Racial and gender variation in use of diagnostic colonic procedures in the Michigan Medicare population.

BACKGROUND: There is accumulating evidence that screening programs can alter the natural history of colorectal cancer, a significant cause of mortality and morbidity in the US. Understanding how the technology to diagnose colonic diseases is utilized in the population provides insight into both the access and processes of care. METHOD: Using Medicare Part B billing files from the state of Michigan from 1986 to 1989 we identified all procedures used to diagnose colorectal disease. We utilized the Medicare Beneficiary File and the Area Resource File to identify beneficiary-specific and community-sociodemographic characteristics. The beneficiary and sociodemographic characteristics were, then, used in multiple regression analyses to identify their association with procedure utilization. RESULTS: Sigmoidoscopic use declined dramatically with the increasing age cohorts of Medicare beneficiaries. Urban areas and communities with higher education levels had more sigmoidoscopic use. Among procedures used to examine the entire colon, isolated barium enema was used more frequently in African Americans, the elderly, and females. The combination of barium enema and sigmoidoscopy was used more frequently among females and the newest technology, colonoscopy, was used most frequently among White males. CONCLUSION: The existence of race, gender, and socioeconomic disparities in the use of colorectal technologies in a group of patients with near-universal insurance coverage demonstrates the necessity of understanding the reason(s) for these observed differences to improve access to appropriate technologies to all segments in our society.

Use of community versus individual socioeconomic data in predicting variation in hospital use.

OBJECTIVES: (1) To examine the association of socioeconomic characteristics (SES) with hospitalization by age group, and when using measures of SES at the community as opposed to the individual level. (2) Thus, to support the inference that socioeconomic factors are important in the analysis of small area utilization data and address potential criticisms of this conclusion. DATA SOURCES: The 1989 Michigan Inpatient Database (MIDB), the 1990 U.S. Census, the 1989 Area Resource File (ARF), and the 1990 National Health Interview Survey (NHIS). STUDY DESIGN: A qualitative comparison of socioeconomic predictors of hospitalization in two cross-sectional analyses when using community as opposed to individual socioeconomic characteristics was done. DATA EXTRACTION. Hospitalizations (excluding delivery) were extracted by county from the MIDB and by individual from the NHIS. SES variables were extracted from the U.S. Census for communities and from the NHIS for individuals. Measures of employment for communities were from the ARF and information on health insurance and health status of individuals from the NHIS. PRINCIPAL FINDINGS: Both analyses show similar age-specific patterns for income and education. The effects were greatest in young adults, and diminished with increasing age. Accounting for multiple admissions did not change these conclusions. In the individual-level data the addition of variables representing health and insurance status substantially diminished the size of the coefficients for the socioeconomic variables. CONCLUSIONS: By comparison to parallel individual-level analyses, small area analyses with community-level SES characteristics appear to represent the effect of individual-level characteristics. They are also not substantially affected by the inability to track individuals with multiple readmissions across hospitals. We conclude that the impact of SES characteristics on hospitalization rates is consistent when measured by individual or community-level measures and varies substantially by age. These variables should be included in analyses of small area variation.

Hospitalization for gastrointestinal and liver diseases: the effect of socioeconomic and medical supply factors.

A growing body of research has documented significant variation in health care use between communities. As the health care system is transformed, providers and payers should understand the interaction between a community, its sociodemographic characteristics, and its use of health resources. We describe the association between a population's demographic, socioeconomic, and medical resources and hospital use related to gastrointestinal and liver diseases. We used an all-payer hospital discharge database for Michigan from 1986 to 1988. We identified all medical and surgical hospital admissions during this period from two of the Diagnostic Related Group, Major Diagnostic Categories: No. 6, Diseases and Disorders of the Digestive System; and No. 7, Diseases and Disorders of the Hepatobiliary System and Pancreas. We analyzed age- and sex-specific use rates. Finally, we analyzed the influence of sociodemographic variables from the Area Resource File at the county level, on hospital use, using a Poisson regression model. We noted a significant association between increased hospitalizations and increased age in a community. Hospital beds per capita did not influence admission rates overall, although more hospital beds were associated with more medical admissions. Overall, the total physician supply was associated with more admissions. Finally, the most important socioeconomic variable was education. As the level of education of a county increased, hospital admissions decreased dramatically. The transformation of the health care delivery system presents opportunities and challenges. Understanding the underlying epidemiology of disease and how it interacts with a community's socioeconomic and medical resources or medical supply characteristics will be necessary to meet the community's health needs and to ensure the financial viability of providers. This is especially true when payers use a standard payment in a region, such as Medicare's managed care payment, without adjustments for the underlying population characteristics known to influence use.

