Model-driven execution of phenotype algorithms – introduction of the Terminology- and Ontology-based Phenotyping Framework

Abstract

Automated detection of patients who meet certain clinical criteria or who experience adverse events plays a key role in improving patient care. Rule-based systems are one approach to enable standardised modelling of algorithms to detect patients with certain observable characteristics (phenotypes). Another important aspect is the retrieval of data from electronic health records in hospital information systems and clinical research database systems. In this work, we propose a new rule-based system, suitable for domain experts without IT background that allows automatic retrieval of patient data on execution by translating search queries into source-specific query languages.As a basis for rule-based structured phenotype algorithms, we use a model derived from the Core Ontology of Phenotypes. The model allows phenotype algorithms to be defined with properties such as code system relationships to improve interoperability. With a relatively simple mapping, IT specialists can provide a specification to connect these models to a local hospital information system.We developed an interactive web application, the Terminology- and Ontology-based Phenotyping Framework, to support domain experts in modelling and executing phenotype algorithms. Depending on the data-holding source system, an appropriate adapter is used to translate and execute queries in the source system-specific language. The framework was tested against a SQL database and a FHIR server, both initialised with randomly generated data. Generated queries yielded identical result sets.The proposed computer-assisted approach can improve disease detection and clinical trial recruitment and enables a clear division of tasks between domain experts and IT specialists.

Keywords

• phenotype
• algorithms
• eligibility determination
• health information interoperability
• terminology as topic