Fig. 2.
Pictorial representation of the multi-state model. There are six transition specific hazards, qrs(t), where r and s are contained within the state set (1 no impairment, 2 slight impairment, 3 moderate/severe impairment, 4 death) and t represents time. The three cognitive lifestyle covariates plus sex and the time-dependent variable age are linked to each hazard via log-linear regression