for general hospitals in Israel was introduced in July 1990. The new system specified that for 15 selected procedures, hospitals would be paid by the insurers prospectively, rather than by the traditional perdiem arrangement. The rates were determined by the Ministry of Health. Henceforth, the number of selected procedures has increased and by now 40 procedures are included. In line with the ever-lasting interest in the effect of financial incentives on suppliers of medical care, the purpose of this paper is to examine the first-year effect of this change on the volume of activity, length of stay, quality of care, and hospitals’ real income. We focused on five selected procedures (cholecystectomy, hysterectomy, hip replacement, operations on lens and heart surgeries) performed in the four largest Israeli medical centers (Sheba, Sorasky, Rambam, and Hadassah). The analysis includes more than 17,000 hospitalizations occurring during two years prior to the change (July 1988–June 1990) and the first year after its implementation (July 1990–June 1991). We, therefore, examined, the short-term effects, wherein changes in the hospitals’ behavior are reflected mainly in the above-mentioned hospitalization characteristics. Further analysis will be required to examine the long-run implications of the change as well as its effect on the rest of the general inpatient sector in Israel. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Per-diem; Prospective payments; General inpatient care; Israel
Introduction On July 1st, 1990 a new reimbursement system for general hospitals in Israel was enacted. The new arrangement introduced fixed prospective payments (PP) to hospitals for 15 selected procedures. This arrangement was revolutionary in that it replaced the
*Corresponding author. Department of Health Management, School of Public Health, The Hebrew UniversityHadassah, PO Box 12272, 91120 Jerusalem, Israel. Fax: +972-643-5083. E-mail address: [email protected] (A. Shmueli).
traditional cost-based per-diem payment by prospective payments for these procedures. At the time of the change (1990), Israel had 43 general hospitals. Thirteen were government hospitals; 8 were owned by the General Sick Fund; 14 were public notfor-profit centers; and 8 were private for-profit institutions. There were 12,193 general beds. Almost half (46%) of the beds were owned by the government; 31% by the General Sick Fund; 20% by public hospitals; and 3% of the beds were in private hospitals. There were 253 beds per 100,000 population. The acute admission rate was 15.6 per 100 and inpatient days rate was 79 per 100 population. Average occupancy was 89% and average length of stay was 5.3 days (MOH, 1998–1991).
0277-9536/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 0 1 ) 0 0 2 3 3 - 7
982
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
Inpatient care was paid for by the sick funds. In 1990, there were four non-profit sick funds, providing individual voluntary health insurance to more than 95% of the population. In reality, however, during the research period (and the first 3 years of operation), the new payments were not fully prospective. The hospitals were paid completely prospectively for stay in the operating and intensive care departments. Stay in other departments (such as internal or geriatric) was reimbursed retrospectively, based on per-diem payment. In addition, all readmissionsFwhenever they occurredFwere reimbursed according to the per-diem rate. Only in 1993 did the payments become fully prospective, including no further reimbursement for readmissions within 7 days upon discharge. The PP rates were (as is the per-diem rate) determined by the Ministry of Health. They were based on some costing exercises done for the procedures in the large government hospitals, without specific length of stay targets (cost plus). These rates were applied to all admissions from all payers. The expected effects of the introduction of PP on hospitals’ behavior are well known. Facing zero marginal revenue, hospitals are expected to reduce resource use, including length of stay in all admissions. However, limiting resource use by the hospitals might result in lower quality of care, leading many to recommend a mixed retrospective–prospective system (Ellis & McGuire, 1986; Ma, 1994). Israel is not the pioneer of change from retrospective to prospective payments for inpatient care. US Medicare introduced the prospective payment system (PPS) in 1983 and several other systems have followed. In all cases, the main goals of the change were inpatient care cost containment and enhancing hospital efficiency. The implications of introducing prospective payments to hospitals have been the subject of a huge amount of research. It is generally agreed that in the short run, average length of stay considerably decreases in relation to its trend. It dropped 9% in the first year of Medicare PPS (DRG) and a total of 15% between 1982 and 1985 (Feinglass & Holloway, 1991; Lave & Frank, 1990). Kahn et al. (1990) found that among Medicare patients hospitalized for heart disease, heart attacks, pneumonia, CVA, and hip fracture, average length of stay dropped by 24% between 1981–1982 (‘‘before’’) and 1985–1986 (‘‘after’’). We note that in the Israeli case, the case-mix index is expected to rise, due to referral of below-average cases by the sick funds to other (cheaper) sources of care. Henceforth, a drop in length of stay can be reliably attributed to the introduction of PP rather than to changes in case-mix. The effect of introducing prospective payments on the number admissions is theoretically ambiguous. Medicare’s experience shows that in the long run the volume
of activity has tended to fall. Some researchers argue, however, that this is not a true decrease in admissions (and social costs), since Medicare transferred the more severe cases to facilities not paid by PPS, and the less severe cases to outpatient care (Ellis & McGuire, 1996). It is well documented that, in general, the average casemix of patients paid under PPS has gone up, namely, the average severity of patients has increased (McLellan, 1997). However, part of that increase is attributed to managerial strategies (‘‘DRG-creep’’) to increase revenues. Although there has been some concern about possible reduction in quality of care following the introduction of PP, the evidence tends to refute such fear (Feinglass & Holloway, 1991; Kahn et al., 1990). However, these analysts did find some evidence of ‘‘cost shifting’’, namely, that hospitals tend to assign more cases to other forms of treatment, including ambulatory, nursing, and family care. It should be noted that such shifts are sometimes socially desirable. The level of cost-saving efforts and the financial conditions of hospitals under prospective payments depend not only on the marginal revenue but also on the average revenue, namely, on the level of the prospective rates and on the intensity of care provided. The American PPS experience began with generous rates so that average margins were high. These rates were eroded by inflation later on and margins fell steadily and have tended to be negative, causing some hospitals to exit the market (Cutler, 1995). The American PPS is currently based on over 700 DRGs. The classification of patients into groups is based on primary and secondary diagnosis as well as on the age and sex of the patient. While such a system might induce ‘‘DRG creep’’, namely, an incentive for the hospital to classify patients as being sicker than they really are, it nonetheless provides a good classification according to risk. The Israeli PP groups are very large and each includes a very heterogeneous group of treatments and patients. Partial ‘‘insurance’’ against high-risk patients was provided (until 1993) by reimbursing the hospital per-diem for stay in wards other than the operating one or intensive care units (a mixedpayment system). Such an arrangement clearly provides an incentive for the hospital to transfer patients to other, cheaper, departments as early as possible and to reduce the proportion of each stay in the operating departments out of the total stay. Focusing on treatment, rather than on diagnosis, the Israeli PP provides incentives to perform invasive and surgical procedures, rather than providing conservative and non-surgical treatments for the particular clinical problem. This is partially true for the US as well, because although the American Medicare PPS is based on diagnoses (DRG), over 40% of the groups are actually not related to diagnoses, but to specific intensive procedures (McLellan, 1997). In
983
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
fact, the main failure of the American DRG system lies in increasing the average intensity of inpatient care, resulting in ever-increasing expenditures on health. In this study we focused on three government hospitals and one non-profit center. Naturally, one may question the strength of financial considerations (moving from per-diem to prospective reimbursement) in the behavior of these hospitals. A study of hospital competition in Israel during the early 1990s (Chinitz & Rosen, 1993) concluded that although government hospitals were budgeted, they were intensively involved in competitive strategies aimed to attract patients. These strategies included developing links with community medicine, developing new services, competing for key staff, and improving facilities, equipment, customer service and patient satisfaction. High levels of admissions and turnover were used by the hospitals’ directors to justify higher next-year budgets in the Ministry of Health’s budget allocation amongst its centers. In the early 1990s, two further developments underlined the importance of financial considerations in hospitals’ behavior. First, to shorten queues, the MOH allowed the hospitals to perform ‘‘afternoon sessions’’ managed by extra-budgeted Research Funds (see below). Second, in 1990 the Netanyahu State Commission of Inquiry into the Functioning and Efficiency of the Israeli Health Care System recommended, among other things, transforming government hospitals into self-managing ‘‘Trusts’’, a recommendation which has boosted business-like behavior among the large government centers. Following the American experience, several nations have introduced prospective payments for general hospitals and experienced similar changes. For example, the introduction of a DRG-based hospital financing system in Italy in 1995 caused an increase in number of admissions, a drop in length of stay, and a rise in severity of illness. No change was found in mortality and readmission rates (Louis et al., 1999). The goal of the present investigation is to examine the effect of the introduction of PP for selected procedures in Israel, on the number of admissions, length of stay, quality of care, and hospitals’ real incomes during its first year. It is assumed that in the short run these changes reflect the main modifications in hospitals’ behavior in response to the new reimbursement system.
