Accuracy of preferred language data in a multi-hospital electronic health record in Toronto, Canada
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
Accurate preferred language data is a prerequisite for providing high-quality care. We investigated the accuracy of preferred language data in the electronic health record (EHR) of a large community hospital network in Toronto, Canada. We conducted a point-prevalence audit of patients admitted to intensive care, internal medicine, and nephrology services at three hospitals. We asked each patient "What is your preferred language for health care communication?" and reported on agreement (with 95% confidence intervals [CI]) between interview-based and EHR-based preferred language. We used Bayesian multilevel logistic regression to analyze the association between patient factors and the accuracy of the EHR for patients who preferred a non-English language. Between June 17, 2024, and July 19, 2024, we interviewed 323 patients, of whom 124 (38%) preferred a non-English language. Median age was 77 years and 46% were female. EHR accuracy was 86% for all patients. The probability of the EHR correctly identifying a patient with non-English preferred language (sensitivity) was 69% (CI 60-77), specificity was 97% (CI 94-99), positive predictive value was 95% (CI 88-98), and negative predictive value was 83% (CI 79-87). There were 26 different non-English preferred languages, most commonly Cantonese (27%) and Tamil (14%). Accuracy was better for patients who were female or older, and varied by hospital and medical service. Mechanisms to improve accuracy for language preference data are needed to improve the validity of research studying preferred language, mitigate algorithmic bias, and overcome language-based inequities.
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 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| 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 it