Why the study?
Few studies quantitatively estimate CHD risk on-site as an auxiliary diagnosis, and machine learning models require validation using real clinical data before adoption in real clinical settings.
Does a two-layer machine learning model improve the accuracy of on-site coronary heart disease prediction compared to pooled cohort equations in high-risk adults?
Population
20,821 patients with CHD and 9,796 controls from a single hospital
Comparison
Two-layer machine learning model vs pooled cohort equations
Design
Retrospective hospital-based cohort study
Key result
A two-layer machine learning model achieved an on-site coronary heart disease prediction accuracy of 0.79 (95% CI 0.79-0.80), outperforming pooled cohort equations (accuracy 0.59).
Authors
Loading...
May enhance on-site CHD prediction in high-risk adults; leaves open prospective validation before clinical adoption.
Cohort (n=30,617)
No
Does a two-layer machine learning model improve the accuracy of on-site coronary heart disease prediction compared to pooled cohort equations in high-risk adults?
Absolute Event Rate: 0.79% vs 0.59%
A two-layer machine learning model using routine clinical data significantly improves the accuracy of on-site coronary heart disease prediction compared to traditional pooled cohort equations.
Mo et al. (2025) conducted a cohort in Coronary heart disease (n=30,617). Two-layer machine learning model (TLML) vs. Pooled cohort equations (PCEs) was evaluated on On-site CHD prediction accuracy (95% CI 0.79-0.80). A two-layer machine learning model achieved an on-site coronary heart disease prediction accuracy of 0.79 (95% CI 0.79-0.80), outperforming pooled cohort equations (accuracy 0.59).