Vascular surgeons' resource use at a university hospital related to diagnostic-related group and source of admission.

PURPOSE: To determine whether differences in the conduct of individual practices of attending vascular surgeons account for variations in resource use at a university hospital. METHODS: The practice patterns of six attending vascular surgeons at the University of Michigan Hospital were assessed for patient length of stay (LOS), ancillary service use, and the number of nursing hours required. Included in the study were 1930 hospitalized patients who had one of the 10 most frequently encountered diagnostic related groups (DRGs). Statistical analyses of variables that were thought likely to affect resource use included multiple regression models. RESULTS: Patient age, sex, insurance, source of admission (direct admission or transfer admission), surgeon, and DRG category together accounted for 22% of LOS variation, 27.7% of variation in ancillary service use, and 29.4% of variation in nursing hours. In no model did the individual surgeon's practice significantly effect the LOS, ancillary use, or nursing hours. Patients transferred from other hospitals had increased resource use in all models. The DRG category alone explained 20.9% of the variance in LOS, 25.2% of the variation in ancillary service use, and 21.2% of the variance in nursing hours. CONCLUSION: Differences in the conduct of individual vascular surgeons' practices accounted for less than 1% variation in hospital resource use. The most important influences on resource use were the DRG category and the source of patient admission. Modification of the frequency and manner of accepting transfer patients to vascular surgery services of a university hospital may have a major impact on hospital resource use.

Comparing the hospitalizations of transfer and non-transfer patients in an academic medical center.

BACKGROUND: By accepting and caring for patients transferred from other institutions, academic medical centers have been able to develop comprehensive training and research programs. Whether academic institutions can continue to do this in the future is questionable. To the extent that transfer patients are more complex and severely ill than non-transfer patients, they are likely to consume more resources, and in managed care payment systems, they could place accepting hospitals in financial jeopardy. METHOD: Between July 1989 and December 1993, the internal medicine, surgery, and pediatrics services of the 880-bed University Hospital of the University of Michigan accepted 8,740 patients from other hospitals. The hospitalizations of these patients were compared with those of the 76,047 non-transfer patients on these services. The statistical methods used were Student's t-test, chi-square, Cochran-Mantel-Haenszel chi-square, and analysis of variance. RESULTS: The hospitalizations of the transfer patients were more complex and resource-use intensive. The transfer patients were more likely (p<.0000) to be length-of-stay outliers as defined by Medicare standards (28% vs 10%) and to suffer in-hospital death (9.4% vs 2.5%). After case-mix adjustment and exclusion of length-of-stay outliers, transfer patients on the three services (surgery, medicine, and pediatrics) remained in the hospital 1.62, 1.15, and 0.84 days longer (p<.0001) than non-transfer patients. Ancillary-service resource use was assessed using a relative-value-unit (RVU) scale based on direct-cost dollars. The transfer patients' case-mix-adjusted resource use exceeded that of the non-transfer patients by 1,155,850 and 957 RVUs for surgery, pediatrics, and medicine (p<.0001). Although the transfer patients were more likely to have Medicaid insurance, the differences in lengths of stay and use of ancillary services persisted throughout all insurance groups. Indeed, transfer status, compared with age, sex, and insurance status, was the best predictor of high resource use. CONCLUSION: The transfer patients stayed longer and consumed more hospital resources than did the non-transfer patients. Age, sex, case-mix, and insurance status did not account for these differences. To limit the financial liability that transfer patients pose, academic medical centers could be forced to abandon their traditional role of caring for such patients. The consequences of this possibility should be explored.

Comparing the use of diagnostic tests in Canadian and US hospitals.

