Computerized detection of third heart sounds improves sensitivity for the emergency department diagnosis of heart failure

Abstract

Study objectives: We determined the diagnostic test characteristics of physician auscultation versus computerized detection of a third heart sound (S3) in emergency department (ED) patients with possible heart failure. We hypothesized that computerized S3 detection would be more sensitive and specific than physician auscultation for a primary discharge diagnosis of heart failure. Methods: Patients were enrolled at 2 community EDs and 1 tertiary care ED if they presented with symptoms of heart failure, including dyspnea. Diagnostic test characteristics of an S3 were described for 2 methods, both performed shortly after ED arrival and before treatment: (1) a validated algorithm from an acoustic heart sound recording (Audicor, Inovise Medical, Inc.); and (2) physician auscultation and notation of findings on a structured case report form. A primary diagnosis of heart failure was based on the final hospital discharge summary. Data are reported with 95% confidence intervals (CIs). Results: Of 135 patients enrolled, 46 patients had a primary diagnosis of heart failure, and 68 patients had no heart failure. Black patients composed 56.3% and male patients 46.9% of the sample. Median age category was 61 to 65 years; 46 patients were younger than 55 years, and 13 patients were older than 85 years. For a primary diagnosis of heart failure, sensitivity of S3 by physician auscultation was 17.8 (95% CI 9.3 to 31.3), whereas for the Audicor algorithm it was 41.3 (95% CI 28.3 to 55.7). Specificity was 98.5 (95% CI 91.9 to 99.7) and 86.8 (95% CI 76.7 to 92.9), respectively. Physician auscultation of an S3 had a positive predictive value of 88.9 (95% CI 56.5 to 98.0). For the Audicor algorithm, it was 67.9 (95% CI 49.3 to 82.1). Negative predictive value was similar for both, 63.7 (95% CI 54.1 to 72.4) and 68.6 (95% CI 58.2 to 77.4), respectively. Conclusion: This pilot study suggests that the S3 detected by the Audicor algorithm is more sensitive to heart failure than physician auscultation. The specificity is slightly reduced. Such data, along with further refinements in this new technology, may prove valuable in the ED evaluation of patients with possible heart failure.