Latent Class Analysis with Semi-parametric Proportional Hazards Submodel for Time-to-event Data

TL;DR

This paper introduces a semi-parametric latent class analysis method for time-to-event data, providing validated inference for covariate effects and hazards, with improved prediction over Cox models, demonstrated through simulations and real data.

Contribution

The paper develops a novel NPMLE-based latent class analysis approach with validated inference for heterogeneity in time-to-event data, enhancing predictive accuracy.

Findings

Improved predictive performance over Cox regression.

Validated inference procedures for covariate effects.

Successful application to MCI cohort data.

Abstract

Latent class analysis (LCA) is a useful tool to investigate the heterogeneity of a disease population with time-to-event data. We propose a new method based on non-parametric maximum likelihood estimator (NPMLE), which facilitates theoretically validated inference procedure for covariate effects and cumulative hazard functions. We assess the proposed method via extensive simulation studies and demonstrate improved predictive performance over standard Cox regression model. We further illustrate the practical utility of the proposed method through an application to a mild cognitive impairment (MCI) cohort dataset.