Online Plausibility Checks for Patient Pathways: Medical ontologies in online process mining.

Problem Statement. A patient pathway describes all events recorded during
the treatment of one specific patient. In the field of PODS4H, data quality issues
(e.g., completeness, timeliness, or plausibility) remain a major challenge. In
an online setting, the complete-
ness of the patient pathway is unknown. Hence, other quality checks, e.g., the
verification of guideline compliance, become more arduous.
Objective. In this poster, we discuss how to measure the plausibility of an online
patient pathway, i.e., an ongoing patient encounter. We want to validate if the
currently observed pathway remains plausible whenever new events are recorded.
Approach. We utilize the Shapes Constraint Language (SHACL3) to contin-
uously validate patient pathways. The pathway, that is, the online process in-
stance, is modelled as continuously growing Resource Description Framework
(RDF4) data graph. This data graph is validated against a shapes graph, a
publicly available Web Ontology Language (OWL5)-based medical ontology. We
plan to use SNOMED-CT or parts of it as shapes graph. This approach is based
on previous work in the industrial domain.
Discussion. We discuss this vision and expected challenges with the PODS4H
community, specifically: (1) how to model the data graph, i.e., the online process
instance, (2) how to construct generic SHACL constraints, and (3) whether the
expressiveness of SNOMED-CT is sufficient for this use case.