|
Activity Number:
|
80
- Contributed Poster Presentations: Mental Health Statistics Section
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Mental Health Statistics Section
|
|
Abstract #312904
|
|
|
Title:
|
Addressing Informative Observation Times and Dependent Censoring in Electronic Health Records Data
|
|
Author(s):
|
Alexandra Klomhaus* and Hilary Aralis and Catherine A. Sugar and Jessica Jeffrey and Mark Grossman and Patricia Lester
|
|
Companies:
|
University of California, Los Angeles and UCLA Department of Biostatistics, Semel Institute for Neuroscience & Human Behavior and University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles
|
|
Keywords:
|
Informative observation times;
Dependent censoring;
Longitudinal data analysis;
Shared random effects model;
Electronic health records
|
|
Abstract:
|
Medical records give rise to complicated data structures, due to the tendency of both patients to seek and end treatment based on how they feel, and physicians to determine and update treatment schedules in response to patients’ symptoms. The longitudinal evaluation of a medical outcome (“repeated measure”) when frequency of visits to a health provider (“informative observations”) and treatment termination (“dependent censoring”) are associated with that outcome can lead to biased parameter estimates when such dependencies are ignored. Building on prior research demonstrating the utility of correlated random effects, our proposed joint model evaluates patient symptoms over time while accounting for informatively-timed observations and censoring. It incorporates several extensions to make it broadly suitable for electronic health record (EHR) data and can be implemented in standard statistical software. We apply this model to EHR data obtained from Behavioral Health Associates, a UCLA Health primary and behavioral health collaborative care system, to model depression symptom trajectories and identify factors affecting rates of improvement.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2020 program
|
