Required reading for wearables in healthcare
Paul Sonnier reposted a Nature article on the Digital Health group recently that I recommend as required reading for anyone considering how wearables can impact healthcare. This is the first healthcare oriented article I have seen that talks about how to use wearables as sensors in healthcare solutions. Neil Savage provides an engineering-like description of how some are searching for new measures of health and patient behavior with devices that others have deemed “clinically irrelevant.”
The article’s focus is how to use the data generated by mobile technology and wearables, but I took away three key points from a systems engineering perspective that I believe inform the “how to do” with wearables as opposed to just the “what to do”.
The first point is the "signal to noise" discussion. Savage calls this out in the highlight box – the much larger and diverse data sets that widespread smart phone connected devices generate allows researchers to “find signals in noisy data.” The massive amount of data that wearables and mobile phones generate creates a “signal analysis” environment for researchers that previously was not easily accomplished with targeted trials. Wearable data can be accessed through tools like the Apple and Google app stores, as mentioned in the article, which means that the data very quickly comes from a population as opposed to a study group. The mass of data supports statistical analysis that is analogous to traditional signal analysis techniques used in the engineering and design of sensor based control systems.
Second, when the researchers consider the behaviors and physiological characteristics of interest, we see that these devices can be used to measure patient parameters in both real time and over time. The article mentions using audio streams from phones to muscle tone via voice stability, mental states via the combination of voice quality and social interactions, and behavioral changes through the signatures of both wearable data and interactions with apps on the smart phones. Again, these are classic “spectral analysis” and “edge detection” techniques wherein the engineer breaks down signals into components to better understand the system. In this case researchers are breaking down the raw activity signals provided by the devices to derive behaviors or even physical characteristics of the patient that they can then correlate to health conditions.
Finally, the discussion about designed experiments combining laboratory instruments with consumer wearables, smart phones and prompted user feedback is analogous to creating computer aided design (CAD) models for complex systems and processes that in turn allow the design, test, and deployment of control systems, in this case chronic care therapies. The researchers at Northeastern University are using wearables combined with laboratory instruments to create new models of the patient that then allow them to use those sensors to inform the design of the therapy. Of particular interest in this case is the combination of “over time” data from the wearables with “real time” from the laboratory. Clearly the volume of “over time” data carries great value in the understanding of the patient and their condition.
As Northeastern’s Stephen Intille stated, “It's totally different from the way we've dealt with health and medicine in the past.” Sounds like real innovation is on the way.