Evaluating the Prototype of a Clinical Decision Support System in Primary Care: Qualitative Study
Bibliographic record
Abstract
Background: General practitioners are confronted with a wide variety of diseases and sometimes diagnostic uncertainty. Clinical decision support systems could be valuable to improve diagnosis, but existing tools are not adapted to the requirements and workflow in the primary setting. In project SATURN (Smart physician portal for patients with unclear disease), the prototype of a clinical decision support system based on artificial intelligence is being developed together with users specifically for primary care in Germany. It aims to reduce diagnostic uncertainty in cases of unclear and rare diseases and focuses on 3 medical fields. A user-centered design approach is applied for prototype development and evaluation. Objective: This study evaluates the usability of a high-fidelity prototype and explores aspects of user experience like the subjective impression, satisfaction, and areas of improvement. Methods: A total of 5 general practitioners participated in the evaluation, which consisted of (1) a remote think-aloud test, (2) a postsession interview, and (3) a survey with the System Usability Scale. During the think-aloud tests, the participants verbalized their thoughts and actions and solved several vignette-based tasks. Remarkable observations were logged, transcribed with quotes, and analyzed for usability problems and positive findings. All observations and interview responses were deductively assigned to the following categories: (1) content, (2) comprehensibility, (3) user-friendliness, (4) layout, (5) feedback, and (6) navigation. Usability problems were described in detail and solutions for improvement proposed. Median and total scores were calculated for all questionnaire items. Results: The evaluation detected both strengths and areas for improvement. The participants particularly liked the clear and well-structured layout of the prototype. Key issues identified were content-related limitations, such as the inability to enter unlisted symptoms, medications, and examination findings. Also, participants found the terminology for laboratory not suitable to their needs. Another key issue was a lack of user-friendliness concerning the time required to input medication plans and lab values. Participants expressed a need for faster data entry, potentially through direct imports from practice management systems or laboratory files. Adding symptom duration, weighting symptoms, and incorporating hereditary factors were suggestions made for improvement. Overall, the SATURN prototype was deemed useful and promising for future clinical use, despite the need for further refinements, particularly in the areas of data entry, as this is a key obstacle to its use. Conclusions: The usability evaluation methods combined proved to be location independent and easy to use. They provided important findings on usability issues and improvements that will be implemented in a second high-fidelity prototype, which will also be tested by users. Technically demanding user requirements, such as direct data transfer from the practice management system and entry options that require complex data models, were beyond the scope of this project, but should be considered in future development projects.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.035 | 0.055 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.005 | 0.006 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".