A discrete survival model was used
to evaluate the probability of future disability onset adjusted for age, sex, and covariates
representing the current presence of other disabilities in activities of daily living. The
survival model is derived
from a discrete hazard rate. which is the probability that a disability-free person will
develop disability within 2 years.30.3 1
The discrete hazard rate (DH) is analogous to the Cox proportional hazard rate for continuous
data and has the form
where Xk is a covariate measured at interview k, DHk+ 1 is the discrete hazard in the subsequent
2-year interval, and b is the regression coefficient. Advantages of a discrete hazard model include
use of longitudinal information on individuals up to the time of interview, when incident
disability is reported, and incorporation of time-dependent covariates. The discrete hazard model
was estimated by means of
the GUM package32 to fit a generalized
linear model with a complementary log log link. Binary variables were created for each activity
to indicate whether or not