Online Program Home
My Program

Abstract Details

Activity Number: 344 - Semiparametric Modeling
Type: Contributed
Date/Time: Tuesday, July 31, 2018 : 10:30 AM to 12:20 PM
Sponsor: Biometrics Section
Abstract #330373 Presentation
Title: Sample-Weighted Semiparametric Estimates of Cause-Specific Cumulative Incidence Using Left-/Interval Censored Data from Electronic Health Records
Author(s): Noorie Hyun* and Hormuzd A. Katki and Barry Ira Graubard
Companies: Medical College of Wisconsin and Biostatistics Branch, Division of Cancer Epidemiology & Genetics, National Cancer Institute and National Cancer Institute
Keywords: Competing risks; Interval censoring; Stratified random sample; Semiparametric models; Bootstrap; Electronic health records

Electronic health records (EHRs) are cost-effective data source for developing disease-specific risk models while accounting for competing outcomes. However, there are challenges in applying existing risk models to EHRs: left-censoring of prevalent disease, interval-censoring of incident disease and uncertainty of disease prevalence when definitive disease ascertainment is not conducted at baseline. Furthermore, expensive or novel bio-assay tests cannot be conducted on everyone but only on stratified random subsamples from EHRs with varying sampling fraction. We propose a family of semiparametric mixture models for estimating cause-specific cumulative risk/incidence conditional on covariates. We adopt the iterative convex minorant algorithms to nonparametrically estimate cumulative incidence functions of competing outcomes and employ a smoothed bootstrap sample-weighted procedure for consistent confidence interval estimation of a cumulative risk/incidence, of which the asymptotic distribution is not normal. We applied the proposed method to stratified subsamples from EHRs at Kaiser Permanente Northern California to calculate cumulative incidences for competing outcomes.

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

Back to the full JSM 2018 program