Developing and evaluating an automated appendicitis risk stratification algorithm for pediatric patients in the...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support

Background:

• Pediatric abdominal pain is a common emergency department (ED) presentation.
• Accurate and timely risk stratification for appendicitis is crucial for effective patient management.
• Current methods rely on manual chart review, which can be time-consuming.

Purpose of the Study:

• To evaluate a novel automated method for risk stratifying pediatric patients with abdominal pain.
• To assess the performance of a natural language processing (NLP) and machine learning system using electronic health record (EHR) data.
• To compare the automated system's accuracy against physician expert performance.

Main Methods:

• Analysis of EHRs from 2100 pediatric ED patients with abdominal pain.
• Development of an automated system to extract data from physician notes and lab values.
• Risk categorization (high, equivocal, low) based on the Pediatric Appendicitis Score.
• Performance evaluation using recall, specificity, and precision against a physician-created gold standard.

Main Results:

• The automated system achieved an average F-measure of 0.867 for appendicitis risk classification.
• Performance was comparable to that of physician experts.
• High recall and precision were observed across low, high, and equivocal risk categories.
• Essential input data was available within the first 4 hours of the ED visit.

Conclusions:

• Automated risk categorization of appendicitis using EHR content, including clinical notes, is feasible.
• The system demonstrates performance comparable to manual chart reviews by physicians.
• This approach offers a promising avenue for computerized decision support to enhance evidence-based medicine at the point of care.