Cover Image

PAPERBACK
$136.25



View/Hide Left Panel
Click for next page ( 355


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 354
For-Profit Enterprise in Health Care. 1986. National Academy Press, Washington, D.C. Medicare Patient Outcomes and Hospital Organizational Mission Gary Gaumer The objective of this paper is to examine the relationship between indicators of quality of care and ownership status in a large sa~nple of U. S. hospitals. Very little is known about the determinants of interhospital differences in pa- tient outcomes and even less about the rela- tionship between organizational mission and patient care. Work by Lust (1980), Scott et al. (1979), and Bunker et al. (1969) has established that significant disparities in hospital mortality rates exist which are, in part, associated with hospital organizational attributes such as size, nurse staffing, teaching mission, physician ex- penence, and administrative span of control. No studies have yet examined the relationship between proprietary status and patient out- comes, nor has the influence of multihospital system affiliation on outcomes been studied. This paper examines several types of mea- sures which, taken together, may be indicative of differences in patient care practices and quality of care across groups of hospitals with differing organizational missions. The mea- sures we analyze include ~ Post-operative mortality for Medicare elective surgical admissions measured over the stay and within 180 days of admission ~ 90-day post-discharge readmission rates for Medicare elective surgery admissions Stabs regarding acccredita~on by Me Joint Commission on Accreditation of Hospitals UCAH) Two measures of Medicare case mix. Mr. Gaumer is Vice President of health care re- search with Abt Associates, Inc., Cambridge, Massa- chusetts. 354 METHODS The hospital sample used for our analyses includes two components: 1. A 25 percent simple random sample of all continental U.S. hospitals with a median length of stay of 15 days or less over the 1970- 1978 period 2. All other similarly defined short-term hospitals in the 15 states with prospective pay- ment programs. ~ The surgical admissions we studied included inguinal hernia repair, hysterectomy, chole- cystectomy, he m orrhoide ctomy , op en pro s - tatectomy, transurethral resection of prostate, excision of bladder lesion, and mastectomy. Patient data were taken from the Health Care Financing Administration's (HCFA) MED- PAR file, containing clinical and billing data on Medicare inpatient stays for 20 percent of Medicare beneficiaries. Data were gathered for the 1974-1981 period on each sample hos- pital. Table 1 shows the resultant sample sizes. Table 2 shows characteristics of sample hos- pitals. Constructing measures of readmission rates and post-discharge fatality rates required spe- cial processing of the HCFA files. A 90-day readmission indicator (0, 1) was created for each study patient by scanning the entire MEDPAR file across years to see if patients discharged fiom study hospitals were read- mitted to any hospitals within 90 days of the date of discharge. For 180-day post-admission death rates, a similar process was used, except that HCFA's Medicare eligibility files were used to determine if beneficiary death oc- curred within 180 days of the admission. Unpublished accreditation data were ob- tained from JCAH. Status codes were obtained that reflect full (2- or 3-year) accreditation and

OCR for page 354
355 Cal to 3 Cal C) ~0 Z U) _ Con - Cal Con U: Ct ._ O O 5- "D 3 Z ~O ~ 0= LIZ 04 so -0 'Cal ~ Z C >- '~- C) =5 sol _1= eC ._ o -8'~ C.) O =. Z Cal .o D" O Ed ._ Cal ~ C) Z Bet .~ so O ~ Go, Ct i== C; Pa O Cal - ~ CO 00 '$4 can ~ ~ C.D 0 _ Cal oo Cal CO O _ ~ ~ _ _ - CD Cal Cal O O ~ US Cal ~ ~ Cal Cal ~ Cal ~ {D Cal CC) _ Co CO O - 0 1~ 00 CD 0 ~4 0 o0 ~ ~ 00 00 oo CO Cal Cal oC) - Cad 0 ~ o3 ~ O ~ C~ oc ~ C] c~ _ ~ e~ _ ot c~ _ _ _ _ _ _ _ ~ CO ~ ~ 0O ~ C~ U. CD 0t ~ CD oO ~ CS' O C53 ~ O _ O C~ CO =4 _ _ _ _ _ _ C~ e~ C~ 00 C~ O O C~ ~ ~ O - C~ ~ ~ C~ C] ~ C~ C~ _ _ _ _ _ _ _ _ C5~ c~ c~ o~ _ ~ c~ _ ~ ~ ~ C~ ~ C~ u, _ a, C~ ~ 00^- C1^ ~ ~ CO ~ 00 C~ C~ - C~ CO ~ CD 00 ep - ~ 00 CS _ ~ C~ C~ - C~ C~ - CD - O C~ d4 ~ ~ ) ~ ~ 1O 00 _ C~ C~ - O O ~ - - C~ ~ C~ U: ~ C~ ~ C~ ~ ~ C;, ~ C~ ~ ~ _ O) ~_ _ ~ ~ C~ ~ O U: C~ _ C~ _ C~ C~ ~ ~ C ~_ b.0 0 C ~O O S: V: C) ^~>% _ _ ~ ~ o~ ~ ~ g3 c~ ~ ~ E = ~ 5! ~ ~ >- ~ 5 >~ _ ~ ~ c~ ~ 0 ~ ~ ~ 5 o ~ ~ ~ ~ C~ ~ C~ CD ~ 0O ~ 0 ~ o~ ~ ~ o o o (D ~ ~ 0 0 - oo 0

