ePCRN-IDEA2: an agent-based system for large-scale clinical trial recruitment

Abstract

Clinical trials are a key method of evaluating the efficacy of healthcare interventions prior to public release. To allow clinical trials to take place, however, it is necessary to recruit sufficient patients for participation. Unfortunately, such recruitment poses a significant challenge though. In this paper, we discuss a novel agent-based system designed to enable patient recruitment; our system, ePCRN-IDEA, has been implemented and is currently under deployment in the UK healthcare system. Through this deployment, however, we have found a number of challenges relating to scalability; consequently, this paper focuses on an extension called ePCRN-IDEA2 that addresses these problems. Specifically, we place agents on General Practitioners’ (GP) machines to dynamically compute patient eligibility in real-time during consultations, thereby enabling GUI notifications and immediate recruitment. In ePCRN-IDEA, all agents attempted to compute patient eligibility over all trials, resulting in a huge burden for a large-scale deployment (e.g. 100,000 trials). Therefore, in eCPRN-IDEA2, we have embedded the necessary intelligence in agents to dynamically compute eligibility over the trials that are most likely to match the clinic’s and patient’s characteristics. Through simulations, we evaluate the approach to show that our decentralised trial selection algorithm can achieve comparable performance to a global knowledge benchmark with far greater scalability