Length of hospital stay for cerebrovascular disease in the United States: Professional activity study, 1963–1991

-

-

JOURNAL OF THE

NEUROLOGICAL SCIENCES ELSEVIER

Journal of the Neurological Sciences 127 (1994) 214-220

Length of hospital stay for cerebrovascular disease in the United States: Professional Activity Study, 1963-1991 Douglas J. Lanska

*

Departments of Neurology, Preventive Medicine and Environmental Health, and the Sanders Brown Center on Aging, University of Kentucky Medical Center, 800 Rose Street, MS-129, Lexington, KY40536-0084, USA, Neurology Service, Department of Veterans Affairs Medical Center, Lexington, KZ, USA Received 25 April 1994; revised 1 August 1994; accepted 6 August 1994

Abstract Objective: To assess the temporal and spatial variation in length of hospital stay for cerebrovascular disease in the United States over three decades. Design: Age-, region-, and stroke type-specific length-of-hospital-stay data for nearly 4 million patients admitted with cerebrovascular disease were obtained for the Professional Activity Study of the Commission on Professional and Hospital Activities for the period 1963-1991. Main outcome measure: Weighted averages and standard errors of length of stay were calculated for aggregate diagnosis groups within the category of cerebrovascular disease. Averages were age-adjusted by the direct method. Results: Average length of hospital stay declined from a peak of 18 days in 1967 to 8 days in 1991. The decline accelerated sharply from 1982 to 1986 coincident with implementation of the Medicare prospective payment system. Similar declines were observed within each age group, each cerebrovascular disease diagnosis group, and each census region. There were marked and persistent differences in average length of stay between regions, with longer stays in the Northeast and shorter stays in the West. The large interregional variation was not explained by differences in age or cerebrovascular disease diagnoses. Conclusions: Implementation of the Medicare prospective payment system produced a marked decline in length of hospital stay for cerebrovascular disease, which was superimposed on a preexisting, but slower, decline. Much of the marked persistent interregional variation probably results from persistent widespread variation in patient management. Keywords: Stroke; Cerebrovascular disease; Hospitalization; United States; Diagnosis related groups; Medicare; Prospective

payment

I. Introduction

2. Methods

Although the average length of hospital stay for cerebrovascular disease has declined since the 1960s (Kahn et al. 1990; Pokras 1986; Lagoe 1987), the reasons for this decline are not entirely clear, and considerable practice variation persists (Kahn et al. 1990; Pokras 1986; Lagoe 1987; Commission on Professional and Hospital Activities 1991). This study explores the temporal and spatial variation in length of hospital stsy for cerebrovascular disease in the United States for the period 1963-1991.

Age-, region-, and stroke type-specific length-ofhospital-stay data for patients admitted with cerebrovascular disease were abstracted from more than 100 volumes of national and regional statistics published by the Commission on Professional and Hospital Activities (CPHA) for the period 1963-1991. C P H A was founded in 1955 to oversee the Professional Activity Study (PAS), a hospital discharge abstract system begun three years earlier (Kuehn 1973; Kincaid 1977). C P H A published national data on length of stay irregularly for the period 1963-1967. Since 1969, C P H A has published annual national and regional data. Data tabulated by C P H A are from general, nonfederal, short-term hospitals in the United States, exclusive of hospitals in U.S. territories and Puerto Rico.

* Tel.: (606) 257-5341; Fax: (606)258-1040. Elsevier Science B.V.