OCR for page 354
356 FOR-PROFIT ENTERPRISE IN HEALTH CARE TABLE 2 Characteristics of Sample Hospitals,a 1981 Data Proprie- Non Proprie- tary Non-profit tary Inde- profitsInde VariableChain pendent Chainpendent Number of hospitals in sample107 117 4351,310 Average number of beds staffed148 103 199185 Minimum40 11 158 Maximum638 400 872969 Average Medicare case-mix index1.00 0.97 1.02102 Minimum0.79 0.71 0.720.56 Maximum1.20 1.22 1.391.75 Percent fully JCAH accreditedC0.79 0.50 0.740.55 Percent covered by state prospective payment program0.17 0.44 0.570.64 Percent located in SMSAd0.78 0.68 0.610.60 County Characteristics (unweighted mean) Percent population on AFDCe3.9 4.3 4.24.2 Births per 100,000 population1,675 1,568 1,6581,506 Per capita income8,756 8,540 8,3078,423 Percent population with health insurance75.5 81.5 82.287.2 Percent population on Medicare11.5 12.8 13.013.4 Number of hospital beds per 100,000 population537 638 787770 M.D.s per 100,000 population161 153 144145 Percent M.D.s who are specialists0.81 0.75 0.710.75 HMO penetration rate0.065 0.041 0.0380.035 Percent population white0.80 0.81 0.880.89 Median years of education11.9 11.S 11.911.9 Unemployment rate7.2 7.3 7.67.7 - aExcludes government hospitals and members ofthe Council of Teathing Hos- pitals. bAmerican Hospital Association data for 1981-1982 were used to assign this ownership status indicating membership in a multibospital system. CJoint Commission on Accreditation of Hospitals. Standard metropolitan statistical area. eAid to Families with Dependent Children.

OCR for page 354
MEDICARE PATIENT OUTCOMES TABLE 3 Unadjusted In-Hospital Mortality Rates by Hospital Type Proprietary Propnetary Nonprofita Nonprofit Chain Independent Chain Independent All eight elective procedures 1975 0.010 0.013 0.012 0.013 1977 0.009 0.009 0.014 0.012 1979 0.010 0.007 0.011 0.011 1981 0.011 0.004 0.011 0.012 1981 by procedure Hemorrhoidectomy 0.000 0.000 0.000 0.004 Cholecystectomy 0.042 0.014 0.032 0.032 Inguinal hernia 0.004 0.004 0.006 0.004 Open prostatectomy 0.000 0.000 0.014 0.015 Transurethral resection of prostate 0.009 0.000 0.009 0.010 Hysterectomy 0.000 0.013 0.002 0.006 Excision of bladder lesion 0.000 0.000 0.010 0.012 Mastectomy 0.000 0.012 0.015 0.006 aAmencan Hospital Association data for 1981-1982 were used to assign this ownership status indicating membership in a multihospital system. only 1-year accreditation as a result of the sur- vey. For analytic purposes, we used an indi- cator of whether the hospital was fully (2- or 3-year) accredited or not. 2 Medicare case-mix data for 1981 were coded from the September 30, 1983 Federal Regis- ter. 357 Several measures of organizational mission were developed from AMA's annual survey designation of proprietary status. An indicator of affiliation with a proprietary hospital system was coded Tom 1974-1981 annual directories of He Federation of American Hospitals. AMA's 1982 annual survey data were used to deter TABLE 4 Unadjusted 180-Day Mortality Rates by Hospital Type _ . . . Proprietary Proprietary Nonprofita Nonprofit Chain Independent Chain Independent All eight elective procedures 1975 0.049 0.045 0.050 0.013 1977 0.063 0.057 0.058 0.012 1979 0.062 0.070 0.054 0.011 1981 0.055 0.041 0.056 0.054 1981 by procedure Hemorrhoidectomy 0.020 0.000 0.029 0.025 Cholecystectomy 0.059 0.067 0.083 0.076 Inguinal hernia 0.027 0 043 0.029 0.026 Open prostatectomy 0.051 0.032 0.055 0.040 Transurethral resection of prostate 0.077 0.041 0.063 0.064 Hysterectomy 0.000 0.026 0.019 0.019 Excision of bladder lesion 0.095 0.039 0.083 0.095 Mastectomy 0.033 0.037 0.047 0.037 aAmencan Hospital Association data for 1981-1982 were used to assign this ownership status indicating membership in a muldhospital system.

