Interpretable AI in EHR-Based Clinical Decision Support: A Scoping Review of Models, Methods, and Trends
Bibliographic record
Abstract
Background: Electronic Health Records (EHRs) generate vast volumes of structured and unstructured patient data, and artificial intelligence (AI) is increasingly proposed to convert that data into actionable clinical insight. Yet the opacity of many AI models — their so-called "black box" character — continues to limit clinician trust and real-world adoption, which is precisely why interpretable, or explainable, AI has drawn so much recent attention. Aims: This scoping review set out to map how AI models and interpretability methods have been applied within EHR-based Clinical Decision Support Systems (CDSS), and to characterize the clinical domains, countries, and outcomes represented in this literature. Methods: Following PRISMA-ScR guidance, we searched PubMed, Scopus, Web of Science, and IEEE Xplore for peer-reviewed studies published between 2018 and 2023. After screening, 36 studies met inclusion criteria and were charted for AI model type, interpretability method, clinical domain, country, sample size, and reported outcome. Results: Neural networks were the most frequently applied model (13 studies), followed by random forest (9), XGBoost (8), and logistic regression (6). SHAP was the dominant interpretability method (15 studies), ahead of attention-based visualization (12) and LIME (9). Cardiology was the most studied domain (10 studies), and contributions were evenly spread across the USA, Canada, the UK, India, Germany, and Australia (6 each). Sample sizes ranged from 430 to 1,350 patients. Conclusion: Interpretable AI is steadily, if unevenly, being woven into EHR-based CDSS research. Closing the gap in underrepresented domains and standardizing how interpretability itself is evaluated remain the field's clearest next steps.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.003 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 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 teacher head, 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".