After identifying a potential problem space for further investigation, we began interviewing physical therapists and patients and ultimately found a disconnect between patients’ prescribed routines and their confidence in their ability to do these exercises correctly at home. We crafted the following mission statement to guide us through our research.

To get a better understanding of the factors which impact a patient’s compliance with their at-home exercise routine, we began interviewing physical therapists and patients, as well as conducting site visits to PT clinics. After conducting over 30 interviews and site visits, we identified a few key obstacles in compliance:

We distilled our learnings into an extensive Patient Journey Map, which guided us throughout the entirety of the design process. The journey map can be found to the left. Click on it to zoom and look at each step, as it is quite thorough.

We continued by conducting an extensive journal review, investigating studies in compliance which had already been conducted, in addition to articles detailing areas for future development. We identified three key areas of need for a potential solution, which guided our next steps for development, as we explored potential avenues for increasing compliance.

Based on these needs identified through earlier interviews, we decided to develop two separate solutions which could potentially work together: a wall mounted exercise device, and a novel method for capturing and defining compliance through remote monitoring.

We conducted thorough research into existing methods for monitoring form of exercises, and learned that while there were existing methods with proven success, there were several issues which prevented them from adequately meeting the needs of our user base. These options were expensive, which prohibits their potential integration for at-home use. Additionally, many of these sensors were large and attached to the user’s body, which served as a distraction from doing the exercises. It is important for the experiences of exercising with and without monitoring to be as similar as possible.

Ultimately we identified two methods of capturing data which we believed to be promising, with current gaps in market for a potential solution to make an impact:

Upon identifying these methods, we set out to test further which specific data streams are best at capturing relevant information for analyzing compliance. We had to balance devices which were capturing data at different frequencies, as well as determine how to synchronize this data. Ultimately, we acquired three separate devices (pictured below) to use in our next round of testing.

We selected these devices because they require minimal effort on behalf of the user to operate. The balance mat is advantageous as the user does not need to put on any additional motion capturing devices on their body, and the user can conduct exercises as normal.

We crafted detailed testing plans to assess whether these different methods could accurately identify compliance. We went through four separate rounds of testing:

Each of these rounds of testing meant not only creating targeted testing protocols to answer specific questions, but also spending the time to conduct the actual testing (we did this testing in our group as opposed to with recruited users) as well as data analytics. In between each round was a necessary pivot, changing our tests to ensure they aligned with our intentions for the project.

The first two rounds of testing were done so that we as a team could understand how our devices worked, as well as to ensure the accuracy of their data. We separated the balance mat data and biometric data for these rounds to simplify testing, as keeping everything time synchronized proved to be difficult. In this round of testing we were sure to iron out any kinks in the data recording process in preparation for extensive proof of concept testing. These rounds of testing served as a barometer for us, to ensure that we were continuing down the right path before we did more advanced proof of concept testing.

After performing benchmark testing for the balance mat, where we were able to prove a visible difference in center of pressure path between correct and incorrect, we learned that we needed more control over the data. We began using the “BodiTrak Pro” software to capture data, which is intended for developers and data scientists. We had previously been using the “BodiTrak Balance” software, intended for use by physical therapists.

In the exercise breadth testing, we tested four separate exercises to assess which created the clearest results for our proof of concept. Ultimately the most promising exercise was birddog, an exercise where the user is on all fours and extends their opposing arm and leg.

Once we identified birddog as our most promising exercise, we began to do more extensive testing to analyze if it was possible to identify how the user was doing their exercise incorrectly. We went back and conducted another site visit and observation to identify failure modes for birddog. We found three distinct ways to break compliance for this exercise which we went on to test.

For this round we tested four separate ways of doing the exercise: birddog with correct form, with core unengaged, with incomplete arm/leg extension, and with the exercise interrupted by falling.