Multivariable risk prediction models for postoperative cardiac injury in adults undergoing non-cardiac surgery: a...

Postoperative cardiac injury, encompassing myocardial infarction (MI) and myocardial injury after non-cardiac surgery (MINS), is a major perioperative complication associated with substantial morbidity and mortality. While numerous prediction models have been developed using traditional statistical and machine learning approaches, their comparative performance, calibration quality and methodological rigour remain unclear. This protocol outlines a systematic review and meta-analysis to comprehensively evaluate multivariable risk prediction models for postoperative cardiac injury in adults undergoing non-cardiac surgery.

Following Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols guidelines, this PROSPERO-registered systematic review will search PubMed, Embase, Web of Science, Cochrane Library, Scopus, grey literature and trial registries for studies developing, validating or updating multivariable prediction models for postoperative cardiac injury (MI or MINS) occurring within 72 hours of non-cardiac surgery. Two reviewers will independently extract data and assess quality using CHARMS (Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies) and PROBAST (Prediction Model Risk of Bias Assessment Tool) tools. The primary outcome is discriminative performance (area under the receiver operating characteristic curve), with calibration metrics and diagnostic accuracy measures as secondary outcomes. Random-effects meta-analyses will pool performance estimates for models validated in multiple cohorts. Heterogeneity will be explored through subgroup analyses and meta-regression, examining factors including model methodology (regression-based vs machine learning), predictor types and validation contexts. Sensitivity analyses will test the robustness of findings.

Ethical approval is not required for this study, as it is a systematic review and meta-analysis based on previously published data.

CRD420251041628.