OCR for page 354
358 mine if the hospital was a member of a mul- tihospital system. Using these data, for-profit affiliations were available for all hospital years, but nonprofit chain affiliation was only cap- tured at one point in time. DESCRIPTIVE STATISTICS Tables 3, 4, and 5 contain means on patient outcome measures.3 The unadjusted outcome data often show large differences across types of ownership, and considerable instability over time no doubt arising, in part, from the rel- atively small samples of patients for some pro- cedure groups. Mortality During the Stay For the aggregate of eight elective proce- dures, death rates averaged about 1.1 percent. No significant trends are apparent. Death rates generally appear lower in proprietary hospi- tals, particularly in the independent investor- owned hospitals. Post-operative death rates are often zero for procedures such as open pros- tatectomy and excision of bladder lesion. Mor- tality rates for the specific procedures show the instability in measures that arises from in- frequent deaths and relatively small samples. FOR-PROFIT ENTERPRISE IN HEALTH CARE Mortality Rates Within 180 Days of Admission Six-month mortality rates average about 5 percent for all elective surgery, often some- what lower in proprietary hospitals. No trends are apparent. Among the specific procedures, bladder lesions and gall bladder removals have highest death rates. Patterns of mortality by hospital organizational types are very incon- sistent across the specific procedures. Readmission Rates Within 90 Days of Discharge Readmission rates for the eight procedures are about 9 percent. There is a slight upward trend in these rates. Readmission rates show similar performance of nonprofit and proprie- taries, although on balance, mean rates for proprietaries are slightly higher. Readmission rates are low for hernia repair, hysterectomy, and mastectomy, and highest for procedures relating to the bladder and prostate. The descriptive data suggest that it may not be meaningful to statistically examine many of the specific procedures separately due to the combined consequences of rare outcomes and small samples of cases for particular ownership TABLE 5 Unadjustec! 90-Day Readmission Rates by Hospital Type Proprietary Proprietary NonprofitQ Nonprofit Chain Independent Chain Independent All eight elective procedures 1975 0.067 0.083 0.084 0.081 1977 0.104 0.101 0.103 0.094 1979 0.088 0.093 0.084 0.083 1981 0.111 0.089 0.093 0.089 1981 by procedure Hemorrhoidectomy 0.140 0.056 0.079 0.073 Cholecystectomy 0.083 0.072 0.092 0.087 Inguinal hernia 0.077 0.075 0.058 0.052 Open prostatectomy 0.098 0.081 0.110 0.081 Transurethral resection of prostate 0.136 0.129 0.115 0.113 Hysterectomy 0.063 0.013 0.057 0.064 Excision of bladder lesion 0.190 0.099 0.114 0.139 Mastectomy 0.076 0.063 0.076 0.047 America Hospital Association data for 1981-1982 were used to assign this ownership status indicating membership in a multihospital system.

OCR for page 354
MEDICARE PATIENT OUTCOMES categories. We examined standardized out- comes for the following types of procedures: All elective procedures taken together Cholecystectomy Hernia repair Transurethral resection of prostate. Figures 1, 2, 3, and 4 show farther detail on the unadjusted patient outcome rates for these four procedure categones. Data on length of stay are also included for reference pur- poses. All tables show means weighted by number of admissions in the particular surgical category. Ike tables show four bed-size cat- egories. Size of Hospital Patient outcomes for elective procedures seem to vary systematically with respect to hospital size. Larger hospitals have longer lengths of stay, higherin-hospitalfatality rates, and lower readmission rates. Inguinal hernia repair is the only exception, where no pattern in readmission rates is apparent. The pattern is different for mortality rates within 180 days of admission, generally, the fatality rates are much lower for very small hospitals. Chole- cystectomy is an exception, where no size pat- tern is seen. Medicare Case Mix Hospitals were classified on the basis of the HCFA-published 1981 Medicare diagnosis-re- lated group (DRG) case-mix index. As with size, patterns are evident in He raw data. Readmission patterns are quite uniform, hos- pitals with a more complex (costly) Medicare case mix have lower readmission rates and shorter lengths of stay for these elective pro- cedures. Mortality rate patterns are not so uni- form. Excepting cholestectomy, the in-hospital and 180~day mortality rates are generally higher in hospitals with higher case-m~x values. No pattern is evident for cholecystectomy. Organization of Hospital Looking first at differences between pro 359 prietary and voluntary hospitals, the weighted means show that proprietary hospitals have comparable or lower mortality rates. In-hos- pital mortality rates are lower for all eight pro- cedures taken together (.008 compared to .012) and for transurethral resection of the prostate (.004 compared to .0101. Proprietary hospitals have lower 180-day mortality rates for all eight procedures together (.049 compared to .055) and for cholecystectomy (.061 versus .076~. The only observed instance where proprietary hos- pitals have much higher mortality is hernia repair, where 180-day post-admission mortal- ity is 3. 7 percent compared to 2.7 percent for nonprofit hospitals. Readmission rates are gen- erally much higher in proprietary hospitals, excepting cholecystectomy, where the reverse is true. Length-ostay differences are nonex- istent or quite small, though for the aggregate of eight procedures and prostate surgery the proprietary hospitals have slightly lower lengths of stay. When subgroups of investor-owned hospi- tals are compared to their voluntary chain and nonchain counterparts, some patterns emerge. Independent proprietary hospitals have the lowest values on both mortality rate measures of all four groups of hospitals except for cho- lecystectomy; for this procedure, in-hospital fatality rates are equal to their voluntary coun- terparts and 180-day mortality is higher than in other types of hospitals. Readmission rates and length of stay for independent proprie- taries are generally equal to or higher than for independent voluntary hospitals. The excep- tion is hernia repair, where independent pro- prietary hospitals have the lowest readmission rate. The pattern for proprietary chains is similar; for readmission rates, proprietary chains have higher rates than voluntary chains, except for hernia repair. But unlike the independent pro- prietaries, the investor-owned chain hospitals have lower lengths of stay than the voluntary chain hospitals. For chain hospitals there is no pattern whatsoever for mortality rates across procedure categones; for the aggregate of eight procedures, proprietary chains have mortality rates about equal to those observed in vol- untary chain hospitals.

