Antiretroviral therapy (ARV) can intensely reduce HIV transmission by suppressing viral load in infected individuals. Hence, it is important to retain patients in care to ensure they are taking necessary medication, for the purposes of preserving their health and preventing new infections. We aim at using a data-driven approach to estimate the impact of starting ARV at different time of care on the retention of patient across the follow-up period.
We apply Bayesian additive regression tree to model the evolvement of patient retention status at approximately one year, and evaluate patient engagement status under certain treatment policy using Monte Carlo simulation. Comparison of engagement rate posterior distributions is given to see how patient retention differs when treatment starts at different time of care.