AI-Driven Injury Reporting in Pediatric Emergency Departments
Devin Singh1,2,3, Alper Celik4, Evangeline W J Zhang1
1Division of Pediatric Emergency Medicine, Hospital for Sick Children (SickKids), Toronto, Ontario, Canada.
Insights
Natural language processing (NLP) transformer models can automate injury case detection in emergency department (ED) records. This improves public health surveillance by reducing manual review and enabling faster data analysis.
Area of Science:
- Public Health
- Medical Informatics
- Artificial Intelligence
Background:
- Injury surveillance is crucial for child morbidity and mortality prevention.
- Manual review of emergency department (ED) records for injury tracking is time-consuming and strained by high patient volumes.
- Timely data analysis is compromised, delaying the detection of public health risks.
Purpose of the Study:
- To evaluate natural language processing (NLP) transformer models for automating injury case detection in ED medical records.
- To assess the potential for improving the Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP) workflow efficiency.
- To maintain high sensitivity in identifying injury-related cases.
Main Methods:
- A prognostic study analyzed 217,173 pediatric ED visits from 2017-2023 at a tertiary pediatric center.
- Two transformer-based NLP models (DistilBERT and BERT-large) were fine-tuned using supervised learning.
- Models classified medical records for CHIRPP reporting, with outcomes including true positive rate, area under the curve, and reduction in manual review.
Main Results:
- Both NLP models demonstrated high accuracy, with AUROC of 0.983 and AUPRC around 0.93.
- Models achieved a true positive rate of 0.90, identifying 90% of injury cases.
- Manual review burden was reduced from 100% to 17% of ED visits, significantly improving efficiency.
Conclusions:
- NLP transformer models effectively automate injury case detection in ED medical records.
- This automation has the potential to enable real-time injury surveillance monitoring.
- The findings suggest a significant improvement in public health risk detection and response capabilities.
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.
More Related Videos
09:49Assessing Changes in Synaptic Plasticity Using an Awake Closed-Head Injury Model of Mild Traumatic Brain Injury
Published on: January 20, 2023
08:13Development and Implementation of a Multi-Disciplinary Technology Enhanced Care Pathway for Youth and Adults with Concussion
Published on: January 20, 2019
Related Concept Videos
01:21Types of Reports II: Incident or Occurrence Report
Purposes:
In the healthcare industry, reports play a crucial role in documenting incidents within an agency. The primary objective of these reports is to ensure patient safety, uphold the...
01:26Types of Reports III: Telephone and Verbal Reports
Here's an overview of each type:
Telephone Orders
01:24Data Reporting and Recording