Managing care, incentives, and information: an exploratory look inside the "black box" of hospital efficiency.

OBJECTIVE: We sought to estimate the impact of individual dimensions of hospitals' managed care strategies on the cost per hospital discharge. STUDY SETTING/DATA SOURCES: Thirty-seven member hospitals of seven health systems in the Pacific, Rocky Mountain, and Southwest regions of the United States were studied. STUDY DESIGN: Separate cross-sectional regression analyses of 21,135 inpatient discharges were performed in 1991 and 23,262 discharges in 1992. The multivariate model was estimated with hospital cost per discharge as the dependent variable. Model robustness was checked by comparing regression results at the individual discharge level with those at the level of the hospital/clinical condition pair. DATA COLLECTION/EXTRACTION METHODS: Information on hospitals' managed care strategies was provided by mail and phone survey of key informants in 1991 and 1992. Other hospital characteristics were collected from AHA Annual Survey data, and discharge data from hospital abstracting systems. PRINCIPAL FINDINGS: The pooled discharge analysis indicated three dimensions of hospital managed care strategy that consistently related to lower costs per hospital discharge: the proportion of hospital revenues derived from per case or capitation payment, the hospital's mechanisms for sharing information on resource consumption with clinicians, and the use of formalized, systematic care coordination mechanisms. CONCLUSIONS: Three strategies appear to hold promise for enhancing the efficiency of inpatient resource use: (1) "fixed price" hospital payment incentives, (2) hospital approaches to sharing resource use information with clinicians, and (3) the application of formal care management mechanisms for specific clinical conditions.

An alterative method for hospital partition determination using hierarchical cluster analysis.

The classification of short-term hospitals into homogeneous groups has become an integral part of many systems designed to abate continuing cost inflation in the hospital industry. This paper describes one approach which was developed to identify homogeneous groups of short-term hospitals. The approach, based on hierarchical cluster analysis, defines an objective measure (called expected distinctiveness) to evaluate any group of hospitals identified by a hierarchical grouping structure or dendrogram. Using this measure, an efficient algorithm is developed which finds the hospital partition from the identified groups which maximizes total expected distinctiveness. A numerical example illustrates the application and extensions.

A current reappraisal of Berry's hospital typology.

In an earlier article, Berry published the results of a study that examined possible relationships among hospital facilities and services. In general, Berry found that hospitals behaved in a reasonably consistent fashion; that is, hospitals added facilities and services in a well-defined order, and these facilities and services were added in distinct groups. Berry defined five such groups and hypothesized that each group represents an increasing level of case-mix complexity. Given the current interest in using Berry's results to measure hospital case mix, this study attempted to replicate Berry's results using recent data from the American Hospital Association. Using a number of statistical methodologies (including cluster analysis and Guttman scale analysis), we found that hospitals continue to add facilities and services in a well-defined order (although this order contains some notable differences from the order found by Berry). However, our results indicate that hospitals no longer add facilities and services in well-defined groups but in a more continuous fashion. Thus, it would appear that hospitals have become more differentiated.

The determination of alternative hospital classifications.

The problem of determining alternative hospital classifications when explicit constraints are recognized in the classification process is discussed. It is argued that such constraints are an integral part of most hospital classification processes and should be handled directly. A methodology based on cluster analysis, which can directly accommodate any number of specific constraints that might be imposed on the classification process, is described. A number of these possible constraints are presented. To illustrate the process, the methodology is applied to a data set of 200 short-term general hospitals; empirical results are presented that demonstrate the potential effects of various constraints on the hospital classification process.

On the measurement of hospital case mix.

This article discusses a number of issues related to the measurement of hospital diagnostic case mix. We initially examine a number of previous attempts to measure case mix based on surrogate measures (e.g., facilities and services) and information from predetermined discharge-classification systems. Since a number of researchers have attempted to reduce diagnostic classification data into a single-valued (i.e., scalar) case mix index, we then discuss a number of concepts and assumptions implicit in the construction of such indices. Among these assumptions is the property of functional homogeneity; this property and a methodology baed on Q-type factor analysis for testing for the presence of this property are defined. In order to illustrate the use of the methodology, it is applied to data from 153 hospitals in downstate New York.