OCR for page 354
360 6.0 5.0 4.0 3.0 1.2 1.0 id ~ 0.8 111 0.6 0.4 12.0 .0.0 8.0 6.0 9 con 7 75- 200 200 400 400+ SIZE (beds) i FIGURE 1 Eight elective procedures, 1981 data FOR-PROFIT ENTERI'RISE IN HEALTH CARE 180 Day Mortality Mortality During Stay Hi _ ~ _ ~ l it, Readmission J; ength of Stay (live discharges) .9- Not .9 1.1 1.1 + For M ED ICAR E CASEM IX Profit For Profit ~1 Chn Ind Chn Ind Propr Volun ORGANIZATION

OCR for page 354
MEDICARE: PATIENT OUTCOMES 9.0 8.0 Z 7.0 UJ 6.0 5.0 4.0 4.5 4.0 3.5 3.0 c: cry LU 2.5 2.0 t.5 1.0 16.0 14.0 z 12.0 cay cr: us 10.0 8.0 6.0 16 14 12 10 180 Day Mortality I. Mortality During Stay Readmission .N L Length of Stay (live discharges) ~ 75- 200- .9- Not For Chn Ind Chn Ind <75 200 400 400+ .9 1.1 1.1+ For Profit Propr Volun Profit SIZE (beds) MEDICARE ORGANIZATION CASEM iX FIGURE 2 Cholecystectomy outcomes, 1981 data. 361

OCR for page 354
362 FOR-PROFIT ENTERTRISE IN HEALTH CARE 8.0 7.D Be c.' 6.0 LL 5.0 4.0 1 .0 0.8 0.6 LL 0.4 0.2 o 16.0 '_ 14.0 of C: 12.0 G 10.0 8.0 12 10 180 Day Mortality i/ ~ . Mortality During Stay . ~ Readmission _ . l ~ Length of Stay (live discharges) 75- 200- 9- Not For Chn Ind Chn Ind <~75 200400 400+ .9 1.1 1.1t For Profit Propr Volun Profit SIZE (beds) MEDICARE ORGANIZATION CASEM IX FIGURE 3 Inguinal hernia repair, 1981 data.

OCR for page 354
MEDICARE PATIENT OUTCOMES 8.0 7.0 z Cal 6.0 G UJ C~ 5.0 4.0 1.0 0.8 0.6 0.4 0.2 o 16.0 '_ 14.0 at up ~ 12.0 cr 10.0 8.0 12 In 6 10 SIZE (beds) MEDICARE CASEM I X FIGURE 4 Transurethral resection of prostate, 1981 data. 363 180 Day Mortality ~ l OCR for page 354
364 Accreditation Trends The trends in accreditation status have not been pronounced, although there are clear dif- ferences across groups of hospitals. Table 6 and Figure 5 show the unadjusted data on hill ICAH accreditation rates (percent finely accredited). There is a clear demarcation between chain and nonchain hospitals, but little difference between proprietary and nonprofit indepen- dents. For the chain affiliates, about 65 to 75 percent of hospitals maintain full accredita- tion. For independents, hill accreditation rates in 1981 were 50 to 55 percent, with the pro- prietaries up from 40 percent in the early 1970s. Table 6 also shows the specific accreditation status of hospital years being studied. Lower accreditation rates for proprietary indepen- dents are apparently due to lower rates of par- ticipation in the ICAH program. STATISTICAL RESULTS Mortality During the Stay Table 7 shows the results of statistical tests on the ratio of actual to expected death rates FOR-PROFIT ENTERPRISE IN HEALTH CARE at discharge. 4 The table shows various differ- ences of interest (rows) for each of the aggre- gate and individual elective procedures (columns). If probability is less than 0.10 the coefficient is reported which, because of the log form, is interpreted as the difference be- tween the indicated groups of hospitals ex- pressed in percentage terms (e.g., mortality rates are x percent higher in the first-listed group of hospitals than in the other). If the coefficient was not significant (p <.lO), only the direction of differences is reported. For example, the first cell in the table (.119) in- dicates that chain-~liated hospitals have in- hospital mortality rates for all elective proce- dures combined that are ll.9 percent higher than for nonchain hospitals, ret par. The dominant conclusion Tom these tests is that hospital ownership may not be a strong or consistent imBuence on postoperative mor- tality rates; significant results are not propa- gated across all procedures categories. The data are not without patterns, Cough they may fail to be fillly persistent; proprietary status is De- quently found to be associated with lower in- hospital mortality, and chain affiliation is oRen associated with higher mortality. There is no TABLE 6 Trends in Accreditation by the loins Commission on Accreditation of Hospitals UCAH) Percent with 2-Year Accreditationa Proprietary Proprietary Nonprofitb Nonprofit Year Chain Independent Chain Independent 1974 0.660 0.399 0.724 0.605 1975 0.688 0.387 0.679 0.586 1976 0.649 0.400 0.623 0.505 1977 0.688 0.399 0.640 0.514 1978 0.729 0.415 0.662 0.492 1979 0.663 0.398 0.710 0.522 1980 0.676 0.496 0.730 0.547 1981 0.785 0.496 0.736 0.547 Accreditation Status: All Years Contained Percent full 69 42 Percent provisional 10 10 Percent other 21 48 69 54 14 12 18 34 aThe balance of hospitals either did not seek ICAH accreditation, had status of unaccredited, or were accredited for only 1 year. bAmerican Hospital Association data for 1981-1982 were used to assign this ownership status for nonprofits, indicating membership in a multihospital system.

