A statistically based method for identifying hospital classification criteria

The development of valid classification criteria for U.S. hospitals. A number of bills recently introduced in the U.S. Congress call for the linkage of classification criteria to cost limits for hospitals. Such proposals have not indicated how the classification criteria should be validated or tested.A research project was therefore undertaken to determine whether 87 large community hospitals could be classified into interpretable and reproducible homogenous groups. By means of an automatic interaction detector (AID), a set of unique classification criteria were identified. These included residency and internship education programs, medical school affiliation, renal dialysis, and organ bank facilities. Application of the criteria to 1970 and 1971 data for the 87 hospitals resulted in five reproducible and stable groups of hospitals. The criteria were validated by several tests involving different types of cost comparisons and ratios.The research results indicate that an AID-based classification structure is a feasible model for grouping or classifying large hospitals for comparative purposes. Only a small number of variables are necessary to classify large hospitals, and the criteria do not need to be overly complex. Many of the variables traditionally used may be omitted.