Continuous-time probabilistic models for longitudinal electronic health records

Alan D Kaplan; Uttara Tipnis; Jean C Beckham; Nathan A Kimbrel; David W Oslin; Benjamin H McMahon; MVP Suicide Exemplar Workgroup

doi:10.1016/j.jbi.2022.104084

Continuous-time probabilistic models for longitudinal electronic health records

J Biomed Inform. 2022 Jun:130:104084. doi: 10.1016/j.jbi.2022.104084. Epub 2022 May 7.

Authors

Alan D Kaplan¹, Uttara Tipnis², Jean C Beckham³, Nathan A Kimbrel⁴, David W Oslin⁵, Benjamin H McMahon⁶; MVP Suicide Exemplar Workgroup

Affiliations

¹ Computational Engineering Division, Lawrence Livermore National Laboratory, 7000 East Ave., Livermore, CA 94550, USA. Electronic address: kaplan7@llnl.gov.
² Computational Engineering Division, Lawrence Livermore National Laboratory, 7000 East Ave., Livermore, CA 94550, USA.
³ Durham Veterans Affairs (VA) Health Care System, Durham, NC, USA; VA Mid-Atlantic Mental Illness Research, Education and Clinical Center, Durham, NC, USA; Department of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Durham, NC, USA.
⁴ Durham Veterans Affairs (VA) Health Care System, Durham, NC, USA; VA Mid-Atlantic Mental Illness Research, Education and Clinical Center, Durham, NC, USA; Department of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Durham, NC, USA; VA Health Services Research and Development Center of Innovation to Accelerate Discovery and Practice Transformation, Durham, NC, USA.
⁵ VISN 4 Mental Illness Research, Education, and Clinical Center, Center of Excellence, Corporal Michael J. Crescenz VA Medical Center, Philadelphia, PA, USA; Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, PA, USA.
⁶ Theoretical Biology and Biophysics, Los Alamos National Laboratory, Los Alamos, NM, USA.

PMID: 35533991
DOI: 10.1016/j.jbi.2022.104084

Abstract

Analysis of longitudinal Electronic Health Record (EHR) data is an important goal for precision medicine. Difficulty in applying Machine Learning (ML) methods, either predictive or unsupervised, stems in part from the heterogeneity and irregular sampling of EHR data. We present an unsupervised probabilistic model that captures nonlinear relationships between variables over continuous-time. This method works with arbitrary sampling patterns and captures the joint probability distribution between variable measurements and the time intervals between them. Inference algorithms are derived that can be used to evaluate the likelihood of future using under a trained model. As an example, we consider data from the United States Veterans Health Administration (VHA) in the areas of diabetes and depression. Likelihood ratio maps are produced showing the likelihood of risk for moderate-severe vs minimal depression as measured by the Patient Health Questionnaire-9 (PHQ-9).

Keywords: Electronic health records; Mixture models; Probabilistic models; Time-dependent modeling.

Publication types

Research Support, Non-U.S. Gov't
Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

Algorithms
Electronic Health Records*
Humans
Machine Learning*
Models, Statistical
Probability