Influence of patient preferences and local health system characteristics on the place of death. SUPPORT Investigators. Study to Understand Prognoses and Preferences for Risks and Outcomes of Treatment

OBJECTIVE

To examine the degree to which variation in place of death is explained by differences in the characteristics of patients, including preferences for dying at home, and by differences in the characteristics of local health systems.

DESIGN

We drew on a clinically rich database to carry out a prospective study using data from the observational phase of the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT component). We used administrative databases for the Medicare program to carry out a national cross-sectional analysis of Medicare enrollees place of death (Medicare component).

SETTING

Five teaching hospitals (SUPPORT); All U.S. Hospital Referral Regions (Medicare).

STUDY POPULATIONS

Patients dying after the enrollment hospitalization in the observational phase of SUPPORT for whom place of death and preferences were known. Medicare beneficiaries who died in 1992 or 1993.

MAIN OUTCOME MEASURES

Place of death (hospital vs non-hospital).

RESULTS

In SUPPORT, most patients expressed a preference for dying at home, yet most died in the hospital. The percent of SUPPORT patients dying in-hospital varied by greater than 2-fold across the five SUPPORT sites (29 to 66%). For Medicare beneficiaries, the percent dying in-hospital varied from 23 to 54% across U.S. Hospital Referral Regions (HRRs). In SUPPORT, variations in place of death across site were not explained by sociodemographic or clinical characteristics or patient preferences. Patient level (SUPPORT) and national cross-sectional (Medicare) multivariate models gave consistent results. The risk of in-hospital death was increased for residents of regions with greater hospital bed availability and use; the risk of in-hospital death was decreased in regions with greater nursing home and hospice availability and use. Measures of hospital bed availability and use were the most powerful predictors of place of death across HRRs.

CONCLUSIONS

Whether people die in the hospital or not is powerfully influenced by characteristics of the local health system but not by patient preferences or other patient characteristics. These findings may explain the failure of the SUPPORT intervention to alter care patterns for seriously ill and dying patients. Reforming the care of dying patients may require modification of local resource availability and provider routines.

Diagnostic testing following screening mammography in the elderly

BACKGROUND

To provide some sense of the general frequency and timing of diagnostic testing following screening mammography in the United States, we investigated the experience of women screened in the Medicare population.

METHODS

By use of Medicare's National Claims History System, we identified a cohort (n=23172) of women 65 years old or older screened during the period from January 1, 1995, through April 30, 1995, and tracked each woman over the subsequent 8 months for the performance of additional breast imaging and biopsy procedures. Using two claims-based definitions for newly detected breast cancer, we also estimated the positive predictive value of screening mammography.

RESULTS

For every 1000 women aged 65-69 years who underwent screening, 85 (95% confidence interval [CI]=79-91) had follow-up testing in the subsequent 8 months; 76 (95% CI=71-82) had additional breast imaging, and 23 (95% CI=20-26) had biopsy procedures. Corresponding numbers for women aged 70 years or more were similar. Some women underwent repeated examinations; 13% of those receiving diagnostic mammograms had more than one; 11% of those undergoing biopsy procedures had more than one. About half of the women who underwent a biopsy had the procedure more than 3 weeks after the imaging test upon which the decision to perform a biopsy was presumably made. The estimated positive predictive value of an abnormal screening mammogram (defined as a mammogram that engendered additional testing) was 0.08 (95% CI=0.06-0.10) for women aged 65-69 years and 0.14 (95% CI=0.12-0.16) for women aged 70 years or more.

CONCLUSION

Additional testing is a frequent consequence of screening mammography and may require a considerable period of time to come to closure. The need for additional testing, however, is weakly predictive of cancer.

Variation in carotid endarterectomy mortality in the Medicare population: trial hospitals, volume, and patient characteristics

CONTEXT

The North American Symptomatic Carotid Endarterectomy Trial (NASCET) and the Asymptomatic Carotid Atherosclerosis Study (ACAS) demonstrated the efficacy of carotid endarterectomy (CEA) in reducing the risk of stroke and death in selected patients when surgery was performed in institutions whose participation depended on demonstrated excellence. Thirty-day mortality rates in the trials were very low: 0.6% in NASCET and 0.1% in ACAS.

OBJECTIVE

To assess perioperative mortality among Medicare patients undergoing CEA in all nonfederal institutional settings.

DESIGN

Retrospective national cohort study.

