Estimation of multiple treatment effects in observational survival data is complicated due to confounding, heterogeneity, and selection bias. To address this, we define individual treatment effect (ITE) and average treatment effect (ATE) directly through comparison of survival under counterfactual treatment assignments. Our requirement is overlap, which permits only individuals who are eligible for treatments to be included in such comparisons. Because not all treatments will have clearly defined eligibility criteria, we propose new random forest methods to estimate individual treatment eligibility. These methods possess the unique feature of being able to incorporate external expert knowledge either in a fully supervised way (i.e., we have a strong belief that knowledge is correct), or in a minimally-supervised fashion (i.e., knowledge is not considered gold-standard). We directly estimate ITE using an extension to random survival forests we call virtual twin random survival forest interaction. Motivation arose from the problem of developing treatment decision strategies for ischemic cardiomyopathy using a large data set comparing four well established therapies.