Background: Extended-duration ambulatory ECG monitoring enables detection of transient arrhythmias but is affected by signal degradation over time. Characterizing how ECG signal quality evolves across wear is essential for interpreting downstream algorithm performance.
Methods: Fourteen-day continuous single-lead ECG (Zio Patch; iRhythm, San Francisco) was collected from patients who participated in R01HL137617. ECG recordings were segmented into 10-second intervals, with the first 5 minutes of each hour analyzed (n = 720/day). Signal quality for each segment was assessed using a CNN–LSTM model, which we previously validated. Signal quality was quantified as the proportion of interpretable segments relative to total segments and summarized using segment-level weighting. Temporal trends in signal quality were evaluated at the daily and subject levels. Days with substantial signal degradation were defined as those with >20% poor-quality segments. Longitudinal changes in signal quality were assessed using linear mixed-effects models with day as a fixed effect and patient as a random intercept to account for within-subject correlation. Pairwise within-patient comparisons with Wilcoxon signed-rank test were used to compare key time points (Day 1, Day 7, and Day 14).
Results: Among 23 patients contributing 229,380 ECG segments, the overall weighted mean for interpretable segments was 94.5%, and 2.8% of days (9/322) exhibited >20% poor-quality segments. Mean daily signal quality remained high but declined over time, from Day 1 98.2 ± 3.6%, Day 7 97.2 ± 4.2%, and Day 14 92.1 ± 8.2%. In longitudinal models, signal quality decreased significantly over time (β = −0.69% per day, p < 0.001). Within-patient comparisons showed no significant difference between Day 1 and Day 7 (mean change −1.13%, p = 0.051), but signal quality was lower on Day 14 than on Day 1 (mean change −8.85%, p < 0.001)
Discussion: Extended-duration ambulatory ECG monitoring maintains high signal quality though declines modestly by day 14.