MBF4CR: A model-based framework for supporting an automated cancer registry system

The Cancer Registry of Norway (CRN) collects medical information (e.g., laboratory results, clinical procedures and treatment) of cancer patients from different medical entities, for all cancer patients in Norway. The collected data are checked for validity and correctness (i.e., validation) and is the basis for the registration of cancer cases (i.e., aggregation) by employing more than a thousand of medical rules. However, the current practice of CRN lacks of a systematic way to capture the domain knowledge and maintain medical rules at a proper level of abstraction. To tackle these challenges, this paper proposes a model-based framework (named as MBF4CR) for capturing the domain knowledge, formalizing medical rules, automating rule selection, and enabling data (cancer messages and cancer cases) validation and aggregation using Unified Modeling Language (UML) and Object Constraint Language (OCL). MBF4CR systematically captures domain knowledge (e.g., cancer messages) as an UML class diagram and formally specifies medical rules as OCL constraints. By associating tags to OCL constraints, MBF4CR enables an automated rule selection process with tool support. We employed a case study from CRN that consists of 187 medical rules to evaluate MBF4CR from two aspects: Performance in terms of selecting and executing rules, and Correctness in terms of producing correct validation and aggregation results. Results show that MBF4CR can facilitate the current practice by complying with the medical domain knowledge with an acceptable performance, while reducing the maintenance effort.