Do I sound sick? Self-perceived and objectively measured voice impairment in acutely decompensated heart failure: fist insights from the AHF-Voice study

Abstract Introduction Acute heart failure (AHF) is a common reason for recurrent hospitalization. Daily smartphone-based recording of vocal biomarkers appears promising for the early detection of impeding AHF. However, the value of patient-reported outcome measures (PROMs), i.e. self-perceived voice alterations during decompensation, is unknown. Methods AHF-Voice is an ongoing monocentric prospective cohort study focusing on voice alterations in patients hospitalized for AHF. Exclusion criteria: listed for heart transplantation, high output HF, cardiogenic shock, history of vocal fold disease or surgery, or life-expectancy <6 months. During hospitalization, medical history & echocardiogram were taken and routine blood sampling was done. Daily voice recordings including sustained vowels and standardized reading were collected using a specially designed smartphone app under supervision. At admission and discharge, the self-perceived level of voice alteration was assessed by the Voice–Handicap Index-12 (VHI-12, 0-48 points) questionnaire and a global voice scale (Likert scale, 0-3 points). For both scales, higher values indicate a more severe voice impairment. Hypervolemia was quantified by body composition analysis. Preprocessing and feature extraction of voice recordings were performed using Praat software. The association of self-perceived voice impairment with surrogates of congestion and vocal biomarkers was assessed by bivariate correlation and linear regression. Results Between Apr 2023 and Feb 2024, 72 hospitalized patients with AHF were included. Four patients withdrew consent, 5 patients were excluded due to limited voice recording quality. Hence, 63 patients were considered for analysis: mean age 75±10 years, 67% men, 91% were in NYHA class III/IV, 49% de novo HF, mean LVEF was 45±16%, and median value of Charlson comorbidity score was 2 [1, 3]. Both measures of self-perceived voice impairment improved between admission and discharge (Table 1A and Figure). Concomitantly, body weight, hypervolemia, NT-proBNP, and vocal biomarkers (total reading time and shimmer) decreased. Bivariate correlation showed that both self-perceived scales of voice impairment were closely correlated, as expected (Spearman’s rho 0.36). Linear regression identified surrogates of congestion and objective voice impairments as predictors for change in both PROMs (Table 1B). However, the explained variance was low, e.g. for change in volume status and NT-proBNP 14% and 1% (VHI-12), and 4% and 7% (Global voice scale), respectively. Conclusion This preliminary analysis in patients with AHF showed that self-perceived voice impairment was sensitive to change during recompensation. Objectively generated voice measures and congestion only partially explained the observed improvement of these PROMs. This might indicate that the self-perceived voice impairment constitutes a complimentary domain of patient well-being and indicator for congestion that should be explored further.