Statistical Computation and Its Application in Hospital Performance Management: A Performance Appraisal Perspective

Abstract

Assessing hospital performance during quality improvement initiatives relies heavily on statistical methodologies aimed at benchmarking. This multi-institutional study evaluates established statistical approaches for their applicability in improving healthcare quality. Using a dataset encompassing 48,521 patients undergoing craniotomy surgery from 2019 to 2023 across 180 hospitals, 6 statistical techniques were analyzed. Non-contraction methods included indirect standardization without hospital effects, with fixed effects, and direct standardization with fixed effects. Contraction methods comprised direct standardization with random effects, indirect standardization with random effects, and Exponential Smoothing methods. Adjusted rates, rankings, and performance outliers are the main emphasis of this study, which evaluates hospital performance concerning the rate of operation deaths and severe complications or deaths among approaches. Contraction methods significantly reduced inter-hospital variation in death rate(adjusted: 1%2%; observed: 0%-10%) and severe complications or death (observed: 3%-35%; adjusted: 7%-17%). Hospital rates were effectively brought closer to the collective mean using these techniques. Direct standardization with hospital random effects resulted in significant changes in death rate quintile rankings for 17% to 39% of hospitals when compared to fixed effects. Indirect standardization with random effects revealed no performance outliers for death rates in compact and moderate-sized hospitals, but logistic and fixed effect approaches revealed outliers. The choice of statistical approach has a considerable impact on hospital rankings and the identification of performance outliers. These insights are critical for appropriately evaluating the hospital's effectiveness in quality improvement initiatives.