A multisensory algorithm for detection of upcoming congestion in chronic heart failure patients

Abstract Background Heart failure (HF) is characterized by relatively stable periods that are interrupted by burdensome episodes of fluid retention requiring hospitalization. These admissions are a strong marker for disease progression and pose a burden on patients and healthcare resources. Timely detection of fluid retention, to enable pharmacological intervention, is the pillar in preventing hospitalizations. The multisensory cardiac implantable electronic device (CIED) based HeartLogic™ algorithm can alert in case of upcoming congestion. The aim of the current analysis is to evaluate the performance of the HeartLogicTM algorithm in a real-world ambulant heart failure population. Methods Consecutive HF patients with an activated HeartLogic™ algorithm on their CIED were included for analysis. All patients were included from 01-01-2018 until 01-02-2023, and followed-up according to the HF carepath (figure 1). The cumulative HeartLogic™ index automatically generated an alert if the index surpassed the preset threshold (≥16). An alert was deemed true positive (≥2 signs/symptoms of congestion on top of the alert) or be false positive (≤1 signs/symptoms). Without an alert a patient was either true negative (≤1 signs/symptoms) or false negative (≥2 signs/symptoms). To determine the algorithm performance in daily practice a logistic regression model with linear mixed models was used. Furthermore, patients with ≥65% true positive alerts and ≤1 false positive alert per year were compared to patients without alerts to identify characteristics of patients who benefit most from the HeartLogicTM algorithm supported management. Results Of all 145 eligible patients 139 patients could be included for analysis (6 patients had an LVAD), 77% was male and median age of 70[60-77] year. The majority of the patients had a CRT-D (65%) and 49% had an ischemic etiology of HF. Follow-up consisted of 330 patient years (median 2.4 years). During the study period, 289 HeartLogicTM alerts were observed, 15 were excluded (incomplete clinical information). The remaining 274 alerts were included for analysis. Majority (n=200, 73%) of alerts were true positive for fluid retention. The remaining 74 alerts were false positive. The positive predictive value to detect upcoming congestion was 72%, the negative predictive value 98%. The sensitivity was 86% and the specificity 88%. Patients with HeartLogic™ alerts had a significantly lower left ventricular ejection fraction (p<0.05), higher degree of mitral regurgitation (p<0.05) and higher levels of NT-proBNP (p<005), when compared to patients without alerts (Figure 2). No differential response was observed based on age, gender or etiology of heart failure. Conclusions HeartLogic™ algorithm supported care adequately detects impending fluid retention in a real-world heart failure population. Patients who benefited most had a lower left ventricular ejection fraction, more severe mitral regurgitation and higher NT-proBNP levels at baseline.The heart failure carepathLVEF, MR grade and NT-proBNP levels