A new comprehensive study reveals that a significant majority, two-thirds, of epidemiologists participating in specialized Field Epidemiology Training Programs (FETPs) are actively incorporating artificial intelligence (AI) tools into their professional work. This widespread adoption, however, starkly contrasts with the availability of formal training, as only one in five of these professionals has received any structured AI instruction. Furthermore, a notable quarter of the surveyed epidemiologists expressed considerable ethical concerns regarding the accuracy, potential biases, and overall reliability of AI-generated content, underscoring a critical gap between technological integration and preparedness within the field. This indicates a pressing need for updated educational frameworks to address both practical application and the responsible use of AI in public health epidemiology.
Troubleshooting coding errors
Among the 105 epidemiologists surveyed, a substantial 66% reported utilizing AI in their daily field epidemiology tasks. This usage was particularly high among fellows in the European program (89%), compared to 54% in both the US and Canadian programs. The frequency of AI use was considerable, with 42% of users engaging with the technology weekly and 30% on a daily basis, while only 26% used it occasionally. A strong majority felt comfortable or somewhat comfortable using AI (84%). ChatGPT emerged as the dominant platform, used by 87% of respondents, with institutional AI tools also seeing moderate use (25%). The primary applications of AI included efficiently troubleshooting coding errors (91%), assisting in writing new code (75%), and generally enhancing work efficiencies (42%). Qualitative feedback highlighted AI's effectiveness in reducing time spent on coding, facilitating the learning of new coding techniques, simplifying existing code, enabling code transfer across different software, and generating innovative ideas for data analysis. Beyond coding, AI also proved beneficial in improving writing quality, streamlining administrative tasks like meeting minutes and project timelines, and aiding in background research and information summarization, thereby contributing to overall operational effectiveness.
Training lags behind adoption
Despite the widespread adoption, a significant proportion (41%) of AI users identified various obstacles to fully integrating the technology. These barriers encompassed technical limitations, restricted access to advanced AI tools, and a lack of clear institutional policies or guidelines regarding AI usage. Moreover, substantial concerns were raised about the accuracy and reproducibility of AI-generated information, feeding into broader ethical considerations. A quarter of AI users specifically voiced worries about data privacy, inherent biases within AI algorithms, environmental impacts associated with large-scale AI operations, and the potential for overreliance on AI, which could diminish critical human judgment. A striking finding was that only one in five of the surveyed fellows had received any form of structured AI training, whether formal or informal. This highlights a significant disparity between the rapid embrace of AI tools and the preparedness of the workforce. Respondents expressed a strong desire for practical, hands-on instruction in using AI for core epidemiological functions such as coding, data analysis, scientific writing, data visualization, and outbreak detection. They also emphasized the need for comprehensive training on ethical considerations and effective prompting strategies to maximize AI utility while mitigating risks. Additionally, a clear demand existed for education on AI's inherent limitations and common pitfalls.
Curricula should promote AI literacy
The study's authors strongly recommend that Field Epidemiology Training Programs (FETPs) proactively update their curricula to foster robust AI literacy among participants. This proposed educational framework should not only cover the practical applications of AI but also delve deeply into its limitations, potential biases, and critical ethical implications. Such an update is deemed essential given AI's established role as a routine tool in contemporary field epidemiology practice for a substantial number of trainees. Specific recommendations for effective training include the implementation of hands-on workshops, the integration of real-world case studies to provide practical experience, and ensuring accessible access to various AI tools and platforms. The objective is to equip learners with comprehensive experience in leveraging AI across fundamental epidemiological tasks, including coding, study design, sophisticated data analysis, scientific writing, and crucial areas like surveillance and outbreak detection, ensuring they can apply AI responsibly and effectively.
