Making wearables useful for digital health

The trouble with a lot of wearable tech and accompanying apps is the bone-jarring lack of clinical context and prosaic health application. They provide no context; how does your exercise pattern specially affect your hypertension? What do your EP & GP think? How reliable and accurate is the data being provided?

These days the proliferation of cheap wearables reveals even the dumbest marketing deadbeat to be enamoured with heart rate training zones completing a glossy green tick for wearables packaging. Of course, accurate and reliable 12 cord ECG equivalent heart rate data has far more value than simple training load management. We are of the view that basic exercise data and blood pressure and glucose measures from FDA-approved devices is a reasonable basis for information feedback to at-risk individuals and their clinicians. We also believe that the higher the fidelity of the data source for example heart rate data, the more useful the potential assessments may be. Which for us, means a measure for heart rate variability (HRV) that has an appropriate level of fidelity and can be integrated into our machine learning engine such that our clinicians can be assisted in making better informed, in situ adjustments to prescriptive plans.

HRV

Heart rate variability is a measurement more often associated with high performance sport than those at risk of developing metabolic disorders. A groundswell of recent research may change this. Heart rate variability (HRV) informs of physiological functioning, and in particular balance in the Autonomic Nervous System (ANS). The vagus nerve, the tenth of 12 paired cranial nerves controls parasympathetic innervation of the heart and will act to lower heart rate. Some of the calculated HRV parameters are measures of sympathetic and parasympathetic tone.

Parasympathetic tone in particular is of interest, since it correlates with several physical, mental and emotional parameters. A high vagal tone is reflected in:

 ? better working memory

 ? ability to focus attention (concentration)

 ? better blood sugar control

 ? emotional stabiliy

 ? sociability

A low vagal tone is reflected in:

 ? diabetes

 ? cardiovascular disease.

That's right, I have alluded to an association between vagal tone and diabetes. [5] in a landmark longitudinal study of Chinese workers spanning 4 years found participants with faster heart rates, suggesting lower automatic function, had increased risk of diabetes, pre-diabetes, and conversion from pre-diabetes to diabetes. Each additional 10 beats per minute was associated with 23 percent increased risk of diabetes, similar to the effects of a 3 kilogram per meter square increase in body mass index.

This is the tip of an iceberg of evidence pointing to the association between HRV and susceptibility to type 2 diabetes. [4] in another significant study with over 9,000 diabetic subjects determined that

Cardiac autonomic impairment appears to be present at early stages of diabetic metabolic impairment, and progressive worsening of autonomic cardiac function over 9 years was observed in diabetic subjects.

The gotcha with tracking heart rate variability currently is the lack of support afforded by wrist wearables such as the best selling fitbit range. It's easy enough to incorporate the necessary cardiac data from chest-strapped based device providers such as POLAR who provide a sound solution. We are planning a HRV function in 2.0 but we have yet to decide the best data source for it. 

Resting Heart Rate

Meanwhile, there is much we can do with our current AI-based analytic capability and resting heart rate (RHR). In a meta-review of RHR and T2DM, [1] found

a strong positive association between high resting heart rate and the risk of type 2 diabetes. As a non-invasive marker of type 2 diabetes risk, resting heart rate may have potential in the clinical setting, especially for interventions aimed at lowering the risk of type 2 diabetes.

In our experience designing digital health solutions the facility to include a simple RHR tracker with predictive capability can help

a) alter the individual and the clinician to adverse patterns and

b) train ML algorithms to better identify at-risk trends.

Behavioural Design Considerations

Whilst we have cranked out a reasonable solution to this problem space using Athrú I wont be sharing those secrets here. I will though share these user scenarios and questions that help shape our thinking and may kickstart some soul searching on your part.

Causal forensics

What's your link to the individual's clinical assessment data? How do you establish and manage personal and digital linkages with the clinician? Is this web of relationships confined to the primary care network or is there a role to be played by a key allied health practitioner and insurer?

How is test data entered and updated? How clean and reliable is the glucometer API? Hold on; a glucometer, is that necessary? Good question exactly what is the diagnosis and what role if any will a BT-enabled glucometer have in any medical intervention? 

How did the individual reach this clinical state? What have been the triggers (internal and external)? What is their level of motivation to change? What is their level of necessary competency around lifestyle, diet, exercise and stress management? How ready are they to embark on change? 

What are you measuring?

You think you are designing an app - God bless you....you are likely wrong. You are in fact designing a trusted nurse, coach, problem solver that adapts as the user's individual emotional, social and physical indicators play out in undulating waves over time. It's not a 12 week fix. 

What is personalisation?

We all think Recommender engines these days; understandably. The underlying algorithms are well tested and broadly used. However, we are really looking to ensure a truly adaptive UX right down to the persuasive design elements that are most effective for each individual. You need to ally your measurement strategy with your personalisation functionality; [2] and [3]. In situ measurement of behavioural outcomes of system features is crucial to adaptive personalisation profiling.

What place does "nudge" have to do with any of this?

I'm at best a half-arsed part fan of this populist junk-food style science but it has real application in some public health interventions and of course persuading those who can ill afford it to consume more crap. In terms of the whole RHR-HRV Type 2 diabetes mellitus and impaired glucose tolerance app design space there are elements of nudge that can be gainfully applied; visual framing around social proof; anchoring and priming and loss aversion to name a handful. 

Stay tuned - the next post will be around Personal Savings Orientation, TTM and shifting towards a better FIN-tech app for savings.

References

[1] Aune, D., B. ó Hartaigh, and L.J. Vatten. "Resting Heart Rate And The Risk Of Type 2 Diabetes: A Systematic Review And Dose–Response Meta-Analysis Of Cohort Studies". Nutrition, Metabolism and Cardiovascular Diseases 25.6 (2015): 526-534. Web.

[2]Kaptein, M. and van Halteren, A. (2012). Adaptive persuasive messaging to increase service retention: using persuasion profiles to increase the effectiveness of email reminders. Personal and Ubiquitous Computing, 17(6), pp.1173-1185.

[3]Kaptein, M., Markopoulos, P., de Ruyter, B. and Aarts, E. (2015). Personalizing persuasive technologies: Explicit and implicit personalization using persuasion profiles. International Journal of Human-Computer Studies, 77, pp.38-51.

[4] Schroeder, E. B. et al. "Diabetes, Glucose, Insulin, And Heart Rate Variability: The Atherosclerosis Risk In Communities (ARIC) Study". Diabetes Care 28.3 (2005): 668-674. Web.

[5] Wang, L. et al. "Resting Heart Rate And The Risk Of Developing Impaired Fasting Glucose And Diabetes: The Kailuan Prospective Study". International Journal of Epidemiology 44.2 (2015): 689-699. Web.

[6] Zhang, X. et al. "Resting Heart Rate And Risk Of Type 2 Diabetes In Women". International Journal of Epidemiology 39.3 (2010): 900-906. Web.