Introduction: Digital medicine consists of the integration of active pharmaceuticals and wearable/ingestible sensors combined with mobile and web-based tools to improve effectiveness of clinical decision making by providing actionable feedback on treatment adherence and patient response. Primary drivers for unobserved ingestions can play a critical component in communications between the patient and their care team.
Methods: Ingestion and survey data from two clinical trials (NCT02722967, NCT02219009) with digital medicine were used. The fraction of times a patient responded with available survey responses over the treatment duration was used to assess natural groupings of response patterns across patients. A density-based clustering algorithm was applied after t-distributed stochastic neighbor embedding projection to the representative two-dimensional space in order to identify possible response-pattern groups in the population.
Results: Eight (8) patient survey response groups were identified, displaying varying rates of response utilization during treatment duration, with one group consistently responding, ‘I took my dose, but it didn’t register’. Patients who provided multiple responses demonstrated better temporal correction than those who did not. Additionally, for patients who only responded one way, there were multiple types of behavior observed on the day prior.
Conclusions: These results demonstrate that there appear to be reporting preferences of patients for primary drivers of unobserved ingestions. Additionally, more variety in reported survey responses was associated with higher ingestion success immediately following the survey prompt. Of the patients who only reported system error, there is evidence to support multiple potential driving motives.