Owing to the number of levels to the factor x-ray type (34), this could be due to insufficient data.
The x-rays were classified by the part of the body irradiated (x-ray type), such as chest x-rays.
Furthermore, it does not allow for the effects of the x-ray type on the performances of individual junior doctors.
The data collected included the date, patient's age, x-ray type (eg, scaphoid), radiographer's triage result, junior doctor's diagnosis and reference diagnosis.
In contrast, x-ray type was not a significant covariate for either logit sensitivity (LRT=34.4, df=24, p=0.078) or logit specificity (LRT=23.3, df=33, p=0.89).
This would suggest that any differences in performance between the high- and low-prevalence populations are unlikely to be due to differences in the relative proportions of x-ray type.
Features considered abnormal on an x-ray depended on the x-ray type and included fractures (skeletal x-rays), cardiomegaly (chest x-rays) and dilated bowel (abdominal x-rays), thus covering a range of target disorders and are detailed elsewhere.
As any effects of prevalence on performance may be explained by differences in performance across different x-ray types, the interaction between the prevalence and x-ray type was evaluated.
Since the x-ray type is an indicator of the type of target disorder and therefore patient, this implies that the diagnostic performance does depend to some degree on both the type of x-ray being interpreted and the target disorder being sought.
Instead, physicians and diagnosticians have to print out the digital images on X-ray-type film or laser paper.
