Beacons and Adherence Part 2

I used the word app in my previous post, but that’s a bit grand.

Rather, I’m going to be using a tool called Automate to create a pretotype of my adherence app.

I’m a big believer in pretotyping. It’s a fast way to test out a concept. More than a few “AI-powered” products were pretotypes powered by humans working behind the scenes.

I’ll be using Automate’s drag-and-drop flow builder to model my app’s logic. My two key needs are the ability to detect Bluetooth Low Energy (BLE) beacons and trigger notifications in the Android notification drawer.

Beacons and Adherence

I recently started taking medicine to help manage my cholesterol levels. I decided this was better than thinking I’d manage it via diet and exercise alone.

Twenty-four days after my first 30-day fill, I have 11 pills left. If I have perfect adherence and refill when I’m out of supply, I’ll have a PDC of 86%. As adherence goes, not bad. But if I extrapolate based on my (short) history, I’ll be just under that 80% adherence threshold.

Years ago, a colleague and I had discussed using beacons to help improve adherence due to forgetfulness. Simple time-based reminders can work, but our thinking was that a contextual reminder using time and proximity to the pill bottle would be more effective. In the days before perpetual work-from-home, location could also be used to trigger an urgent alert if you were about to head out to the office without taking your medication.

COVID-19 Digital Surveillance

Last month I came across an article in Nature assessing the use of social media as an early warning for COVID. Naturally, I was intrigued.

But as I read it, I couldn’t help but think back to the Google Flu debacle. Google touted–in what was later labeled as “big data hubris”–its ability to predict flu outbreaks simply by using its vast amounts of search data.

As we learned, they couldn’t.

COVID-19 and Game Theory

Over the holiday break, I took advantage of the downtime to finally do some reading. I skew towards non-fiction and found a lightweight book on game theory that refreshed my thinking on a fascinating course I took 20 years ago.

Thus primed, when following the news on COVID-19 and individuals’ responses to social distancing, it dawned on me that the choice to follow mandates or ignore them was a version of the prisoner’s dilemma. I decided to explore this a bit further with a basic scenario.

AI-Powered COVID Screening

The healthcare industry has seen a number of innovations which center on using maching learning on audio recordings to triage or diagnose a condition or disease.

My first exposure to this field was Winterlight Labs, who developed technology that uses voice recordings to diagnose Alzheimer’s. Similar efforts by other organizations have assessed loneliness, psychosis, PTSD, and depression.

Recently, news broke of maching learning being used to identify COVID-19 in a subject simply by listening to their cough.