Temporal risk of heart failure hospitalisation during multisensor alert state: a self-controlled case series analysis of 1.7 million ambulatory person days
L Witten Davodian, D Davodian, F H Fussing, N D Johansen, D Modin, T Bekfani, P Kumarathurai, N Risum, C Chien, M Fudim, A Hernandez, B Claggett, L Koeber, C M Stolen, T Biering-SoerensenAbstract
Background
A novel implanted cardiac device-based multisensor monitoring algorithm has proven predictive of heart failure (HF) decompensation, but the temporal risk of hospitalisation during state of alert has yet to be quantified.
Purpose
To quantify the within-patient temporal risk of acute HF hospitalisation during multisensor alert state.
Methods
In a retrospective analysis, we studied nationwide data from 1,749 patients implanted with a specific multisensor algorithm compatible ICD or CRT-D device. The algorithm daily generates a composite HF index based on multiple sensor inputs. When exceeding a fixed, prespecified threshold (16), an alert is triggered and once again resolved when the index recovers below an another fixed, prespecified threshold (6). States of alert were defined as continuous periods from alert onset to resolution. Each patient's alert days were compared to the same patient's non-alert days, eliminating confounding by any fixed patient characteristic (age, sex, comorbidity, device type). Alert state was not displayed to patients or clinicians and did not trigger any clinical action. Alert state was modelled as a time-varying exposure on ambulatory person-days. The primary endpoint was acute HF hospitalisation. Within-patient conditional Poisson regression with patient-level stratification (self-controlled case-series approach), cluster-robust standard errors, and log person-time offset was used. A log person-time offset ensured that incidence rate ratios (IRR) reflect risk per day, making comparisons across risk windows of different length valid. Prespecified windows from alert onset (days 0-3, 4-7, 8-14, 15-30, ≥31) quantified temporal risk (Figure 1).
Results
Among 1,729,459 ambulatory person-days, 8.3% occurred during active alert episodes, and 277 acute HF admissions were observed. The overall incidence rate ratio (IRR) during alert versus non-alert was (IRR 3.3, 95% CI 2.9–3.7). Risk was highest within 0–3 days of alert onset (IRR 3.9, 95% CI 2.4–6.4), attenuated during days 4–7 (IRR 2.2, 95% CI 1.1–4.4) and 8–14 (IRR 2.1, 95% CI 1.2–3.5), and remained elevated during days 15–30 (IRR 3.1, 95% CI 2.2–4.2). Among persistent alerts (≥31 days), the IRR remained high (IRR 4.2, 95% CI 3.4–5.2)(Figure 1B). Eighteen percent (n=23) of alert-state hospitalisations occurred within the first 7 days after alert onset, 33% (n=43) between 8-30 days, and 49% (n=63) after ≥ 31 days.
Conclusions
Multisensor alert states define high-risk periods for acute HF hospitalisation, with risk concentrated in the first days after alert onset. Sustained elevation among persistent alerts likely reflects unresolved decompensation. These findings support the potential utility of alert-triggered workflows prioritising rapid assessment after onset and continued follow-up for non-resolving alerts.Figure 1Temporal Risk