Can Fitbits be used in the clinical arena? It depends on what data points you’re looking for, and how you use them.
Fitbit, iWatch, Pebble, Garmin, Spire, Withings… As I continue to add trackers to my wrists and pockets, I wonder: What’s the point?
For telehealth, the answer seems implicit – care from a distance ￼￼￼￼requires information to be gathered remotely to provide a complete representation of a patient’s condition. And, logically, if some data are good, more are better. Health trackers promise more data points, more access to healthcare in remote areas and greater granularity to detect changes in health measures like heart rate, blood pressure or weight. However, few medical conditions are diagnosed or managed purely by a number, and some compare these devices to a “hammer looking for nails” within the healthcare industry. With standards of measure that differ between manufacturers and some which misread during aggressive activities or with changing skin conditions, it’s debatable whether these devices may ever be considered medical grade. In turn, fitness companies are wary of their use outside of activity tracking, mostly for liability reasons. So it is no surprise that we have yet to see these devices make an impact on the hundreds-of-billion-dollar industry of chronic health conditions like heart disease, asthma or stroke, let alone acute care. Furthermore, with minimal gains realized in the years since their implementation, and the slowing growth of users, are we arriving at the cusp of the first contraction of the wearables field?
Despite all of these hurdles, a recent case report brought new life to the argument and suggested a model for wearables’ use in medical care. In the March 29, 2016, issue of Annals of Emergency Medicine, a FitBit heart-rate tracker was used to determine the time of onset of atrial fibrillation in a patient who otherwise could not remember the timing of his symptoms due to a grand-mal seizure. The emergency physicians were able to utilize the data from his phone to identify when he likely went into fibrillation, allowing for safe cardioversion in the ED – an intervention that would not have been possible without the data provided by his heart rate monitor. This case study introduced the potential role of wearable devices’ data as a valuable element to be mined from a patient’s history. So, what was different?
Instead of attempting to predict a medical condition with machine learning, or alter a behavior, the device, and its user interface were used to help answer a clinical question: when did the patient’s heart rate change? Rarely is artificial intelligence, appification, or best-fit modeling able to tease apart the nuanced complexities of a patient’s conditions as a physician can in mere moments spent in the room. No individual reading gave a diagnosis, but the comparison of pre- and post-seizure made identification of an accurate therapeutic window possible. I am reminded of patients who bring in sheets of paper with multiple hand-written readings of blood pressure and weight from home, using their personal devices and those from the pharmacy. Although unfamiliar with the machines used to measure these points, I can gain access to the patient’s natural history and pattern of disease, beyond their current state or vital signs. This insight helps me to narrow differentials, suggest treatment plans or prevent decompensation. Similarly, while the above physician knew little of the specifics of the device, he looked for a particular change in reading and found it, enabling him to provide safe personalized care.
Extrapolating from here one might wonder, how many steps is the coronary artery disease patient taking before tiring due to chest pain? How much weight has a CHF patient gained in the last week? How long was an epileptic patient seizing before arriving at the hospital? These are only three of the many questions that arise every day in the emergency department. In attempting to answer them, I am often left pursuing invasive and expensive modalities like CT scans or admission to the hospital for EEG. Preventing just one CHF exacerbation could mean a savings of almost $40,000. With the correct contextualization, data points supplied by health trackers could curtail rising healthcare costs or be key factors between life and death. However, we can only intervene if we know why the patient gained weight or decreased her walking distance. It could be a CHF exacerbation, or she could merely be on vacation. The data alone can be misleading, no matter its accuracy or precision.
But if we stop expecting the data to bring about a paradigm shift in healthcare on its own, and instead allow these into our clinical workflow as and when needed (even one patient at a time), we might begin to realize the benefits of the wearable revolution we have been expecting. Just like physicians are trained to understand the patient’s verbal narrative, perhaps we should incorporate the information from these devices into that history. Their impact on patient care, and the medical field as a whole, will depend on the person wearing it and why they are at our office.
Some data are sparse but accurate (ie. Individual patient visits over years) while others are regular but less precise (ie. Home blood pressure and heart rate monitoring). Either way, a numbers-only approach to health is difficult. However, if physicians are cognizant of the technologies available and understand their strengths and limitations, we might be able to interpret this data in much the same way a vague complaint is utilized. While it may not be accurate or precise, we can determine its utility based on the scenario at hand.
Much has been made about whether these devices are able to be used in the treatment of disease, and generally, fitness companies are wary of their use outside of activity tracking. Their standards of measure differ between manufacturers and may misread during activities or when skin conditions change. However, much like choosing to listen to a heart with a stethoscope or to ultrasound at the bedside, the usefulness of data is often found in the clinical scenario. The numbers these devices supply alone do not enable us to understand where a patient is along their spectrum of disease, however, given the right clinical scenario they can be amazingly powerful.
Ultimately, the utility of wearable health trackers to affect meaningful clinical outcomes will depend on who is interpreting their data and why. The context is king!
Great article! We need to focus on data that is portable (doc to patient, doc to other doc, patient to doc, etc) and that is secure. We also need to see evidence that new tech interventions are increasing our patients’ health.
The team at Mt. Sinai that built the asthma app were not just able to increase the “n” of their studies, but they were also able to increase medication compliance and schedule timely patient visits based on their data findings. (http://apps.icahn.mssm.edu/asthma/) Similarly, the team at UCSF behind Health eHeart is focused on the best way to use massive cardio data to direct individual care. (https://www.health-eheartstudy.org/study)