Home-use devices that record a brief single-lead ECG during blood pressure measurement can flag rhythm irregularities labeled as “possible AF”. The clinical pathway from a device alert to patient presentation and diagnostic workup is not well characterized. This sub-analysis of the Omron Heart Study evaluated rates, timing, and correlates of hospital visits following device-generated possible AF alerts in older adults with treated hypertension.
The parent study was a prospective, multicenter, observational trial conducted in Japan from April 2022 to July 2023. Enrollment targeted adults aged ≥60 years with physician-treated hypertension. Recruitment used a decentralized approach: invitations through the OMRON Connect smartphone application and additional enrollment via a study website to capture non-app users. Electronic informed consent was obtained online, and a Complete device (Omron Healthcare) was mailed to each participant.
Participants were instructed to perform two paired blood pressure and 30-second single-lead ECG recordings each morning and evening for 3 months. The OMRON Connect app displayed results; when the device algorithm detected a rhythm consistent with possible AF, the application showed a possible AF alert and advised consulting a physician. The study implemented no active clinical follow-up or alert confirmation for participants; weekly adherence reports were sent automatically, and call-center staff contacted participants only if adherence fell to 0% for three consecutive days.
All device-flagged 30-second tracings were adjudicated after the 3-month recording window by six board-certified cardiologists working in blinded review. Each tracing was read independently by three randomly assigned cardiologists; AF was confirmed if at least two adjudications indicated AF. Discordant cases were reread and resolved by consensus. Adjudication results were not communicated to participants and thus did not influence their immediate care-seeking behavior.
Primary outcomes for this analysis were (1) self-reported visit to a healthcare facility within 3 months prompted by a device alert and (2) additional diagnostic testing performed. Questionnaire responses at 3 and 12 months captured visits, testing, new AF diagnosis, and subsequent interventions. Univariable and multivariable logistic regression models assessed patient-intrinsic predictors of visiting a healthcare facility; device-extrinsic factors such as alert frequency were analyzed using univariable approaches. Time from first alert to visit was modeled with Poisson or negative binomial regression when appropriate.
Of 4,078 enrollees, 258 were excluded for insufficient measurements, leaving 3,820 participants in the analysis set. During the 3-month recording window, 1,700 participants received at least one device-generated possible AF notification. Physician adjudication confirmed AF in 220 of these 1,700 alerted participants (12.9%).
At 3 months, questionnaire response among alerted participants was 98.9% (1,682 respondents). Within 3 months, 399 participants (23.5%) reported visiting a healthcare facility in response to the device alert; 289 (17.0%) reported additional diagnostic testing; 35 (2.1%) reported a new AF diagnosis; and 18 (1.1%) reported initiation of anticoagulation. By 12 months, an additional 80 participants reported new AF diagnoses, bringing the total with a clinical AF diagnosis to 115—equivalent to 52.3% of the 220 adjudicated AF cases. Between 3 and 12 months there were also small numbers reporting electrical cardioversion (2 participants), catheter ablation (14 participants), and further anticoagulation starts (10 participants).
In multivariable logistic regression examining patient-intrinsic baseline characteristics, the following factors were associated with significantly lower odds of visiting a healthcare facility within 3 months: age <65 years (OR 0.75, 95% CI 0.59–0.95), current alcohol use (OR 0.73, 95% CI 0.57–0.95), absence of daily palpitations (OR 0.69, 95% CI 0.50–0.95), and absence of palpitations on the alert day (OR 0.51, 95% CI 0.30–0.85). Past smoking (versus never) was associated with higher likelihood of visiting. The authors evaluated interactions among key variables but found none that were both statistically significant and clinically interpretable.
Analyses of device-extrinsic factors showed that the number of possible AF detections on the first alert day was positively associated with visiting within 3 months (OR 1.25 per additional detection). Presentation of possible AF on two consecutive days was associated with greater odds of visiting (OR 2.22). These alert-frequency metrics did not show a significant association with time from alert to visit.
Among 399 participants who reported visiting a healthcare facility within 3 months, 175 provided valid visit dates within the observation window. In this subset, the median time from first device alert to medical visit was 28 days and the mean was 36.4 days. No candidate covariates were significantly associated with the time to visit in the models applied.
Despite device advice to consult a physician after a possible AF alert, fewer than one in four participants actually reported visiting a healthcare facility within 3 months, and fewer than one in five underwent additional testing. The median delay to presentation was roughly one month. Only 12.9% of device alerts were adjudicated as AF, which frames the low follow-up rate in context: many alerts were not confirmed as AF. The population studied was recruited in part from OMRON Connect app users and volunteers, which the authors note may represent relatively health-conscious individuals; real-world follow-up rates in broader consumer populations could be lower.
The study was observational and designed to capture naturalistic care-seeking behavior; no active reminder or clinical follow-up was provided following alerts. ECG adjudication occurred after the recording window and results were not returned to participants, so adjudication did not influence behavior. Data underlying the analysis cannot be publicly shared owing to privacy constraints, per the approving Ethics Review Board. The time-to-visit analyses were limited by missing or out-of-window visit dates, leaving 175 participants with valid dates out of 399 who reported visits.
In this cohort of older adults with treated hypertension using a home BP monitor with ECG function, follow-up after device possible AF alerts was low: 23.5% reported a healthcare visit and 17.0% underwent further testing within 3 months. Younger age, alcohol consumption, and absence of palpitations predicted lower likelihood of seeking care, while greater alert frequency increased the likelihood of visit. A substantial fraction of adjudicated AF cases lacked a clinical diagnosis at 12 months. The results underscore gaps between mHealth detection and clinical action and suggest a need to improve the positive predictive value of alerts and develop strategies to link device alerts to timely clinical evaluation.