Text

The eQS programme has the goal to assure and improve quality of care in German health care. This programme is based on around 300 quality indicators. These indicators are often (risk-adjusted) complication rates or rates that count deviations from medical guidelines. They are computed using clinical or administrative data or patient reported data (PROMs/PREMs). The indicators are used to detect performance outliers and to initiate quality improvement measures. They are also publicly reported for each hospital in order to inform patients.

There are however two key challenges when reporting the results of indicators to support patient decision-making:

1. Reports typically account for large variations in hospital case numbers by adding uncertainty intervals. However, non-statisticians tend to ignore these intervals, focusing only on point estimates.
2. A single indicator often does not capture the full complexity of quality of care. Yet it is even more difficult to compare multiple indicators, especially if they are measured on different scales.

To address these challenges, IQTIG has developed a Bayesian approach. This approach calculates for each indicator the posterior probability that a given hospital meets a predefined threshold. These probabilities are averaged over multiple indicators to obtain a composite index aggregating all quality information.

In this talk, we discuss technical details of this approach, including the choice of prior distributions and thresholds, as well as some consequences towards decision-making.

The talk presents joint work and ideas developped by various persons at IQTIG for different projects commissioned by the Federal Joint Committee (G-BA) and the Federal Ministry of Health (BMG), most notably the Qualitätsportal [1] and the Bundes-Klinik-Atlas.

The author declares that he has no competing interests.

The author declares that an ethics committee vote is not required.

References