PT - JOURNAL ARTICLE AU - Elphick, H E AU - Lancaster, G A AU - Solis, A AU - Majumdar, A AU - Gupta, R AU - Smyth, R L TI - Validity and reliability of acoustic analysis of respiratory sounds in infants AID - 10.1136/adc.2003.046458 DP - 2004 Nov 01 TA - Archives of Disease in Childhood PG - 1059--1063 VI - 89 IP - 11 4099 - http://adc.bmj.com/content/89/11/1059.short 4100 - http://adc.bmj.com/content/89/11/1059.full SO - Arch Dis Child2004 Nov 01; 89 AB - Objective: To investigate the validity and reliability of computerised acoustic analysis in the detection of abnormal respiratory noises in infants. Methods: Blinded, prospective comparison of acoustic analysis with stethoscope examination. Validity and reliability of acoustic analysis were assessed by calculating the degree of observer agreement using the κ statistic with 95% confidence intervals (CI). Results: 102 infants under 18 months were recruited. Convergent validity for agreement between stethoscope examination and acoustic analysis was poor for wheeze (κ = 0.07 (95% CI, −0.13 to 0.26)) and rattles (κ = 0.11 (−0.05 to 0.27)) and fair for crackles (κ = 0.36 (0.18 to 0.54)). Both the stethoscope and acoustic analysis distinguished well between sounds (discriminant validity). Agreement between observers for the presence of wheeze was poor for both stethoscope examination and acoustic analysis. Agreement for rattles was moderate for the stethoscope but poor for acoustic analysis. Agreement for crackles was moderate using both techniques. Within-observer reliability for all sounds using acoustic analysis was moderate to good. Conclusions: The stethoscope is unreliable for assessing respiratory sounds in infants. This has important implications for its use as a diagnostic tool for lung disorders in infants, and confirms that it cannot be used as a gold standard. Because of the unreliability of the stethoscope, the validity of acoustic analysis could not be demonstrated, although it could discriminate between sounds well and showed good within-observer reliability. For acoustic analysis, targeted training and the development of computerised pattern recognition systems may improve reliability so that it can be used in clinical practice.