Although Americans pay much more for a day in the hospital than Canadians, we know little about whether inpatient physician practice patterns might explain some of this difference. The authors compared the utilization of all diagnostic imaging (plain radiographs, computed tomography (CT) and magnetic resonance imaging (MRI) scanning, ultrasound, nuclear medicine and vascular studies) and selected laboratory tests (hematology, basic biochemistry, and advanced biochemistry) for all patients discharged with selected medical and surgical diagnoses in 1990 and 1991 from four university hospitals and four community hospitals in Canada (n = 6,491) and the United States (n = 7,980). Overall, US medical patients received 22% more diagnostic tests than their Canadian counterparts (544.2 relative value units [RVUs] vs. 446.5 RVUs in Canada, P < 0.001), which was mainly the result of higher radiology use. Although mean radiology use was 40% higher in the United States (370.0 vs. 264.5 RVUs in Canada, P < 0.05), there was little difference in the use of laboratory tests between countries (174.2 vs. 182.4 RVUs in Canada, P = 0.3). Within radiology, only CT and MRI use differed significantly between countries (US patients received 119% more tests than Canadians). These findings were consistent after adjustments for age, gender, diagnosis-related group, and university status. Differences in test use between countries were mainly the result of more testing among the US elderly than counterparts in Canada. Among surgical patients, there was little difference between countries for radiology (76.3 vs. 67.3 RVUs in Canada, P < 0.05) and laboratory (83.6 vs. 91.4 RVUs in Canada, P < 0.05). Comparable inpatients admitted to US hospitals received more diagnostic tests than their Canadian counterparts even in hospitals with similar availability of technology. Differences between countries were larger for high-cost tests than for lower-cost tests. Much of the difference in test use is explained by more intensive use for the elderly in the United States.

The integrated inpatient management model. Lessons for managed care.

The Integrated Inpatient Management Model was a 2.5-year controlled prospective trial of using a clinical information system to direct and monitor physician and hospital practice on general medicine services of an 880-bed university hospital. For the over 2,000 admissions on both a control service and the intervention service, the mean length of stay (LOS) decreased when compared with historic norms (0.68 and 0.95 days respectively; P < 0.01 for both). This difference in mean LOS represents a savings of 580 hospital days for the intervention over the control service; (95% confidence interval, 300 to 1420 days). There also was a trend for the intervention service to have fewer LOS outliers than expected (P = 0.14). Ancillary service use decreased by 17% on both control and intervention services (a trend that disappeared after the study was terminated), while other internal medicine services experienced a 29% increase in this measure of resource use. The intervention service experienced fewer preventable deaths (P = 0.04), but there were no differences in global quality of care measures, readmission and mortality rates, and patient satisfaction. This use of a clinical information system is a prototype for the systems that will be needed for all forms of managed care.

Do attending or resident physician practice styles account for variations in hospital resource use?

Prospective payment has created incentives for hospitals to identify physicians who are responsible for high or excessive rates of resource use. However, at teaching hospitals it is unclear whether individual attending or resident physicians account for a substantial portion of the observed variations in hospital resource use. To explore this issue, case-mix adjusted hospital length of stay and ancillary resource use at a university teaching hospital for 7,667 consecutive discharges on general medicine wards and 7,566 discharges on medical subspecialty wards were evaluated. After controlling for case mix and patient characteristics (patients' age, sex, marital status, insurance status, and ward service), only 2% of the length of stay variance (log transformed) was attributable to the attending physician on general medicine wards (P = 0.06) and 1% on subspecialty medicine wards (P < 0.01). For total ancillary resource use, about 2% of the variance was attributable to general medicine and subspecialty ward attendings. Similar associations were found for resident physicians, although the overlap of attending and resident physicians' month-long rotations prevented critical appraisal of their independent contributions to resource use. Furthermore, labeling attending physicians as high or low hospital resource utilizers based on data from one month of attending duty (mean admissions = 33 +/- 7) would be scarcely better than randomly classifying them (kappas ranged from -0.05 for length of stay on subspecialty services to 0.18 for pharmacy use on general medicine services). In conclusion, in this university teaching hospital, attendings and residents account or a small, although statistically significant, amount of the variation in hospital resource use. It would be impractical for the hospital to reliably profile the resource use intensity of individual physicians.

The integrated inpatient management model's clinical management information system.

The rising cost of health care has increased the call for cost control. The pressing need to control cost, coupled with the increase in managed care and prospective payment, has placed new urgency on administrators and clinicians to work collaboratively in providing efficient and effective care. We have developed the Integrated Inpatient Management Model (IIMM) to assist in this collaborative effort. We describe the IIMM's clinical information system that provides decision support to both administrators and clinicians. This clinical information system is the information backbone for the development and monitoring of practice guidelines or critical pathways. An integrated information system of this type is essential if hospitals are to prosper during the next decade.

