The Hidden Role Of AI In Treatment Selection: Friend Or Foe To The Study And Practice Of Pharmacoepidemiology?

Background: The future of prescribing is likely to be guided by artificial intelligence (AI). AI can utilise a patient’s medical history to personalise treatment decisions, potentially improving patient outcomes and preventing costly adverse events and drug failures. Examples of such AI-based tools include treatment success predictor tools and medication review tools. The rise of these tools offers both challenges and opportunities to the study and practice of pharmacoepidemiology and real world evidence generation. This is a collaborative proposal by members of the Digital Epidemiology SIG, the RWE Collaborative SIG and the Medical Device SIG.

It is challenging to measure the impact of these tools as it is often difficult to know how and when these tools are being used. Many tools exist in ‘research-only mode’, have poor reporting, or are only used fleetingly. Such activities could result in strange artefacts in the real-world surveillance of medicines, as these tools will influence prescribing trends.

For marketing authorisation holders (MAH) this topic is especially pertinent as MAH are likely to have little to no control over which tools are paired with their medication in a medical setting, yet they may determine market access on a global scale. The real world data underpinning these tools is also likely to be of interest (or contention). These tools may also be slower to reflect important medicine label changes, which could put patient safety at risk.

Objectives: This symposium will provide a high-level understanding of what these AI-based prescribing tools do (including a real-life example of how these tools can have an impact on prescription trends) and include discussion of limitations such as the potential exacerbation of healthcare inequities. It is intended for researchers and MAH involved in drug utilisation studies and those involved in regulatory decision making. We hope to use this symposium to spark deeper conversations around the role of pharmacoepidemiologists in designing and studying these tools.

Description: - AI-based treatment selection and its impact on the study and practice of pharmacoepidemiology (Harriet Dickinson – 15 minutes)
- Regulatory expectations for Software as a Medical Device and Clinical Decision Support Systems (Danica Marinac-Dabic, FDA)– 15 minutes)
- How these tools can exacerbate inequalities in medication access (Enriqueta Vallejo-Yagüe)- 15 minutes)
- Collaborating for success: why patients, providers, medical device experts, software developers and pharmacoepidemiologists all need a seat at the table (Katoo Muylle 15 minutes)
- Case study: how AI tools can restrict access to much needed pain medication (Juan M. Hincapie-Castillo - 15 minutes)
- Panel discussion (15 minutes).