The future of healthcare could be even more data-driven and personalized than it is today. Instead of determining that a patient has a light fever based on a thermometer reading of 37°C, a diagnosis would be based on a comparison of an elevated resting body temperature against a similarly observed baseline from that individual. Rather than an arbitrary (or worse, financially driven) model where patients visit their doctors for sparsely scheduled checkups, people may receive care based on a cadence optimized to their own unique situations.
Such a utopian future requires both patients and doctors to be equipped to more closely monitor and communicate any changes in chronic conditions and also be more proactive about preventive care.
With the array of continuous-monitoring wearables already on the market or in a prototyping or R&D phase1, technologists are certainly continuing to push the envelope in terms of the types of data that could be collected non-intrusively. And medical professionals are stepping into the digital health world. In a recent study on the efficacy of wearable health monitoring, researchers who monitored 100 patients who underwent major surgeries at Cedars-Sinai Medical Center in Los Angeles, California, found that activity monitors improved accuracy in step count assessment. Higher step count (up to 1000 steps) on the day after the surgery was, in turn, associated with 63% lower odds of a prolonged hospital stay, according to the research, which was covered this spring by the American Medical Association (AMA).2 In fact, the AMA has created a Digital Health Implementation Playbook that provides step-by-step guidance, best practices, and resources to, in its words, "accelerate the adoption and scale of digital health solutions."
However, we're still only taking baby steps towards personalized, decentralized healthcare. After all, it takes a lot of recorded data, associated with ground truths, before this information can be distilled into insights that could be applied to future data. For example, to develop algorithms that accurately interpret a given data set as steps, researchers had to collect data from pedestrians on a variety of parameters: ages, weights, and stride lengths; different types of shoes worn; and their walking patterns on different surfaces and inclines and at different speeds. In addition, they had to gather other data not associated with any steps to make sure that the algorithms won't falsely interpret them as steps.
Building trust for wearables in the healthcare world involves increasing the understanding among doctors of how the devices themselves work and also of data analytics principles.
"Garbage in, garbage out" may be a common adage, but it is very much applicable when it comes to continuously monitored data. To say that we can get personal health insight from continuous monitoring presumes a "chain of trust." In other words:
- The interpretation of any data must not only be accurate but reliable. The challenge lies in handling "borderline" data. Any interpreting strategy or algorithm faces data sets that it finds ambiguous. For an algorithm to be reliable, users must be able to quantitatively understand its detection limits and error characteristics.
- The data and/or its interpretation must reliably reach the decision maker for it to become actionable.
- The data must be correctly associated with historical records of the patient for it to have context.
- The data must be proven to be authentic to trigger any meaningful action.
The remainder of this article takes a look at what is needed in order for wearable healthcare devices to attain a level of trust so that it could augment clinical data to inform healthcare decisions. The article also highlights what needs to happen from a healthcare infrastructure standpoint to enable this trust.
Ensuring Healthcare Wearables Are Worn, Not Abandoned
According to projections by IHS, global wearable product shipments are expected to exceed 200 million units in 2019—doubling their volume over a six-year period. More significantly, the value of wearable devices is increasing as the total revenue growth of the market is faster than its unit volume growth.3 Applications have expanded from fitness and exercise monitoring to continuous monitoring of blood-oxygen levels, blood-glucose levels, and electrodermal activities (for epilepsy management).
From a device standpoint, a number of issues related to system and mechanical design must be addressed before wearables became ingrained into the daily lives of even more people. The biosensors must be designed to overcome measurement challenges stemming from factors such as skin tone, sweat, motion, blood perfusion, and ambient light. These factors may be unique to the form factor and use case for each variable. For example, compared to the wrist, the outer ear canal has higher blood perfusion and is less prone to vigorous motion. So an in-ear device like an earbud can be more effective in capturing more optical heart-rate signals than, say, a wrist-worn device. However, comfort for long-term wear and a consistently good sensor-to-tissue contact may present a different set of challenges.