Table 1 presents the definitions of these procedures. While Cholecystectomy includes only one type of intervention, all other procedures include a variety of treatments all for which the hospital is prospectively reimbursed equally. Hysterectomy includes three types of operations (subtotal, total abdominal and vaginal). Joint replacements include several operations; we selected the types performing hip and head of femur replacements. Operations on lens include mainly (90%) cataract extraction and insertion of lens prosthesis. Heart surgeries include many types of interventions (not included are cardiac catheterizations and angioplasty which fall under separate group rates). Upon consulting a cardiologist, these were classified into four main subgroups: revascularization, operations on valves, operations on structures adjacent to valves and other operations. The types of operations within procedures were used to control for case-mix. Amongst the general hospitals, the four largest medical centers were selected for the study. Three are government/municipal hospitals (Sheba, Sorasky, and Rambam) and the fourth is a public non-profit center (Hadassah). These four centers include 26% of all general beds and host 23% of all general inpatient days. Specifically, 20% of all lens operations, 30% of all hip replacements, 60% of all heart surgeries, 12% of all hysterectomies, and 30% of all cholecystectomies were performed in these centers in 1991 (MOH, unpublished data). It is clear, however, that these four medical centers do not represent the Israeli general inpatient care sector. All of them are big teaching hospitals, located in the large cities of Tel Aviv, Haifa, and Jerusalem. The choice of the four centers was shaped by several considerations. First, the General Sick Fund and the private centers refused to participate in the study. Second, 10 out of the
Table 1 Definitions of procedures and types Procedure
Type
5122
Cholecystectomy Hysterctomy
Methods
Hip replacement
Following the methodology used by Kahn et al. (1990), we focused on five selected procedures to examine the effect of the introduction of PP: Cholecystectomy, hysterectomy, hip replacement, operations on lens (cataract), and heart surgeries. These procedures were chosen primarily because of the existing potential reduction in LOS compared to other countries.
Operations on lens Heart surgeries
ICD-9
Subtotal abdominal Total abdominal Vaginal Hip replacement Rep. of head of femur
683 684 685 8151-2, 8159 8161-2 1300–1390
Revascularization Operations on valves Operations on structures Adjacent to valves Other
3610–3630 3551–3599 3500–3539 3691–3923
984
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
14 non-profit hospitals are small charity hospitals (owned by the Christian Church) or hospitals serving only the Arab population in the Jerusalem district. These hospitals did not have good, reliable, updated databases. Similar data problems were encountered in the relatively small peripheral government hospitals. We believe that the four centers studied represent the Israeli governmental and public non-peripheral general inpatient subsector, where about half of the Israeli general inpatient care is provided. A total of 17,400 hospitalization records were identified by the relevant ICD-9 codes (Table 1) during the periods July 1988–June 1990 (‘‘before’’) and July 1990–June 1991 (‘‘after’’). Quality of care was measured by the 60-day postdischarge readmission rate and in-hospital, 60- and 365days post-discharge mortality (mortality data was obtained by matching the identity numbers in our data with the Ministry of Interiors population file. For 6.5% of the cases no match was found). The effect of the introduction of PP (indicated by a dummy variable where 1=after the change) on length of stay (both total and prospectively paid) was estimated
using regression analysis. The other explanatory variables included case-mix (indicated by type of operation, age, and sex) and the operating hospital fixed effect. We used an ordinary least squares (OLS) regression corrected for heteroskedasticity with the dependent variable being the natural logarithm of the stay. This specification was found superior in terms of fit and significance to the linear model. A Poisson regression, specifically accounting for the measurement nature of length of stay, provided exactly the same estimates. The effects of the introduction of PP (a dummy variable where 1=after the change) on readmission and mortality rates were estimated using Logit regressions. The set of explanatory variables was the same as above.
Results Admissions Table 2 presents the yearly number of admissions before and after the change. While the number of admissions for cholecystectomy and hysterectomy re-
Table 2 Number of cases and total LOS Yearly n
Mean (M)
Median
SD (S)
% over M+2S
% over M+3S
Cholecystectomy Before After % Change n
849 849 0.0 2547
12.5 12.4 0.8
10.0 9.0 10.0
8.3 12.8 54.2
3.5 2.0
1.5 1.2
Hysterectomy Before After % Change n
854 860 0.7 2568
11.2 11.0 1.8
9.0 9.0 0.0
6.8 8.0 17.6
4.0 2.2
2.3 1.2
Hip replacement Before After % Change n
588 839 42.7 2015
27.2 23.9 12.1
19.0 16.0 15.8
23.6 23.9 1.3
4.3 3.0
1.6 1.2
Operations on lens Before After % Change n
2,375 2,096 11.7 6846
5.3 4.7 11.3
5.0 4.0 20.0
5.5 8.7 58.2
0.7 0.2
0.3 0.2
Heart surgeries Before After % Change n
934 1,446 54.8 3314
20.7 19.5 5.8
16.0 13.0 18.8
17.6 22.7 28.0
3.2 3.4
1.4 1.8
985
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
mained constant, the number of cataract operations performed in these hospitals dropped by 12%, the number of hip replacements increased by 43%, and the number of heart surgeries increased by 55%.
expected from the 1991 Israeli PP incentives structure, the prospectively paid LOS decreased and the retrospectively paid (per-diem) LOS increased. Total LOS, however, decreased.
Length of stay
Sixty days readmission rate
Several summary measures of the distributions of the observed (unadjusted) total length of stay (LOS) are presented in Table 2. For all procedures, both the mean and the median LOS dropped, as is expected from the incentives originated in PP. The standard deviation of LOS, however, increased with the PP. For cholecystectomy and cataract treatments that increase reached 54– 58%. This means that the drop in LOS was not uniform across different levels of LOS before the change. However, the rate of outliers (LOS above the mean plus 2 standard deviations) dropped (except for heart surgeries). Table 3 presents the adjusted rates of change in total LOS and in the prospectively paid LOS from the regression analysis. The full estimated regressions are presented in Appendix A. In all five procedures total LOS dropped considerably, with a 7% decrease for cholecystectomy, a 3% drop for hysterctomies, and 10– 18% drop for the other procedures. The sharpest decrease occurred for cataract operations. Prospectively paid LOS dropped even more, with an 8% decrease for cholecystectomy, a 6% drop for hysterctomy and a 15– 22% drop for the other procedures. For lens operations, there is no stay outside the ophthalmology ward, so that the prospectively paid LOS is identical to total LOS. These results indicate that the proportion of prospectively paid LOS in total LOS decreased. Actually, for all procedures, the LOS reimbursed by per-diem payments did not decrease. In cataract treatment, cholecystectomy, and hysterectomy, LOS in other departments is negligible. However, in hip replacement, the LOS, which is reimbursed per-diem, increased from 7.2 to 7.9 days, a 10% increase. In heart surgeries, the retrospectively paid LOS increased from 3.7 to 4.3 days, a 16% increase. As
Table 4 reports the readmission rates within 60 days from (live) discharge. The ‘‘before’’ and ‘‘after’’ rates are the unadjusted ones. The third column reports the adjusted odds-ratios (OR) from the Logit regression. Appendix A reports the full set of estimates. In all five procedures the readmission rates increased, although the increase was only significant for cholecysectomy, hip replacement, and heart surgeries. The adjusted OR indicates a 50–80% increase in readmission rate for these procedures.
Table 3 Adjusted rates of change in LOSa,b
Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries
Total LOS
PP LOSc
7.3 3.1 17.6 18.1 9.7
7.7 5.6 21.7 F 14.9
a Dependent variables are ln(LOS). Adjusted for age, sex, type of procedure and operating hospitals. b See Tables 8 and 9 for the full estimated regressions. c Prospectively paid LOS.