OCR for page 354
MEDICARE PATIENT OUTCOMES 0.7 0.6 0.5 OA 0.3 FIGURE 5 Accreditation trends by hospital status. significant evidence that proprietary chains are significantly different from other ~ror~rietarv hospitals. We are concerned that the pattern (higher mortality in chains; lower mortality in pro- prietaries) may be confounded with severity. Although mortality rates were standardized for the age/sex/co-morbid status and covariates provide control for aggregate case-mix and size differences, we still worry that the differences in mortality may be partially due to severity differences which are not captured in our stan- dardization approach. In part, our concern stems from the pattern of case-mix results be- low showing that proprietary hospitals may have a less complex case mix. 180-Day Post-Admission Mortality The statistical results for tests on this mea- sure generally tend to be the reverse of those found for in-hospital mortality, though usually less consistent. Compared to voluntary hos- pitals, chain hospitals are not consistently dif- ferent from their independent counterparts, though they have significantly lower mortality rates for prostate surgery. Investor-owned hospitals have higher 180-day mortality rates, though these differences are statistically sig- nificant only in the hernia repair and prostate models. The effect of switching to proprietary status is measured directly as Test 3; here we find no significant differences to support the 365 ~ 0.8 _ G Z 0.7 _ -__ o 0.6 _ - - _ o: ^ ' _ Cry if: I Cal J J LL Proprietary -_ _ I_ Independent Proprietary _ _ ~ Nonprofit Chain _ _ Nonprofit Independent ~ 1 1 1 1 1974 1975 1976 1977 1978 YEAR 1979 1980 1981 pattern of higher 180-day mortality in inves- tor-owned hospitals. Results are also inconsistent on the differ- ences between independent proprietary hos- pitals and their chain-affiliated counterparts. There is some indication that switching from independent to chain status is associated with lower 180-day mortality rates; this pattern is statistically significant for the aggregate of eight procedures, and the signs are consistent for the other procedure-specific tests. 90-Day Readmission Rates The readmission tests in Table 7 show only one consistent pattern. Chain affiliates are of- ten found to have higher readmission rates. There is also some evidence that proprietary chains have higher readmission rates than other proprietary hospitals. Accreditation Rates Table 8 shows the statistical results for ac- creditation rates. The likelihood of hill (2-year) accreditation with JCAH is consistently re- lated to the organizational measures; no doubt this reflects, in part, mission and image dif- ferences which differentially affect the pro- pensities of hospitals to seek JCAH accreditation. All the tested models show that chain-a~liated hospitals are more often ac

OCR for page 354
Inhospital mortality Chain vs. independent Proprietary vs. voluntary Switch from voluntary to proprietary Proprietary chain vs. proprietary independent Switch from proprietary independent to proprietary chain 180-day mortality Chain vs. independent Proprietary vs. voluntary Switch from voluntary to proprietary Proprietary chain vs. proprietary independent Switch from proprietary independent to proprietary chain 90-day readmission rate Chain vs. independent Proprietary vs. voluntary Switch from voluntary to proprietary Proprietary chain vs. + -.248 (.099) proprietary independent .204 (.057) Switch from proprietary independent to proprietary chain - 366 FOR-PROFIT ENTERPRISE IN HEALTH CARE TABLE 7 Statistical Results on Patient Outcomes: Percentage Differences in Adverse Outcome Ratea (p value in parentheses) Test Statistic All Eight Electives Cholecystectomy Inguinal Hernia Repair Transurethral Resection of Prostate .119 (.033) 120 (.033) -.401 (.000) + + + + . 124 (.056) - -.181 (.096) + .181 (.002) .172 (.001) .171 (.000) + _ + .197 (.0~01) + + + -.420 (.024) .873 (.035) .764 (.063) + .157 (.062) aDependent variable is natural log of ratio of actual to expected rate. Only coefficients with p <.10 are shown (p value in parentheses). Positive values indicate higher adverse outcome rates in the first-listed category of hospital. credited than others. Results also indicate that, after standardization, proprietary chains have significantly higher accreditation rates than other investor-owned hospitals. Both models that test the effect of switching status show similar patterns, though differences are not statistically significant. Case Mix Case-mix differences were examined by means of two measures made on Medicare ad missions. One measure, available only for 1981, is the Medicare case-mix index based on ICD9 DRGs which was developed for the Tax Equity and Fiscal Responsibility Act (TEFRA). The other measure is the expected 180-day post- admission mortality rate for the aggregate of Medicare elective surgery cases.5 The former measure is a form of resource use index; the latter is a measure of severity or prognosis. Results are not Filly consistent, though gen- erally both measures show that proprietary hospitals have a simpler case mix than others, cet par.

