Mapping Cardiovascular Prediction Models in the General Population: Where Are We Now and Where Next?

This scoping review aimed to comprehensively map the landscape of prediction models for predicting cardiovascular disease (CVD) risk in the general population over the past two decades, including model development, validation, updating, and impact assessment.

PubMed, Web of Science, and Embase databases were searched from January 2006 to November 2025 for studies that developed, validated, updated, or assessed the impact of at least one CVD risk prediction model in the general population. Two reviewers independently screened studies and extracted data on general characteristics, justification, methodological characteristics, and reporting characteristics. Findings were summarized descriptively.

We included 322 studies, comprising 220 development studies (yielding 298 models), 94 independent validation or update studies, and eight studies providing nine impact assessments involving seven models. Of the 220 development studies, 90.0% cited at least one existing model; the most frequently cited justification was insufficient predictors (44.4%). Among the 298 models, 86.6% had undergone internal validation, while external validation (43.0%), updating (8.7%), and impact assessment (2.3%) remained low. Reporting completeness varied: 82.3% reported both discrimination and calibration, 49.4% provided full model specification, and 51.1% offered a practical tool. Notably, only 24.3% of models published after 2015 cited the TRIPOD guideline. Only 11.1% were reported to be recommended in clinical guidelines.

Despite a substantial increase in CVD prediction model development, critical gaps persist in external validation, reporting transparency, and clinical impact assessment. Future research should prioritize strengthening the evidence base of existing models through rigorous validation, transparent reporting, and implementation research.