Since 2005, Pangaea Information Technologies has been working as a partner with Rush University Medical Center under the leadership of Dr. Dino Rumoro DO, MPH, FACEP, Chairman Department of Emergency Medicine to advance the science of disease surveillance. Through this collaboration, we have developed the GUARDIAN Disease Surveillance System – an automated medical chart analysis and decision support system that utilizes state-of-the-art artificial intelligence and natural language processing algorithms to search for dozens of known infectious diseases, in real-time, across all patients currently within a hospital’s emergency department. One of the very first diseases that was encoded into the GUARDIAN system was Viral Hemorrhagic Fever (VHF) – the class of viruses which includes Ebola. Were a patient with Ebola to enter one of the emergency departments in Chicago, which is currently monitored by the GUARDIAN system, they would be flagged by the system and the attending doctor would be alerted as early as 15 minutes after the patient walked through the emergency department door (as we have demonstrated through simulation tests of the system at Rush). Early detection coupled with automated alerting is critical for the containment of a highly contagious disease like Ebola. By performing complete medical chart review on every patient, the GUARDIAN system is able to detect a likely VHF case even if the patient does not disclose information that would otherwise alert the hospital staff (such as recent travel to regions affected by the outbreak).
GUARDIAN differs from traditional disease surveillance systems in that it is not trend-based but instead reviews every patient individually. Most systems monitor for changes in the rates at which broad categories of symptoms (e.g. “respiratory,” “fever,” “neurological,” etc.) are detected. This approach is well suited to detecting the beginning of the flu season or detecting widespread food contamination, but it is poorly fitted for diseases, such as Ebola, where a single case is a significant event. GUARDIAN’s models offer increased sensitivity, specificity and accuracy over the trend based models. GUARDIAN analyzes the entire patient chart including labs and radiology as opposed to merely looking at chief complaint. GUARDIAN is able to parse out meaningful information from largely unstructured free text data in real time and subsequently analyze the data using a rules based approach using a semi-naïve Bayesian network. GUARDIAN also is able to identify any key missing data that would increase the confidence of the model’s output. In the event that multiple cases are discovered, GUARDIAN performs spatial-temporal clustering of those cases to identify the loci of outbreaks as quickly as possible.
What happens when a potential outbreak is discovered?
The GUARDIAN system automatically sends out an alert to the pre-determined chain of command including public health once any patient exceeds the probabilistic threshold for any of the diseases that we are monitoring. The system also has Response Packages for each disease which clearly outlines the approved procedures for each disease including Ebola. Each hospital can customize the response packages based on their operations but in the event that a disease like Ebola is detected, they would implement patient isolation, masking and restrict access as well as alerting public health and the proper chain of command.
What’s next for disease surveillance?
Medical, health, genomic, consumption, and related data are increasingly available, accessible and aggregated through hospital, government, and private sector information systems, but this wealth of data needs to be mined, cleaned and analyzed to provide meaningful information used to improve the decision making process.
In the past, this has been up to astute public health professionals reviewing the data using statistical packages such as SAS or SPSS. Moving forward, disease surveillance will leverage more advanced AI based “big data” techniques for identifying the signal within the noise and will continue to aggregate even richer datasets into this ecosystem. Additionally, data sources that were historically not available, such as social media, can alert doctors and intelligent computer systems to adjust their sensitivities and thresholds according to a potential threat being discussed in the public. This takes the place of the slower traditional method of news media reporting to public health and public health reporting down to the hospitals to adjust their monitoring procedures accordingly. With all of these data sources, there is uncertainty and error, so any system which hopes to pull meaning from them needs to leverage state-of-the-art statistical and analytical techniques to prevent false positives that would erode public confidence in these types of systems.
Next steps
The key to success for these systems is adoption. The U.S. Army Medical Research and Materiel Command (USAMRMC), Telemedicine & Advanced Technology Research Center (TATRC) under the leadership of Maj. Gen. Brian C. Lein and Senator Richard Durbin (IL) have been the largest proponents for developing and implementing the GUARDIAN system to provide real time surveillance and situational awareness of potential outbreaks before they spread.
Currently, the GUARDIAN system is in the Rush hospitals, with a plan to implement region wide by leveraging our relationships with Health Information Exchanges (HIEs) as well as partnerships with Public Health and leading Health Information System vendors.
About RUMC
Rush is a leading not-for-profit health care, education and research enterprise comprising Rush University Medical Center, Rush University, Rush Oak Park Hospital and Rush Health.A unique combination of research and patient care has earned Rush national rankings in nine of 16 specialty areas in U.S. News & World Report’s 2013-14 America’s Best Hospitals issue, among other recognitions of our quality of care and accreditations.
About Pangaea
Pangaea Information Technologies, Ltd. is a Chicago-based business focusing on real time analysis of big data to improve decision making in the healthcare, military, homeland security and private sectors.