Though Artificial Intelligence (AI) has been largely in the background for many decades, with the recent introduction of ChatGPT and other applications, AI has launched into the mainstream. Much of the recent discourse has focused on the fascinating and sometimes controversial ways AI is encroaching on our daily lives – from cheating on homework to online fraud. However, AI has also been making tremendous advancements in the healthcare realm and may soon have applications that can directly assist with infectious disease and outbreak forecasting.
A recent article in The New England Journal of Medicine provides a review of some of latest developments with AI technology applied to public health and infectious disease. In the paper, Brownstein, et al. highlight a few of the novel ways machine learning and other technologies are being applied to help expand surveillance capabilities from early warning tools to using data mining to better track down infection sources (Brownstein et al., 2023).
Early Warning Systems: With the abundance of information and real-time communications occurring digitally, there is a trove of information that can assist public health experts in identifying the initial stages of a potential outbreak. However, due to its voluminous nature and the broad spectrum of relevance, it is more than can be processed by humans, yet inconsequential for AI. As highlighted in the article, HealthMap is a prime example of how effective surveillance can be achieved as it monitors the internet for relevant indicators of a real-time outbreak developments versus other disease related material (the authors use journal articles and vaccination campaigns as key examples). Even more impressive, the technology can be utilized in nine languages and is able to connect geographic locations to better track and coalesce developments.
Source Tracking: When an outbreak does occur, AI can execute elaborate contact tracing to identify the initial source. For example, Enhanced Detection System for Healthcare-Associated Transmission (EDS-HAT) can examine a hospital’s electronic medical records and coupled with whole-genome sequencing, can analyze the myriad of complex connections to help discern the infection source. Developed by University of Pittsburgh, the system has already demonstrated its keen detective work by solving several challenging outbreak scenarios, including one spanning six patients over the course of seven months that may otherwise have never been unraveled.
The authors shed light on other ways AI is leading improvements in public health including pathogen identification, antimicrobial resistance trends, as well as response guidance, which applies AI to help to monitor and provide improved risk assessment for government and agencies, such as the deployment of Eva in Greece, which helped identify high-risk travelers for COVID-19 testing.
While the article features a number of impressive case studies and emphasize there are many more, they also acknowledge there are still many issues and challenges to adopting AI. Viruses are nimble and algorithms will need to be updated to stay up to date with the changing variants, symptoms, and patient demographics. And AI will need to be supplemented by molecular tests to ensure it can distinguish between simultaneously present pathogens. The article goes on to elucidate other concerns such as privacy and issues with source data that can undermine AI’s conclusions. In summary, while AI is not new the realm of public health, it is offering innovations at a rapid pace, but as the article suggests, AI should “complement rather than replace high-quality traditional surveillance.”
Learn more: Artificial Intelligence and Machine Learning: Applying Advanced Tools for Public Health
Reference:
Brownstein, J. S., Rader, B., Astley, C. M., & Tian, H. (2023). Advances in Artificial Intelligence for Infectious-Disease Surveillance. N Engl J Med, 388(17), 1597-1607. https://doi.org/10.1056/NEJMra2119215
Nicole is a Research Associate for the SPECTRE Program