Assessing Treatment Effects Overtime on Mortality of Opioid Use Disorders Under a Generalized Cox Proportional Hazards Model

• Activity Number: 349 - Lifetime Data Science Student Awards
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Lifetime Data Science Section
• Author(s): Trevor Thomson* and Joan X. Hu and Bohdan Nosyk
• Companies: Simon Fraser University and Simon Fraser University and Simon Fraser University
• Keywords: Change point analysis; Opioid agonist treatment; Stratified Cox regression model; Time-dependent covariates; Time-dependent stratification

Abstract

Preliminary studies indicate that retention on opioid agonist treatment can result in substantial reductions in the risk of mortality for people with opioid use disorder. This protective effect may vary by stages of opioid agonist treatment use. This motivated us to propose a survival model where some model components vary according to treatment history. An example is the time-dependent stratified Cox proportional hazards model, which includes the treatment status as a covariate. Since such a covariate is time-dependent and internal, likelihood/partial likelihood methods do not directly apply. We provide an estimating function based procedure for estimating the model parameters. A pseudo-score test statistic is proposed to test hypotheses on the model parameters. This testing procedure can be utilized in a sequential fashion to finalize the strata of individuals. We examine the proposed approach both asymptotically and numerically via simulation. We illustrate the proposed approach to an administrative dataset capturing opioid agonist treatment dispensations and deaths in British Columbia, Canada. This is joint work with X. Joan Hu (SFU) and Bohdan Nosyk (St. Paul's Hospital, SFU)