Abstract
Background: The TriageHF algorithm provides remote monitoring for heart failure (HF), but its clinical implementation is limited by a high rate of false positive alerts. Objective: To assess whether incorporating alarm duration and individual time at risk can improve the diagnostic performance of the TriageHF algorithm. Methods: This was a single-center prospective cohort study in which 37 patients with Medtronic ICDs implanted between January 2020 and June 2022 were enrolled. HF events were defined as episodes requiring intravenous diuretics or hospitalization. A total of 609 TriageHF alerts were analyzed. Two strategies were analyzed: a standard approach (high-risk and moderate-risk alerts > 7 days considered positive) and a modified approach (high-risk and moderate-risk alerts > 15 days considered positive). The relationship between time spent in a low-risk state and the algorithm's positive predictive value (PPV) was also assessed. Results: In the standard configuration, sensitivity was 96.7% and specificity was 76.4%, with 81.9% of false positives. The modified approach showed improved specificity (85.6%) and PPV (24.3%), with minimal impact on sensitivity (87%) and negative predictive value (99.1%). There was a significant inverse correlation between time spent at low-risk and individual PPV (R(2) = 0.64, p = 0.018). Conclusions: Using a ≥15-day threshold improved the specificity and PPV of the TriageHF algorithm. Incorporating individual time at risk may further refine risk stratification and enhance cost-effectiveness in HF remote monitoring strategies.