Implementation of Emergency Medical Text Classifier for syndromic surveillance

Area of Science:

• Public Health
• Infectious Disease Epidemiology
• Health Informatics

Background:

• Syndromic surveillance systems are crucial for early detection of infectious disease outbreaks.
• Emergency department clinical notes offer valuable data but pose challenges for accurate concept extraction.
• Existing methods require improvement for efficient and effective public health surveillance.

Purpose of the Study:

• To implement and evaluate the Emergency Medical Text Classifier (EMT-C) system in a production environment for syndromic surveillance.
• To assess the performance of EMT-C compared to a baseline system in extracting relevant concepts from clinical notes.
• To gauge user satisfaction with the EMT-C system, focusing on workload manageability.

Main Methods:

• Development and implementation of the Emergency Medical Text Classifier (EMT-C) system.
• Evaluation of system performance using key metrics, including positive predictive value and false positive rates.
• User satisfaction surveys and feedback collection within the production surveillance setting.

Main Results:

• The EMT-C system demonstrated superior performance across all evaluated metrics compared to the baseline system.
• The system successfully maximized positive predictive value and minimized false positives, aligning with user preferences.
• Users reported slightly higher satisfaction with the EMT-C system, indicating improved usability and workload management.

Conclusions:

• The Emergency Medical Text Classifier (EMT-C) is an effective tool for enhancing syndromic surveillance using emergency department clinical notes.
• Incorporating user input and testing in a production environment are vital for successful implementation of new surveillance technologies.
• EMT-C offers a promising advancement in the timely and accurate detection of public health emergencies.