Toward Naturalistic Self-Monitoring of Medicine Intake

Since drug actions are dose- and time-dependent, adherence to prescribed medications is essential for the effectiveness of therapies. Unfortunately, several studies show that when patients are responsible for treatment administration, poor adherence is prevalent. Hence, it is necessary to devise effective methods to remotely assess medication compliance and support self-administration of drugs. Existing methods include electronic reminders such as short message service reminders and pill reminder apps. Although those tools may help increasing adherence, they interfere with the normal routine of patients by providing unnecessary reminders, or providing the reminder at an unfortunate time. More sophisticated solutions include the use of smart packaging and ingestible sensors to quantify and monitor drug intake. While those solutions do not interfere with normal routines, currently they are restricted to patients involved in a few clinical studies. In this paper, we introduce a novel system to support self-administration of drugs without interfering with the patient's routines. The system is based on a combination of cheap sensors and a smartphone. Tiny Bluetooth low energy sensors attached to medicine boxes communicate motion data to an app running on the patient's smartphone. Thanks to a machine learning algorithm, the app detects intake events, and reminds the user only when needed. Active learning is used to improve recognition rates thanks to the user's feedback. Preliminary experiments with a dataset acquired from volunteers show that the algorithm can detect most intake events with a few false positives. At the time of writing, we have developed a working prototype of the system, and we are beginning an experimental evaluation with a group of patients of an Italian hospital.