Photoplethysmography-based smartwatches for atrial fibrillation detection in stroke survivors: The Liverpool-Huawei Stroke Study
S H M Lam, S L Harrison, Y Zheng, A Hill, T Hliang, Y Guo, D A Lane, G Y H LipAbstract
Background
Detecting atrial fibrillation (AF) in stroke survivors is critical for preventing recurrent stroke. However, the diagnostic accuracy of photoplethysmography (PPG)-based smartwatches in this high-risk population remains uncertain. The Liverpool-Huawei Stroke Study evaluated the performance of Huawei smartbands for AF detection following an ischaemic stroke.
Purpose
1. To determine the clinical effectiveness of Huawei smartband by assessing the negative predictive value (NPV) and positive predictive value (PPV) for identifying incident and known AF in a post-stroke population. 2. To identify clinical predictors associated with AF detection.
Methods
Between June 2022 and June 2024, patients hospitalised with a recent ischaemic stroke were recruited from 8 National Health Service Trusts. Participants wore a Huawei smartband paired with an Android smartphone for one month. The device automatically screened for AF at regular intervals, transmitting data via the Huawei Health App. Following alerts generated by the device, clinical evaluation with ECG and 5-14 days Holter or ECG patch monitoring were used to confirm AF diagnosis although these were not done concurrent with the alert.
Results
491 patients (median age: 62 [QR 54, 72]; 35% female) were recruited (Table). During the one-month monitoring period, 192 (39.1%) individual AF alerts were received, 15 (7.8%) were confirmed as incident AF, 114 (59.4%) were confirmed AF-negative, and 39 (7.9%) identified as known AF. (Figure). The resulting NPV and PPV were 0.89 (0.85, 0.93) and 0.32 (95% CI: [0.25, 0.40]), respectively. Multivariable logistic regression indicated that age (OR: 1.06, 95%CI [1.03, 1.11]), current smoker (0.35 [0.10, 0.96]) and systolic blood pressure (0.98 [0.96, 0.99]) were independently associated with AF detection (Figure).
Conclusions
The Huawei smartband revealed incident AF in 15 (7.8%) patients post-stroke with a high NPV (89%), but a low PPV (32.1%). This may reflect that ECG monitoring was not done concurrent with the alert and ectopic beats (common in this population) were flagged as an alert due to irregular PPG rhythm; however, the absence of an alert likely reflects the absence of AF post-stroke. Baseline age, smoking status, and systolic blood pressure were independent predictors for AF detection.Figure