Semi-Markov Multistate Modeling Approaches for Multicohort Event History Data

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Complex event history processes require advanced statistical modeling.
• Multicohort studies present unique analytical challenges.
• Understanding COVID-19 patient trajectories necessitates robust methodologies.

Observation:

• Two Cox-based multistate modeling approaches were evaluated.
• Cohort information was incorporated as a fixed covariate or a stratum variable.
• The Markov property was assessed, with semi-Markov adjustments made when necessary.

Findings:

• Both the cohort-covariate and stratum-cohort models effectively analyzed multicohort event histories.
• The cohort-covariate approach facilitates direct estimation and interpretation of cohort-specific effects.
• The stratum-cohort approach provides greater flexibility in estimating transition probabilities across different cohorts.

Implications:

• The choice of model depends on the specific research question and inferential goals.
• These methods offer valuable tools for analyzing longitudinal health data, particularly in infectious disease research.
• The study provides insights into modeling patient pathways in the context of COVID-19 hospitalization data.