By Anind K. Dey, Carnegie Mellon University, Embedded Assessment Principal Investigator.
As part of our project to detect cognitive decline in elders, we are embedding sensors to observe how they perform everyday activities and how these performances change over time.
In recruiting participants for our project, we conduct interviews with potential subjects, observe and videotape how they perform the daily activities of interest (medicine taking, meal preparation, and telephone use), and discuss whether we can realistically build a customized solution for monitoring the performance of these activities. This is of course in addition to whether we think they would be a good subject (at risk for cognitive decline but is not currently experiencing decline, lives alone, performs the activities reliably ...).
In our attempt to recruit subjects, we have already faced two different types of setbacks: technology-related and people-related.
We have deployed instrumented pillboxes, bed and chair sensors, and telephone use sensors in the apartments of two pilot participants. We will be visiting participants every two weeks so technologies are designed to last at least that long. While replacing batteries, pillboxes, or sensors on a frequent basis might be a pain, we decided that designing technology that had to last longer wasn’t a good use of resources. So far the technology had been working great and after troubleshooting a few anomalies, we are receiving data on our server.
Our biggest struggle is customizing the technology to each individual’s habits. As you can imagine, elders perform activities in any number of ways. In our fieldwork, we observed some elders taking pills using a pillbox that they filled or had filled once a week, while others took their pills straight from the pill bottles multiple times a day. While both ways are perfectly acceptable, they each call for a unique sensing solution and from an engineering standpoint they are very different. This is just one example, but it applies to all the observations of daily living we are interested in, and all the facets of those daily activities as well: e.g., swallowing pills using bottled water vs. getting water from the tap, making instant coffee vs. using a coffeemaker ...
It has become clear that if we could recruit subjects that all performed their daily activities in the same way, our lives would be A LOT easier, but the systems we build would be less realistic, as this doesn't mimic real life behavior.
On the human side, we are lucky to have the support of Presbyterian SeniorCare and the Allegheny Country Area Agency on Aging to help us interview, assess, and finally recruit subjects that meet our stringent criteria. After interviewing and videotaping a woman making breakfast and taking pills we thought she would be a great match for the project. Unfortunately she passed away of natural causes before we could recruit her. It is easy to get caught up in the goals of our research and the design of the technology, but it is also important to remember that we are dealing with real people who have real issues and lives outside of our study. While this can be obvious to others, we sometimes need to be reminded of it.
In our attempt to get to real observations of daily living, we have some significant obstacles to overcome.