Predicting Chronic Disease Hospitalizations from Electronic Health Records: An Interpretable Classification Approach

TL;DR

This paper introduces interpretable machine learning methods to predict hospitalizations for chronic diseases like heart disease and diabetes using electronic health records, balancing accuracy with interpretability.

Contribution

It proposes two novel methods, K-LRT and JCC, for improved prediction and interpretability, with theoretical guarantees for the JCC approach.

Findings

Validated algorithms on large Boston Medical Center datasets.

Demonstrated effectiveness of proposed methods in clinical prediction.

Provided theoretical guarantees for the JCC method.

Abstract

Urban living in modern large cities has significant adverse effects on health, increasing the risk of several chronic diseases. We focus on the two leading clusters of chronic disease, heart disease and diabetes, and develop data-driven methods to predict hospitalizations due to these conditions. We base these predictions on the patients' medical history, recent and more distant, as described in their Electronic Health Records (EHR). We formulate the prediction problem as a binary classification problem and consider a variety of machine learning methods, including kernelized and sparse Support Vector Machines (SVM), sparse logistic regression, and random forests. To strike a balance between accuracy and interpretability of the prediction, which is important in a medical setting, we propose two novel methods: K-LRT, a likelihood ratio test-based method, and a Joint Clustering and Classification (JCC) method which identifies hidden patient clusters and adapts classifiers to each cluster. We develop theoretical out-of-sample guarantees for the latter method. We validate our algorithms on large datasets from the Boston Medical Center, the largest safety-net hospital system in New England.