OCR for page 354
MEDICARE PATIENT OUTCOMES TABLE 8 Statistical Results for Accreditation and Case Mix Percentage Dilierence* Expected 180-Day Full Accredi- Mortality for 1981 Medicare tation Rate Elective Surgery Case-Mix Index Chain versus independent Proprietary versus voluntary Switch from voluntary to proprietary Proprietary chain versus proprietary .087 (.~) + -.114 (.000) independent .170 (.000) Switch from proprietary independent to proprietary chain + -.001 (.078) -.019 (.001) NA + HA *Coefficients are interpreted as the percentage difference between the indi- cated groups. Only coefficients with p <.10 are shown. SUMMARY AND DISCUSSION Findings on Investor-Owned Hospitals The analysis of quality-of-care indicators provides no evidence for concluding that the profit motive, in the aggregate, has compro- mised patient care to the point of causing large and systematic differences in post-operative mortality or readmission. Results indicate that Medicare post-operative mortality rates are of- ten lower in proprietary hospitals, after stan- dar~li~ing for patient age, sex, and co-morbidity; and controlling for interhospital differences in hospital size and other hospital and market characteristics. This pattern does not persist for mortality rates within 180 days of admis- sion; proprietary hospitals have somewhat higher mortality rates although they are not sufficiently higher than the levels in voluntary hospitals to be considered statistically sig,nifi- cant. The findings on readmission rates admit to no pattern whatsoever again providing no support for the view that investor-owned hos- pitals have poorer patient outcomes. In sum, the patient outcome findings do not show any persisting pattern that would support a con- clusion that proprietary hospitals are providing poorer quality of care. 367 Analysis of accreditation status with JCAH shows that proprietaries are consistently less likely to be fully (2-year) accredited. Whether this finding has a bearing on quality of care is, of course, problematic due to the nature ofthe ICAH survey and the fact that the accredita- tion program is voluntary. The data indicate that lower rates of participation in the ICAH program are largely responsible for the lower rates of full accreditation, implying that the statistical findings may have little or no bearing on the quality of the patient care process in investor-owned hospitals. Findings on Chain Sequences for l~vestor-Owned Hospitals There is very inconsistent evidence on the issue of how proprietary chains may be dis- tinctive from independent proprietaries in teens of quality indicators. No significant dif- ferences or pattern of results is seen when mortality rates are compared directly. Switch- ing from independent to chain affiliation is wealdy associated with higher in-hospital mor- tality and lower 180-day mortality. Patterns in readmission differences are not seen. JCAH accreditation rates are also found to be more favorable in proprietary chains than in other

OCR for page 354
368 investor-owned hospitals. Finally, chain affil- iation is associated with more complex case mix. These results are, without question, not a conclusive test on the issue of quality of care. Aside from the usual caveats about correla- tional analysis and a nonrandomized design and imperfect indicators of quality, we are con- cerned about several statistical issues. Deaths (and even readmissions) are relatively rare events for elective surgery. Given small sam- ple sizes per hospital, we observe fairly ex- treme variability in our measures. While the use of weighted least squares and the conti- nuity correction factor will tend to minimize this problem, we still observe quite "noisy" data tending to make it less likely to reject the null hypothesis. That is, differences be- tween groups need to be quite large before it is likely that they are considered statistically significant. While we use a liberal critical con- fidence limit (p<0. 10), it still appears that in- tergroup differences in outcome rates smaller than 10 to 12 percent or so are not detectable. The sample size issue is most germane to consideration of the use of the results stem- ming from the four-way design. This difficulty is unfortunate because this approach essen- tially allows each hospital to be its own control, which is a strong hedge against confounding stemming from omitted determinants of out- comes that are unique to individual hospitals. The number of hospitals that switch status is quite small. Estimates of the ejects of switch- ing from proprietary independent to propne- tary chain status are based on about 1.2 percent of the sample of hospital years (N = 15,422 in total). Estimates of the effects of switching to proprietary status (versus nonprofit) are based on only about 3.7 percent of the sample, where such switches occur. Consequently, the re- sults of the four-way model tests are not likely to be reliable. FOR-PROFIT ENTERPRISE IN HEALTH CARE lbe figure to detect persistent patterns may, in part, derive from these statistical power considerations. We can conclude, however, that no apparent pattern of large ownership differ- ences exist for serious patient outcomes fol- lowing elective surgery. NOTES These states are Anzona, Colorado, Connecticut, Indiana, Kentucky, Maryland, Massachusetts, Min- nesota, Nebraska, New Jersey, New York, Pennsyl- vania (western), Rhode Island, Washington, and Wisconsin. 2We followed Dowling et al. (1976) who reported that separating 1-year from full accreditation status of- fers a more sensitive measure of accreditation status. 3Hospi~1 year means have been weighed by number of admissions in specific surgical categories. 4The ratio of actual to expected outcomes is used to standardize outcomes for interhospital differences in patient age, sex, and co-morbid status. Appendix A describes the standardization approach as well as other aspects of the statistical method. Appendix B contains the full results of the statistical models that are summarized in this section. sThese rates are developed on the basis of patient age, sex, procedure, and existence (or not) of a second diagnosis on admission. REFERENCES Bunker, J., W. Forrest, F. Mosteller, and L. Vandam, eds. (1969) The National Halothane Study, National Institute of General Medical Sciences, Bethesda, Md. Dowling, William et al. (1976) The Evaluation of Blue Cross in Medicaid Prospective Reimbursement Systems in Downstate New York. Final Report, DHEW Contract HEW-OS-74-248. Lust, H. (1980) The relationship between surgical volume and mortality: An exploration of causal factors and alternative models. Medical Care, 18~9), Septem- ber. Scott, W., B. Flood, and W. Ewy (1979) Organiza- tional determinants of services, quality, and cost of care in hospitals, Mudbank Memorial Fund Quarterly, 57~2~.

