By Anind K. Dey, Carnegie Mellon University, Embedded Assessment Principal Investigator.Two weeks ago, I attended the second Project HealthDesign workshop at the Vanderbilt Center for Better Health, with the Project HealthDesign team and the other grantees. It was another valuable workshop. Clinicians from each of the grantee projects attended, and there was a wonderful presentation by Dr. Kevin Johnson from Vanderbilt, a first round grantee.I not only learned more about the projects of my fellow grantees, but I also learned a lot more about our Embedded Assessment project. Up until this point, I thought I understood the observations of daily living (ODLs) that our project wanted to focus on, but I realized that I had something to learn. My team's focus is on using sensors embedded into everyday objects that can be used to monitor how well an elder performs everyday activities, and therefore be used to perform in situ assessment of cognitive decline. In consultation with our occupational therapists on the team, Linda Kent from Presbyterian SeniorCare and Diane Collins from the University of Pittsburgh, my graduate student, Matthew Lee, and I came up with a list of ODLs: medicine taking, phone use and meal preparation. But, what we realized during conversations with the other attendees are that these ODLs are really meta-ODLs or observable ODLs (not a great name, I know!) or proxies. The ODLs that we really care about are the things that these ODLs consist of: change in fine and gross motor control, ability to initiate and sequence activities, ability to situate oneself, etc. I'll call these elemental ODLs, for lack of a better name. While we had been considering our meta-ODLs somewhat independently, I think we have an opportunity to look at them more holistically, and treat all sequencing actions (in meal preparation and medicine taking, for example) as the same set of activities.
This is particularly exciting because it could allow us to answer the "why" question that has alluded us up until now. When we show potential users of our technology sample graphs produced by the system, they may reveal an increase in how much time they spend on a particular task over a month we will inevitably get asked "Why is that?". We can't answer that question without ground truth data and observations of all activity inside their residences. Perhaps it's due to a cognitive decline, but perhaps it's because they always call a friend in the middle of the activity. One is clearly a cause for action, while the other is not. But, if we can now collect and visualize data about the meta-ODLS, we can see how patients perform on elemental ODLs within these. For example, we could show that a patient has difficulty with sequencing tasks, both for the medicine taking and the meal preparation tasks. These kind of correlated activities would go a long way at answering the "why" question and helping our patients and clinicians trust the data our system can produce.
Obviously we're really excited about this new understanding. It's going to be a lot more work to explore these ideas, but I think it'll be really fruitful in the end!