Using language models to integrate clinical decision support and note taking: a qualitative study
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
<bold>Introduction:</bold> Prior to designing a novel user interface which would integrate clinical decision support from the Acute COPD Exacerbation Prediction Tool (ACCEPT) with physician note-taking, we conducted an exploratory qualitative study to understand physician attitudes, experiences, and expectations for this kind of functionality. <bold>Methods:</bold> We conducted semi-structured interviews with practicing physicians recruited through convenience sampling from five teaching hospitals in Canada and the US. We asked questions about how formal risk assessment is used, experience and trust in prediction models, EHR integration, and dictation practices. We qualitatively analyzed interview transcripts using affinity diagrams to produce themes highlighting relevant topics. <bold>Results:</bold> We interviewed ten physicians (70% female, 4 respirologists, 4 cardiologists, 2 residents) whose clinical experience ranged from 2 to 38 years. Our affinity analysis revealed five high-level themes. All participants were routinely but infrequently using prediction models that required manual data entry outside of the EHR system, mostly to determine malignancy risk for pulmonary nodules, pre-operative mortality risk assessment, or statins eligibility. Reasons for using models included risk communication with patients or colleagues, and automating mundane tasks. Frustration with data organization, retrieval, and letter writing in the EHR system was common. <bold>Conclusions:</bold> Participants were receptive to decision support software interfaces that would simplify data retrieval and letter writing, but were also willing to accept some inconvenience to use a truly useful risk score, especially if they only used it occasionally.
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.
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 | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 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 it