OCR for page 354
APPENDIX A Statistical Methods CONTROLLING FOR SEVERITY A principal analytic issue in evaluating in- terhospital differences is controlling for prog- nosis or expected outcome on admission. Our approach was to adjust or normalize the mea- sured outcome rate for factors known to affect prognosis. The ratio of the measured outcome to the expected outcome provides a measure of seventy-adjusted outcomes for a given hos- pital year. These "expected" rates are the cell means on each outcome measure (fatality and read- mission rates) for a 20 percent sample of Med- icare admissions patients in all years (1974- 1981) from a 25 percent random sample of short- term U. S. hospitals. The 96 cells were defined by procedure (eight categories), age (three cat- egories 65-74, 75-84, over 84), sex (two cate- gories), and presence of a second diagnosis (two categories). Two sets of norms (means) were used: 1974-1978 and 1979-1981. This was done to acknowledge the shift to ICD9-CM coding on Medicare files which began in 1979. After the patient file for each year was scored with appropriate means, both the actual and "expected" outcome rates were computed for each hospital year observation by summing across patients and dividing by the number of cases. SPECIFYING OWNERS H I ~ MEASURES Three basic forms of ownership influence were examined in the study. The first two specify the ownership influence within the twos way or treatment-control design: Model 1: Prop (= 1 if proprietary, 0 other- wise), and Chain (= 1 if member of chain, 0 otherwise) Model 2: Prop (= 1 if proprietary, 0 other- wise), and Chain (= 1 if member of chain, 0 otherwise), and Prochain (= 1 if proprietary chain, O otherwise) The coefficients on the organization mea- sures in the Model 1 specification tests for differences between chain and nonchain hos- pitals and between proprietary and nonpro- prietary hospitals. In Mode} 2, the coefficient on the Prochain variable tests whether pro- prietary chain hospitals are different from other proprietary hospitals. Both models test for dif- ferences in means between groups of hospitals (e. g., proprietary versus other), controlling for differences in means between groups on other covariates we include. A third specification uses a different test: measuring the difference between groups in their pre/post change. This four-way (pre/post- treatment/control) design reduces the risk that unmeasured baseline differences between the groups of institutions are confounding the re- sults. This design is intrinsically preferable to the treatment control approach used in the first two models, but supers here due to several factors: the small number of hospitals that ac- tually change status (between profit and non- profit and between chain and independent) over the study period; and the inability to gather tine series data on nonprofit chain affiliation. Hence, the hypothesis test Ho: (Qpos~ - Qpre)PnOPR (Qpos' Qpre)NONPROP = 0 is highly leveraged on those few cases where ownership status changed. We do test this ap- proach, aclmowledging the problem, allowing the results to be considered by the reader as part of the overall pattern of findings. The specification is Mode! 3: Prop (see other), and DProchain (see other), and DChain (= 1 if ever chain affili- ated)i Prop (= 1 in years when proprie- tary, 0 otherwise), and DProp (= 1 if ever proprietary), and 369

OCR for page 354
370 Chain (= 1 for years when affiliated with a chain)) In Mocle} 4, the coefficient on Prop tests whether the change to proprietary status in- fluenced outcomes. The coefficient on Pro- chain in Model 3 tests whether a change from an independent proprietary to a chain-affili- ated proprietary influenced the outcomes. COVARL\TES The characteristics of hospitals in various ownership groups of interest are different. Of course, without random assignment this prob- lem of noncomparability is always a problem, precluding simple hypothesis tests of cli$er- ences in means. We use a multivanate (regres- sion) approach to standardize groups for differences in many of these hospital charac- teristics. Table 2 presents descriptive data on the values of covariates across ownership groups. In addition to those measures shown in the table, the regression models include covariate as Whether binding review by a professional standards review organization (PSRO) was conducted in the hospital for the year Whether the hospital was subject to state certificate-of-need (CON) authority for each year Percentage increase in the CPI for the SMSA County population County population per square mile A set of year-specific, dummy (0,1) vari- ables A set of region-specific dummy variables Number of staffed beds following Luft (1980~. This measure captures scale effects on outcomes. ESTIMATION For the patient outcome analyses, a weighted regression (OLS) was used to estimate the basic models. This is done to remove heteroscedas- tic (systematic) variances in residuals across hospital year observations which, if uncor- rected, make all parameter estimates ineffi- cient (though unbiased). These patterns are likely in our model because our hospital year FOR-PROFIT ENTERPRISE IN HEALTH CARE outcome measures are based on samples of patients of widely varying sizes, and while the 20 percent sample Is probably random, the variances of these estimates are likely to be systematically related to the number of cases in each hospital year on which the mean was computed. CONTINUITY CORRECTION FACTOR Another estimation problem is that, for some measures, we can expect to observe a cluster of zeros for outpatient outcome variables. With small samples per hospital year, for example, it is plausible that no deaths or readmissions will be observed even though some cases were admitted. The number of zeros in the data approaches 50 percent for some types of elec- tive surgery. Consequently, we observe a bi- modal distribution on our measures. The approach for dealing with the clusters of cases at zero (and at high extreme values) is the use of a "continuity correction factor." This ap- proach is a standard technique for smoothing the "ends" of a distribution on abmary variable (e.g., values approaching 0 and values ap- proachir~g 17.2 SIGNIFICANCE LEVELS In all statistical analyses we report coeffi- cients that are significant at the p = .10 level or better (two-tailed test). There are No rea- sons for reporting at a level of significance somewhat lower than the p = .05 that is cus- tomary in the literature. First, we are con- cerned that measurement errors in the MEDPAR file may elevate the standard errors in the morlel, causing significant differences to be overlooked if tolerance levels are too stringent. Second, we believe that the policy applications of this work require that we allow less than the usual chance of committing false negative (type II) elTors (ignoring an adverse difference because a critical significance level is too stringent). NOTES ZAHA data for 1981-1982 were used to assign this ownership status for nonprofits indicating membership in a multihospital system.

