A decision-theoretic planning approach for clinical practice guideline modelling

Current formalisms for modelling a Clinical Practice Guideline (CPG) as a Computer-Interpretable Guideline (CIG) do not support the explicit representation of uncertainty due to the unpredictability of treatment outcomes and the incomplete knowledge of the actual patient condition. Given this limitation, the existing CIG approaches support only the computation of patient-specific diagnostic and treatment recommendations for the immediate next step in the patient care-pathway from the clinical perspective. Some existing approaches allow the computation of patient-specific care plans accounting for resource and time constraints, however they do not explicitly model outcomes and clinical data uncertainty. In this paper we propose a decision-theoretic planning approach to modelling and automatic computation of patient-specific care plans that addresses these limitations, whereby adherence to CPGs is maximised given specific, but uncertain, patient data and clinical outcomes. The premise of this work is that both patient and clinician decision making are better modelled and supported by computing care plans instead of single point-of-care recommendations. In this paper we introduce and motivate the clinical problem, present a formal mathematical specification of the approach and we compare our approach with existing ones from the point of view of CIG modelling. Finally we outline further research and discuss aspects that would prove crucial for the implementation of the approach.