This study aimed to provide an overview,
as accurate as possible,
of the current mean skill level of radiology residents in reading chest radiograph.
These results show that a majority of radiology residents have a general background in CXR allowing the achievement of less elaborate tasks such as distinguishing normal from abnormal findings.
However,
the rather poor performances of residents to provide a diagnosis,
especially for abnormal CXR (47.4% of true diagnoses),
is a first signal suggesting room for improvement in teaching chest radiography.
Residents performed better in the interpretation than in the detection CXR category to determine the normality of CXR and to decide on the need of further CT study.
This may be explained by the obvious aspect of pathologic findings in the interpretation CXR category,
while abnormalities in the detection CXR category were subtler.
However,
when assessing the rate of true diagnoses,
we observed no significant difference,
and even an inverse tendency with only 45.7% of CXR truly diagnosed in the interpretation category.
These findings indicate an overall insufficient level of knowledge of CXR semiology but need to be investigated with regard to training characteristics of residents.
The third level of analysis of our study,
i.e.
the capacity to provide a true diagnosis,
offered the most discriminating tool to study factors influencing residents’ skills.
Advanced residents performed better than novice,
especially for detection task,
which appears consistent with previous studies showing that skill level grew with experience [10–14].
Previous attendance at CXR training course during residency improved residents’ skills related to CXR interpretation,
while it did not improve performances related to CXR detection.
These findings confirm the idea that,
beyond the detection task,
CXR reading involves an interpretation task which is mainly driven by prior learning of radiographic patterns as well as understanding of the rules of image formation.
The link between theoretical course and skill related to CXR interpretation task,
already demonstrated by others [13],
should encourage further thought about the means of promoting CXR teaching during residency.
In our study,
only 18% of residents had benefitted of such training course,
and the great majority of them (more than 91%) expressed a need for further CXR education.
We did not find any influence of the number of CXR read per week,
especially on residents’ detection skills.
In conclusion,
this study warns us about radiology residents’ lack of theoretical background in CXR reading,
and demonstrates the role of training course in improving residents’ skill related to interpretation task.
Conversely,
the number of CXR read per week seems not to influence residents’ skill.
These results may suggest promoting systematic CXR theoretical training course in the curriculum of radiology residents.