Mortality Table 5 presents the changes in the unadjusted rates of in-hospital mortality, within 60 and 365 days, upon (live) discharge. The Logit analysis revealed that for all procedures, the adjusted before–after OR’s are not significantly different from one for the three measures
Table 4 Readmissions within 60 days (%)a
Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries
Before
After
Adjusted ORb,c
5.2 6.3 6.2 4.8 4.8
9.9 7.0 10.1 5.1 7.7
1.8 (1.1) 1.7 (1.1) 1.5
a
Percent out of live discharges. Logistic regression. Adjusted for age, sex, type of procedure and operating hospitals. See Table 10 for the full estimated regression. c Values in parentheses are not significantly different from one. b
Table 5 Unadjusted mortality rates (%) In hospital
Within 60 Da
Within 365 Da
Before After Before After Before After Cholecystectomy 1.1 Hysterectomy 0.1 Hip replacement 4.1 Operations on lens 0.0 Heart surgeries 6.6 a
1.6 0.1 1.5 0.0 5.3
Percent of live discharges.
0.4 0.2 2.5 0.5 1.2
0.5 0.1 2.8 0.2 1.1
3.0 1.7 10.2 3.4 3.7
4.1 2.1 9.7 2.6 3.6
986
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
of mortality, namely, that no significant change in mortality was introduced by the PP.
heart surgeries, the income per day was twice as large as the per-diem rate.
Hospitals’ real incomes Table 6 provides general information on the reimbursement rates for inpatient care before and after the introduction of PP for the five selected procedures. For international comparisons, the values are given in 1991 USD adjusted for purchasing power ($GDP PPP). The rates are the average rates over the respective periods. Table 7 presents the changes in hospitals’ real revenues from the five procedures. Before the change, revenue per day is the per-diem rate. After the introduction of PP, revenue per day was calculated as the prospective rate divided by the ‘‘after’’ median total length of stay. Before the change, revenue per case was calculated by the per-diem rate multiplied by the ‘‘before’’ median total length of stay. After the change, the revenue per case is the prospective rate. For cholecystectomy, hysterectomy and lens operations, real income dropped with the introduction of PP. This is true for both incomes per day and per case. For cataract surgeries, revenue per day dropped by 44% and per case by 55%. For hip replacement, revenue per day increased by 15%, but total revenue per case decreased by 3%. For heart surgeries, real incomes increased dramatically (179% per day and 126% per case). These changes are reflected in the ratio of the per-diem rate to revenue per day after the introduction of PP. For lens operations, the perdiem rate was twice as large as the income per day. For
Discussion
Table 6 Reimbursment rates for inpatient care (1991 $GDP PPP)
Per-diem Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries
Before
After
230 F F F F F
244 1732 1617 3754 519 8085
Table 7 Percent changes in hospitals’ real revenues
Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries a
After.
Per day
Per case
Per-diem/Rev. per daya
16 22 15 44 179
25 22 3 55 126
1.3 1.4 0.9 1.9 0.4
The results show that the introduction of PP for selected procedures to general hospitals in Israel caused the ‘‘classical’’ expected short-term changes in the hospitalizations’ characteristics. Controlling for age, sex, type of procedure, and the operating hospital, total length of stay decreased. The drop in LOS in heart surgeries, hip replacements and cataracts was higher than thrice the mean annual decrease in LOS in the respective wards over the period 1975–1989. The drop in the prospectively paid LOS (in the operating department and in intensive care units) was even sharper. While the new reimbursement system did not cause any significant change in in-hospital, within 60 and 365 days upon discharge mortality rates, the odds for 60 days readmission increased. We note that at the time (1991), all readmissions were reimbursed, regardless of their timing. Further research will explore whether the increase in readmission rates was indeed an indicator of lower quality of care (and of premature discharges), or whether it resulted from strategic behavior of the hospitals to increase revenues (Cutler, 1995). The new PP introduced changes in the hospitals’ revenues per case and revenues per day. These changes were determined both by the levels of the prospective rates and the changes in length of stay. In performing cataract surgeries, the hospitals’ revenues per day dropped by 44% and revenue per case by 55%. The rate (519 USD adjusted for 1991 GDP PPP) was low and even with the 18% drop in length of stay, the perdiem rate was twice as large as the revenue per day. These changes clearly provide an incentive for the hospitals to avoid cataract patients. Indeed, the results show a 12% drop in the number of cataract surgeries performed (Table 2). Revenues for hysterectomies and cholecystectomies dropped by 22–25%. However, the number of cases remained constant. Although revenues for a hip replacement surgery dropped by 3%, the number of cases increased by 43%. Apparently, for these three procedures, the ‘‘before’’ LOS, which determined the revenue per case before the introduction of PP, was long enough to assure such high margins, that even the decrease in revenue per case left the hospitals with satisfactory ones. Finally, the sharp increase of 55% in the number of heart surgeries is clearly consistent with the 126% increase in the hospitals’ revenues per heart surgery (the PP rate was set at 8085 $PPP). The increase in the number of admissions to hip replacement and heart surgeries is consistent with the Italian experience (Louis et al., 1999) but contradicts the
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
US experience, where the anticipated increase did not materialize (Hodgkin & McGuire, 1994). For heart surgeries, the increase in utilization is related to the generous rate per case (and see below). For hip replacement, while the revenue per case dropped slightly, the revenue per day increased by 15% and was 11% higher than the per-diem rate. Consequently, hip replacement patients were still more desirable than other patients paid for by the per-diem arrangement. The changes in the number of admissions and length of stay induce a change in load of total inpatient days in the relevant wards. According to our findings, while there was no change in cholecystectomies (general surgery), nor in hysterectomies (gynecology), the volume of hip replacement hospitalization days increased by 30% in the orthopedic ward, the volume of heart surgeries inpatient days increased by 50% in the chest surgeries ward, and cataract inpatient days dropped by 23% in the ophthalmology department. Departmental data on number of beds and occupancy rates over the (calendar) years 1989–1991 (MOH, 1989–1991) for the four hospitals included in the study, indicate that in three of the four hospitals, the number of beds in the ophthalmology ward dropped by about 20% between 1989 and 1991. Although the trend to perform eye surgeries in outpatient facilities began prior to the introduction of the PP (see below), the relatively unattractive prospective rate probably accelerated that trend. Although hip replacement hospitalization days increased by 30%, there is no parallel increase in terms of occupancy in the orthopedic ward. Furthermore, occupancy rates in these wards actually decreased between 1989 and 1990 in all four hospitals. Nor was an increase in occupancy was found in rehabilitation wards. It seems that hip replacement days substituted inpatient days for other orthopedic procedures (no data is available to explore this issue further). All hospitals witnessed an increase in 1990–1991 in occupancy in the cardiology department and/or in the cardiac intensive care units. However, that increase should also be attributed to the sharp increase in cardiac catheterizations and angioplasty following the introduction of the PP to these procedures. As in most natural experiments, there are several other factors that should be considered with regard to the identification of the effects of the introduction of PP. Note, however, that we examined changes from one year to the following year, so that the effects of general time trends in the data are minimized. First, during that period, laparascopy proceduresFwith shorter staysF became popular, replacing the traditional cholecystectomies. However, the use of laparascopy in the four medical centers included in this investigation did not begin before the end of 1991. Consequently, the 7–8% drop in total length of stay cannot be attributed to that technological change. Second, there were three organiza-
987
tional–structural issues that should be noted. Since 1989, the sick funds have exercised ‘‘length of stay control’’ in order to contain their inpatient retrospectively paid costs. Under pure PP, no such control is needed (actually, a need for quality control is more necessary). As was mentioned earlier, the retrospectively paid LOS increased following the introduction of the new reimbursement system (although total LOS decreased), so that the sick funds’ pressure cannot explain the drop in LOS. Another initiative of the sick funds, which started in that period, was to shift cataract patients to other forms of care, mainly ambulatory care facilities. This may explain the 12% drop in the number of cases. However, as was previously mentioned, if that shift was selectiveFwith lower risk patients being shiftedFthe cataract patients case-mix was supposed to rise and the 18% drop in LOS could not result from that shift. The low level of the prospective payment for cataract surgery set by the Ministry of Health might have had a similar purpose, namely, to encourage that shift. The generous rate for heart surgeries was probably part of the Ministry of Health’s initiative to reduce waiting times for these surgeries. In early 1990, under public pressure, the MOH launched a ‘‘shortened queues’’ campaign, which included special payments to hospitals and their operating staff for afternoon sessions. That campaign induced a sharp increase in the number of surgeries performed. The July 1990 introduction of the PP for heart surgeries, with its generous level, enhanced that earlier trend. This investigation focusedFfor the first timeFon the implications of the introduction of the Israeli PP during its first year to the governmental and public, non-profit, non-peripheral, general inpatient subsector. While more experience has accumulated since, data (in)availability poses severe difficulty in analyzing that experience. Further research is clearly needed, however, to evaluate the change (and its expansion) in subsequent years and in the entire Israeli inpatient sector. Acknowledgements We would like to thank Gabi Barabash and Nira Shamai for their assistance at various stages of the research, and the management of the participating hospitals for providing us with the data. We also thank the referees for their useful comments on an earlier draft.
