The effect of underreporting response bias on the assessment of psychopathology

A recent review offered the conclusion that the utility of considering response bias--in particular, underreporting, or defensiveness--in assessing psychopathology with structured inventories has not been demonstrated in practical criterion-related situations. The present research tested for the presence of suppressor and moderator effects in 3 such mental health classification tasks (disordered participants vs. control participants) using relevant predictors. Regression analyses showed the presence of one or both effects for all 3 tasks. The practical effects of these response biases were then illustrated by examining 4 measures of predictive accuracy at different levels of the response-bias measure: predictor/criterion correlations, hit rates, sensitivity/specificity, and positive/negative predictive power. It was concluded that there is clear utility in taking account of underreporting response bias in practical situations, particularly where extreme response bias is likely to be present. Confirmation of these findings with additional response-bias measures is needed.