SSDI 0022-5 10X(94)00199-5

D.J. Lanska /Journal of the Neurological Sciences 127 (1994) 214-220

CPHA defined short-term hospitals as those with a median stay for all admissions of less than 30 days. Most of the data are from hospitals participating in PAS, but CPHA also incorporated available data from various hospital associations, prepaid medical plans, and individual non-PAS hospitals. CPHA excluded data on length of stay for individual cases for the following reasons: in-hospital death, admission from or transfer to another short-term hospital, discharge against medical advice, and length of stay greater than 99 days. For the present study, average length-of-stay data were abstracted for cerebrovascular disease hospitalizations coded using adaptations of three revisions of the International Classification of Diseases: ICDA (1962), years 1963-1967, rubrics 330-334; H-ICDA (Commission on Professional and Hospital Activities. 1968), years 1969-1978, rubrics 430-438; and ICD-9CM (U.S.Department of Health and Human Services 1989), years 1979-1991, rubrics 430-438. Weighted averages of length of stay for patients admitted for cerebrovascular disease were calculated for the period 1963-1991 for all patients and within age groups, and for the period 1969-1989 by region, using the 4 regions designated by the U.S. Bureau of the Census. For the period 1979-1991, weighted averages of length of stay were calculated by aggregate diagnosis groups within the category of cerebrovascular disease: subarachnoid hemorrhage (ICD-9-CM rubrics 430 with or without coded paralysis, 342 or 344); other intracranial hemorrhage (ICD-9-CM rubrics 431-432 with or without coded paralysis, 342 or 344); occlusion and stenosis of the precerebral arteries (ICD-9-CM rubric 433 with or without coded paralysis, 342 or 344), occlusion and stenosis of the cerebral arteries (ICD-9-CM rubric 434 with or without coded paralysis, 342 or 344), transient cerebral ischemia (ICD-9-CM rubric 435), and other, ill-defined, and late effects of cerebrovascular disease (ICD-9-CM rubrics 436 with or without coded paralysis, and 437-438). Standard errors for the aggregated means were estimated from the total sum of squares and the aggregate sample size, by first calculating the within-groups and between-groups sums of squares from the individual component group variances and sample sizes. Age-adjustment was done by the direct method, using as the standard the age distribution of hospital discharges for cerebrovascular disease from PAS hospitals across the country in 1990.

3. Results

During the period of this study, the PAS collected data on approximately 3.9 million cerebrovascular disease hospitalizations, with more than 150000 cerebrovascular disease hospitalizations per year in the period from 1973 to 1986. Average length of hospital

20

,

,

r

I

'

'

'

I

'

'

215

'

I

'

'

'

I

'

'

'

I

'

'

'

I

'

'

'

18

16

14

12

,-r.

,-I

10

8

6 t~ 4

--CRUDE ..... AGE-ADJUSTED

2

0 1964

1968

1972

1976

1980

1984

1988

992

YEAR

Fig. 1. Average length of hospital stay in days. Data are for patients admitted for cerebrovascular disease to PAS hospitals in the United States from 1963 to 1991.

stay for cerebrovascular disease declined considerably from a peak of 18 days in 1967 to 8 days in 1991 (Fig. 1). The decline accelerated sharply from 1982 to 1986, followed by several years of approximately stationary average lengths of stay. Lengths of stay for all years were greater in patients with multiple coded diagnoses and in operated patients (not shown). Age-adjustment did not appreciably influence the findings. The smooth curves and the small standard errors (< 0.5% of the plotted values for all years) indicate that the temporal changes were not due to chance. From 1969 through 1984, similar declines were observed within each age group (Fig. 2). Over this interval, average lengths of hospital stay were similar among the younger age groups. Average length of hospital stay was approximately 2 days longer in those 65 years and older compared with younger age groups. The decline accelerated for each age group from 1982 to 1986. For this period, the absolute and relative declines were greatest in those aged 65 and over, so that by the late 1980s the length of stay in the oldest age group was shorter than that for the youngest age groups. From 1979 to 1991, similar declines were observed within each cerebrovascular disease diagnosis group, with accelerated declines from 1982 to 1986 (Fig. 3). Throughout this period, the average length of hospital stay by diagnosis group maintained the same rank order, which was (from longest to shortest): subarachnoid hemorrhage; other intracranial hemorrhage; occlusion and stenosis of cerebral arteries; other, ill-defined, and late effects of cerebrovascular disease; oc-

D,J. Lanska /Journal of the Neurological Sciences 127 (1994) 214-220

216

20

I

18 16

I

I

I

[

I

-

24 .....

~ \

.......... .,::: ,,

.'

14

g

~.~.

cn

:=

I

20

":':"?~rk-

....."

10 8

6 <

4

- -

65+

---

50-64

........ .....

