Anind Dey, Ph.D., dwellSense Principal Investigator, Carnegie Mellon University
Throughout our project, we have had issues with recruiting subjects for our study. Our goal was to recruit 30 elders, along with their caregivers and doctors. However, we have struggled to recruit 20 subjects. There are a number of possible reasons for this. Some are due to not being targeted enough in recruiting, and some are due to not having enough constraints. To perform our study, we require a great deal of access to subjects and their apartments in order to perform assessments, to regularly check the state of the display and sensors, and to replace batteries.
From a time and manpower perspective, we needed to recruit subjects within a small geographic area. Otherwise, the commute time to and from our participants' apartments would have been unmanageable. So, we made a choice to identify a small number of locations from which to recruit subjects. That drastically limited our subject pool, but having subjects spread around the city would have reduced the number of subjects we could have actually worked with, too. It was not easy to identify the potential locations or potential subject populations centered in an area, particularly because we weren't looking for individuals with particular disabilities or diseases. Instead, we were looking for elders who were "at risk" but not necessarily exhibiting signs of cognitive decline.
Once we found suitable locations with large potential subject pools, we faced other difficulties in recruiting. One issue was the amount of time we wanted participants to commit to being part of our study. Because we expect the personal health information we collect via sensors to provide value over time rather than immediately, and because we wanted enough time for participants to get over the novelty factor of our embedded assessment appliances and displays, we needed participants to commit for several months. Many potential subjects had a hard time making this long-term commitment.
We also wanted to maximize the number of assessments we could make for each participant. This meant we were looking for subjects who regularly perform the three activities we monitor: coffee making, making/receiving phone calls, and taking medicine. Some potential subjects didn't work out because they drank tea instead of coffee, or made instant coffee rather than using a coffee maker. These activities, though perfectly reasonable, made them inappropriate for our study. Others had digital phone lines that were not compatible with our phone sensing hardware, which works with analog phone lines. Furthermore, some took pills from the original pill bottles rather than from pill boxes. We were afraid that changing someone's habits (from using pill bottles to using pill boxes) would impact their behavior too much and introduce errors in actions, so we removed these individuals from the potential subject pool. In the end, we chose to accept subjects if they used pill boxes and satisfied the requirement for either the coffee making or the phone use.
Further complications arose from the nature of the research itself. When they heard our description of the study, some subjects declined to participate. When asked why, it was clear that participating in such a study and using such a system that tracks and represents health information served as another reminder of their declining physical and/or cognitive health. The last thing they wanted was a "constant" notification of how they were performing everyday activities.
We are continuing to recruit subjects, but this is a slower process than we had expected. We’ve definitely learned a lot about recruiting for this type of study.
Creative Commons photo by jcestnik.