Joint Research Conference
June 24-26, 2014
Intensity Estimation for Poisson Processes Used to Model a Real-Time Medication Event MonitorIt is challenge to model medication taking behavior and eventually to predict the change of behavior, especially for those vulnerable patients (e.g. substance abusers). With the help of modern technology (e.g. Wisepill device), we are able to monitor the medication taking behavior in real time. In the meanwhile, after a few minutes delay of detecting the events and analyzing the events, a real time intervention can be generated to communicate with patients directly by the packages of Python, such as text messages to the mobile phone of the patients. The statistical methods that can be used to model the complexity of the medication taking events and to predict the future missed dosage are important for designing the personalized module of the device and therefore personalized intervention. This intervention not only will improve the medication adherence, but the personalized intervention will help to build up the correct medication behavior in the future. In this talk, we use Non-Homogeneous Poisson Process (NHPP) to model the sequence of medication taking during the day. The six hour window (3 hours before and after the dose schedule time) is considered as valid medication taking event. The daily real time medication events were monitored for more than 16 week by Wisepill device. We will model the extreme event of medication taking (outside the designed medication event window). Eventually, we use the model to predict the non-adherence events (missing medication event of the day). Therefore a customized intervention for adherence of medication could be performed on patients in the real-time.