A Framework for the Detection of Conflicts between Clinical Pathways using Constraint Solvers

Clinical pathways (CPs) use flowchart-like notation to present the recommended treatment steps for patients suffering from single diseases. Increasingly many patients suffer from multiple concurrent conditions (multimorbidity), leading to potential conflicts when multiple pathways are followed in parallel, and the need for automated methods to detect and resolve such conflicts.

Formal notations have been proposed for modelling CPs. BPMN+V, based on Business Process Model and Notation (BPMN) and endowed with a formal syntax and semantics, allows CPs to be rigorously specified and analysed with respect to their interaction with data. Using a state-space exploration method the effectiveness of BPMN+V has been demonstrated for modelling CPs and discovering conflicts between them. Such methods are however potentially inefficient and unsuited to richer models (e.g. temporal or stochastic) or computational search or optimisation to resolve the discovered conflicts.

We describe a specification of the syntax and semantics of BPMN+V as sets of constraints which allows efficient modelling and conflict detection using constraint solvers, and also provides a flexible foundation for future modelling and optimisation. We implement a prototype of this approach using the Z3 SMT solver and demonstrate its effectiveness to discover conflicts between clinical pathways for several major morbidities. We outline how the method could be extended to optimise over CPs.