Systematic review and meta-analysis of external validation studies of multivariable diagnostic or prognostic models: a primer

Article type
Year
Authors
Debray T1, Damen J1, Snell K2, Ensor J3, Hooft L1, Riley R3, Reitsma J1, Moons K1
1Julius Center for Health Sciences and Primary Care, Netherlands
2University of Birmingham, UK
3Keele University, UK
Abstract
Background: External validation of prediction models is highly recommended and increasingly common in the literature. Review of such studies may help to identify whether an existing model is sufficiently accurate across different settings, and how it could be improved further.

Objectives: The aim of this study was to:
1. provide guidance for systematically reviewing external validation studies of a prediction model;
2. discuss ‘good practice’ when summarizing validation study results, and
3. provide recommendations for interpreting meta-analysis estimates of model performance.
Here we present several key steps: preparing the study protocol, formulating the search strategy, critical appraisal and risk of bias assessment, quantitative data extraction and preparation, meta-analysis, investigating heterogeneity across studies and reporting of results.

Methods: We illustrate each step in an exemplar review where we summarize the discrimination and calibration performance of the EuroSCORE for predicting operative mortality in patients undergoing coronary artery bypass grafting.

Conclusions: Systematic review and meta-analysis of external validation studies may help to interpret the potential applicability and generalizability of a prediction model.