JMIR Publications

ABSTRACT

Background:

Clinical decision support systems (CDSS) have shown promise in improving diagnosis in primary care, particularly for chronic diseases. The SATURN project developed a CDSS prototype for primary care in Germany that uses artificial intelligence (AI) to reduce diagnostic uncertainty in unclear and rare diseases. It generates recommendations based on clinical data from university hospitals stored in a standardised common data model. However, integrating primary care data in Germany remains challenging due to the use of country-specific vocabularies and heterogeneous data structures. Therefore, integration of medical concepts into General Practitioners’ (GPs) user interfaces and improved workflow design is needed.

Objective:

This study investigates how the user interface of a CDSS for primary care should be designed to facilitate the user-friendly entry of medical concepts.

Methods:

A structured, iterative user-centred design process of five steps was applied: (1) conceptualisation with the analysis of requirements and objectives, (2) implementation of the CDSS user interface, (3) analysis of user and expert feedback, (4) workshop with primary care electronic health record (EHR) system experts and (5) development of an extended concept for the CDSS user interface.

Results:

The study identified requirements and options for supporting data entry in the user interface of a CDSS for primary care, thus providing a framework for future implementation and prototype development. Us-ability testing revealed the need to refine input support, particularly the language used to make the sys-tem easy for GPs to use. Further effort is required to map this work. In addition, we identified a system connection to the primary care electronic patient record systems as essential.

Conclusions:

This paper demonstrates how standardised medical terminologies can be integrated into a CDSS user interface for primary care. The user-centred design proved effective for developing CDSS user interfaces, interfaces that align with clinicians’ workflow and provide a user-friendly experience. Although the current technical infrastructure presents significant challenges when connecting to primary care EHR systems, collaboration with experts enabled us to identify potential interface solutions. Overall, the results indicate a need for optimised physician language input support and interoperability, including standardised data entry, automated tools and AI- based approaches to enhance data quality and usability.