Watching Patients — Incorporating Wearables Into Care
TL;DR
•Digital health is a growing area that enables more home-based testing and care, convenience, and more scalable access to knowledge
•ECG, sleep assessments, computer vision, and virtual triage represent initial digital tools patients are likely to use and bring to internists’ attention
• Performance varies by tool and modality, but over the last 5 years tools have increasingly advanced in their accuracy
•Data biases increasingly will be important to recognize and address
•Regulatory bodies are ‘catching up’ but do not cover all technologies today
Tech industry trends show that in the 2020, more than 66 million smart watches were sold. That’s on top of nearly 60 million in 2019. Those figures describe a trend that we are seeing in the clinic — people increasingly have access to devices that combine hardware (sensors) and software (algorithms) that turn these watches into capable clinical devices.
At Froedtert & Medical College of Wisconsin Health Network, we had begun to integrate device data through smartphones, namely Apple Healthkit and Google Fit. This allows patients to share blood pressure, heart rate, and glucose logs back to clinicians. We will plan on expanding these capabilities in the near future.
Within the last couple of years, patients are now making use of wearable devices beyond just taking and reporting these straightforward measurements. In the clinic, we are now seeing patients come in with notifications from wearables.
Here, we present few cases of wearables in the clinic.* (*These cases are composite cases from ones we have seen or are constructed as learning examples.) A version of this was presented at the MCW General Internal Medicine CME Conference in 2020.
Part 1 — Hearts A Flutter
Our cases begins with a57 y/o man who presents to his PCP after his watch alerted him to the possibility of an abnormal heart rhythm. He had been doing a brisk walk with his dog when he received the alert. He was asymptomatic at the time, other than a report of general fatigue. He had no palpitations, shortness of breath, or chest pain.
His past medical history is notable for hypertension, for which he takes lisinopril 10mg daily.
On examination, his BMI is 38 kg/m2. His blood pressure is 142/84, heart rate is 80, and oxygen saturation is 99% on ambient air. His heart sounds are regular, without murmurs. He has no elevated jugular pressure
How heart rhythm wearables work
Wearables such as Apple’s watch work by one or more of two ways. To track heart rates, and also the rhythm or cadence of them, photoplethysmography is used. This technology is used in a similar fashion in pulse oximeters. Light is emitted of a certain wavelength, and then is reflected off of red cells as they come nearer a s photodiode sensor as the pulse wave from systole moves down past the wrist. The computer in the watch then is capable of translating these sensor signals into heart rate, and it can then also process whether the signal is normal (regular rhythm) or abnormal (irregular rhythm, such as in atrial fibrillation).
The second way, which is present on newer models, is to create a single lead electrocardiogram (ECG). The bottom of the watch serves as one end of the circuit, and pressing a finger of the contralateral arm to the watch crown is able to complete the circuit. The computer in the watch then is able to reduce the noise and smooth the signal into a single lead ECG. Algorithms can then assess whether the tracing is that of normal sinus rhythm or atrial fibrillation.
The Research
Questions came out early about how this technology would be used. Would it be accurate, or would it provoke more anxiety in people unnecessarily? Would it actually lead to overdiagnosis of atrial fibrillation (where people who might have rare episodes and at less risk of stroke would be diagnosed and treated, without any clear benefit)? Or would it help diagnose cases early before strokes arose?
A study from Stanford provided the first large scale look at how Apple’s watch worked. The study invited watch owners to enroll as long as they did not have pre-existing atrial fibrillation, ultimately enrolling 419,297 patients.
If a patient received an irregular pulse notification (like our case example, and which occurred in half a percent of all participants, but more in older patients), they received an additional ECG patch to wear for seven days. Of these, atrial fibrillation was present in 34% of users.
In summary, a watch notification was fairly predictive, especially if a second one occurred (in the study population, people with a first notification received a patch, and then if they had a second notification that could be correlated using the ECG patch the positive predictive value was 0.84 (95% CI 0.76–0.92)
Essentially, as pre-test probability rises (increased age as just one, but also risk factors like hypertension like our patient has), the likelihood that the result is a true positive rises.
A study published in February 2021 in JAMA Cardiology randomized sensor wearing (via a patch though, instead of a watch) among older patients. This study showed a ten fold increase in detection of AF compared with a control group that measured blood pressure twice daily (and where the cuff had an ability to detect an abnormal heart rhythm).
While longer term studies are needed to say whether these screenings reduce incident stroke or lower mortality, we can say that these tools are fairly accurate in the right clinical settings and with the right pre-test probability.
Check back for Part 2 — Skin Cancer Screenings
