Using AI as a Tool for Infectious Disease Surveillance

Recently, conversations surrounding artificial intelligence (AI) have been growing, as sites like ChatGPT have become a hub for information via chatbot. Not surprisingly, the potential for AI assistants like ChatGPT to play a role in health, through medicine or public health, have increasingly come into discussion and larger notions of the pros and cons.

A recent study in JAMA Internal Medicine evaluated such a tool in medicine, comparing “written responses from physicians and those from ChatGPT to real-world health questions. A panel of licensed health care professionals preferred ChatGPT's responses 79% of the time and rated ChatGPT's responses as higher quality and more empathetic.”

In this context though, how could AI be harnessed for something that is all too important and..well relevant given our experiences in the last three years: infectious disease surveillance?

A recent publication in the New England Journal of Medicine, posed this very question. Noting the success of various AI applications, from early-warning systems, epidemiological forecasting, and even resource allocation. The authors sought to assess how AI could support disease surveillance through early-warning tools and differentiate between various circulating infectious diseases, but also how such a tool could backtrack outbreaks to their source.

The authors provided a detailed review of various AI functions in disease surveillance, ranging from risk assessments to source identification and hotspot detection, which can strengthen not only outbreak response, but reinforce a stronger preparedness system. A particularly interesting examples provided was source detection, citing a study from the University of Pittsburgh that developed EDS-HAT, which is the Enhanced Detection System for Healthcare-Associated Transmission, which combines whole-genome surveillance sequencing and machine learning to extract data from patient electronic medical records during outbreaks.

The authors noted that “The algorithm was trained by means of a case–control method that parsed the EMR data from patients known to have infections from the same outbreak (cases) and EMR data from other patients in the hospital (controls used to establish baseline levels of exposure relatedness). This form of learning guided the algorithm to identify EMR similarities (e.g., procedures, clinicians, and rooms) of cases with linked infections. Analysis of EDS-HAT determined that real-time machine learning based on EMRs in combination with whole-genome sequencing could prevent up to 40% of hospital-borne infections in the nine locations studied and potentially save money.”

What makes something like this so exciting? The capacity to identify outbreaks we might be missing, such as a drug-resistant bacterial infection in two patients who underwent the same type of imaging by the same technician. The authors shared another example. “In another instance, the source of a Pseudomonas aeruginosa outbreak among six patients in multiple units of a hospital over a period of 7 months was missed because of the wide separation of time and space. Genome surveillance suggested that the cases were all connected, and the machine-learning algorithm identified a contaminated gastroscope as the likely source of the outbreak—an easy target for intervention.”

These are just a handful of the applications for a single category that was reviewed, ultimately pointing to a larger capacity for AI to support outbreak response and rapid surveillance efforts. The authors do note several challenges that need addressing, such as awareness that these tools should serve as a supplement and tool rather than a replacement for traditional high-quality surveillance, but also that AI cannot replace coordination across jurisdictions or cross-functional partners.

There is a goal of building stronger analytic and collaborative surveillance networks though, which may improve this capacity. Ultimately, emerging tech, like AI, has a role within disease surveillance and outbreak science and now is the time to critically build the infrastructure and integration with existing systems.

Watch our recent interview with a physician who is the chief medical information officer for a major health system weighs in on AI, including how it could potentially be applied to journal articles and how clinicians are already using it.