OCR for page 354
MEDICARE PATIENT OUTCOMES 2 Following techniques suggested by Cox (1970) and Bishop et al. (1975), we can use an adjustment of the form: Ai + 1/6 ~ = Ni + 113 where Ai = the observed number of adverse outcomes for the hospital year; Ni = the number of cases; and = the adjusted adverse outcome rate. The constants (1/6, 1/3), which define the value of this Taylor expan- sion on n are suggested by Mosteller and Tukey (1977). The choice remains somewhat arbitrary, though larger fractions tend to have more dramatic "smoothing" ef- fects. APPENDIX B Mode! Results 371 REFERENCES Bishop Y., S. Fienberg, and P. Holland (1975) Dis- crete Multivariate Analysis. Cambridge, Mass.: MIT Press. Cox, D. R. (1970) Analysis of Binary Data. London: Chapman and Hall. Luff, H. (1980) The relationship between surgical volume and mortality. Medical Care 18, September. Mosteller, F., and J. Tukey (1977) Data Analysis and Regression. Reading, Mass.: Addison Wesley. TABLE B-1 Statistical Results on Post-Operative Mortality Rates: Percentage Difference in Mortality Rate During the Staya Surgical Category Model All Elective Inguinal Surgery Cholecystectomy Hernia Repair Transurethral Resection of Prostate Model 1 Chainb Proprietary Model 2 Chain Proprietary Prochaind Model 3 DChaine Prochain Proprietary Model 4 DProprietaryf Proprietary Chain O. 119 (0.033) 0.120 (0.033) -0.401 (0.000) 0.124 (0.056) - - 0.181 (0.096) 0.129 (0.023) 0.116 (0.031) + + - -0.274 (0.035) - + 0.129 (0.032) 0.121 (0.032) ~ - t a Dependent variable is natural log of ratio of actual to expected mortality rate at discharge. Only coefficients with p <.10 are shown (p value in parentheses). b Chain = 1 if chain affiliated in year. Proprietary = 1 if proprietary in year. dProchain = 1 if proprietary and chain. eDChain = 1 for all years if ever chain affiliated. fDProprietary = 1 for all years if ever proprietary.

OCR for page 354
372 FOR-PROFIT ENTERPRISE IN HEALTH CARE TABLE B-2 Statistical Results on Readmission Rate: Percentage Difference in Readmission Ratea Surgical Category All Elective Surgery Cholecystectomy Model Inguinal Transurethral Resection Hernia Repair of Prostate Mastectomy Model 1 Chamb Proprietary Model 2 Chain Proprietary Prochain~ Model 3 DChaine Prochain Proprietary Model 4 + 0.204 (0.057) - DProprietaryJ Proprietary Chain 0.197 (0.001) + 0.2~ (0.000) 0.268 (0.045) - - 0.420 (0.024) - 1. 167 (0.004) 0.873 (0.035) -1.176 (0.004) 0.764 (0.063) 0.382 (0.000) + aDependent variable is natural log of ratio of actual to expected rate. Only coefficients with p (p value in parentheses). bChain = 1 if chain affiliated in year. Proprietary = 1 if proprietary in year. Prochain = 1 if proprietary and chain. eDChain = 1 for all years if ever chain affiliated. fD proprietary = 1 for all years if ever proprietary. <.10 are shown

OCR for page 354
MEDICARE PATIENT OUTCOMES TABLE B-3 Statistic Results on 180-Day Morbid Rate: Percentage Deference in Mortality Ratea . 373 Surgical Category All Elective Inguinal Transurethral Resection Model Surgery Cholecystectomy Hernia Repair of Prostate ~. Mastectomy Model 1 Chainb - + + - 0.081 (0.002) PropnetaryC + ~0.172 (0.001) 0.171 (0.000) Model 2 Chain - + + -0.089 (0.001) Proprietary + + 0.141 (0.004) + Prochaind - + + 0.157 (0.062) Model 3 DChaine + ~0.116 (0.021) - 0.430 (0.033) Prochain -0.248 (0.099) - - - Proprietary + + 0.116 (0.021) + Model 4 DPropnetaryf + + 0.337 (0.035) + Proprietary + + Chain - + + -0.081 (0.002) a Dependent variable is natural log of ratio of acutal to expected rate. Only coefficients with p <.10 are shown (p value in parentheses). bChain = 1 if chain affiliated in year. Proprietary = 1 if proprietary in year. dProchain = 1 if proprietary and chain. eDChain = 1 for all years if ever chain affiliated. fDProprietary = 1 for all years if ever proprietary.

OCR for page 354
374 FOR-PROFIT ENTERPRISE IN HEALTH CARE TA;BLE B-4 Statistical Results on Case Mix and SCAM Measuresa lCAH Status (1 = 2 Year, O Otherwise Expected 180-Day Mortality Rate for All Elective Surgeries 1981 Medicare Case-Mix Index Model 1 Chain0.087 (0.000) Proprietary-0.114 (0.000) - 0.001 (0.078) - 0.019 (0.001) Model 2 Chain0.066 (0.000) + + Proprietary-0.177 (0.000) - -0.024 (0.001) Prochain0.170 (0.000) - + Model 3 Proprietary DChain DProchain Model 4 D Proprietary Proprietary Chain - O. 182 (0.000) 0.199 (0.000) + -0.085 (0.008) 0.088 (o.ooo) + NA NA _ . . aCoefFicients indicate differences for the indicated group expressed in per- centage terms: numbers in parentheses are p values. bModels of the Joint Commission on Accreditation of Hospitals UCAH) used ordinary least-squares estimation.