Inferring multimodal latent topics from electronic health records

Yue Li,Janet E Olson,Manolis Kellis,Yan Miao,Yuri Ahuja,Pratheeksha Nair,Xing Han Lu,Yuening Wang,Weiqi Liu,Amir Ardalan Kalantari Dehaghi,Mark A Frye,Aihua Liu,Zhi Wen,Tamas Ordog,Liming Guo,Jose Davila-Velderrain,Euijung Ryu,Ariane Marelli,Joanna M Biernacka

doi:10.1038/s41467-020-16378-3

Abstract

Electronic health records (EHR) are rich heterogeneous collections of patient health information, whose broad adoption provides clinicians and researchers unprecedented opportunities for health informatics, disease-risk prediction, actionable clinical recommendations, and precision medicine. However, EHRs present several modeling challenges, including highly sparse data matrices, noisy irregular clinical notes, arbitrary biases in billing code assignment, diagnosis-driven lab tests, and heterogeneous data types. To address these challenges, we present MixEHR, a multi-view Bayesian topic model. We demonstrate MixEHR on MIMIC-III, Mayo Clinic Bipolar Disorder, and Quebec Congenital Heart Disease EHR datasets. Qualitatively, MixEHR disease topics reveal meaningful combinations of clinical features across heterogeneous data types. Quantitatively, we observe superior prediction accuracy of diagnostic codes and lab test imputations compared to the state-of-art methods. We leverage the inferred patient topic mixtures to classify target diseases and predict mortality of patients in critical conditions. In all comparison, MixEHR confers competitive performance and reveals meaningful disease-related topics.

Highlights

Electronic health records (EHR) are rich heterogeneous collections of patient health information, whose broad adoption provides clinicians and researchers unprecedented opportunities for health informatics, disease-risk prediction, actionable clinical recommendations, and precision medicine
Using the Medical Information Mart for Intensive Care (MIMIC)-III data as an particular example, we learned seven basis matrices corresponding to clinical notes, ICD-9 billing codes, prescriptions, Diagnosis-Related Groups (DRG) billing codes, current procedural terminology (CPT) procedural codes, lab tests, and lab results from the MIMIC-III dataset
We present MixEHR as an unsupervised Bayesian framework to efficiently model the distribution of heterogeneous highdimensional EHR data

Summary

Introduction

Electronic health records (EHR) are rich heterogeneous collections of patient health information, whose broad adoption provides clinicians and researchers unprecedented opportunities for health informatics, disease-risk prediction, actionable clinical recommendations, and precision medicine. EHRs present several modeling challenges, including highly sparse data matrices, noisy irregular clinical notes, arbitrary biases in billing code assignment, diagnosis-driven lab tests, and heterogeneous data types. To address these challenges, we present MixEHR, a multi-view Bayesian topic model. MixEHR disease topics reveal meaningful combinations of clinical features across heterogeneous data types. The main objective of our approach is twofold: (1) distill meaningful disease topics from otherwise highly sparse, biased, and heterogeneous EHR data; and (2) provide clinical recommendations by predicting undiagnosed patient phenotypes based on their disease mixture membership.

Methods

Results

Conclusion