
Multi-State Models for Disease Natural History 387
18.5 Discussion
Natural histor y of disease can be modeled using a variety of appr oaches that
fall under the general framework of multi-state models , including Markov pro -
cesses, nonhomogeneous Markov processe s, semi-Markov processes, and hid-
den Markov processes. Simple a pproaches such as the Markov process facili-
tate estimation but make use of strong assumptions that may b e unrealis tic
for certain diseases. More complex models such as se mi- Markov processes may
more accurately describe the disease proces s, but at the cost of substantially
complicating estimation, especially when data