Socioeconomic influence on small area hospital utilization.

Health care policy makers, concerned with the rising cost of health care, have focused on the observed variation in the use of hospitals as a potential area in which to lower health care costs, i.e., if hospital utilization can be decreased, health care costs may also decline. However, it is crucial that the reasons for the observed variation in the current practice be understood or attempts to reduce costs may lead to policies that harm groups of patients and the providers and institutions currently delivering care. Using hospital discharge data from 59 hospital market communities in the lower peninsula of Michigan in 1984-86, the authors examined possible associations between socioeconomic characteristics and the observed small area variation in hospital discharge rates. First, a series of Poisson regressions was used for each of five covariates and 112 modified diagnosis-related groups (DRGs). Then, multiple regressions were examined, utilizing the five socioeconomic characteristics, after excluding statistically influential communities. The results indicate that community characteristics, including education, poverty, and unemployment, have a statistically significant association with the observed small area hospital discharge rate for many DRGs. Moreover, the direction of the effect is consistent across multiple disease categories. In multiple regressions, the five selected socioeconomic variables explained 48% of the variance for medical admissions and 19% for surgical admissions. For most DRGs, high educational levels were associated with lower hospitalization rates. The authors also identified statistically influential communities whose hospital utilization profile was different from that of most communities in Michigan.

An evaluation of generic screens for poor quality of hospital care on a general medicine service.

In this study, 675 general medicine admissions at a university teaching hospital were reviewed to evaluate six potential generic quality screens: 1) in-hospital death; 2) 28-day early readmission; 3) low patient satisfaction; 4) worsening severity of illness (as determined by an increase in Laboratory Acute Physiology and Chronic Health Evaluation APACHE-L); and 5) deviations from expected hospital length of stay; and 6) expected ancillary resource use. The quality of care for a stratified random sample of admissions were evaluated using structured implicit review (inter-rate reliability, Kappa = 0.5). Patients who died in-hospital were substantially more likely than those who were discharged alive to be rated as having had substandard care (30% vs. 10%; P < 0.001). In contrast, cases who had subsequent early readmissions did not have poorer quality ratings. Similarly, lower patient satisfaction was not associated with poorer ratings of technical process of care. Cases with lower-than-expected ancillary resource use (case-mix adjusted for diagnosis-related group) were more likely to be rated as having received substandard care than those with higher-than-expected resource use (16% vs. 6%; P < 0.05), and there was a similar trend for cases with shorter than expected length of stays. Associations between worsening severity of illness, as determined by APACHE-L scores, and quality were confounded because such patients were more likely to have died in-hospital.(ABSTRACT TRUNCATED AT 250 WORDS)

Evaluating the care of general medicine inpatients: how good is implicit review?

OBJECTIVE: Peer review often consists of implicit evaluations by physician reviewers of the quality and appropriateness of care. This study evaluated the ability of implicit review to measure reliably various aspects of care on a general medicine inpatient service. DESIGN: Retrospective review of patients' charts, using structured implicit review, of a stratified random sample of consecutive admissions to a general medicine ward. SETTING: A university teaching hospital. PATIENTS: Twelve internists were trained in structured implicit review and reviewed 675 patient admissions (with 20% duplicate reviews for a total of 846 reviews). RESULTS: Although inter-rater reliabilities for assessments of overall quality of care and preventable deaths (kappa = 0.5) were adequate for aggregate comparisons (for example, comparing mean ratings on two hospital wards), they were inadequate for reliable evaluations of single patients using one or two reviewers. Reviewers' agreement about most focused quality problems (for example, timeliness of diagnostic evaluation and clinical readiness at time of discharge) and about the appropriateness of hospital ancillary resource use was poor (kappa < or = 0.2). For most focused implicit measures, bias due to specific reviewers who were systematically more harsh or lenient (particularly for evaluation of resource-use appropriateness) accounted for much of the variation in reviewers' assessments, but this was not a substantial problem for the measure of overall quality. Reviewers rarely reported being unable to evaluate the quality of care because of deficiencies in documentation in the patient's chart. CONCLUSION: For assessment of overall quality and preventable deaths of general medicine inpatients, implicit review by peers had moderate degrees of reliability, but for most other specific aspects of care, physician reviewers could not agree. Implicit review was particularly unreliable at evaluating the appropriateness of hospital resource use and the patient's readiness for discharge, two areas where this type of review is often used.

