Landmark Multi-State Modeling for Complex Disease Trajectories
Analyzing complex disease trajectories using multi-state models (MSMs) requires handling both complex event structures and the computational demands of large-scale registries. We introduce an MSM framework extending beyond simple illness-death structures to model time-to-recurrence (relapse or reinfection) by integrating data from initial events.
We address two challenges: defining recurrence at fixed post-first infection intervals via landmark constraints, and managing a population-level sample of 400,000 individuals, a scale exceeding conventional tools like MSMpred. Applied to COVID-19 data in the Basque Country, our methodology utilizes Cox cause-specific hazard models. While the initial model uses population registries, the second is restricted to individuals at risk after a specific threshold. We propose and compare several approaches incorporating baseline and time-dependent covariates at the landmark time. This framework is adaptable to cancer and other infectious diseases.
Palabras clave: Disease recurrence Landmark analysis arge-scale registries Multi-state models Survival Analysis