Towards an AI Planning-Based Pipeline for the Management of Multimorbid Patients

Abstract

Treatment of patients with multimorbidity is one of the greatest challenges for clinical decision support. While evidence-based management of specific diseases is supported by clinical practice guidelines, concurrent application of multiple guidelines requires checking for possible adverse interactions between interventions and mitigating them, before a management plan is constructed. In earlier work, we developed an approach that casts the problem of multimorbidity management as an AI planning problem. In this paper we build on this earlier work and make progress towards creating a pipeline that inputs disease and patient-specific information and outputs a management plan. We describe research focused on selected aspects of pipeline development and illustrate these aspects with a clinical case implemented using the PDDL planning language and the OPTIC planner.