SETTING AND PATIENTS

All 113300 Medicare patients undergoing CEA during 1992 and 1993 in "trial hospitals" (those participating in NASCET and ACAS, n=86) and "nontrial hospitals" (all other nonfederal institutions performing CEAs, n=2613). Nontrial hospitals were stratified into terciles based on volume of CEAs performed.

MAIN OUTCOME MEASURES

Crude and adjusted perioperative (30 day) mortality rates.

RESULTS

The perioperative mortality rate was 1.4% (95% confidence interval [CI], 1.2%-1.7%) at trial hospitals; mortality in nontrial hospitals was higher: 1.7% (95% CI, 1.6%-1.8%) (high volume); 1.9% (95% CI, 1.7%-2.1 %) (average volume); 2.5% (95% CI, 2.0%-2.9%) (low volume); (P for trend, <.001). In multivariate modeling, patients undergoing their procedures at trial hospitals had a mortality risk reduction of 15% (95% CI, 0%-31%) compared with high-volume nontrial hospitals, 25% (95% CI, 7%-40%) compared with average-volume hospitals, and 43% (95% CI, 25%-56%) compared with low-volume hospitals (P for trend, <.001).

CONCLUSION

Medicare patients' perioperative mortality following CEA is substantially higher than that reported in the trials, even in those institutions that participated in the randomized studies. Caution is advised in translating the efficacy of carefully controlled studies of CEA to effectiveness in everyday practice.

Managing conflicts over limited resources

Many conflicts in health care, such as disputes over physician compensation or plan coverage decisions, recur frequently and can be viewed as processes of production. This article proposes a model of the negotiation process and suggests an approach to the management and improvement of common conflicts over limited resources.

Perceived adequacy of tangible social support and health outcomes in patients with coronary artery disease

OBJECTIVE

Health outcomes of patients with chronic disease might be influenced by assistance from others in performing daily activities. We examined whether perceived adequacy of such tangible support was associated with prognosis in a cohort of patients with coronary artery disease.

DESIGN

Longitudinal cohort study.

SETTING/PARTICIPANTS

In spring 1993, a cohort of 1,468 patients with chronic artery disease was identified using claims data. The cohort consisted of all surviving residents of Manitoba, Canada, who had been hospitalized for acute myocardial infarction from 1991 to 1992: 820 patients completed the initial survey, and 734 completed a follow-up survey approximately 1 year later.

MEASUREMENTS AND MAIN RESULTS

Adequacy of tangible support was assessed by asking if respondents needed help at home because of health problems, and whether these needs were met. We examined the association between perceived adequacy of tangible support and health outcomes at 1 year (mortality, physical function). Of 820 participants, 74% perceived no need for help, 13% had sufficient help, 9% needed more help, and 5% needed much more help; 31 patients died during follow-up. After adjustment for age and initial health status, odds ratios (95% confidence interval) for death were: sufficient help 1.8 (0.61, 5.8); need more help 3.2 (1.1, 9.4); and need much more help 6.5 (2.0, 21.6) compared with respondents with no perceived need. Decline in physical function was also linearly related to perceiving less-adequate tangible support. Sensitivity analyses indicated it is highly improbable that results were due to selection bias.

CONCLUSIONS

Perceived lack of needed assistance was related to mortality and to decline in physical functioning. Adequacy of tangible support was an important prognostic factor for these patients with coronary artery disease and may be a determinant of health outcomes.

Treatment and health outcomes of women and men in a cohort with coronary artery disease

BACKGROUND

Women with coronary artery disease are treated differently than men. Although mortality has been studied, functional outcomes for women and men have not been prospectively compared.

METHODS

The Manitoba Health Reform Impact Study used hospital databases to identify all residents aged 45 years and older in Manitoba who were hospitalized for a myocardial infarction between October 1, 1991, and September 30, 1992. Cohort members were interviewed twice, an average of 16 and 25 months after hospitalization. Baseline and follow-up measures included treatments (eg, physician visits, diagnostic testing, revascularization, and cardiac medications), physical health status (physical component summary [PCS] score derived from the Medical Outcomes Study Short Form 36), reinfarction, and mortality.