M

YEARS

YEARS

35-49

YEARS

0

YEARS

-34

4

I

O T H E R II'¢TRACRACHANIAL H E M O R R H A G E

-

- SII~OS~S A N D O ~ C L U m O N Or CFaEBP~L ~ A ~ I ~ ~ , ILL-DEfINED. AND [ATE EPI~X~/S Or CER~BROVASCU[AR DISFA~E

2

S'n~OSJS AND OCCLUSION OF PHI~C£RI~SRAL ARTERII~

O

0

I

1964

L

1968

I

1972

I

L

1976

1980

I

1984

cJ TRA~SI~n" C]CREBRALISCHEMIA

L

1988

U

1992

L

1978

I

1980

I

1982

I984

I

I

I

1986

1988

1990

1 ~92

YEAR

Fig. 2. Average length of hospital stay in days by age group. Data are for patients admitted for cerebrovascular disease to PAS hospitals in the United States from 1963 to 1991.

clusion and stenosis of precerebral arteries; and transient cerebral ischemia. Age adjustment did not appreciably influence the findings (not shown). From 1969 through 1991, similar declines were observed within each census region (Fig. 4). However, there were marked differences in average length of stay between regions. The North Central and Southern regions closely approximated the average national experience, whereas the Northeastern region persistently had average lengths of stay approximately 3 days longer, and the Western region had average lengths of stay approximately 3 days shorter than the nation as a whole. All of the regions experienced accelerated declines from 1982 to 1986, but the Northeastern region and to a lesser extent the Southern region experienced partial rebound increases in average length of stay from 1986 to 1988. Age adjustment did not appreciably influence the findings (not shown). The regional differences in average length of stay were not attributable to different distributions of types of cerebrovascular disease in the different regions, since the distributions of stroke types were very similar across regions (Fig. 5), and since the regional differences in average length of stay generally held for each type of cerebrovascular disease (Fig. 6). Since the late 1970s and early 1980s there was a considerable decrease in the proportion of cerebrovascular disease hospitalizations with principal diagnoses of ill-defined or late effects of cerebrovascular disease (ICD-9-CM rubrics 436-438); over this same period, there was a compensatory increase in the proportion of more specific cerebrovascular disease diagnoses, and particu-

YEAR

Fig. 3. Average length of hospital stay in days by diagnosis group. Data are for patients admitted for cerebrovascular disease to PAS hospitals in the United States from 1979 to 1991.

larly those coded to ICD-9-CM rubric 434 ("occlusion of cerebral arteries"), which includes cerebral thrombosis and cerebral embolism. Despite the changing profile of coded cerebrovascular disease diagnoses since the late 1970s, the distributions of diagnoses across census regions at any one point in time were similar (see Fig. 5). Also, at any one time, for each type of

22

~

,

,

I

r

,

I

'

r

F

'

I

'

'

/

'

'

I

'

'

I

2O

<

"-__

14

:i~

.

%

N N

', " < ~ : i _ _ 8

.

6 - -

NORTHEASTERN

---

NORTH

REGION

4

2 1968

CENTRAL

- SOUTHERN

---

1971

WESTERN

1974

REGION

REGION

[

REGION

1977

[ 1980

983

1986

1989

YEAR

Fig. 4. Average length of hospital stay in days by U.S. census region. Data are for patients admitted for cerebrovascular disease to PAS hospitals in the United States from 1969 to 1991.

2]7

D.J. Lanska /Journal of the Neurological Sciences 127 (1994) 214-220

BEFORE PROSPECTIVE PAYMENT

TRANSITION

AFTER PROSPECTIVE PAYMENT

1979-1982

1985-1985

1986-1989

50 40 .30 20 10 0 SAH

I

ICH

ORA OCA

TIA OTH

] Northeost

~

SAH

ICH

OPA OCA

North Centrol

~

TIA OTH

West

SAH

~

ICH O P A

South

OCA

TIA OTH

~

All

Fig. 5. Grouped bar charts of percentage of each cerebrovascular disease discharge diagnosis by U.S. census region. Data are for patients admitted for cerebrovascular disease to PAS hospitals in the United States from 1969 to 1991. Abbreviations: SAH, subaraehnoid hemorrhage (ICD-9-CM rubric 430); ICH, intracerebral and other and unspecified intracranial hemorrhage (ICD-9-CM rubrics 431-432); OPA, occlusion of precerebral arteries (ICD-9-CM rubric 433); OCA, occlusion of cerebral arteries (ICD-9-CM rubric 434); TIA, transient ischemic attack (ICD-9-CM rubric 435); OTH, other, ill-defined, and late-effects of cerebrovascular disease (ICD-9-CM rubrics 436-438).

cerebrovascular disease, the Northeast region had longer average lengths of stay than the national average, the West had shorter stays than the national average, and the North Central and Southern regions generally approximated the national average (see Fig. 6). Except for subarachnoid hemorrhage, which represents only about 1% of cerebrovascular disease discharge diagnoses, the Northeast had the longest average length of stay and the West had the shortest average length of stay for each cerebrovascular disease

BEFORE PROSPECTIVE PAYMENT 1979-1982

type for all time periods studied in the interval from 1979 to 1989.

