A Forest Approach for Analyzing Heterogeneous Transitions of Health States: Implications for Organ Transplantation

M. Zhang and G. Wang. Under Review.

The state transition is an essential concept to model the change in states, such as a patient's health states. Although the existing literature extensively uses this concept in various decision-making models, they treat all units in a sample as a collective entity when modeling state transition. Numerous instances show that state transition may vary across units in diverse scenarios such as disease progression among patients. In this study, we first develop a new heterogeneous transition forest (called "HT forest") approach to capture the heterogeneous transitions of states across units. Then, we apply the proposed approach to an organ transplantation setting and estimate the heterogeneous transition probabilities of health states for patients awaiting organ transplantation. Our results show the transition of health states is heterogeneous across patients with different features. Finally, we develop a model of the Markov decision process to assess the value of leveraging heterogeneous transitions to facilitate personalized decision-making for patients. Our results show patients will increase their total expected life by a sizeable amount if they follow the optimal decisions derived from the patient-specific transition probabilities.