RESULTS

Of the 820 patients who completed the initial survey, 31 died during the follow-up period, and 734 completed the follow-up survey. Data were complete for the primary outcome (PCS score) and all relevant covariates for the 677 patients who were included in this study Women constituted 34% of this cohort. Although women had more physician visits during follow-up, they were less likely to have undergone treadmill testing or angiography (odds ratio, 0.68; 95% confidence interval, 0.46-0.99). Women were equally likely to report taking beta-adrenergic blocking agents, but were less likely than men to report the use of aspirin (odds ratio, 0.69; 95% confidence interval, 0.48-0.98). After adjusting for baseline differences in PCS scores, age, income, social supports, and the levels of angina and dyspnea, the PCS score for women declined by 1.4 points, while the score for men improved by 0.2 points (P = .03). During the follow-up period, reinfarction and mortality rates were low overall, but were not different in men and women.

CONCLUSIONS

In this cohort of patients with known coronary artery disease, we found less aggressive treatment of coronary artery disease and less use of aspirin among women than among men during 1 year of observation. After controlling for baseline differences, women with coronary artery disease experienced a more rapid decline in physical health status than did men during 1 year of follow-up.

The 3H-leucine tracer: its use in kinetic studies of plasma lipoproteins

3H-leucine administered as a bolus has been widely used as a tracer in kinetic investigations of protein synthesis and secretion. After intravenous injection, plasma specific radioactivity decays over several orders of magnitude during the first half-day, followed by a slow decay lasting a number of weeks that results from recycling of the leucine tracer as proteins are degraded and 3H-leucine reenters the plasma pool. In studies in which kinetic data are analyzed by mathematical compartmental modeling, plasma leucine activity is generally used as a forcing function to drive the input of 3H-leucine into the protein synthesis pathway. 3H-leucine is an excellent tracer during the initial hours of rapidly decreasing plasma activity; thereafter, reincorporation of recycled tracer into new protein synthesis obscures the tracer data from proteins with slower turnover rates. Thus, for proteins such as plasma albumin and apolipoprotein (apo) A-I, this tracer is unsatisfactory for measuring fractional catabolic (FCR) and turnover rates. By contrast, the kinetics of plasma very-low-density lipoprotein (VLDL)-apoB, a protein with a residence time of approximately 5 hours, are readily measured, since kinetic parameters of this protein can be determined by the time plasma leucine recycling becomes established. However, measurement of VLDL-apoB specific radioactivity extending up to 2 weeks provides further data on the kinetic tail of VLDL-apoB. Were plasma leucine a direct precursor for the leucine in VLDL-apoB, the kinetics of the plasma tracer should determine the kinetics of the protein. However, this is not the case, and the deviations from linearity are interpreted in terms of (1) the dilution of plasma leucine in the liver by unlabeled dietary leucine; (2) the recycling of hepatocellular leucine from proteins within the liver, where recycled cellular leucine does not equilibrate with plasma leucine; and (3) a "hump" in the kinetic data of VLDL-apoB, which we interpret to reflect recycling or retention of a portion of the apoB protein within the hepatocyte, with its subsequent secretion. Because hepatocellular tRNA is the immediate precursor for synthesis of these secretory proteins, its kinetics should be used as the forcing function to drive the modeling of this system. The VLDL-apoB tail contains the information needed to modify the plasma leucine data, to provide an appropriate forcing function when using 3H-leucine as a tracer of apolipoprotein metabolism. This correction is essential when using 3H-leucine as a tracer for measuring low-density lipoprotein (LDL)-apoB kinetics. The 3H-leucine tracer also highlights the importance of recognizing the difference between plasma and system residence times, the latter including the time the tracer resides within exchanging extravascular pools. The inability to determine these fractional exchange coefficients for apoA-I and albumin explains the failure of this tracer in kinetic studies of these proteins. For apoB-containing lipoproteins, plasma residence times are generally determined, and these measurements can be made satisfactorily with 3H-leucine.

Where do elderly veterans obtain care for acute myocardial infarction: Department of Veterans Affairs or Medicare?

OBJECTIVE

To examine Department of Veterans Affairs (VA) and Medicare hospitalizations for elderly veterans with acute myocardial infarction (AMI), their use of cardiac procedures in both systems, and patient mortality.

DATA SOURCES

Merging of inpatient discharge abstracts obtained from VA Patient Treatment Files (PTF) and Medicare MedPAR Part A files.

