Deep Learning Models to Predict Diagnostic and Billing Codes Following Visits to a Family Medicine Practice: Development and Validation Study
Bibliographic record
Abstract
Background: Despite significant time spent on billing, family physicians routinely make errors and miss billing opportunities. In other disciplines, machine learning models have predicted Current Procedural Terminology codes with high accuracy. Objective: Our objective was to derive machine learning models capable of predicting diagnostic and billing codes from notes recorded in the electronic medical record. Methods: We conducted a retrospective algorithm development and validation study involving an academic family medicine practice. Visits between July 1, 2015, and June 30, 2020, containing a physician-authored note and an invoice in the electronic medical record were eligible for inclusion. We trained 2 deep learning models and compared their predictions to codes submitted for reimbursement. We calculated accuracy, recall, precision, F1-score, and area under the receiver operating characteristic curve. Results: Of the 245,045 visits eligible for inclusion, 198,802 (81%) were included in model development. Accuracy was 99.8% and 99.5% for the diagnostic and billing code models, respectively. Recall was 49.4% and 70.3% for the diagnostic and billing code models, respectively. Precision was 55.3% and 76.7% for the diagnostic and billing code models, respectively. The area under the receiver operating characteristic curve was 0.983 for the diagnostic code model and 0.993 for the billing code model. Conclusions: We developed models capable of predicting diagnostic and billing codes from electronic notes following visits to a family medicine practice. The billing code model outperformed the diagnostic code model in terms of recall and precision, likely due to fewer codes being predicted. Work is underway to further enhance model performance and assess the generalizability of these models to other family medicine practices.
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 | 0.003 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| 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".