4. Discussion Since hospitals participate voluntarily, PAS does not have a defined population base, nor is it necessarily a representative sample of U.S. hospitalizations. Nevertheless, the PAS system is the largest source of U.S.

1985-1985

AFTER PROSPECTIVE PAYMENT 1986-1989

SAH ICH OPA OCA TIA OTH

SAH ICH OPA OCA TIA OTH

TRANSITION

50

£ o

25

>-

2O

09

15

ZE I,(_3 Z

10 5 0 SAH ICH OPA OCA TIA OTH I Northeost

North Centrol

~

West

South

~

All

Fig. 6. Grouped bar charts of average length of hospital stay in days by U.S. census region, stratified by the type of cerebrovascular disease listed as principal discharge diagnosis. Data are for patients admitted for cerebrovaseular disease to PAS hospitals in the United States from 1969 to 1991. For abbreviations see the legend for Fig. 5.

218

D.J. Lanska /Journal of the Neurological Sciences 127 (1994) 214-220

hospitalization data spanning the past several decades. During the peak participation years from 1972 to 1986, 25-30% of U.S. short-term, nonfederal, general hospitals were included in the tabulations of length of stay, representing collectively 7-14 million discharges per year. While both the proportion of participating U.S. hospitals and the number of discharges have declined since the late 1970s, the tabulated data still represent 16-22% of U.S. short-term, nonfederal, general hospitals and 5-6 million discharges per year. An independent study by the Institute of Medicine of the National Academy of Sciences in 1974 showed that the cerebrovascular disease diagnoses on hospital discharge abstracts compiled by private abstracting services (including the CPHA) were accurately coded: reabstraction and critical review of hospital records by an Institute of Medicine field team altered the principal diagnosis in only 15% of cerebrovascular disease diagnoses (in many cases because of erroneous selection of the principal diagnosis from among all diagnoses, rather than because of mistakes in assigning a code number) and produced no significant change in the average length of stay (Institute of Medicine et al. 1977). PAS data indicate that the average length of hospital stay for cerebrovascular disease increased slightly after enactment of Medicare in 1965, and then declined - by more than 50% - from the late 1960s. The decline has been fairly steady except for a sharp acceleration coincident with the change in Medicare reimbursement from fee-for-service to prospective payment in 1983 (Kahn et al. 1990; Pokras 1986). From 1981-82 to 1985-86, average length of hospital stay for cerebrovascular disease declined from 12.5 days to 9.0 days - a drop of 28%. Independent estimates using Medicare hospital claims data suggest that length of hospital stay for cerebrovascular disease fell by 32% over these intervals as a result of implementation of Medicare prospective payment (Kahn et al. 1990). While the Medicare data are a useful resource for studies of length of hospital stay, they do not provide information on the general population under age 65, they are not available for as long a period, summary results by disease and demographic group are not readily available in hardcopy or machine-readable form, and the public-use computer files are very expensive (particularly for an extended temporal period), difficult to obtain, and expensive and time-consuming to process (11 million discharge records per year of data). As a result, temporal and spatial patterns of length of hospital stay for stroke in the Medicare population remain to be fully elucidated. Annual tabulations of estimated hospital length of stay data by disease category have also been published for the National Hospital Discharge Survey, an ongoing survey of approximately 200000 inpatient records per year obtained from national multi-stage stratified

