Developing Knowledge-enhanced Chronic Disease Risk Prediction Models from Regional EHR Repositories

TL;DR

This paper presents a method for developing localized disease risk prediction models by tuning from regional EHR data and injecting domain knowledge, significantly improving prediction accuracy for ASCVD in diabetic populations.

Contribution

It introduces a knowledge-enhanced approach for localizing risk models using regional EHR data and knowledge injection, improving prediction performance over standard models.

Findings

Knowledge-enhanced models achieve higher AUC (0.723) compared to baseline (0.653).

Localization improves prediction accuracy for regional populations.

Knowledge injection enhances model performance in disease risk prediction.

Abstract

Precision medicine requires the precision disease risk prediction models. In literature, there have been a lot well-established (inter-)national risk models, but when applying them into the local population, the prediction performance becomes unsatisfactory. To address the localization issue, this paper exploits the way to develop knowledge-enhanced localized risk models. On the one hand, we tune models by learning from regional Electronic Health Record (EHR) repositories, and on the other hand, we propose knowledge injection into the EHR data learning process. For experiments, we leverage the Pooled Cohort Equations (PCE, as recommended in ACC/AHA guidelines to estimate the risk of ASCVD) to develop a localized ASCVD risk prediction model in diabetes. The experimental results show that, if directly using the PCE algorithm on our cohort, the AUC is only 0.653, while our knowledge-enhanced localized risk model can achieve higher prediction performance with AUC of 0.723 (improved by 10.7%).