Model Performance Metrics for Cardiovascular Disease Prediction in Hypertensive Patients

A single-center retrospective study enrolled 3,588 hypertensive patients free of cardiovascular disease (CVD) at baseline and followed them for a mean of 8.3 years. The dataset, containing 155 variables enriched with trend variables and latent class analysis, was used to train an XGBoost model for predicting new-onset CVD. The model, achieving 86% ROC AUC, was developed by author Enrique Rodilla and shared on figshare in May 2026.

Use Cases

• Benchmarking traditional risk scores (like SCORE2) against AI models based on predictive performance metrics.
• Investigating the predictive power of the number of antihypertensive drugs for cardiovascular events.
• Developing feature selection and engineering pipelines for clinical datasets with 155 variables.
• Training and validating XGBoost models for binary classification of new-onset cardiovascular disease.

Strengths

• Includes performance metrics for a validated XGBoost model (86% ROC AUC, 81% sensitivity, 78% specificity).
• Based on a clinical cohort of 3,588 patients with a mean follow-up of 8.3 years.
• Dataset was enriched using statistical measures, Principal Component Analysis (PCA), and Latent Class Analysis (LCA).
• Model identified a specific, clinically relevant predictor (number of antihypertensive drugs) as having the strongest power.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data is from a single-center study, which may reflect geographic or institutional bias.

Provenance

Source

Enrique Rodilla via figshare.

Collection Method

Single-center retrospective study of consecutive hypertensive patients.

Time Range

Follow-up period averaged 8.3 years; specific start/end years are not provided.

Freshness

Last updated 2026-05-21 17:23:19; freshness should be verified.