Table 2.

Summary performance estimates given for the independent significant covariate, prevalence. Also given are the estimates of sensitivity for each level of the covariate x-ray group, which was significant for the dependent variable logit (sensitivity)

Model estimates of performance characteristics in significant covariates
	High prevalence	Low prevalence
Sensitivity (%)
Soft tissue x-rays	93.7 (79.5 to 98.3)	68.3 (44.3 to 85.3)
Appendicular x-rays	97.3 (93.3 to 99.0)	84.0 (70.3 to 92.2)
Axial skeletal x-rays	58.6 (17.3 to 90.5)	17.0 (2.4 to 63.1)
Summary	95.8 (91.1 to 98.1)	78.3 (65.7 to 87.2)
Specificity (%)
Summary	56.0 (41.9 to 69.2)	92.3 (90.0 to 94.2)
Positive likelihood ratio
Summary	2.2 (1.6 to 3.0)	10.2 (7.6 to 13.8)
Negative likelihood ratio
Summary	0.07 (0.03 to 0.17)	0.23 (0.14 to 0.38)
Diagnostic Odds ratio
Summary	37.3 (3.6 to 101.3)	36.1 (21.0 to 62.3)

All estimates are derived from the hierarchical regression model and take into account variation in performance between individual doctors and different x-ray groups. The covariate x-ray group has three levels: soft tissue (chest and abdominal x-rays), appendicular (limbs, hands and feet) and axial (skull, spine and sacrum). Interaction terms were not significant. 95% CIs are shown in brackets.