Artikel
Feasibility Study of Using a Guideline Derived Expert System for Transfusion Indication
Suche in Medline nach
Autoren
Veröffentlicht: | 19. August 2022 |
---|
Gliederung
Text
Introduction: Blood donations in Germany in 2020 fell by 2.7% compared to 2019. This is around 1.3 million fewer donations than in 2011 [1]. Due to this steady decline, efficient handling and use of blood products is essential to compensate this gap. The AutoPiLoT [2] project therefore aims to reduce waste of blood products by using guideline-based transfusion indication to impede unnecessary orders and by better prediction of blood product demand, the first part being described within this paper.
Method: To develop a guideline-compliant recommendation system, first the Cross-Sectional Guidelines for Therapy with Blood Components and Plasma Derivatives of the German Medical Association [3] needed to be converted into a decision tree to make it operable. The resulting algorithm was then transferred into a prototype with a web-based user interface and tested on a ward of the department hematology of the University Hospital Essen. Within the test period, blood transfusions were administered in 50 cases, for which we retrospectively compared the recommendation of the prototype based on the patients' health parameters with an expert’s actual decision. The resulting deviations were assessed with the help of a questionnaire.
Results: The expert system prototype achieved a 100% accuracy and specificity for packed red blood cells and a 100% accuracy but 63% specificity for platelet concentrates. This shows a high applicability for guideline-based recommendations. The few deviations were very patient-specific and not due to an error in the decision algorithm. It could be integrated to the hospital information system.
Discussion: A translation of guidelines into a recommendation algorithm was found to be feasible and indicated to correspond to real-life decisions in everyday medical practice. Therefore, a subsequent trial on a larger scale is needed to investigate possible improvements on guideline compliant ordering. Automation of data extraction and notifications is prepared using FHIR and will be implemented.
The authors declare that they have no competing interests.
The authors declare that a positive ethics committee vote has been obtained. Ethic Committee: Medizinische Fakultät der Universität Duisburg-Essen - 20-9386-BO.
References
- 1.
- Henseler O. Tabellen Gewinnung, Herstellung, Import, Export und Verbrauch 2019 und Auswertungen über mehrere Jahre. Langen: Paul-Ehrlich-Institut; 2020 Oct 26. Available from: https://www.pei.de/DE/newsroom/pflichtberichte/21tfg/21-tfg-berichte-node.html
- 2.
- Intelligente Zuordnung von Blutprodukten (AutoPiLoT). Bundesgesundheitsministerium; 2022 [cited 2022 May 14]. Available from: https://www.bundesgesundheitsministerium.de/ministerium/ressortforschung-1/handlungsfelder/forschungsschwerpunkte/digitale-innovation/modul-2-smarte-datennutzung/autopilot.html
- 3.
- Bundesärztekammer, editor. Querschnitts-Leitlinien zur Therapie mit Blutkomponenten und Plasmaderivaten. 5. aktualisierte und überarbeitete Aufl. Köln: Deutscher Ärzteverlag; 2021.