Intra- and interobserver variation in lung sound classification. Effect of training.

Abstract

This study explores how last year medical students at the University of Tromsø, the Arctic University of Norway, interpret and describe different lung sounds. This is done by measuring intra- and interobserver variation in agreement among 16 students, when reporting abnormal lung sounds after listening to audio recordings. Agreement with a reference standard is included, and testing of effect on training on these agreements. To test the training effect the students were separated in two groups, one of them having an intervention, a 3 hour course. The results serves to inform the medical society about the inconsistency in reporting lung sounds in this particular population, and hopefully also help finding measures to obtain better agreement. Cohens kappa have been used to measure intraobserver agreement and agreement with the reference standard, Fleiss kappa to measure interobserver agreement. An “exact” Mann-Whitney U test for testing the effect of the course. The kappa level of agreement set to define acceptable agreement is “moderate”, with a lower limit of .41. The results indicate highly acceptable intraobserver agreement, and the agreement tended to improve in both the intervention and the control group. The agreement with the reference standard was also highly acceptable for the category wheezes and acceptable for crackles and the abnormal category. A tendency to positive change in the intervention group when compared to the control group was found, but the difference was only statistically significant for the abnormal category in the agreement against reference standard. The interobserver agreement did not reach the limit of acceptable, except for wheezes. Summarized, a weak effect of the intervention was observed.