Whatever the form factor, wearable sensors must be able to perform reliably while overcoming the effects of common noise and error sources. For a better user experience, the devices should run for long periods between charges and dissipate little heat. A truly non-invasive health monitor must also be comfortable to wear and convenient to use. A product that is bulky, unattractive, or requires the wearer to acquire special postures or new habits in order to capture measurements is more likely to be shoved into a drawer than worn continuously. This means that the underlying technologies must be miniaturized, low power, and have some level of automation and/or simplicity in how data is collected and analyzed.
Then there's the regulatory approval matter. In the U.S., for example, devices intended to diagnose, prevent, cure, or mitigate disease will need approval from the country's Food and Drug Administration (FDA). So far, there really are just a handful of devices that have passed the test. However, the FDA has launched a digital health software precertification pilot program that could eventually help streamline the approval process. It should be noted that FDA approval by itself may not be the definitive seal of quality. Wearable devices that provide continuous monitoring are subjected to environments far different from clinical settings, and even the subjects' physiological conditions could vary considerably over different activities. As the saying goes, "Your results may vary."
Building the Chain of Trust
One of the first steps toward building the chain of trust is to establish baselines that can serve as guidance of accuracy. This would involve comparing data collected by wearables to data gathered using clinical equipment in a controlled setting for a population of people. The clinically derived data would serve as the truth to validate the accuracy of the wearable data. This step is essential, yet fraught with challenges.
For one, clinical equipment is designed for use under clinical settings only. Using clinical equipment to capture vital signs that are representative of the wearable use cases is often difficult and sometimes inaccurate. Careful selection of the population of test subjects as well as construction of the representative use cases are both very important. If these are not truly representative, there's the risk of unleashing a wave of false positives or false negatives.
Also, the key to personalized healthcare is to compare data from the patient's own history or baseline. However, unlike for machinery, the baseline for a patient is not static. Factors such as growth, aging, and medication could all materially alter these baselines. Our current lack of understanding of the normal variations and the significance of deviations from these baselines could erode our ability to trust continuously monitored healthcare data.
One of the advantages of data collected by wearable devices is that it provides a continuous record over a period of time.
Data authentication and privacy are big, yet unresolved issues. Patients need to trust that their sensitive information will be used as intended and not get into the wrong hands. Their doctors, in turn, need to know that data on which they base their diagnosis are trustworthy and the limits of any potential inaccuracies are well understood. We have addressed this problem in our traditional healthcare institutes. Doctors, hospitals, testing facilities, and equipment are well certified and operate in well-defined parameters. When a patient brings files from one doctor to another, the receiving doctor has some notion on how much he or she could trust the information presented.
Alas, this is not yet true for most wearable devices. Only recently have disposable monitoring devices been adopted for pre- and post-operative care. They represent the first steps toward continuous, personalized medical monitoring on a limited basis.4
Embarking on a More Holistic Approach to Healthcare
Doctors have traditionally made their diagnoses based on tests as well as information gleaned from talking to and observing their patients. To fully incorporate data that has been continuously collected by wearables into this landscape, doctors will need at least a basic understanding of engineering principles (to help them trust the devices) and data analysis (to help them trust the data). On the medical education side, some of this work is underway. For example, Stanford Medicine conducted a pilot project called Humanwide from January to December 2018. In this study, primary care teams at the school's Primary Care 2.0 clinic worked with 50 patients who were monitored holistically, including via mobile monitoring devices. An outcome of this program was to demonstrate how feasible it is to provide healthcare in a comprehensive, patient-centered, data-driven manner.
In the near-term, it's likely that continuous monitoring will be incorporated into more patient care programs for short durations and targeting very specific objectives. Indeed, from an algorithmic development standard, it is more realistic at this time to look for a specific signal (an expected change) from the sea of data collected. Doing so, however, can lay the groundwork for a future where wearables take on a more integral role in patient health.