Odds ratio estimation in a steady-state population.

A three-state Markov process is used to model a steady-state population. The classical P = ID equation is derived, as well as a number of identities originally reported by Miettinen (Am J Epidemiol 1976; 103: 226-235) and Freeman and Hutchison (Am J Epidemiol 1980; 112: 707-723. Ibid. 1986; 124: 134-149). Using theoretical and empirical methods it is demonstrated that the odds ratio derived from a prevalence-prevalence case-control study conducted in a steady-state population will generally be a biased estimate of the incidence density ratio. However the bias will not likely be of practical importance unless the incidence density ratio is at least 10 or more.