Using AI/ML to Improve Patient Adherence

In a recent video interview with Applied Clinical Trials, Dominique Demolle, CEO of Cognivia, discussed challenges in improving patient adherence in clinical trials. She highlighted the reliability of adherence tracking methods, such as pill counts and ingestible sensors, and the need for tailored patient engagement strategies. Some potential solutions to these challenges include the use of artificial intelligence and designing more patient-centric studies.

ACT: Is there potential for the use of artificial intelligence/machine learning (AI/ML) to improve adherence? If so, how?

Demolle: I would say definitely, yes. When we start thinking about the adherence, the magical trick would be to be able to predict if a patient is going to be or not adherent when entering the clinical trial or at risk of dropping out or not being adherent. Of course, to achieve that, you need in your recipe two sets of ingredients. The first is to understand the drivers of the non-adherence, and the second one, because it's a multi factorial source of elements, you need to integrate that information and understand the importance of all those different elements, and then you come to machine learning and AI, so yes, definitely it is a way to move forward to better adherence in clinical trials, but you have to capture from the patient themselves. What are the triggers? And there are multiple dimensions, some traits of personality, but health literacy, for example, think about health literacy, know that we have to implement a diversity action plan. Of course, patients coming from a minority might be less educated regarding to medical support, or what is a clinical trial? Why it is important to take drugs at such a such point of time. So, per se, it's all a concept that you have to integrate in your AI model, and then all the other factors, like the motivations, the stress, some patients are stressed about adverse events and will disengage very early, so the idea would be to succeed, to integrate that information using machine learning, using the modeling, and predict a risk for a patient to not be adherent, and then you have that to engage your patient with a more tailored patient engagement strategy. This is how I view AI and how it could improve the retention, this is all what we are looking for. Patients are difficult to recruit. We don't want to screen them out. This would also cost a lot, this would not reflect real world, so we want them to enter the trial, and we want them to keep them in the trial until the end. It's really a win-win for the sponsor, the patients, the investigator, and the stakeholder.