Trends in Consumer Wearable Devices with Cardiac Sensors in a Primary Care Cohort

Wearable devices are common (1) and often record data relevant to cardiac health, including physical activity, heart rate, electrocardiograms, and atrial fibrillation (AF) episodes (2). The degree to which wearables are discussed in healthcare settings remains unknown.

We assessed mention of devices in provider notes as a surrogate for how often they are discussed in healthcare settings, using a cohort of over 520,000 longitudinal primary care patients defined by presence of two primary care visits between 1 and 3 years apart (3). We included all outpatient notes with >250 characters between 2005 and 10/28/2019, excluding duplicates, notes with copied-forward device mentions, and notes for select participants in a trial utilizing AliveCor devices (4). We used regular expressions to search for devices including AliveCor Kardia, Apple Watch, Fitbit, Garmin, and generic terms “smartwatch,” “fitness tracker,” and “step tracker”. Search terms were chosen after review of the literature and manufacturer marketing materials.

Patient-level factors associated with device mention were assessed using logistic regression, with covariates including: age at cohort entry, sex, race, ethnicity, median zip code income (5), and baseline comorbidities (≥2 International Classification of Diseases-9 or -10 [ICD] codes for AF/atrial flutter [AFL], other arrhythmias, pacemaker/implantable cardioverter-defibrillator [PPM/ICD], coronary artery disease [CAD], peripheral arterial disease [PAD], heart failure [HF], stroke/transient ischemic attack [TIA], diabetes mellitus [DM], hypertension, hyperlipidemia, and obesity). Complete cases were analyzed (n = 4,744 excluded for missing zip code). Encounter-level factors associated with device mention were assessed using mixed effects logistic regression for all notes belonging to patients with ≥1 device mention, with a random effect per patient and fixed effects for same-day cardiology or primary care encounters, calendar year, and same-day ICD codes for the above comorbidities. Odds ratios were adjusted for all covariates. The study was approved by the Mass General Brigham Institutional Review Board. Data supporting these findings are available from the corresponding author upon reasonable request.

Of 492,192 patients with 21,013,729 notes, there were 298,820 (60.7%) female patients and 373,745 (75.9%) White patients. Mean age was 48.4 ± 16.9 years. Common comorbidities included hypertension (n = 159,022, 32.3%) and hyperlipidemia (121, 492, 24.7%). AF/AFL was present in 15,964 (3.2%) patients and 36,862 (7.5%) patients had other arrhythmias. Annual mention of wearables increased from 1/620,872 (0.0002%) notes in 1/143,789 (0.0006%) patients in 2005 to 2,863/2,064,535 (0.14%) notes in 2,222/248,498 (0.89%) patients in 2019 (Figure 1). Fitbit was most common (n = 6,366 patients, 74.1% of patients with devices) and accounted for a rapid rise in device mention between 2012 and 2016, before declining in 2017.

Figure 1. Temporal Trends in Patients with Wearable Device Mention.

A) Annual number of total patients in the cohort with at least one note. B) Percent of patients with at least one note mentioning a wearable device, with release years of select devices/features annotated. ECG: electrocardiogram; OTC: over-the-counter; Rx: prescription.

Patient-level factors associated with device mention were female sex (odds ratio [OR] 1.43 [95% CI, 1.36–1.50]), higher income (OR 1.27 per $50,000 [95% CI, 1.27–1.31]), obesity (OR 2.34 [95% CI, 2.21–2.48]), AF/AFL (OR 1.72 [95% CI, 1.54–1.93]), hyperlipidemia (OR 1.40 [95% CI, 1.33–1.48]), other arrhythmias (OR 1.29 [95% CI, 1.19–1.39]), and hypertension (OR 1.28 [95% CI, 1.21–1.35]). Relative to self-identified White individuals, other races had lower odds of device mention: Asian OR 0.59 (95% CI 0.51, 0.67), Black OR 0.54 (95% CI 0.49, 0.60), Not Recorded OR 0.45 (95% CI 0.31, 0.66), and Other OR 0.41 (95% CI 0.28, 0.60). Self-identified Hispanic individuals had lower odds of device mention compared to non-Hispanics (OR 0.65 [95% CI 0.46, 0.92]). Other negative associations included HF (OR 0.56 [95% CI, 0.47–0.66]), PPM/ICD (OR 0.70 [95% CI, 0.50–0.98]), PAD (OR 0.71 [95% CI, 0.61–0.82]), stroke/TIA (OR 0.78 [95% CI, 0.67–0.91]), and age (OR 0.92 per 10 years [95% CI, 0.90–0.93]).

Note-level factors associated with device mention were cardiology (OR 4.74 [95% CI, 4.43–5.06]) or primary care (OR 1.88 [95% CI, 1.80–1.96]) encounters, calendar year (OR 1.33 per year [95% CI, 1.31–1.34]), obesity (OR 2.94 [95% CI, 2.79–3.09]), AF/AFL (OR 1.46 [95% CI, 1.36–1.56]), other arrhythmias (OR 1.67 [95% CI, 1.55–1.79]), hypertension (OR 1.17 [95% CI, 1.12–1.23]), and hyperlipidemia (OR 1.48 [95% CI 1.40–1.57]). Negative associations included PPM/ICD (OR 0.39 [95% CI, 0.29–0.52]), HF (OR 0.56 [95% CI, 0.48–0.65]), CAD (OR 0.75 [95% CI, 0.62–0.90]), and stroke/TIA (OR 0.75 [95% CI, 0.62–0.90]). In an analysis of devices with AF detection algorithms during the study period (AliveCor and Apple Watch), the association was stronger for calendar year (OR 1.75 per year [95% CI, 1.69–1.80]), cardiology encounters (OR 7.45 [95% CI, 6.59–8.43]), AF/AFL (OR 1.75 [95% CI, 1.56–1.96]), other arrhythmias (OR 2.07 [95% CI, 1.86–2.31]), and weaker for stroke/TIA (OR 1.01 [95% CI, 0.73–1.39]), hyperlipidemia (OR 1.16 [95% CI, 1.02–1.32]), and obesity (OR 1.09 [95% CI, 0.90–1.33]).

Of 120 notes reviewed (20 per device) for algorithm performance, positive predictive value was 92.5%. Four invalid matches were for “Garmin” and the remainder were ambiguous or present only in note metadata. Device mentions predominantly related to activity monitoring/weight loss (42, 37.8%) and suspected arrhythmia (30, 27.0%).

Among primary care patients in an academic healthcare system, notes referencing wearables increased between 2005 and 2019 with nearly 1% of patients having documentation of a device by 2019. Device mention was less frequent among older patients with more comorbidities, and was more frequent among White, non-Hispanic, and female patients, and those with higher estimated income.

Our study is observational and does not distinguish between patient-initiated (e.g., reported abnormal Apple Watch ECG) or provider-initiated (e.g., recommendation of Fibit for weight loss) device mentions. Additionally, we do not have granular data on specific device models or features. Future study is warranted to assess the impact of wearables on healthcare costs, disparities, and outcomes.

Abbreviations:

AF

Atrial Fibrillation

AFL

Atrial Flutter

CAD

Coronary Artery Disease

DM

Diabetes Mellitus

FDA

Food and Drug Administration

HF

Heart Failure

ICD

International Classification of Diseases

OR

Odds Ratio

PAD

Peripheral Arterial Disease

PPM/ICD

Pacemaker/Implantable Cardioverter-Defibrillator

TIA

Transient Ischemic Attack