Changing patterns of hospital use for patients with musculoskeletal disease in Michigan, 1980 to 1987.

Over the past 10 years there have been dramatic changes in health care financing in the United States, such as Medicare's Prospective Payment System for hospitalized Medicare beneficiaries, and in health services delivery, such as the growth in health maintenance organizations and other forms of managed care. These changes have occurred largely in response to payors' concerns about the rising cost of health care. A study of such changes in financing and delivery, and how specific groups of patients are affected is necessary so that the effects of these changes on patients' health can be determined. We examined the hospitalization rates for patients with musculoskeletal diseases in Michigan from 1980 through 1987. During this period, the overall age-adjusted hospitalization rates decreased 7.0% per year (p = 0.001). The decrease occurred less for surgical discharges (6.0% per year) than for medical discharges (8.6% per year) (p < 0.001). While these overall trends are of interest, they obscure disease-specific trends that vary significantly from both the overall, and the medical and surgical trends. For example, while surgical discharges, in general declined, procedures related to major joint and limb reattachment (DRG #209) increased at a rate of 6.3% per year. And while medical discharges in general decreased over this period, discharges for osteomyelitis increased 5.4% per year. The patterns of disease-specific trends offers insight into the possible causes for these changes. Finally, it is important to understand the epidemiology of hospital use to evaluate the effects of new medical care delivery and payment systems on the care of subsets of patients.

APACHE-L: a new severity of illness adjuster for inpatient medical care.

A principal concern regarding Medicare's diagnosis-related group (DRG)-based prospective payment system is whether hospitals caring for more severely ill patients may be undercompensated for the services they provide. Research on possible inequities in hospital payment has been hampered by the absence of an objective, easily obtained, and valid measure of patients' severity of illness. Because laboratory data are objective and computerized in most of our nation's hospitals, a system utilizing such data, if shown to discriminate between patients of differing expected resource use, could prove most helpful in examining possible inequities in prospective payment system hospital payment. At a major teaching hospital, data were used from length of stay inlier patients in the 10 most frequent medical DRGs in the U.S. to develop and evaluate a severity of illness system called APACHE-L. APACHE-L uses the laboratory component of the original APACHE score. Whereas DRGs explained 20% of the variation in length of stay for the top ten DRGs, APACHE-L explained up to an additional 14% of the variation. For ancillary resource use, DRGs explained 10% of the variance, and APACHE-L explained up to an additional 15%. Diagnosis-related group-specific analyses demonstrated that the amount of resource use variance explained by APACHE-L varied widely depending on the DRG (from R2 = .00 for DRG 410, chemotherapy; to R2 = .38 for DRG 320, kidney and urinary tract infections, age greater than 17 years with complications or comorbidities). The APACHE-L score, which is objective and readily available in our nation's hospitals, shows considerable promise as a severity of illness adjuster for a subset of DRGs.

Measuring outcomes of hospital care using multiple risk-adjusted indexes.

Using existing data sources, we developed three risk-adjusted measures of hospital quality: the risk-adjusted mortality index (RAMI), the risk-adjusted readmissions index (RARI), and the risk-adjusted complication index (RACI). We describe the construction and validation of each of these indexes. After these measures were developed, we tested the relationships among the three indexes using a sample of 300 hospitals. Actual numbers of adverse events were observed for each hospital and compared to the number predicted by the RAMI, RARI, and RACI models. Then each hospital was ranked on each index. Our results showed that no relationship existed between a hospital's ranking on any one of these indexes and its ranking on the other two indexes. This result provides some evidence that no measure of quality should be used by itself to represent different aspects of the quality of hospital care. Adequate overall measures of hospital quality will need to include multiple measures in order to be credible and to reflect the complexity of hospital care. The findings suggest that consumers, payers, and policymakers cannot simply choose one hospitalwide measure, such as the mortality rate, to validly represent a hospital's performance: those hospitals with high rankings on their mortality rates do not necessarily rank high on their readmission rates or complication rates.

Empirical evaluation of statistical models for counts or rates.

We consider methods for selecting the joint specification of the mean and variance functions in statistical models for rates or counts. Based on analyses of diagnosis-specific hospital discharge rates in Michigan, we show that a Poisson model with an extra variance component for the systematic variation is superior to several other probability models with regard to specification of the error structure. Further, the deviance residual appears superior to the Pearson residual. The proper specification of such variation is crucial for many types of analyses, such as identification of outliers and regression analyses designed to explain the systematic component of the variation.