Appendix A The full estimates of ln total LOS regressions, ln PP LOS regressions and logistics regression of readmissions within 60 days are given in Tables 8–10.
988
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
Table 8 Ln(total LOS) regressions
PP (1=yes) Age Gender (1=men) Hospital 1 Hospital 2 Hospital 3 Constant Hip replacement types (ICD-9) 8141 8151 8159 8161 Hysterectomy types (ICD-9) 683 684 685 Heart surgery groups Group 2 Group 3 Group 4 ADJ R SQ
Hip replacement
Cholecystectomy
Hysterectomy
Operations on lens
Heart surgery
Coefficient
SE
Coefficient
SE
Coefficient SE
Coefficient
SE
Coefficient SE
0.193 0.007 0.051 0.076 0.413 0.038 2.346
0.01 0.001 0.01 0.019 0.016 0.02 0.036
0.076 0.006 0.074 0.149 0.005 0.142 2.117
0.013 0.001 0.013 0.016 0.016 0.021 0.022
0.031 0.008 F 0.427 0.375 0.502 1.735
0.012 0.001 0.011 0.017 0.019 0.021 0.065
0.102 0.007 0.11 0.734 0.409 0.294 2.422
0.416 0.379 0.349 0.652
0.019 0.022 0.018 0.022
F F F F
F F F
F F F
F F F
F F F
0.103
0.013 0.198 0.001 0.003 0.006 0.029 0.468 0.026 0.016 0.03 0.414 0.122 1.221
F F F F 0.534 0.625 0.511
F F F F
F F F F
0.129 F 0.118 F 0.118 F
F F F
F F F
0.078
F F F
0.105
0.059 0.117 0.084
0.275
0.009 0.001 0.01 0.014 0.016 0.022 0.083
0.027 0.012 0.027
0.148
Table 9 Ln(PP LOS) regressions
PP (1=yes) Age Gender (1=men) Hospital 1 Hospital 2 Hospital 3 Constant Hip replacement types (ICD-9) 8141 8151 8159 8161 Hysterectomy types (ICD-9) 683 684 685 Heart surgery groups Group 2 Group 3 Group 4 ADJ R SQ
Hip replacement
Cholecystectomy
Hysterectomy
Coefficient
SE
Coefficient
SE
Coefficient
0.244 0.004 0.044 0.296 0.114 0.059 2.565
0.012 0.001 0.12 0.02 0.017 0.021 0.041
0.079 0.007 0.006 0.094 0.026 0.158 2.088
0.013 0.001 0.013 0.016 0.016 0.021 0.022
0.058 0.008 F 0.442 0.687 0.522 1.655
0.414 0.431 0.328 0.324
0.022 0.024 0.02 0.026
F F F F
F F F
F F F
F F F
F F F
0.122
0.111
Heart surgery SE 0.014 0.001 0.029 0.029 0.031 0.129
F F F F 0.435 0.69 0.577
SE
0.161 0.006 0.067 0.418 0.311 0.241 2.825
0.011 0.001 0.012 0.016 0.017 0.024 0.098
F F F F 0.139 0.125 0.123 F F F
0.128
Coefficient
F F F 0.114 0.177 0.522 0.172
0.032 0.014 0.038
989
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989 Table 10 Logistic regression of readmissions within 60 days Hip replacement a
Cholecystectomy a
Hysterectomy a
p
Coefficient
p
Coefficient p
Coefficient
p
Coefficient pa
0.09 0.02 0 0 0.37 0.05 0
0.095 0.028 F 7.529 0.281 0.651 4.043
0.095 0.017 0.166 1.489 0.503 0.142 1.654
0.28 0 0.15 0 0.09 0.53 0
0.405 0.013 0.188 2.195 0.799 0.133 2.277
0.531 0.007 0.281 2.407 0.554 0.984 2.356
0 0.21 0.11 0.01 0.19 0.01 0
0.578 0.011 0.431 0.714 0.338 0.598 3.078
0.532 0.337 0.179 0.076
0.12 0.4 0.58 0.87
F F F F
F F F F
F F F
F F F
F 0.62 F
F F F
F F F
F F F
CHI SQ
88.02
p¼0
66.4
p¼0
75.96
0.53 0 0.34 0.14 0.02 0
0
a
Heart surgery
Coefficient PP (1=yes) Age Gender (1=men) Hospital 1 Hospital 2 Hospital 3 Constant Hip replacement types (ICD-9) 8141 8151 8159 8161 Hysterectomy types (ICD-9) 683 684 685 Heart surgery groups Group 2 Group 3 Group 4
a
Operations on lens
0.02 0.01 0.29 0 0.04 0.78 0
F F F F
F F F F
F F F
F F F
F F F
0.537 0.66 1.509
0.09 0.04 0
214.23
p¼0
p ¼ 0 166.51
p¼0
Wald test.
References Cutler, D. (1995). The incidence of adverse medical outcomes under prospective payment. Econometrica, 63, 29–50. Chinitz, D., & Rosen, B. (1993). A tale of two markets: Hospital competition in Israel. Brookdale Institute RR-30-93, Jerusalem, September 1993. Ellis, R. P., & McGuire, T. G. (1986). Provider behavior under prospective reimbursement. Journal of Health Economics, 5, 129–151. Ellis, R. P., & McGuire, T. G. (1996). Hospital response to prospective payment: Moral hazard, selection, and practice-style effects. Journal of Health Economics, 15, 257–277. Feinglass, J., & Holloway, J. J. (1991). The initial impact of the Medicare PPS on US health care: A review of the literature. Medical Care Review, 48, 91–115. Hodgkin, D., & McGuire, T. (1994). Payment levels and hospital response to prospective payment. Journal of health Economics, 13, 1–29.
Kahn, K. L., Keelr, E. B., & Sherwood, M. J., et al. (1990). Comparing outcomes of care before and after implementation of the DRG-based prospective payment system. Journal of the American Medical Association, 264, 1984–1989. Lave, J. R., & Frank, R. G. (1990). Hospital supply response to prospective payment as measured by length of stay. Advances in Health Economics and Health Services Research, 11, 1–26. Louis, D. Z., Yuen, E. J., Braga, M., & Cicchetti, A., et al. (1999). Impact of a DRG-based hospital financing system on quality and outcomes of care in Italy. Health Services Research, 34, 405–415. Ma, C. A. (1994). Health care payment systems: Cost and quality incentives. Journal of Economics & Management Strategy, 3, 93–112. McLellan, M. (1997). Hospital reimbursement incentives: An empirical analysis. Journal of Economics & Management Strategy, 6, 91–128. Ministry of Health (1989–1991). Inpatient care institutions in Israel. Jerusalem, Hebrew.
