Improved cardiovascular risk prediction using nonparametric regression and electronic health record data

Area of Science:

• Health Informatics
• Biostatistics
• Cardiovascular Epidemiology

Background:

• Electronic Health Records (EHR) are increasingly used, yet their potential for improving clinical risk prediction via advanced statistical analysis is underexplored.
• The Veterans Health Administration's extensive EHR implementation offers a unique opportunity to investigate these methods.

Purpose of the Study:

• To compare the predictive performance of various statistical approaches for cerebrovascular and cardiovascular (CCV) death risk.
• To evaluate the utility of traditional risk predictors versus comprehensive EHR data in CCV risk prediction.

Main Methods:

• Retrospective cohort study of Veterans Health Administration patients (2003-2007).
• Risk prediction models included Framingham risk score, logistic regression, generalized additive modeling, and gradient tree boosting.
• Performance assessed using area under the receiver operating characteristic curve (AUC), Hosmer-Lemeshow test, and reclassification tables, with cross-validation.

Main Results:

• Regression methods surpassed the Framingham risk score, improving AUC from 71% to 73%.
• Incorporating additional EHR-derived variables further boosted AUC to 78% with significant net reclassification improvement (0.29).
• Nonparametric regression enhanced calibration and discrimination compared to logistic regression.

Conclusions:

• Flexible statistical methods and EHR-derived data can substantially improve clinical risk prediction for cerebrovascular and cardiovascular death.
• Even with data quality limitations, health systems can leverage internal EHR data for better patient risk assessment.