Empirical Bayes methods for monitoring health care quality

TL;DR

This paper applies empirical Bayes methods to monitor health care quality across Dutch gynecological centres, introducing new estimates and predictive techniques for performance assessment over time.

Contribution

It introduces a simple crude estimate of centre effects in logistic regression and methods for predicting future performance, extending previous empirical Bayes approaches.

Findings

New crude estimate of centre effects in logistic regression

Method for predicting next year's performance

Measure of rankability of health care centres

Abstract

The paper discusses empirical Bayes methodology for repeated quality comparisons of health care institutions using data from the Dutch VOKS study that annually monitors the relative performance and quality of nearly all Dutch gynecological centres with respect to different aspects of the childbirths taking place in these centres. This paper can be seen as an extension of the pioneering work of Thomas, Longford and Rolph and Goldstein and Spiegelhalter. First of all, this paper introduces a new simple crude estimate of the centre effect in a logistic regression setting. Next, a simple estimate of the expected percentile of a centre given all data and a measure of rankability of the centres based on the expected percentiles are presented. Finally, the temporal dimension is explored and methods are discussed to predict next years performance.