[Preoperative risk prediction models for noncardiac surgery patients : Interpret and use risk scores correctly].

Abstract

Risk prediction models are an established component of the preoperative evaluation. In its guidelines the European Society for Cardiology proposes several risk scores but the benefit of these is mostly unclear for clinicians. This article describes the individual steps in the preparation of a valid prediction model with a focus on the parameters, discrimination, calibration and external validation. The clinical benefits of the risk scores proposed in the guidelines with respect to these parameters was investigated. All proposed risk scores appear to show a good discrimination in the validation cohorts. Only a few reliable data for a good calibration could be compiled. The external validity of the individual models is unclear. The general benefit of the risk scores cannot be recommended as data for calibration or discrimination in external cohorts are lacking. A precise estimation of the risk cannot be expected.