The effects of introducing prospective payments to general hospitals on length of stay, quality of care, and hospitals’ income: the early experience of Israel Amir Shmuelia,b,*, Orna Intratorc, Avi Israelid a
Department of Health Management, School of Public Health, The Hebrew University-Hadassah, PO Box 12272, 91120 Jerusalem, Israel b Gertner Institute for Health Policy Research, Israel c Department of Statistics, The Hebrew University, Israel d Hadassah Medical Organization, Israel
Abstract A new reimbursement system for general hospitals in Israel was introduced in July 1990. The new system specified that for 15 selected procedures, hospitals would be paid by the insurers prospectively, rather than by the traditional perdiem arrangement. The rates were determined by the Ministry of Health. Henceforth, the number of selected procedures has increased and by now 40 procedures are included. In line with the ever-lasting interest in the effect of financial incentives on suppliers of medical care, the purpose of this paper is to examine the first-year effect of this change on the volume of activity, length of stay, quality of care, and hospitals’ real income. We focused on five selected procedures (cholecystectomy, hysterectomy, hip replacement, operations on lens and heart surgeries) performed in the four largest Israeli medical centers (Sheba, Sorasky, Rambam, and Hadassah). The analysis includes more than 17,000 hospitalizations occurring during two years prior to the change (July 1988–June 1990) and the first year after its implementation (July 1990–June 1991). We, therefore, examined, the short-term effects, wherein changes in the hospitals’ behavior are reflected mainly in the above-mentioned hospitalization characteristics. Further analysis will be required to examine the long-run implications of the change as well as its effect on the rest of the general inpatient sector in Israel. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Per-diem; Prospective payments; General inpatient care; Israel
Introduction On July 1st, 1990 a new reimbursement system for general hospitals in Israel was enacted. The new arrangement introduced fixed prospective payments (PP) to hospitals for 15 selected procedures. This arrangement was revolutionary in that it replaced the
*Corresponding author. Department of Health Management, School of Public Health, The Hebrew UniversityHadassah, PO Box 12272, 91120 Jerusalem, Israel. Fax: +972-643-5083. E-mail address: [email protected] (A. Shmueli).
traditional cost-based per-diem payment by prospective payments for these procedures. At the time of the change (1990), Israel had 43 general hospitals. Thirteen were government hospitals; 8 were owned by the General Sick Fund; 14 were public notfor-profit centers; and 8 were private for-profit institutions. There were 12,193 general beds. Almost half (46%) of the beds were owned by the government; 31% by the General Sick Fund; 20% by public hospitals; and 3% of the beds were in private hospitals. There were 253 beds per 100,000 population. The acute admission rate was 15.6 per 100 and inpatient days rate was 79 per 100 population. Average occupancy was 89% and average length of stay was 5.3 days (MOH, 1998–1991).
0277-9536/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 0 1 ) 0 0 2 3 3 - 7
982
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
Inpatient care was paid for by the sick funds. In 1990, there were four non-profit sick funds, providing individual voluntary health insurance to more than 95% of the population. In reality, however, during the research period (and the first 3 years of operation), the new payments were not fully prospective. The hospitals were paid completely prospectively for stay in the operating and intensive care departments. Stay in other departments (such as internal or geriatric) was reimbursed retrospectively, based on per-diem payment. In addition, all readmissionsFwhenever they occurredFwere reimbursed according to the per-diem rate. Only in 1993 did the payments become fully prospective, including no further reimbursement for readmissions within 7 days upon discharge. The PP rates were (as is the per-diem rate) determined by the Ministry of Health. They were based on some costing exercises done for the procedures in the large government hospitals, without specific length of stay targets (cost plus). These rates were applied to all admissions from all payers. The expected effects of the introduction of PP on hospitals’ behavior are well known. Facing zero marginal revenue, hospitals are expected to reduce resource use, including length of stay in all admissions. However, limiting resource use by the hospitals might result in lower quality of care, leading many to recommend a mixed retrospective–prospective system (Ellis & McGuire, 1986; Ma, 1994). Israel is not the pioneer of change from retrospective to prospective payments for inpatient care. US Medicare introduced the prospective payment system (PPS) in 1983 and several other systems have followed. In all cases, the main goals of the change were inpatient care cost containment and enhancing hospital efficiency. The implications of introducing prospective payments to hospitals have been the subject of a huge amount of research. It is generally agreed that in the short run, average length of stay considerably decreases in relation to its trend. It dropped 9% in the first year of Medicare PPS (DRG) and a total of 15% between 1982 and 1985 (Feinglass & Holloway, 1991; Lave & Frank, 1990). Kahn et al. (1990) found that among Medicare patients hospitalized for heart disease, heart attacks, pneumonia, CVA, and hip fracture, average length of stay dropped by 24% between 1981–1982 (‘‘before’’) and 1985–1986 (‘‘after’’). We note that in the Israeli case, the case-mix index is expected to rise, due to referral of below-average cases by the sick funds to other (cheaper) sources of care. Henceforth, a drop in length of stay can be reliably attributed to the introduction of PP rather than to changes in case-mix. The effect of introducing prospective payments on the number admissions is theoretically ambiguous. Medicare’s experience shows that in the long run the volume
of activity has tended to fall. Some researchers argue, however, that this is not a true decrease in admissions (and social costs), since Medicare transferred the more severe cases to facilities not paid by PPS, and the less severe cases to outpatient care (Ellis & McGuire, 1996). It is well documented that, in general, the average casemix of patients paid under PPS has gone up, namely, the average severity of patients has increased (McLellan, 1997). However, part of that increase is attributed to managerial strategies (‘‘DRG-creep’’) to increase revenues. Although there has been some concern about possible reduction in quality of care following the introduction of PP, the evidence tends to refute such fear (Feinglass & Holloway, 1991; Kahn et al., 1990). However, these analysts did find some evidence of ‘‘cost shifting’’, namely, that hospitals tend to assign more cases to other forms of treatment, including ambulatory, nursing, and family care. It should be noted that such shifts are sometimes socially desirable. The level of cost-saving efforts and the financial conditions of hospitals under prospective payments depend not only on the marginal revenue but also on the average revenue, namely, on the level of the prospective rates and on the intensity of care provided. The American PPS experience began with generous rates so that average margins were high. These rates were eroded by inflation later on and margins fell steadily and have tended to be negative, causing some hospitals to exit the market (Cutler, 1995). The American PPS is currently based on over 700 DRGs. The classification of patients into groups is based on primary and secondary diagnosis as well as on the age and sex of the patient. While such a system might induce ‘‘DRG creep’’, namely, an incentive for the hospital to classify patients as being sicker than they really are, it nonetheless provides a good classification according to risk. The Israeli PP groups are very large and each includes a very heterogeneous group of treatments and patients. Partial ‘‘insurance’’ against high-risk patients was provided (until 1993) by reimbursing the hospital per-diem for stay in wards other than the operating one or intensive care units (a mixedpayment system). Such an arrangement clearly provides an incentive for the hospital to transfer patients to other, cheaper, departments as early as possible and to reduce the proportion of each stay in the operating departments out of the total stay. Focusing on treatment, rather than on diagnosis, the Israeli PP provides incentives to perform invasive and surgical procedures, rather than providing conservative and non-surgical treatments for the particular clinical problem. This is partially true for the US as well, because although the American Medicare PPS is based on diagnoses (DRG), over 40% of the groups are actually not related to diagnoses, but to specific intensive procedures (McLellan, 1997). In
983
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
fact, the main failure of the American DRG system lies in increasing the average intensity of inpatient care, resulting in ever-increasing expenditures on health. In this study we focused on three government hospitals and one non-profit center. Naturally, one may question the strength of financial considerations (moving from per-diem to prospective reimbursement) in the behavior of these hospitals. A study of hospital competition in Israel during the early 1990s (Chinitz & Rosen, 1993) concluded that although government hospitals were budgeted, they were intensively involved in competitive strategies aimed to attract patients. These strategies included developing links with community medicine, developing new services, competing for key staff, and improving facilities, equipment, customer service and patient satisfaction. High levels of admissions and turnover were used by the hospitals’ directors to justify higher next-year budgets in the Ministry of Health’s budget allocation amongst its centers. In the early 1990s, two further developments underlined the importance of financial considerations in hospitals’ behavior. First, to shorten queues, the MOH allowed the hospitals to perform ‘‘afternoon sessions’’ managed by extra-budgeted Research Funds (see below). Second, in 1990 the Netanyahu State Commission of Inquiry into the Functioning and Efficiency of the Israeli Health Care System recommended, among other things, transforming government hospitals into self-managing ‘‘Trusts’’, a recommendation which has boosted business-like behavior among the large government centers. Following the American experience, several nations have introduced prospective payments for general hospitals and experienced similar changes. For example, the introduction of a DRG-based hospital financing system in Italy in 1995 caused an increase in number of admissions, a drop in length of stay, and a rise in severity of illness. No change was found in mortality and readmission rates (Louis et al., 1999). The goal of the present investigation is to examine the effect of the introduction of PP for selected procedures in Israel, on the number of admissions, length of stay, quality of care, and hospitals’ real incomes during its first year. It is assumed that in the short run these changes reflect the main modifications in hospitals’ behavior in response to the new reimbursement system.
