AI-Driven Injury Reporting in Pediatric Emergency Departments
Bibliographic record
Abstract
Importance: Injury is a leading cause of morbidity and mortality among children worldwide. Prevention strategies rely on timely and accurate injury surveillance. Many national programs, including the Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP), depend on manual review of emergency department (ED) medical records to track injury trends. Rising ED volumes have strained manual processes, delaying data analysis and compromising early detection of public health risks. Objective: To evaluate whether natural language processing (NLP) transformer models can automate injury case detection in ED medical records, improving CHIRPP workflow efficiency while maintaining high sensitivity. Design, Setting, and Participants: Prognostic study of ED visits from January 1, 2017, to December 31, 2023, at The Hospital for Sick Children, a high-volume tertiary pediatric referral center in Toronto, Canada, and a core CHIRPP site. The dataset included pediatric ED visits across all age groups. All medical records were labeled as requiring or not requiring CHIRPP reporting, with no exclusions. Two transformer-based NLP models, DistilBERT-base-uncased (model 1) and BERT-large-uncased (model 2), were fine tuned using supervised learning to classify medical records as CHIRPP-reportable or not. Exposure: Application of fine-tuned NLP transformer models to routine ED visit data to automate classification of injury-related cases for surveillance reporting. Main Outcomes and Measures: Outcomes included true positive rate (TPR), true negative rate (TNR), false positive rate (FPR), false negative rate (FNR), area under the receiver operating characteristic curve (AUROC), and area under the precision-recall curve (AUPRC). Additional outcomes included the proportion of medical records requiring manual review to achieve 90% sensitivity. Results: Among 217 173 pediatric ED visits across all age groups, model 1 achieved an AUROC of 0.983, AUPRC of 0.932, TPR of 0.90, TNR of 0.99, FPR of 0.014, and FNR of 0.10. Model 2 showed similar performance with an AUROC of 0.983, AUPRC of 0.931, TPR of 0.90, TNR of 0.99, FPR of 0.012, and FNR of 0.09. Both models identified 90% of injury cases while reducing manual medical record review from 100% to 17% of ED visits. Conclusions and Relevance: NLP transformer models accurately automated detecting injury cases in ED patient medical records, with the potential of enabling real-time injury surveillance monitoring.
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 | 0.004 | 0.020 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".