AI — Making Decentralised Trials Easier

The rapid transition toward decentralisation in clinical trials is continuing, bringing with it the many benefits of improved recruitment and retention of participants, increased diversity and more opportunities.

However, to fully realise the potential of in-home participation, accelerated by the COVID-19 pandemic, innovative ideas and processes are needed to help trial participants tackle the additional responsibility placed on them. Artificial Intelligence could be the answer — here are three ways in which AI could be a game changer.

Easy Information

A patient participating in a decentralised trial is obviously required to be responsible for gathering important trial data — they complete surveys, keep diaries and conduct other electronic clinical outcome assessments (ECOA’s) at home on a regular basis. Such activity is time consuming, however, and even despite the availability of electronic prompting, it can lead to a patient inputting inaccurate data or becoming overwhelmed and dropping out of the trial altogether.

Reinforcement learning could be the way to address this. A domain of AI, used for sequential decision making, it’s already used to create chess-playing computers which can beat human players. It could potentially be customised to the individual requirements of trial participants and used, for example, to identify what notifications are required and when to send them. Appropriate to the individual in question based on their previous response patterns, for example, it could send motivational messages or reminder deadlines while still ensuring that responses are maximised.

Better Visuals

In cases where patients are required to send evidence videos or photographs that they have recorded themselves, getting the quality just right can be tricky.

Computer Vision is the domain of AI that could change things for the better here — already in use in, for example, by banks who use it to assist customers in taking photos of cheques for electronic lodgements, patients could be coached as they take the photos and videos to make sure that angles are correct, lighting is effective and zoom is just right. Use of computer vision will ensure that the visuals sent back to the researchers is of the appropriate quality, and patients won’t have to take and retake shots to get it right.

Super Sensors

These are often relied upon to gather information but the process of applying multiple sensors all over the body, getting them in the correct configuration and keeping them in place while performing specific mobility assessments can be tricky, time consuming and wide open to errors.

AI models designed for temporal data could be the answer to achieving more accurate and detailed results with fewer sensors. These models can be trained to determine results from just a few, or even single sensors, meaning that mobility outcomes could be measured constantly in a typical living environment.

AI is becoming more and more part of our daily lives and the possibilities it creates are immense. Using it in clinical trials can improve the experience overall for both patient and trial organiser, reducing the burden of participation and resulting in more accurate data to achieve better results for all.