The partially observable Markov decision processes in healthcare: an application to patients with ischemic heart disease (IHD)

Partially observable Markov Decision Process POMDP have been suggested a most popular tool to express sequential decision making under uncertainty. The generality of the standard POMDP model, however, limits practical application of the framework due to the computational complexity of associated solution methods. To alleviate this obstacle we have proposed a specialized POMDP form and our goal is to determine an optimal or nearly optimal policy for the treatment of patients with IHD. This well constructed POMDP model has practical advantage over clinical studies (no risk for the life of patients and low cost), though the latter are necessary to perform before making significant changes to treatment guidelines.