Purpose: Recent advances in technology enable context-sensitive prompting and reminding. The notion of maximizing the efficiency of these reminders is a decision-theoretic optimization problem. In general, it is possible to determine empirically the probability of a desired action with and without reminders. In addition, the cost of reminders is associated with their annoyance. By quantifying these, it is technically possible to determine the expected utility of various types of reminders and therefore find a reminder that maximizes the expected utility at any given time. We examined this framework by developing a context-based reminding to improve patients adherence to a medication administration regimen. As part of a larger study of activity monitoring in the home for older adults, we are currently using unobtrusive sensors to obtain data on an individual’s location data and activities, such as sleep and phone use. This information provides a context for appropriate reminding for taking medications. The purpose of this study was to measure individuals’ utilities on the modalities and formats of reminding devices in order to understand how best to provide context-aware reminding.
Method: We observed medication taking behaviors in a simulated home setting, using a dual-task method for observing interruptions and responses to a variety of devices and strengths of alerts. The devices consisted of watches, cell phones, TV prompts, and beacons. The modalities for the alerting included vibration (for the watch and cell phone), a beep or chime, or voice.
Result: The results of the small empirical study suggest that each subject appears to have reliable “utility” associated with each reminder modality and intensity. On the average, vibrating telephone was found to be the least effective, while the interruption of a TV program was most effective. At the same time, vibration was the least annoying, while auditory alarms were the most annoying.
Conclusion: These results suggest that it is possible to optimize medication taking adherence using decision-theoretic, context-aware approaches. The caveat is that the cost of reminders, e.g. annoyance, is specific to individual patients and thus the reminder system must have the capability to adapt to the individuals’ utilities.
Candidate for the Lee B. Lusted Student Prize Competition