Fig. 4. The juxtaposition of correct (P=ZW∪CW) answers for each sound class for doctors of different specializations and medical students.
The differences for individual sound classes were analyzed based on a Kruskal-Wallis test and post hoc tests with a Holm correction for statistically significant differences. According to the analysis, pulmonologists distinguish the sound classes in a better way when compared to the other groups - as many as three classes are significantly recognized differently by the group of pulmonologists relative to the other specializations: fine crackles (chi2(4) = 11.9, p = 0.018), stridor (chi2 = 14.4, p = 0.006) and rhonchi (chi2 (4) = 11.1, p = 0.026).
Analyzing false positives (defined as cases of indicating a sound class when this class was not indicated in the standard- this is the sum of all the incorrect selections that were not in the given sample, with the individual sound classes being treated independently of each other) the graph shows no differences between groups, except for the prolonged expiratory phase - the only sound class with statistically significant differences (chi2 (4) = 13.1, p = 0.011). This difference can be found in the results obtained by pulmonologists and students.
In the next step, the responses were grouped using the main classes proposed by the European Respiratory Society (ERS) (Pasterkamp et.al, 2016). In practice, this meant that if the participant had indicated any of the subgroups shown in Table 3 and any of them was marked as the correct one in the standard, the answer was treated as a correct one. In the case of the vesicular breath sound or the bronchical one, no other answer that is not part of the subgroup could be concurrently marked. In this way, five main classes of sounds were created.
Table 3. Main classes of respiratory sounds and subclasses that are part of them
main class
subclasses included in the main class
breath sound / bronchial sound
vesicular breath sound
diminished breath sound
louder breath sound
normal bronchial sound
abnormal bronchial sound
no subclasses
crackles
fine crackles
medium crackles
coarse crackles
crepitus
wheezes
inspiratory wheezes
expiratory wheezes
stridor
rhonchi
no subclasses
After the grouping described above, a graph of the correlation between the correct answers for each main class was obtained (Fig. 5). As one can see, for all the classes associated with pathological signals, the pulmonologist group is the most effective one. However, statistically significant differences based on the Kruskal-Wallis test and a comparison of the groups of participants in pairs were obtained only for the class of rhonchi between the group of pulmonologists who scored higher than pediatricians (p = 0.085), other physicians (p = 0.046), interns (p = 0.085) and medical students (p = 0.015). For wheezes, pulmonologists had statistically significantly higher scores than other specializations (p = 0.008). Also interns had better scores than physicians of other specializations (p = 0.026). Generally, it can be stated that this grouping highlighted the advantage of pulmonologists over the rest of the groups in the correct recognition of the respiratory sounds phenomena.