3 min read

Wearables as Early Detection Health Systems

Originally published in FuturEar Podcast


This week’s episode of the Future Ear Podcast features a panel discussion between Andy Bellavia (Knowles Corp), Ryan Kraudel (Valencell), Chris Economos (PhysIQ), and myself. Our conversation this week revolved around the feasibility of using wearables and a variety of sensors to build early detection systems for individual data health sets. The ensemble of guests provided three different perspectives with Andy serving as our hardware expert, Ryan as the expert in data capture, and Chris as an expert in data output & analytics. This served for a fantastic conversation about a topic that I think will be gaining a lot of notoriety soon.

In a world where consumer wearables are proliferating rapidly, the biometric sensors embedded inside said wearables are maturing, and the machine learning algorithms used to make sense of the bio-data being captured are increasingly becoming more sophisticated, mind-blowing possibilities are beginning to be realized. As we explore throughout this conversation, one such possibility is the idea that wearables can be used to alert us to threats in our health. By capturing data with wearables, establishing baseline data sets within our health (what’s normal), and then using machine learning to monitor the data sets to identify anomalies, threats can be detected.

So where are we in terms of the validity and realization of this vision?

Well, as Chris points out, PhysIQ recently conducted a heart failure study sponsored by the US Veterans Association, in which 100 veterans recently discharged from a heart failure hospitalization were outfitted with a wearable and a phone. The first two days that these patients were home, served as the, “learning period,” to establish the baseline for the machine learning algorithm to work off of, followed by, “monitoring mode,” for 90 days. As this was strictly an observational study, PhysIQ operated under the statistical assumption that of the 100 patients being studied, a number would be re-admitted to the hospital (which they were).

So what did they find?

By doing a retrospective analysis on the output of the algorithm, PhysIQ looked at what the algorithm detected and when the detections were made. They found that on average, the algorithm was triggering an alert ten days prior to hospitalization. These results were published in a journal of the American Heart Association called Circulation Health Failure in February of 2020.

Holy shit. Right? In the moment, my mind immediately raced back to a Marc Andreessen podcast interview from late 2018. Here’s what he said about the importance of sensor technology moving into the future:

“The second thing I’d nominate for wearables is the concept of sensors on the body. Here, the Apple Watch is clearly in the lead with what they’re doing with the heartbeat sensor. But I think we’ll have a full complement of medical-grade sensors on our body — in a way that we have chosen to [have them] — over the next five or 10 years. I think we’ll be able to predict things like heart attacks and strokes before they happen. Talk about a killer app. [Laughs.] ‘Beep. I’m going to have a heart attack in four hours. Maybe I should drive to the hospital.’

The survival rate [for heart attack victims] at the hospital is, like, 99 percent. The survival rate for people at home is like 50 percent. There’s an opportunity for a massive increase in quality of life with the sensor platforms people are going to have.”

 MARC ANDREESEN

Another fascinating aspect to this conversation was the notion of, “sensor fusion.” As Andy pointed out, there’s data that can be derived from other modalities, such as our voices (through devices like smart speakers). This beckons back to Andy and I’s conversation with Nikolaj Hviid of Bragi about the sensor, “hive mind.” Both terms relate to the same idea of generating insights from a diverse range of sensors that feed into the same network. As Ryan points out, properly tagging data as to ensure it’s not, “garbage in, garbage out,” is paramount here, but as both Chris and Ryan agreed, the more data the better.

For me, as someone who comes from the hearing aid industry, I can’t help but feel incredibly excited about the idea that one of the strongest use cases for the hearing aids of tomorrow might be tied to this idea of serving as a true, “guardian of health.” Imagine outfitting our aging population with a device that serves so many roles – a communication enhancer, a conduit to audio content and the ambient web, and a preventative health tool used to alert them to threats in their health.

The future is coming a lot faster than we probably think.

-Thanks for Reading-
Dave