Furthermore, timing (inspiratory versus expiratory) and localization errors (absent or incorrect data) were taken into account. The above data were then used to compute a lung sound description inaccuracy score (LSDIS) for each form. One (inaccuracy) point was assigned for each timing error, two for each localization error, and four for each undetected, misclassified or fabricated finding.
Validation of the method
To evaluate the reliability of the assessment of auscultation skills, a group of 12 respiratory medicine residents with 1—4 yrs of internship training were tested. Each subject was asked to report his or her findings on 12 patients, selected and assembled into four groups of three as described above, and the LSDIS was computed for each group as described. On a separate occasion, the auscultation skills of the same residents were evaluated using 12 digitally recorded lung sounds, selected from a published audio tape specifically designed for teaching purposes rate of 11 kHz using a PC IBM-compatible personal computer with 16-bit audio card (SoundBlaster 16, Creative Labs, Singapore), and played through headphones with the same computer using a program written with a simple multimedia authoring system (HSC interactive, HSC Software, Santa Monica, CA, USA). The reports were evaluated as described above, with the obvious exception that sound localization was not considered.
Lung sound session
A multimedia presentation of lung sounds was always given by one author (MR). The presentation lasted approximately 60 min, was held in a small quiet room, and was attended by no more than 10 students at a time. The session consisted of a simple introduction to the basics of lung sounds and presentation of a panel of auscultatory findings previously recorded in various patients. The recording/reproducing system consisted of a custom made Phonopneumograph (EPL, Siena, Italy) based on an Atari STE microcomputer motherboard coupled to a stereo AD/DA converter, capable of a resolution of 16 bits per channel at a sampling rate of 15.5 kHz. One channel was used for the acoustic data, whilst the other was used to store respiratory movements, recorded using a pneumatic belt secured around the chest connected to a pressure transducer. The audio output of the system was connected to a series of analogue and digitals filters, amplified and fed to a speaker located in front of the audience at a distance of 6 feet from the front row at a height of 4 feet. The display was a 14 inch black and white monitor that was located just in front of the audience and was clearly seen by all the students.
Each sound was played several times and its acoustic characteristics were underscored by showing its time expanded waveform together with the respiratory timing tracings and the graph of the acoustic power spectrum.
The origins of the sound were discussed in relation to its
acoustic properties, and the clinical conditions in which it was encountered were briefly reviewed. A minimum of eight sounds was presented in each session, covering normal and abnormal breath sounds (reduced breath sounds, bronchial breath), continuous adventitious sounds (wheezes and rhonchi), different types of crackles (fine, coarse, early inspiratory and late inspiratory). The students were encouraged to actively participate by requesting explanations whenever necessary, posing specific questions, and asking for replay of previous sounds.
Comparison of the number of detected, undetected and misclassified findings reported by each student in the two groups was performed using the Wilcoxon rank test. The homogeneity of the performance of residents and of the different groups of students in the baseline test was tested using Kruskal-Wallis one-way analysis by rank for discrete variables, and by one-way analysis of variance (ANOVA) on the inaccuracy score. The Spearman rank test was used to assess the correlation between LSDIS obtained in patients and on computer transmitted lung sounds. The effect of the multimedia presentation on LSDIS was assessed by means of ANOVA, using the score obtained in the baseline test as a covariate . A two-tailed a of 0.05 was used as the level of significance.
Validation of the method
Multifactor ANOVA of LSDIS obtained by 12 residents on four different groups of patients failed to detect significant differences in LSDIS between the groups of patients (p=O. 17), whilst a significant difference (p=0.03) was present between residents. Average LSDIS scored by individual residents ranged 3.6—16.6, with a mean of 8.1±1.0. These data indicate that our method was able to detect significant differences in auscultation skills among the residents, and that the criteria for selecting and assembling the patient groups were sufficiently reproducible. When these results were compared to the scores obtained by the same doctors evaluating computer recorded sounds, most of them tended to perform slightly better with the latter, and no significant correlation was found between the LSDIS obtained by the same subject with the two methods (Spearman’s rho= 0.41; NS).
Effect of multimedia presentation on students
Fifty-four students agreed to participate in the study, 29 attending the multimedia seminar and 25 the control bedside teaching session; 48 of them completed the study, 27 in the multimedia and 21 in the control group.