Diagnostic performance for heart failure events of HeartLogic algorithm: real-world data
M Pereira Santos, S Campos, M I Oliveira, M Trepa, M J Sousa, C Gomes, B Braganca, M Rajao, B Delgado, A Pinheiro Vieira, H Reis, A Luz, M SantosAbstract
Background
Heart failure (HF) remote monitoring using parameters from cardiac implantable electronic devices (CIED) had shown a remarkable diagnostic performance in their pivotal studies. We aimed to characterize the alert burden in an outpatient remote monitoring HF program and to assess its diagnostic performance in a real-world cohort of patients with HF.
Methods
We conducted a retrospective observational study of patients with an activated HeartLogic® algorithm between February 2022 and July 2023. Data on the number of alerts and duration of continuous days under alert were collected. A composite outcome including all-cause deaths, HF hospitalizations and unplanned emergency room visits due to HF was studied. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the algorithm were calculated.
Results
We studied 40 patients with HF with reduced ejection fraction (mean LVEF 31±9%) who had a CIED (60% CRT-D and 40% ICD) with a mean age of 68 ± 9 years (75% male), 45% had ischemic etiology, 76% were in NYHA functional class II. Over a median follow-up of 1.5 years, there were 38 individual alerts (0.65 alerts per patient-year). Among patients who experienced alerts, the mean duration of the alert status was 38 days and the median Heart Logic Index while in alert was 25. The alert state was 6.8% of the total follow-up duration. HF patients who had an alert had similar baseline features to those without alerts. The overall composite event rate was 23.3 events / 100 person-years. Four events occurred during days on alert (event rate of 8.3% per month vs 1.1% per month during days not in alert), denoting an incidence nearly 8 times higher of adverse events during high-risk days. The algorithm for the composite event had a sensitivity of 63% with a PPV of 28%, a specificity of 60% and a NPV of 86%.
Conclusion
The HeartLogic algorithm generated a low burden of alerts, resulting in low consumption of healthcare professionals’ time. The diagnostic performance of our cohort is consistent with findings from pivotal studies, supporting the algorithm as a valuable tool for HF risk stratification.