The progression of an illness can be modelled with a three-state model: hea
lthy, ill, dead. The model used in this article contains no reversible tran
sitions: it is fully specified hv the transition intensities between the di
fferent states. Such an approach is of value in epidemiology, because the t
ransition intensities to the dead state represent the instantaneous death r
ates for the ill and healthy, while the transition intensity from the healt
hy state to the sick state corresponds to the instantaneous incidence of th
e illness, Smoothed estimates are given for the different transition intens
ities between the different states; the probability of occupying a given st
ate at a given age can be inferred from the intensities, and the prevalence
of the illness is calculated by the ratio between the probability of being
ill and the probability of being alive. This method is applied here to dat
a from the PAQUID study to describe the evolution of serious dependency amo
ng elderly people over 65 living in southwest France. The prevalence of ser
ious dependency increases with age for both sexes. The agr-related increase
in the intensities of the transition to dependency is faster for women tha
n for men.