Table 1 presents the definitions of these procedures. While Cholecystectomy includes only one type of intervention, all other procedures include a variety of treatments all for which the hospital is prospectively reimbursed equally. Hysterectomy includes three types of operations (subtotal, total abdominal and vaginal). Joint replacements include several operations; we selected the types performing hip and head of femur replacements. Operations on lens include mainly (90%) cataract extraction and insertion of lens prosthesis. Heart surgeries include many types of interventions (not included are cardiac catheterizations and angioplasty which fall under separate group rates). Upon consulting a cardiologist, these were classified into four main subgroups: revascularization, operations on valves, operations on structures adjacent to valves and other operations. The types of operations within procedures were used to control for case-mix. Amongst the general hospitals, the four largest medical centers were selected for the study. Three are government/municipal hospitals (Sheba, Sorasky, and Rambam) and the fourth is a public non-profit center (Hadassah). These four centers include 26% of all general beds and host 23% of all general inpatient days. Specifically, 20% of all lens operations, 30% of all hip replacements, 60% of all heart surgeries, 12% of all hysterectomies, and 30% of all cholecystectomies were performed in these centers in 1991 (MOH, unpublished data). It is clear, however, that these four medical centers do not represent the Israeli general inpatient care sector. All of them are big teaching hospitals, located in the large cities of Tel Aviv, Haifa, and Jerusalem. The choice of the four centers was shaped by several considerations. First, the General Sick Fund and the private centers refused to participate in the study. Second, 10 out of the
Table 1 Definitions of procedures and types Procedure
Type
5122
Cholecystectomy Hysterctomy
Methods
Hip replacement
Following the methodology used by Kahn et al. (1990), we focused on five selected procedures to examine the effect of the introduction of PP: Cholecystectomy, hysterectomy, hip replacement, operations on lens (cataract), and heart surgeries. These procedures were chosen primarily because of the existing potential reduction in LOS compared to other countries.
Operations on lens Heart surgeries
ICD-9
Subtotal abdominal Total abdominal Vaginal Hip replacement Rep. of head of femur
683 684 685 8151-2, 8159 8161-2 1300–1390
Revascularization Operations on valves Operations on structures Adjacent to valves Other
3610–3630 3551–3599 3500–3539 3691–3923
984
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
14 non-profit hospitals are small charity hospitals (owned by the Christian Church) or hospitals serving only the Arab population in the Jerusalem district. These hospitals did not have good, reliable, updated databases. Similar data problems were encountered in the relatively small peripheral government hospitals. We believe that the four centers studied represent the Israeli governmental and public non-peripheral general inpatient subsector, where about half of the Israeli general inpatient care is provided. A total of 17,400 hospitalization records were identified by the relevant ICD-9 codes (Table 1) during the periods July 1988–June 1990 (‘‘before’’) and July 1990–June 1991 (‘‘after’’). Quality of care was measured by the 60-day postdischarge readmission rate and in-hospital, 60- and 365days post-discharge mortality (mortality data was obtained by matching the identity numbers in our data with the Ministry of Interiors population file. For 6.5% of the cases no match was found). The effect of the introduction of PP (indicated by a dummy variable where 1=after the change) on length of stay (both total and prospectively paid) was estimated
using regression analysis. The other explanatory variables included case-mix (indicated by type of operation, age, and sex) and the operating hospital fixed effect. We used an ordinary least squares (OLS) regression corrected for heteroskedasticity with the dependent variable being the natural logarithm of the stay. This specification was found superior in terms of fit and significance to the linear model. A Poisson regression, specifically accounting for the measurement nature of length of stay, provided exactly the same estimates. The effects of the introduction of PP (a dummy variable where 1=after the change) on readmission and mortality rates were estimated using Logit regressions. The set of explanatory variables was the same as above.
Results Admissions Table 2 presents the yearly number of admissions before and after the change. While the number of admissions for cholecystectomy and hysterectomy re-
Table 2 Number of cases and total LOS Yearly n
Mean (M)
Median
SD (S)
% over M+2S
% over M+3S
Cholecystectomy Before After % Change n
849 849 0.0 2547
12.5 12.4 0.8
10.0 9.0 10.0
8.3 12.8 54.2
3.5 2.0
1.5 1.2
Hysterectomy Before After % Change n
854 860 0.7 2568
11.2 11.0 1.8
9.0 9.0 0.0
6.8 8.0 17.6
4.0 2.2
2.3 1.2
Hip replacement Before After % Change n
588 839 42.7 2015
27.2 23.9 12.1
19.0 16.0 15.8
23.6 23.9 1.3
4.3 3.0
1.6 1.2
Operations on lens Before After % Change n
2,375 2,096 11.7 6846
5.3 4.7 11.3
5.0 4.0 20.0
5.5 8.7 58.2
0.7 0.2
0.3 0.2
Heart surgeries Before After % Change n
934 1,446 54.8 3314
20.7 19.5 5.8
16.0 13.0 18.8
17.6 22.7 28.0
3.2 3.4
1.4 1.8
985
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
mained constant, the number of cataract operations performed in these hospitals dropped by 12%, the number of hip replacements increased by 43%, and the number of heart surgeries increased by 55%.
expected from the 1991 Israeli PP incentives structure, the prospectively paid LOS decreased and the retrospectively paid (per-diem) LOS increased. Total LOS, however, decreased.
Length of stay
Sixty days readmission rate
Several summary measures of the distributions of the observed (unadjusted) total length of stay (LOS) are presented in Table 2. For all procedures, both the mean and the median LOS dropped, as is expected from the incentives originated in PP. The standard deviation of LOS, however, increased with the PP. For cholecystectomy and cataract treatments that increase reached 54– 58%. This means that the drop in LOS was not uniform across different levels of LOS before the change. However, the rate of outliers (LOS above the mean plus 2 standard deviations) dropped (except for heart surgeries). Table 3 presents the adjusted rates of change in total LOS and in the prospectively paid LOS from the regression analysis. The full estimated regressions are presented in Appendix A. In all five procedures total LOS dropped considerably, with a 7% decrease for cholecystectomy, a 3% drop for hysterctomies, and 10– 18% drop for the other procedures. The sharpest decrease occurred for cataract operations. Prospectively paid LOS dropped even more, with an 8% decrease for cholecystectomy, a 6% drop for hysterctomy and a 15– 22% drop for the other procedures. For lens operations, there is no stay outside the ophthalmology ward, so that the prospectively paid LOS is identical to total LOS. These results indicate that the proportion of prospectively paid LOS in total LOS decreased. Actually, for all procedures, the LOS reimbursed by per-diem payments did not decrease. In cataract treatment, cholecystectomy, and hysterectomy, LOS in other departments is negligible. However, in hip replacement, the LOS, which is reimbursed per-diem, increased from 7.2 to 7.9 days, a 10% increase. In heart surgeries, the retrospectively paid LOS increased from 3.7 to 4.3 days, a 16% increase. As
Table 4 reports the readmission rates within 60 days from (live) discharge. The ‘‘before’’ and ‘‘after’’ rates are the unadjusted ones. The third column reports the adjusted odds-ratios (OR) from the Logit regression. Appendix A reports the full set of estimates. In all five procedures the readmission rates increased, although the increase was only significant for cholecysectomy, hip replacement, and heart surgeries. The adjusted OR indicates a 50–80% increase in readmission rate for these procedures.