samples of approximately 500 short-stay, non-federal, general and specialty hospitals in the United States. Although this tabulated data for length of hospital stay following stroke has not been intensively studied, review of this data suggests that similar patterns of decline in length of hospital stay for cerebrovascular disease were observed in the PAS and the National Hospital Discharge Survey (Pokras 1986; Graves 1984, 1985, 1986, 1987, 1988, 1989, 1991). The tabulated data are not strictly comparable to the PAS data, since the published data for PAS were based on hospitalizations where the patients were discharged alive, while the National Hospital Discharge Survey tabulations are based on all discharges. In addition, while the National Hospital Discharge Survey is based on a probability sample for a single year, the sample size is small (particularly when stratified by disease, region, and demographic variables) and the sample frame has changed over the years, making temporal comparisons problematic. The decline in average length of hospital stay for cerebrovascular disease discharges in the PAS affected hospitalizations for all types of cerebrovascular disease, for patients of all ages, and for all regions of the country. While the decline was most severe in the 65-years-and-older age group directly affected by the changes in Medicare reimbursement, the changes in the Medicare system had a significant impact on all age groups, since other payors (e.g., Medicaid and commercial insurance companies) also attempted to reduce hospital stays by applying the Medicare prospective payment system to their enrollees. Under the Medicare prospective-payment system, hospitalizations are coded to one of presently 490 diagnosis-related groups (DRGs), i.e., aggregations of principal hospital diagnoses for which patients demonstrate similar resource consumption and length-of-stay patterns. Cerebrovascular disease hospitalizations are coded to four separate DRGs: DRG 14 ("specific cerebrovascular disorders except TIA": ICD-9-CM rubrics 430-432, 434, 436, and 437.3), DRG 15 ("transient ischemic attack and precerebral occlusions": ICD-9-CM rubrics 433 and 435), and DRGs 16 and 17 ("nonspecific cerebrovascular disorders" with or without complications or comorbid conditions: ICD-9-CM rubrics 348.3, 348.8, 348.9, 349.89, 349.9, 437.0, 437.1, 437.8, and 437.9). For the majority of patients in each DRG, hospitals receive the same reimbursement regardless of the length of the patient's stay. Limited additional reimbursement is provided for day and cost outliers (e.g., beyond 33-36 days), but the additional reimbursement is well below the actual daily costs incurred in a typical acute care hospital setting (Lagoe et al. 1985). This has provided a strong impetus for hospitals (and physicians) to limit the length of hospital stay whenever possible.

D.J. Lanska / Journal of the Neurological Sciences 127 (1994) 214-220

Several factors probably contributed to the decline in length of stay for cerebrovascular disease in the United States (Commission on Professional and Hospital Activities 1991), in some cases even before institution of the Medicare DRG-based prospective payment system: close scrutiny by hospital utilization review programs of expensive or prolonged hospital stays; discharge of patients sooner to less acute care settings (Kosecoff et al. 1990); provision of more home-based health services as an alternative to continued inpatient care; development of utilization review mechanisms by other payors (Davis et al. 1985); growth of health maintenance organizations and prepaid provider organizations, which closely monitor all medical care to control costs and utilization; and development of new medical technology, which served to identify milder cases which could be evaluated and treated more quickly. In general, outcomes of care have apparently not worsened, and indeed some may have improved, in conjunction with shortening the average length of hospital stay for cerebrovascular disease (Kahn et al. 1990). For example, with the marked decrease in length of hospital stay for stroke as a result of implementation of Medicare prospective payment, 30- and 180-day postadmission mortality declined, and there was no change in the percentage of patients readmitted within one year or in the mean number of in-hospital days within one year, even when these indices were adjusted for changes in sickness at hospital admission (Kahn et al. 1990). Unfortunately, the shorter lengths of hospital stay did not produce a commensurate decrease in hospital bed days for cerebrovascular disease, since the number and rate of cerebrovascular disease hospitalizations increased at least through 1985 (Gillum 1986; Howard et al. 1991; Modan et al. 1992; Office of Surveillance and Analysis 1989; Graves 1984, 1985, 1986, 1987, 1988, 1989, 1991). Several factors probably contributed to the increased rate of cerebrovascular disease hospitalizations (Modan et al. 1992), including a need for hospitals to increase admissions in order to maintain the same income, availability of more sensitive diagnostic tests (particularly computed tomography), and a marked increase in the number of neurologists per capita (Roback et al. 1992; National Center for Health Statistics 1992). Although trends in length of hospital stay were similar in each of the four census regions, marked regional disparities have persisted (Pokras 1986). The Northeastern region has consistently had longer lengths of stay, while the Western region has had shorter lengths of stay in both the PAS and the National Hospital Discharge Survey (Pokras 1986). These regional variations in length of hospital stay for cerebrovascular disease parallel the general pattern of regional variation in length of stay for all hospitalizations

