Use of Markov models to estimate transition probabilities in a follow up study

Abstract

In a follow up study to assess treatment of BV. 207 female adults presenting with BV were treated. Follow up information was obtained at one week. one month. two months and three months. One outcome measurement at each follow up was BV status. Since BV status could be influenced by various habits and practices. the outcome is analyzed by transition first order Markov model. We relate transition probabilities to covariates and previous outcomes through a logistic regression model with Markov structure. This approach allows us to assess not only the baseline characteristics but also the effects of past outcomes in analyzing longitudinal binary data.