What to do with Contradictory Data? Approaches to the Integration of Multiple Malingering Measures

Abstract

This study describes the potential problems and possible solutions to the integration of multiple malingering measures. Multivariate prediction models, using both discriminant function analyses and regression tree approaches, are compared. Study measures, including an abbreviated version of the SIRS (SIRS-A), the MMPI-2, the TOMM and the VIP Verbal subtest, were administered to 29 community members instructed to malinger and 87 psychiatric patients instructed to respond honestly. Predictive accuracy varied substantially across measures and the correlations between tests ranged from .19 to .79. Further, 48% of the psychiatric sample were misclassified as malingering by at least one test and 46% of the malingering sample were classified as honest by at least one test; “unanimous” findings occurred in only half of the cases. Multivariate models identified the SIRS-A as the strongest predictor of malingering, but the MMPI-2, TOMM, and VIP provided significant contributions to these models. The implications of these findings for the problem of multiple, contradictory indicators in general, and the specific problems associated with clinical assessments of malingering in particular, are discussed.