STUDY DESIGN

A retrospective cohort study of male veterans 65 years or older who were prior users of the VA medical system (veteran-users) and who were initially admitted to a VA or Medicare hospital with a primary diagnosis of AMI at some time from January 1, 1988 through December 31, 1990 (N = 25,312). We examined the use of cardiac catheterization, coronary bypass surgery, and percutaneous transluminal coronary angioplasty in the 90 days after initial admission for AMI in both VA and Medicare systems, and survival at 30 days, 90 days, and one year. Other key measures included patient age, race, marital status, comorbidities, cardiac complications, prior utilization, and the availability of cardiac technology at the admitting hospital.

PRINCIPAL FINDINGS

More than half of veteran-users (54 percent) were initially hospitalized in a Medicare hospital when they suffered an AMI. These Medicare index patients were more likely to receive cardiac catheterization (OR 1.24, 95% C.I. 1.17-1.32), coronary bypass surgery (OR 2.01, 95% C.I. 1.83-2.20), and percutaneous transluminal coronary angioplasty (OR 2.56, 95% C.I. 2.30-2.85) than VA index patients. Small proportions of patients crossed over between systems of care for catheterization procedures (VA to Medicare = 3.3%, and Medicare to VA = 5.1%). Many VA index patients crossed over to Medicare hospitals to obtain bypass surgery (27.6 percent) or coronary angioplasty (12.1 percent). Mortality was not significantly different between veteran-users who were initially admitted to VA versus Medicare hospitals.

CONCLUSIONS

Dual-system utilization highlights the need to look at both systems of care when evaluating access, costs, and quality either in VA or in Medicare systems. Policy changes that affect access to and utilization of one system may lead to unpredictable results in the other.

Benchmarking the US physician workforce. An alternative to needs-based or demand-based planning

OBJECTIVE

To propose population-based benchmarking as an alternative to needs- or demand-based planning for estimating a reasonably sized, clinically active physician workforce for the United States and its regional health care markets.

DESIGN

Cross-sectional analysis of 1993 American Medical Association and American Osteopathic Association physician masterfiles.

POPULATION

The resident population of the 306 hospital referral regions in the United States.

MAIN OUTCOME MEASURES

Per capita number of clinically active physicians by specialty adjusted for age and sex population differences and out-of-region health care utilization. The measured physician workforce was compared with 4 benchmarks: the staffing within a large (2.4 million members) health maintenance organization (HMO), a hospital referral region dominated by managed care (Minneapolis, Minn), a hospital referral region dominated by fee-for-service (Wichita, Kan), and the proposed "balanced" physician supply (50% generalists).

RESULTS

The proportion of the US population residing in hospital referral regions with a higher per capita generalist workforce than the benchmark was 96% for the HMO benchmark, 60% for Wichita, and 27% for Minneapolis. The specialist workforce exceeded all 3 benchmarks for 74% of the population. The per capita workforce of generalists was not related to the proportion of generalists among regions (Pearson correlation coefficient=0.06; P=.26).

CONCLUSIONS

Population-based benchmarking offers practical advantages to needs- or demand-based planning for estimating a reasonably sized per capita workforce of clinically active physicians. The physician workforce within the benchmarks of an HMO and health care markets indicates the varying opportunities for regional physician employment and services. The ratio of generalists to specialists does not measure the adequacy of the supply of the generalist workforce either nationally or for specific regions. Research measuring the relationship between physician workforces of different sizes and population outcomes will guide the selection of future regional benchmarks.

Hospital readmission rates for cohorts of Medicare beneficiaries in Boston and New Haven

BACKGROUND

Geographic variations in the use of hospital services are associated with differences in the availability of hospital beds. There continues to be uncertainty about the extent to which unmeasured case-mix differences explain these findings. Previous research showed that the number of occupied beds per capita in Boston was substantially higher than the number of occupied beds per capita in New Haven, Connecticut, and that overall rates of hospital utilization were higher for Boston residents than for New Haven residents.

METHODS

We used Medicare claims data to study cohorts of Medicare beneficiaries 65 years of age or older and residing in Boston or New Haven who were initially hospitalized for one of five indications (acute myocardial infarction, stroke, gastrointestinal bleeding, hip fracture, or potentially curative surgery for breast, colon, or lung cancer). Residents of Boston or New Haven who were discharged between October 1, 1987, and September 30, 1989, were enrolled in the cohort corresponding to the earliest such admission and followed for up to 35 months.

