Expert Systems in Digital Health

Thanks to its technical superiority, the systems’ logic and inner workings are structured in a way that each result can be formally proven by tracing back its logical steps up to the source material. This interpretable reasoning allows the user to verify and understand the produced outcome. 

1. Clinical Decision Support System

Ergobyte has developed an AI-powered service that offers explainable, legally sound medication decision support in real time. The system recommends suitable medications based on patient’s medical conditions and cross-checks them for adverse reactions. It is able to check drug-to-drug and drug-to-disease interactions, suggested medications per disease, detailed treatment posology and relevant advisories in order to produce personalized medical recommendations.

The CDSS works on a knowledge base where all the useful information is stored. Medical rules are created taking into account recent literature and official Summary of Product Characteristics (SmPC) documents. Applicable medication treatments are selected by active ingredient and then adapted to the market’s brand names and specific package concentrations.

Its key features are the following: 

• Rich knowledge base of: 25.000+ rules (based on EMA & FDA marketing authorisations), 2.900 active ingredients, 7.600 package inserts (SmPCs), 40.500 international brand names

• Best-in-class medication safety checks (drug-to-drug/-disease/-food), with high value in cases of polypharmacy and multimorbidity

• Adaptability to specific cases, e.g. psychiatry, palliative care, ICU, rehabilitation, patient adherence

• Service accessible via web/mobile/APIs

Reference: Doulaverakis, Charalampos & Nikolaidis, George & Kleontas, Athanasios & Kompatsiaris, Ioannis. (2012). GalenOWL: Ontology-based drug recommendations discovery. Journal of biomedical semantics. 3. 14. 10.1186/2041-1480-3-14

2. Workflow automation

With its workflow automation component, Ergobyte employs Business Process Modeling Notation (BPMN), an open, well-accepted industry standard, enhances it with custom-built modules that use machine learning to adapt decisions, and couples it with openEHR, an open standard for the structuring of data in electronic health records. 

The system addresses a wide range of needs arising within a healthcare or social facility, such as ward management, therapeutic protocols, nursing care plans, call center flowcharts, etc. Any procedure can be formalized into executable diagrams, with benefits stemming from pervasive decision support and implementation transparency.

Reference: P. Natsiavas, T. G. Stavropoulos, A. Pliatsios, H. Karanikas, G. I. Gavriilidis, V. K. Dimitriadis, G. Nikolaidis, S. Nikolopoulos, P. Skapinakis, E. Thireos and I. Kompatsiaris, “Using Business Process Management Notation to Model Therapeutic Prescription Protocols: the PrescIT Approach”, Medical Informatics Europe (MIE) 2021, Poster Presentation, 29-31 May 2021