Stratified regression analysis of zero-truncated recurrent event data

Anqi A. Chen, X. Joan Hu, Rhonda J. Rosychuk

The Annals of Applied Statistics · 2026 · Volume 20 · Issue 1

Abstract

This paper is motivated by a pediatric mental health care (PMHC) program, which extracted the records of mental health-related emergency department (MHED) visits from population-based administrative databases during 2011–2017. Only information on the subjects with MHED visit experiences is available within a subject-specific time window. We focus on one of the program objectives: understanding how the visit occurrence is associated with the subject’s past as well as their demographic and geographic exposures in the entire population. The available collection of the MHED records is framed as zero-truncated recurrent event data. We introduce an innovative stratified Cox regression model for the event process. The model is intensity-based but requires only a summary of the event history. We propose a new procedure for estimating the model parameters using the zero-truncated data integrated with some relevant population census information. We establish the consistency and asymptotic normality of the proposed estimator and examine its finite sample performance via extensive simulation in contrast with the maximum likelihood estimation based on the zero-truncated data only. The MHED data from the PMHC program are employed to illustrate the proposed approach throughout the paper.

Keywords

conditional intensity function; doubly censored recurrent event data; health administrative data; partially known stratification; supplementary information;

Enable JavaScript for citation tools and additional paper interactions.