Online Program Home
My Program

Abstract Details

Activity Number: 522 - Contributed Poster Presentations: Biometrics Section
Type: Contributed
Date/Time: Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
Sponsor: Biometrics Section
Abstract #304212
Title: Integrative Analysis of Irregularly Measured Biomarkers of Mixed Types in Electronic Health Records
Author(s): Jitong Lou* and Yuanjia Wang and Pengyue Zhang and Lang Li and Donglin Zeng
Companies: University of North Carolina At Chapel Hill and Columbia University and Ohio State University and Ohio State University and UNC Chapel Hill
Keywords: Electronic health records; Generalized linear models; Method of moments; Latent process; Kernel smoothing; Type 2 Diabetes

Electronic health records (EHRs) systematically capture patients' health information in a digital format. To accurately characterize individual disease comorbidity and susceptibility, it is necessary to integrate information of biomarkers over time. However, due to observational data nature, the biomarkers measured over time are multivariate with mixed types. In addition, for each patient, the biomarker data are sparsely measured and the measurement times are irregular and potentially informative. In this paper, we propose multivariate generalized linear models to model different types of biomarkers. We also incorporate latent multivariate Gaussian processes to account for between-biomarker dependence over time. We allow the covariate effects and covariance matrix of the latent processes to be time-dependent. For parameters' estimations, we apply kernel-weight estimating equations based on the first and second moments of the assumed biomarker processes, where our kernel weights account for the heterogeneous intensity of the measurement times. We illustrate our method through extensive simulation studies and then apply our method to analyze an EHR for Type 2 Diabetes (T2D) patients.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program