Missclassification of HIV Disease Stages with Continuous Time Hidden Markov Models

Abstract

The purpose of this study is to explore the simple Markov and Hidden Markov models with continuous time to investigate disease progression of HIV/AIDS patients under ART follow-up at Shashemene Referral Hospital, Ethiopia. The msm R package is used for the analysis. Results from the simple Markov model reveals that the disease progression of the HIV/AIDS patients considered tend to move towards the healthier than the worse state. The mean waiting time for the healthiest state is significantly higher than the other transient states. The total length of time stay in a state declines with severity of the disease stages. Analysis of the misclassification model provides transition rates of the true states. Estimation of the transition rates of the true states are found to be relatively smaller compared to those obtained by the simple Markov model. For the true states compared to observed ones, the conditional probability of moving to the healthiest state from the next worse state grows higher dramatically, while that of moving to next worst state grows slightly lower. The ART based patient care might have positive impacts on the overall progression of the disease. For covariate effects, male patient is more likely to move to worse state than the female does. But age of patient is not significant. The progression of the underlying states of the HIV/AIDS disease behaves similar to that of the generated markers observations except the turning points of the conditional probabilities. The turning points so interesting for be studied further.