RESULTS

The relative rate of readmission in Boston as compared with New Haven was 1.64 (95 percent confidence interval, 1.53 to 1.76) for all cohorts combined, with a similarly elevated rate for each of the five clinical cohorts and each age, sex, and race subgroup examined. Hospital-specific readmission rates varied substantially among the hospitals in Boston and were higher than those in New Haven. No relation was found between mortality (during the first 30 days after discharge or over the entire study period) and either community or hospital-specific readmission rates.

CONCLUSIONS

Regardless of the initial cause of the admission, Medicare beneficiaries who were initially hospitalized in Boston had consistently higher rates of readmission than did Medicare beneficiaries hospitalized in New Haven. Differences in the severity of illness are unlikely to explain these findings. One possible explanation is a threshold effect of hospital-bed availability on decisions to admit patients.

Standardized rates of recurrent outcomes

Longitudinal studies are often concerned with estimating the rate of an event that may recur. Examples are nonmelanoma skin cancer rates, screening rates for breast cancer using mammography and hospital admission rates. We propose simple estimators for directly and indirectly standardized summary rates and relative rates of recurrent events and their variances. We also develop an estimator of the excess rate in an area if the rate in another area applied. For non-recurrent events, the estimators are identical to the usual standardized summary rates. The estimators are independent of the underlying distribution of the event of interest and allow for unequal follow-up times and event rate heterogeneity among individuals. The method is not computationally intensive and does not require specialized software. We illustrate the application of the method in a retrospective cohort study of hospital utilization patterns of Medicare enrollees in Boston and New Haven over a three and a half year period.

Using administrative data to describe casemix: a comparison with the medical record

We compared the coding of comorbid conditions in an administrative database to that found in medical records for 485 men who had undergone a prostatectomy. Only a few specific conditions showed good agreement between charts and claims. Most showed poor agreement and appeared more frequently in the chart. A comorbidity index calculated from each of these sources was used to explore the differences in mortality for patients who had undergone transurethral vs open prostatectomy. The claims-based comorbidity index most often underestimated the index from the chart. Proportional hazards analysis showed that models including either comorbidity index were better than those without an index and models with information from both indices were best. No analysis eliminated the effect of type of prostatectomy on long-term mortality. Claims-based measures of comorbidity tend to underrepresent some conditions but may be an acceptable first step in controlling for differences across patient populations.

Internal validation of Medicare claims data

The Medicare database is commonly used for epidemiology and health services research, but validation of its data by chart review or questionnaire may be difficult and expensive. Since hospital and physician bills are independent in Medicare, however, these two data sources can be used to supplement and corroborate each other. This "internal validation" is illustrated here for hip fracture and prostatectomy. Agreement of the hospital and physician data streams regarding site of hip fracture (neck vs other), treatment of hip fracture (internal fixation vs arthroplasty), and type of prostatectomy (transurethral resection of prostate, open, or radical) was excellent, with percentage of agreement generally between 89% and 99%, and kappa statistics typically between 0.74 and 0.95. When validation with outside data sources is not readily available, such internal validation of Medicare data may be valuable.

Treatment and survival among elderly Americans with hip fractures: a population-based study

OBJECTIVES

This study was undertaken to examine the patterns of treatment and survival among elderly Americans with hip fracture.

METHODS

A 5% national sample of Medicare claims was used to identify patients who sustained hip fractures between 1986 and 1989. In comparing treatment patterns across regions, direct standardization was used to derive age- and race-adjusted percentages. Logistic regression and Cox regression were used to examine short- and long-term survival.

RESULTS

In the United States, 64% of femoral neck fractures were treated with arthroplasty; 90% of pertrochanteric fractures were treated with internal fixation. Higher short- and long-term mortality was associated with being male, being older, residing in a nursing home prior to fracture, having a higher comorbidity score, and having a pertrochanteric fracture. Blacks and Whites had similar 90-day postfracture mortality, but Blacks had a higher mortality later on. For femoral neck fracture, internal fixation has a modestly lower short-term mortality associated with it than arthroplasty has.

CONCLUSION

Variation in the treatment of hip fracture was modest, The increased delayed mortality after hip fracture among Blacks requires further study.

