Bayesian treatment effects due to a subsidized health program: the case of preventive health care utilization in Medellín (Colombia)