Article type
Year
Abstract
Background: Publications on multivariable clinical prediction models have become abundant for both prognostic and diagnostic purposes. Systematic reviews of these studies are increasingly required to identify and critically appraise the existing evidence. There is currently no checklist or tool providing guidance for systematic reviews of studies developing or validating prediction models that can assist reviewers to define the review objectives and appraise study methodology.
Objectives: To develop a checklist to help reviewers framing a well-defined review question, and to determine which details to extract and critically appraise from primary studies on the development or validation of multivariable diagnostic or prognostic prediction models, with a view to assessing the risk of bias and sources of heterogeneity.
Methods: We critically examined existing reporting guidelines and quality assessment tools, key methodological publications on clinical prediction modelling, and tools used in published systematic reviews of multivariable prediction models, to identify the relevant characteristics and domains. The checklist was tested in various systematic reviews.
Results: We identified seven items important for framing the review question (diagnostic versus prognostic model, intended scope of the review, type of prediction modelling studies, target population, outcome to be predicted, time span of prediction, intended moment of using the model), and 11 domains to critically appraise the primary included studies (source of data, participants, outcome, predictors, sample size, missing data, model development, model performance, model evaluation, results, interpretation). Both were combined into the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS).
Conclusions: CHARMS is designed to assist reviewers to help systematic reviewers when framing their review objectives, and to determine which data to extract and critically appraise from primary studies on the development and/or validation of (diagnostic and prognostic) prediction models.
Objectives: To develop a checklist to help reviewers framing a well-defined review question, and to determine which details to extract and critically appraise from primary studies on the development or validation of multivariable diagnostic or prognostic prediction models, with a view to assessing the risk of bias and sources of heterogeneity.
Methods: We critically examined existing reporting guidelines and quality assessment tools, key methodological publications on clinical prediction modelling, and tools used in published systematic reviews of multivariable prediction models, to identify the relevant characteristics and domains. The checklist was tested in various systematic reviews.
Results: We identified seven items important for framing the review question (diagnostic versus prognostic model, intended scope of the review, type of prediction modelling studies, target population, outcome to be predicted, time span of prediction, intended moment of using the model), and 11 domains to critically appraise the primary included studies (source of data, participants, outcome, predictors, sample size, missing data, model development, model performance, model evaluation, results, interpretation). Both were combined into the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS).
Conclusions: CHARMS is designed to assist reviewers to help systematic reviewers when framing their review objectives, and to determine which data to extract and critically appraise from primary studies on the development and/or validation of (diagnostic and prognostic) prediction models.