Improving flexibility and performance of the Acute Care Enhanced Surveillance (ACES) System for public health surveillance: an ensemble of state-of-the-art machine learning and rule-based natural language processing methods

The Acute Care Enhanced Surveillance (ACES) system, based at Kingston, Frontenac, Lennox & Addington (KFL&A) Public Health, is a public health surveillance system that monitors patient activity in near-real- time at 97% of Ontario's acute care hospitals. The system groups patients into health conditions (syndromes) based on their chief complaint in the emergency department. Epidemiologists use ACES to identify disease outbreaks or other potential public health problems from patterns in hospital activity. While useful for tracking COVID-19, key weaknesses in how the system groups patients were found. This project proposes to improve the way the system groups patients by using state-of-the-art computational methods (machine learning and natural language processing) that do not rely on one rigid model only, and that can better classify key pieces of information from the chief complaint. Diverse experts in epidemiology, public health, medicine, and computer science will collaborate to determine how best to define new health conditions, based on these classifications and other available patient information. The ACES system will be updated so that it automatically uses and displays these new health condition groupings, which will compliment those that are already there. These new health condition definitions will make ACES more flexible and relevant, providing public health leaders with an effective tool to better detect and monitor public health threats like opioid consumption issues and new diseases like COVID-19.