219

(Chassin 1983; Pokras 1986; Pokras et al. 1989), which is not explained by differences in age, gender, race, or severity of illness among regions (Gornick 1975, 1982a,b; Blumberg 1982; Chassin 1983). The current results support this general pattern and indicate that the variation in length of hospital stay for stroke is not due to differences in age or distribution of cerebrovascular disease types across census regions. The longer lengths of stay in the Northeast may at least partly be explained by increased hospital bed availability (Ginsberg et al. 1983) and fewer available long-term care beds (Lagoe et al. 1985; Jencks et al. 1989); indeed, the rebound increase in length of stay for cerebrovascular disease in the Northeast from 1986 to 1988 may be a result of saturation of long-term care beds in the area, forcing hospitals in the Northeast to serve as skilled nursing facilities for severely disabled stroke patients (Jencks et al. 1989; Knickman et al. 1984). Reasons for the shorter lengths of stay in the West are not apparent, but greater activity of health maintenance organizations in this region may be one important factor (Luft 1978). Much of the interregional variation, though, likely results from persistent widespread variation in patient management (Wennberg et al. 1984; Lagoe 1987).

Acknowledgements The author gratefully acknowledges the assistance of Raymond Hoffmann, Ph.D., Richard Kryscio, Ph.D., Mary Jo Lanska, M.D., M.S., and Xiafang Mi, M.B., M.P.H. This work was supported in part by a grant from the National Institutes of Health (CIDA K08-NS01549), and by the Office of Research and Development, Department of Veterans Affairs (Research Advisory Group funding).

References International Classification of Diseases (1962) Adapted for indexing hospital records by diseases and operations. U.S. Public Health Service Pub. No. 719 (revised), U.S. Government Printing Office, Washington, DC. Blumberg, M.S. (1982) Regional differences in hospital use standardized by reported morbidity. Med. Care, 20: 931-944. Chassin, M.R. (1983) Variations in hospital length of stay: their relationships to health outcomes. Health Technology Case Study 24, Congress of the United States, Washington, DC. Commission on Professional and Hospital Activities (1968) Hospital adaptation of ICDA, H-ICDA, Commission on Professional and Hospital Activities, Ann Arbor, MI. Commission on Professional and Hospital Activities (1991) Length of stay by diagnosis and operation, United States, 1990, Healthcare Knowledge Resources, Ann Arbor, MI. Davis, K., Anderson, G., Renn, S.C., Rowland, D., Schramm, C.J. and Steinberg, E. (1985) Is cost containment working? Health Aft., 4(3): 81-94.

220

D.J. Lanska /Journal of the Neurological Sciences 127 (1994) 214-220

Gillum, R.F. (1986) Cerebrovascular disease morbidity in the United States, 1970-1983: age, sex, region, and vascular surgery. Stroke, 17: 656-661. Ginsberg, P. and Koretz, D. (1983) Bed availability and hospital utilization: estimates of the "Roemer Effect.". Health Care Fin. Rev., 5: 87. Gornick, M. (1975) Medicare patients: regional differences in length of hospital stays, 1969-71. Soc. Sec. Bull., July. Gornick, M. (1982a) Trends and regional variations in hospital use under Medicare. Health Care Fin. Rev., 3: 41-73. Gornick, M. (1982b) Trends and regional variations in hospital use under Medicare. In: Rothberg, D.L. (Ed.), Regional Variations in Hospital Use: Geographic and temporal patterns of care in the United States, Lexington Books, Lexington, MA, pp. 131-184. Graves, E.J. (1984) Utilization of short-stay hospitals: United States, 1982: annual summary. Vital and Health Statistics, Series 13, No. 78, DHHS Pub. No. (PHS)84-1739, U.S. Department of Health and Human Services, Hyattsville, MD. Graves, E.J. (1985) Utilization of short-stay hospitals: United States, 1983: annual summary. Vital and Health Statistics, Series 13, No. 83, DHHS Pub. No. (PHS)85-1744, U.S. Department of Health and Human Services, Hyattsville, MD. Graves, E.J. (1986) Utilization of short-stay hospitals: United States, 1984: annual summary. Vital and Health Statistics, Series 13, No. 84, DHHS Pub. No. (PHS)86-1745, U.S. Department of Health and Human Services, Hyattsville, MD. Graves, E.J. (1987) Utilization of short-stay hospitals: United States, 1985: annual summary. Vital and Health Statistics, Series 13, No. 91, DHHS Pub. No. (PHS)87-1752, U.S. Department of Health and Human Services, Hyattsville, MD. Graves, E.J. (1988) Utilization of short-stay hospitals: United States, 1986: annual summary. Vital and Health Statistics, Series 13, No. 96, DHHS Pub. No. (PHS)88-1757, U.S. Department of Health and Human Services, Hyattsville, MD. Graves, E.J. (1989) National Hospital Discharge Survey: annual summary, 1987. Vital and Health Statistics, Series 13, No. 99, DHHS Pub. No. (PHS)89-1760, U.S. Department of Health and Human Services, Hyattsville, MD. Graves, E.J. (1991) National Hospital Discharge Survey: annual summary, 1988. Vital and Health Statistics, Series 13, No. 106, DHHS Pub. No. (PHS)91-1767, U.S. Department of Health and Human Services, Hyattsville, MD. Howard, G., Craven, T.E., Sanders, L. and Evans, G.W. (1991) Relationship of hospitalized stroke rate and in-hospital mortality to the decline in US stroke mortality. Neuroepidemiology, 10: 251-259. Institute of Medicine, National Academy of Sciences (1977) Reliability of hospital discharge abstracts, National Academy of Sciences, Washington, DC. Jencks, S.F., Williams, D.K. and Kay, T.L. (1989) Assessing hospitalassociated deaths: in reply. JAMA, 261: 2331. Kahn, K.L., Keeler, E.B., Sherwood, M.J. et al. (1990) Comparing