Table 3 Adjusted rates of change in LOSa,b
Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries
Total LOS
PP LOSc
7.3 3.1 17.6 18.1 9.7
7.7 5.6 21.7 F 14.9
a Dependent variables are ln(LOS). Adjusted for age, sex, type of procedure and operating hospitals. b See Tables 8 and 9 for the full estimated regressions. c Prospectively paid LOS.
Mortality Table 5 presents the changes in the unadjusted rates of in-hospital mortality, within 60 and 365 days, upon (live) discharge. The Logit analysis revealed that for all procedures, the adjusted before–after OR’s are not significantly different from one for the three measures
Table 4 Readmissions within 60 days (%)a
Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries
Before
After
Adjusted ORb,c
5.2 6.3 6.2 4.8 4.8
9.9 7.0 10.1 5.1 7.7
1.8 (1.1) 1.7 (1.1) 1.5
a
Percent out of live discharges. Logistic regression. Adjusted for age, sex, type of procedure and operating hospitals. See Table 10 for the full estimated regression. c Values in parentheses are not significantly different from one. b
Table 5 Unadjusted mortality rates (%) In hospital
Within 60 Da
Within 365 Da
Before After Before After Before After Cholecystectomy 1.1 Hysterectomy 0.1 Hip replacement 4.1 Operations on lens 0.0 Heart surgeries 6.6 a
1.6 0.1 1.5 0.0 5.3
Percent of live discharges.
0.4 0.2 2.5 0.5 1.2
0.5 0.1 2.8 0.2 1.1
3.0 1.7 10.2 3.4 3.7
4.1 2.1 9.7 2.6 3.6
986
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
of mortality, namely, that no significant change in mortality was introduced by the PP.
heart surgeries, the income per day was twice as large as the per-diem rate.
Hospitals’ real incomes Table 6 provides general information on the reimbursement rates for inpatient care before and after the introduction of PP for the five selected procedures. For international comparisons, the values are given in 1991 USD adjusted for purchasing power ($GDP PPP). The rates are the average rates over the respective periods. Table 7 presents the changes in hospitals’ real revenues from the five procedures. Before the change, revenue per day is the per-diem rate. After the introduction of PP, revenue per day was calculated as the prospective rate divided by the ‘‘after’’ median total length of stay. Before the change, revenue per case was calculated by the per-diem rate multiplied by the ‘‘before’’ median total length of stay. After the change, the revenue per case is the prospective rate. For cholecystectomy, hysterectomy and lens operations, real income dropped with the introduction of PP. This is true for both incomes per day and per case. For cataract surgeries, revenue per day dropped by 44% and per case by 55%. For hip replacement, revenue per day increased by 15%, but total revenue per case decreased by 3%. For heart surgeries, real incomes increased dramatically (179% per day and 126% per case). These changes are reflected in the ratio of the per-diem rate to revenue per day after the introduction of PP. For lens operations, the perdiem rate was twice as large as the income per day. For
Discussion
Table 6 Reimbursment rates for inpatient care (1991 $GDP PPP)
Per-diem Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries
Before
After
230 F F F F F
244 1732 1617 3754 519 8085
Table 7 Percent changes in hospitals’ real revenues
Cholecystectomy Hysterectomy Hip replacement Operations on lens Heart surgeries a
After.
Per day
Per case
Per-diem/Rev. per daya
16 22 15 44 179
25 22 3 55 126
1.3 1.4 0.9 1.9 0.4
The results show that the introduction of PP for selected procedures to general hospitals in Israel caused the ‘‘classical’’ expected short-term changes in the hospitalizations’ characteristics. Controlling for age, sex, type of procedure, and the operating hospital, total length of stay decreased. The drop in LOS in heart surgeries, hip replacements and cataracts was higher than thrice the mean annual decrease in LOS in the respective wards over the period 1975–1989. The drop in the prospectively paid LOS (in the operating department and in intensive care units) was even sharper. While the new reimbursement system did not cause any significant change in in-hospital, within 60 and 365 days upon discharge mortality rates, the odds for 60 days readmission increased. We note that at the time (1991), all readmissions were reimbursed, regardless of their timing. Further research will explore whether the increase in readmission rates was indeed an indicator of lower quality of care (and of premature discharges), or whether it resulted from strategic behavior of the hospitals to increase revenues (Cutler, 1995). The new PP introduced changes in the hospitals’ revenues per case and revenues per day. These changes were determined both by the levels of the prospective rates and the changes in length of stay. In performing cataract surgeries, the hospitals’ revenues per day dropped by 44% and revenue per case by 55%. The rate (519 USD adjusted for 1991 GDP PPP) was low and even with the 18% drop in length of stay, the perdiem rate was twice as large as the revenue per day. These changes clearly provide an incentive for the hospitals to avoid cataract patients. Indeed, the results show a 12% drop in the number of cataract surgeries performed (Table 2). Revenues for hysterectomies and cholecystectomies dropped by 22–25%. However, the number of cases remained constant. Although revenues for a hip replacement surgery dropped by 3%, the number of cases increased by 43%. Apparently, for these three procedures, the ‘‘before’’ LOS, which determined the revenue per case before the introduction of PP, was long enough to assure such high margins, that even the decrease in revenue per case left the hospitals with satisfactory ones. Finally, the sharp increase of 55% in the number of heart surgeries is clearly consistent with the 126% increase in the hospitals’ revenues per heart surgery (the PP rate was set at 8085 $PPP). The increase in the number of admissions to hip replacement and heart surgeries is consistent with the Italian experience (Louis et al., 1999) but contradicts the
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
US experience, where the anticipated increase did not materialize (Hodgkin & McGuire, 1994). For heart surgeries, the increase in utilization is related to the generous rate per case (and see below). For hip replacement, while the revenue per case dropped slightly, the revenue per day increased by 15% and was 11% higher than the per-diem rate. Consequently, hip replacement patients were still more desirable than other patients paid for by the per-diem arrangement. The changes in the number of admissions and length of stay induce a change in load of total inpatient days in the relevant wards. According to our findings, while there was no change in cholecystectomies (general surgery), nor in hysterectomies (gynecology), the volume of hip replacement hospitalization days increased by 30% in the orthopedic ward, the volume of heart surgeries inpatient days increased by 50% in the chest surgeries ward, and cataract inpatient days dropped by 23% in the ophthalmology department. Departmental data on number of beds and occupancy rates over the (calendar) years 1989–1991 (MOH, 1989–1991) for the four hospitals included in the study, indicate that in three of the four hospitals, the number of beds in the ophthalmology ward dropped by about 20% between 1989 and 1991. Although the trend to perform eye surgeries in outpatient facilities began prior to the introduction of the PP (see below), the relatively unattractive prospective rate probably accelerated that trend. Although hip replacement hospitalization days increased by 30%, there is no parallel increase in terms of occupancy in the orthopedic ward. Furthermore, occupancy rates in these wards actually decreased between 1989 and 1990 in all four hospitals. Nor was an increase in occupancy was found in rehabilitation wards. It seems that hip replacement days substituted inpatient days for other orthopedic procedures (no data is available to explore this issue further). All hospitals witnessed an increase in 1990–1991 in occupancy in the cardiology department and/or in the cardiac intensive care units. However, that increase should also be attributed to the sharp increase in cardiac catheterizations and angioplasty following the introduction of the PP to these procedures. As in most natural experiments, there are several other factors that should be considered with regard to the identification of the effects of the introduction of PP. Note, however, that we examined changes from one year to the following year, so that the effects of general time trends in the data are minimized. First, during that period, laparascopy proceduresFwith shorter staysF became popular, replacing the traditional cholecystectomies. However, the use of laparascopy in the four medical centers included in this investigation did not begin before the end of 1991. Consequently, the 7–8% drop in total length of stay cannot be attributed to that technological change. Second, there were three organiza-
987
tional–structural issues that should be noted. Since 1989, the sick funds have exercised ‘‘length of stay control’’ in order to contain their inpatient retrospectively paid costs. Under pure PP, no such control is needed (actually, a need for quality control is more necessary). As was mentioned earlier, the retrospectively paid LOS increased following the introduction of the new reimbursement system (although total LOS decreased), so that the sick funds’ pressure cannot explain the drop in LOS. Another initiative of the sick funds, which started in that period, was to shift cataract patients to other forms of care, mainly ambulatory care facilities. This may explain the 12% drop in the number of cases. However, as was previously mentioned, if that shift was selectiveFwith lower risk patients being shiftedFthe cataract patients case-mix was supposed to rise and the 18% drop in LOS could not result from that shift. The low level of the prospective payment for cataract surgery set by the Ministry of Health might have had a similar purpose, namely, to encourage that shift. The generous rate for heart surgeries was probably part of the Ministry of Health’s initiative to reduce waiting times for these surgeries. In early 1990, under public pressure, the MOH launched a ‘‘shortened queues’’ campaign, which included special payments to hospitals and their operating staff for afternoon sessions. That campaign induced a sharp increase in the number of surgeries performed. The July 1990 introduction of the PP for heart surgeries, with its generous level, enhanced that earlier trend. This investigation focusedFfor the first timeFon the implications of the introduction of the Israeli PP during its first year to the governmental and public, non-profit, non-peripheral, general inpatient subsector. While more experience has accumulated since, data (in)availability poses severe difficulty in analyzing that experience. Further research is clearly needed, however, to evaluate the change (and its expansion) in subsequent years and in the entire Israeli inpatient sector. Acknowledgements We would like to thank Gabi Barabash and Nira Shamai for their assistance at various stages of the research, and the management of the participating hospitals for providing us with the data. We also thank the referees for their useful comments on an earlier draft.