Identifying complications of care using administrative data

The Complications Screening Program (CSP) is a method using standard hospital discharge abstract data to identify 27 potentially preventable in-hospital complications, such as post-operative pneumonia, hemorrhage, medication incidents, and wound infection. The CSP was applied to over 1.9 million adult medical/surgical cases using 1988 California discharge abstract data. Cases with complications were significantly older and more likely to die, and they had much higher average total charges and lengths of stay than other cases (P < 0.0001). For most case types, 13 chronic conditions, defined using diagnosis codes, increased the relative risks of having a complication after adjusting for patient age. Cases at larger hospitals and teaching facilities generally had higher complication rates. Logistic regression models to predict complications using demographic, administrative, clinical, and hospital characteristics variables, had modest power (C statistics = 0.64 to 0.70). The CSP requires further evaluation before using it for purposes other than research.

Why are children hospitalized? The role of non-clinical factors in pediatric hospitalizations

OBJECTIVE

Pediatric medical discharge rates vary widely across hospital service areas, beyond differences explained by chance or disease incidence alone. This study examines the relationship between the characteristics of local medical services and the likelihood of hospitalization.

DESIGN

Small area and population-based regression analysis.

SETTING

The 72 hospital service areas of Maine, New Hampshire, and Vermont.

STUDY POPULATION

The 589,290 (1989) children of Maine, New Hampshire, and Vermont < 15 years of age with 120,806 discharges during 1985 through 1989.

MEASUREMENT AND MAIN RESULTS

Using logistic regression and controlling for community income, we found that children residing in zip codes with high per capita bed supply (4.0/1000) had 9% more discharges (odds ratio: 1.09; 99% confidence interval: 1.07, 1.11) compared with children in areas with low per capita bed supply (1.9/1000). Children living 30 minutes from the nearest hospital had 15% fewer medical discharges (odds ratio: 0.849; confidence interval: 0.830, 0.867) than those living in a zip code with a hospital. Residence in one of the three academic medical center hospital service areas resulted in 32% fewer discharges (odds ratio: 0.68; confidence interval: 0.66, 0.70). Similar and statistically significant (P < .01) results were noted for the most common nonperinatal diagnostic categories: asthma/bronchitis (diagnostic related group = 98) and gastroenteritis (diagnostic related group = 184). No effect was noted for femur fracture, a condition for which admission rates equal disease incidence.

CONCLUSIONS

The supply and character of medical care are important influences on the likelihood of hospitalization for pediatric medical conditions for which outpatient alternatives are available.

Academic health centers and health care reform

There is increasing support for the proposition that academic health centers have a duty to accept broad responsibility for the health of their communities. The Health of the Public program has proposed that centers become directly involved in the social-political process as advocates for reform of the health care system. Such engagement raises important issues about the roles and responsibilities of centers and their faculties. To address these issues, the authors draw upon the available literature and their experiences in recent health care reform efforts in Minnesota and Vermont in which academic health center faculty participated. The authors discuss (1) the problematic balance between academic objectivity and social advocacy that faculty must attempt when they engage in the health care reform process; (2) the management of the sometimes divergent interests of academic health centers, some of their faculty, and society (including giving faculty permission to engage in reform efforts and developing a tacit understanding that distinguishes faculty positions on reform issues from the center's position on such issues); and (3) the challenge for centers to develop infrastructure support for health reform activities. The authors maintain that academic health centers' participation in the process of health care reform helps them fulfill the trust of the public that they are obligated to and ultimately depend on.

Estimating the variance of standardized rates of recurrent events, with application to hospitalizations among the elderly in New England

Usual approaches for estimating the variance of a standardized rate may not be applicable to rates of recurrent events. Where individuals are prone to repeated health events, Greenwood and Yule (J R Stat Soc [A], 1920;83:255-79) advocated use of the negative binomial distribution to account for departures from the assumption of randomness of recurrent events required by the Poisson distribution. In this paper, the authors implemented the negative binomial distribution in the computation of annual hospitalization rates within certain hospital market areas. Data used were from 1,549,915 New England residents aged 65 years or more who were enrolled in Medicare between October 1, 1988, and September 30, 1989, and who had 458,593 hospital admissions during that year. New England was partitioned into 170 hospital market areas ranging in population size from 162 to 70,821 elderly Medicare enrollees. The negative binomial distribution demonstrated substantially better fits than the Poisson distribution to the numbers of hospitalizations within hospital market areas. Estimated standard errors for indirectly standardized rates based on the negative binomial distribution were 25-51 percent higher than estimated standard errors that assumed an underlying Poisson distribution. Using regression analysis to smooth overdispersion parameters across hospital market areas produced similar results. The approach described in this paper may be useful in estimation of confidence intervals for standardized rates of recurrent events when these events do not recur randomly.