Biases in the assessment of diagnostic tests

Variance estimation for epidemiologic effect estimates under misclassification

This paper presents basic methods for construction of variance estimators for epidemiologic effect estimates when adjusting for misclassification. Methods are described for differential and non-differential misclassification, and for external and internal estimates of classification rates. The methods take account of sampling variability in both the observed data and the estimated classification rates. The methods are illustrated in a case-control analysis of the association of antibiotic use during pregnancy with sudden infant death syndrome.

A general regression methodology for ROC curve estimation

A method for applying generalized ordinal regression models to categorical rating data to estimate and analyze receiver operating characteristic (ROC) curves is presented. These models permit parsimonious adjustment of ROC curve parameters for relevant covariates through two regression equations that correspond to location and scale. Particular shapes of ROC curves are interpreted in relation to the kind of covariates included in the two regressions. The model is shown to be flexible because it is not restricted to the assumption of binormality that is commonly employed in smoothed ROC curve estimation, although the binormal model is one particular form of the more general model. The new method provides a mechanism for pinpointing the effect that interobserver variability has on the ROC curve. It also allows for the adjustment of ROC curves for temporal variation and case mix, and provides a way to assess the incremental diagnostic value of a test. The new methodology is recommended because it substantially improves the ability to assess diagnostic tests using ROC curves.

Statistical uncertainty due to misclassification: implications for validation substudies

Certain studies incorporate a validation substudy as an integral part of their design. In the substudy, the results of a primary but error-prone measurement are compared with the results of a more accurate (but more difficult or costly) criterion measurement. The results of this substudy are then used to evaluate the impact of errors in the primary measurement on study validity. The present paper shows that the alternative of a fully-validated design (i.e. one that obtains criterion measurements on all subjects) may provide more information per unit cost than a larger study coupled with a validation substudy. Several formulas are provided to aid in detecting such a situation, and illustrated in the design of a case-control study of sudden infant death syndrome (SIDS).