Appendix A The full estimates of ln total LOS regressions, ln PP LOS regressions and logistics regression of readmissions within 60 days are given in Tables 8–10.
988
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989
Table 8 Ln(total LOS) regressions
PP (1=yes) Age Gender (1=men) Hospital 1 Hospital 2 Hospital 3 Constant Hip replacement types (ICD-9) 8141 8151 8159 8161 Hysterectomy types (ICD-9) 683 684 685 Heart surgery groups Group 2 Group 3 Group 4 ADJ R SQ
Hip replacement
Cholecystectomy
Hysterectomy
Operations on lens
Heart surgery
Coefficient
SE
Coefficient
SE
Coefficient SE
Coefficient
SE
Coefficient SE
0.193 0.007 0.051 0.076 0.413 0.038 2.346
0.01 0.001 0.01 0.019 0.016 0.02 0.036
0.076 0.006 0.074 0.149 0.005 0.142 2.117
0.013 0.001 0.013 0.016 0.016 0.021 0.022
0.031 0.008 F 0.427 0.375 0.502 1.735
0.012 0.001 0.011 0.017 0.019 0.021 0.065
0.102 0.007 0.11 0.734 0.409 0.294 2.422
0.416 0.379 0.349 0.652
0.019 0.022 0.018 0.022
F F F F
F F F
F F F
F F F
F F F
0.103
0.013 0.198 0.001 0.003 0.006 0.029 0.468 0.026 0.016 0.03 0.414 0.122 1.221
F F F F 0.534 0.625 0.511
F F F F
F F F F
0.129 F 0.118 F 0.118 F
F F F
F F F
0.078
F F F
0.105
0.059 0.117 0.084
0.275
0.009 0.001 0.01 0.014 0.016 0.022 0.083
0.027 0.012 0.027
0.148
Table 9 Ln(PP LOS) regressions
PP (1=yes) Age Gender (1=men) Hospital 1 Hospital 2 Hospital 3 Constant Hip replacement types (ICD-9) 8141 8151 8159 8161 Hysterectomy types (ICD-9) 683 684 685 Heart surgery groups Group 2 Group 3 Group 4 ADJ R SQ
Hip replacement
Cholecystectomy
Hysterectomy
Coefficient
SE
Coefficient
SE
Coefficient
0.244 0.004 0.044 0.296 0.114 0.059 2.565
0.012 0.001 0.12 0.02 0.017 0.021 0.041
0.079 0.007 0.006 0.094 0.026 0.158 2.088
0.013 0.001 0.013 0.016 0.016 0.021 0.022
0.058 0.008 F 0.442 0.687 0.522 1.655
0.414 0.431 0.328 0.324
0.022 0.024 0.02 0.026
F F F F
F F F
F F F
F F F
F F F
0.122
0.111
Heart surgery SE 0.014 0.001 0.029 0.029 0.031 0.129
F F F F 0.435 0.69 0.577
SE
0.161 0.006 0.067 0.418 0.311 0.241 2.825
0.011 0.001 0.012 0.016 0.017 0.024 0.098
F F F F 0.139 0.125 0.123 F F F
0.128
Coefficient
F F F 0.114 0.177 0.522 0.172
0.032 0.014 0.038
989
A. Shmueli et al. / Social Science & Medicine 55 (2002) 981–989 Table 10 Logistic regression of readmissions within 60 days Hip replacement a
Cholecystectomy a
Hysterectomy a
p
Coefficient
p
Coefficient p
Coefficient
p
Coefficient pa
0.09 0.02 0 0 0.37 0.05 0
0.095 0.028 F 7.529 0.281 0.651 4.043
0.095 0.017 0.166 1.489 0.503 0.142 1.654
0.28 0 0.15 0 0.09 0.53 0
0.405 0.013 0.188 2.195 0.799 0.133 2.277
0.531 0.007 0.281 2.407 0.554 0.984 2.356
0 0.21 0.11 0.01 0.19 0.01 0
0.578 0.011 0.431 0.714 0.338 0.598 3.078
0.532 0.337 0.179 0.076
0.12 0.4 0.58 0.87
F F F F
F F F F
F F F
F F F
F 0.62 F
F F F
F F F
F F F
CHI SQ
88.02
p¼0
66.4
p¼0
75.96
0.53 0 0.34 0.14 0.02 0
0
a
Heart surgery
Coefficient PP (1=yes) Age Gender (1=men) Hospital 1 Hospital 2 Hospital 3 Constant Hip replacement types (ICD-9) 8141 8151 8159 8161 Hysterectomy types (ICD-9) 683 684 685 Heart surgery groups Group 2 Group 3 Group 4
a
Operations on lens
0.02 0.01 0.29 0 0.04 0.78 0
F F F F
F F F F
F F F
F F F
F F F
0.537 0.66 1.509
0.09 0.04 0
214.23
p¼0
p ¼ 0 166.51
p¼0
Wald test.
References Cutler, D. (1995). The incidence of adverse medical outcomes under prospective payment. Econometrica, 63, 29–50. Chinitz, D., & Rosen, B. (1993). A tale of two markets: Hospital competition in Israel. Brookdale Institute RR-30-93, Jerusalem, September 1993. Ellis, R. P., & McGuire, T. G. (1986). Provider behavior under prospective reimbursement. Journal of Health Economics, 5, 129–151. Ellis, R. P., & McGuire, T. G. (1996). Hospital response to prospective payment: Moral hazard, selection, and practice-style effects. Journal of Health Economics, 15, 257–277. Feinglass, J., & Holloway, J. J. (1991). The initial impact of the Medicare PPS on US health care: A review of the literature. Medical Care Review, 48, 91–115. Hodgkin, D., & McGuire, T. (1994). Payment levels and hospital response to prospective payment. Journal of health Economics, 13, 1–29.
Kahn, K. L., Keelr, E. B., & Sherwood, M. J., et al. (1990). Comparing outcomes of care before and after implementation of the DRG-based prospective payment system. Journal of the American Medical Association, 264, 1984–1989. Lave, J. R., & Frank, R. G. (1990). Hospital supply response to prospective payment as measured by length of stay. Advances in Health Economics and Health Services Research, 11, 1–26. Louis, D. Z., Yuen, E. J., Braga, M., & Cicchetti, A., et al. (1999). Impact of a DRG-based hospital financing system on quality and outcomes of care in Italy. Health Services Research, 34, 405–415. Ma, C. A. (1994). Health care payment systems: Cost and quality incentives. Journal of Economics & Management Strategy, 3, 93–112. McLellan, M. (1997). Hospital reimbursement incentives: An empirical analysis. Journal of Economics & Management Strategy, 6, 91–128. Ministry of Health (1989–1991). Inpatient care institutions in Israel. Jerusalem, Hebrew.