Identifying subjects with low adherence to trial visit schedules in the long-term clinical trial
*Ali Falahati, Novo Nordisk A/S 

Keywords: clinical trials; subject retention; predictive modeling; visit window; analysis of adherence; simulations

Subject retention can be a challenge for clinical trial management and, in particular, for the recruiting trial sites. Developing a predictive model to identify individuals with low adherence during the course of a trial would enable the site to take appropriate action to improve subject adherence. Parametric and non-parametric methods were used to fit models around data describing the time elapsed between successive visits. As predictive factors, data collected at screening, as well as adherence records for the subject’s previous completed visits were used. According to the trial protocol, a given visit should occur within a ‘visit window’ assigned to that visit, e.g. 7 days. A gap between the visit window and actual time of visit was assumed to be an indicator of lack of adherence to the visit schedule. Based on the fitted models, predictions can be made regarding which subjects are more likely not to complete their visits within the visit window. This information can then be communicated to each site, allowing the site to take action to ensure adherence to scheduled visits. Different models are fitted at each visit and models are updated on an ongoing basis. Results from the model development will be presented, and the challenges in identifying subjects with low adherence will be discussed.