outcomes of care before and after implementation of the DRGbased prospective payment system. JAMA, 264: 1984-1988. Kincaid, W.H. (1977) Hospital discharge abstract systems: problems facing the clinically-rich (PAS) system. Med. Rec. News, 47(2): 6-24. Knickman, J.R. and Foltz, A.-M. (1984) Regional differences in hospital utilization: How much can be traced to population differences? Med. Care, 22: 971-984. Kosecoff, J., Kahn, K.L., Rogers, W.H. et al. (1990) Prospective payment system and impairment at discharge: the "quicker-andsicker" story revisited. JAMA, 264: 1980-1983. Kuehn, H.R. (1973) The Commission on Professional and Hospital Activities. Bull. Am. Coll. St~rg., 58(7): 7-13. Lagoe, R.J. (1987) Hospital stays by diagnosis related group for neurologic patients treated medically. Neurology, 37: 139-145. Lagoe, R.J. and Lauko, S.J. (1985) Stroke hospitalization under prospective payments: analysis of diagnosis related group 14. Arch. Phys. Med. Rehabil., 66: 773-776. Luft, H.S. (1978) How do health-maintenance organizations achieve their "'savings"? Rhetoric and evidence. New Engl. J. Med., 298: 1336-1343. Modan, B. and Wagener. D.K. (1992) Some epidemiological aspects of stroke: mortality/morbidity trends, age, sex, race, socioeconomic status. Stroke, 23: 1230-1236. National Center for Health Statistics (1992) Health United States 1991 and prevention profile, DHHS Pub. No. (PHS)92-1232, U.S. Department of Health and Human Services, Hyattsville, MD. Office of Surveillance and Analysis, Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control (1989) Hospital discharge rates for cerebrovascular disease United States, 1970-1986. JAMA, 261: 2486. Pokras, R. (1986) Utilization of short-stay hospitals by diagnosis-related groups: United States, 1980-84. Vital and Health Statistics, Series 13, No. 87, DHHS Pub. No. (PHS) 86-1748, U.S. Department of Health and Human Services, Hyattsville, MD. Pokras, R., Kozak, L.J., McCarthy, E. and Graves, E.J. (1989) Trends in hospital utilization: United States, 1965-86. Vital and Health Statistics, Series 13, No. 101, DHHS Pub. No. (PHS)89-1762, U.S. Department of Health and Human Services, Hyattsville, MD. Roback, G., Randolph, L. and Seidman, B. (1992) Physician characteristics and distribution in the U.S. American Medical Association, Chicago, IL. U.S. Department of Health and Human Services (1989) The International Classification of Diseases, 9th revision, Clinical modification: 1CD-9-CM, Vol. 1, DHHS Pub. No. (PHS)89-1260, 3rd edn., U.S. Government Printing Office, Washington, DC. Wennberg, J.E., McPherson, K. and Caper, P. (1984) Will payment based on diagnosis-related groups control hospital costs? New Engl. J. Med., 311: 295-300.