The measurement of mortality. A risk-adjusted variable time window approach.

This paper describes the development of risk-adjusted mortality indices (RAMI) using 1985 MEDPAR data from 657 hospitals. The RAMI methodology is adopted from the Commission on Professional and Hospital Activities, however, both inhospital and post-discharge deaths are counted within time windows that vary by clinical condition. Five different RAMI measures (expected deaths/observed deaths) are developed, compared, and aggregated into various hospital characteristic strata. These measures vary by which discharge is held responsible for deaths within a time window, and whether or not inhospital deaths that occur beyond the time window are included. The RAMIs using varying time windows are compared with the RAMIs based upon inhospital deaths only. The inhospital RAMI was higher for the nonteaching hospitals (.95) as compared with the major and minor teaching institutions (.91 and .89). The RAMIs using the varying time windows, on the other hand, tend to be higher for the teaching institutions (e.g., 1.07 for major teaching hospitals; 0.99 for nonteaching hospitals).

Small area analysis of hospital discharges for musculoskeletal diseases in Michigan: the influence of socioeconomic factors.

PURPOSE: The rise in health care costs has occasioned a number of initiatives in an attempt to reduce the rate of increase. Despite the growth of health maintenance organizations and preferred provider organizations and the introduction of Medicare's prospective payment system, health care costs have continued to increase. Coincident with these efforts, a number of researchers have shown that there exists wide variation in age-adjusted hospital discharge rates, which translate into significant variation in per capita expenditures. Much of the focus on the reasons for hospital admission variability has been on physician practice variation. If most of the variation in hospital discharge rates is due to physician practice style, then payment systems can be developed (e.g., capitation) that limit physician practice variation without harming patients. We examined socioeconomic factors in Michigan communities to assess their association with hospital discharge rates for patients with musculoskeletal diseases. PATIENTS AND METHODS: Data on hospital discharges from 1980 and 1987 were taken from the Michigan Inpatient Data Base. All admissions from the major diagnostic category 8, diagnosis-related group (DRG) 209-256 were included. Zip code-specific hospitalization data were grouped into small geographic areas or hospital market communities (HMCs). Discharge rates were calculated, and profiles of the socioeconomic characteristics of each of the HMCs were developed. A Poisson regression model with an extrasystematic component of variance was used to analyze the association of HMC socioeconomic characteristics with age-adjusted hospital use. RESULTS: We found that four socioeconomic variables, average annual income per capita, percent of the population with four years of college, percent of the population living in an urban area, and percent of families with incomes below the poverty line, explained 26.6% (R2) of the variation in overall hospital discharge rates (p less than 0.001). Moreover, we found that the ability of the model to explain variability was influenced by the type of disease, and that these socioeconomic variables had a consistent effect across the range of DRGs. Finally, we noted that, over the period of 1980 to 1987, socioeconomic factors remained important in explaining hospital use despite the dramatic changes in the delivery of care over this period. CONCLUSION: Socioeconomic factors play a significant role in explaining the observed variation in hospital discharge rates for musculoskeletal diseases. Models utilizing only physician practice variation to account for the population-based differences in discharge rates are overly simplistic. In order to ensure that vulnerable subsets of the population are not harmed by the introduction of cost-containment strategies based on simplistic models, more attention must be paid to the socioeconomic and epidemiologic factors related to hospital use.

Changes in rates of unscheduled hospital readmissions and changes in efficiency following the introduction of the Medicare prospective payment system. An analysis using risk-adjusted data.

The purpose of this study was to analyze changes in rates of unscheduled readmissions and changes in technical efficiency following the introduction of the Medicare Prospective Payment System (PPS). We developed the Risk-Adjusted Readmissions Index (RARI), which allowed us to make comparisons in rates of unanticipated readmissions across hospitals and over time. Data envelopment analysis (DEA), a linear programming technique, was used to measure changes in technical efficiency by comparing the inputs used and the outputs produced across a cohort of hospitals, while adjusting for changes over time in case mix and case complexity. Rates of unscheduled readmissions and efficiency scores were computed for a sample of 245 hospitals for each year. Although both readmission rates and efficiency scores increased for most hospitals, there was no evidence that those hospitals that experienced the greatest increases in efficiency had the largest increases in their rates of unscheduled readmissions.