A Known-Groups Evaluation of the Response Bias Scale in a Neuropsychological Setting

Abstract

We evaluated the Minnesota Multiphasic Personality Inventory-Second Edition (MMPI-2) Response Bias Scale (RBS). Archival data from 83 individuals who were referred for neuropsychological assessment with no formal diagnosis (n = 10), following a known or suspected traumatic brain injury (n = 36), with a psychiatric diagnosis (n = 20), or with a history of both trauma and a psychiatric condition (n = 17) were retrieved. The criteria for malingered neurocognitive dysfunction (MNCD) were applied, and two groups of participants were formed: poor effort (n = 15) and genuine responders (n = 68). Consistent with previous studies, the difference in scores between groups was greatest for the RBS (d = 2.44), followed by two established MMPI-2 validity scales, F (d = 0.25) and K (d = 0.23), and strong significant correlations were found between RBS and F (rs = .48) and RBS and K (r = −.41). When MNCD group membership was predicted using logistic regression, the RBS failed to add incrementally to F. In a separate regression to predict group membership, K added significantly to the RBS. Receiver-operating curve analysis revealed a nonsignificant area under the curve statistic, and at the ideal cutoff in this sample of >12, specificity was moderate (.79), sensitivity was low (.47), and positive and negative predictive power values at a 13% base rate were .25 and .91, respectively. Although the results of this study require replication because of a number of limitations, this study has made an important first attempt to report RBS classification accuracy statistics for predicting poor effort at a range of base rates.

Key words:

• inadequate effort
• malingering
• MMPI-2
• neuropsychological assessment
• Response Bias Scale

ACKNOWLEDGEMENTS

Funding for this project was provided by the School of Psychology and Counseling, Queensland University of Technology. The research protocol was assessed and approved by the Queensland University of Technology Human Research Ethics Committee (QUT Ref. No. 0900001042) and by occupational workplace health and safety officers.

The authors wish to thank Mr. Tony Grima for assistance with data retrieval.

Notes

TBI = traumatic brain injury; PI = personal injury litigation; WC = workers' compensation claim; WC* = 10% of this sample was seeking compensation; DP = disability pension; ACR = avoidance of criminal responsibility; MNCD criteria = diagnostic criteria for malingered neurocognitive dysfunction (Slick et al., 1999); HHI = Henry-Heilbronner Index (Henry, Heilbronner, Mittenberg, & Enders, 2006); FBS = Fake Bad Scale (Lees-Haley, English, & Glenn, 1991); RBS = Response Bias Scale (Gervais, Ben-Porath, & Green, 2007); TOMM = Test of Memory Malingering (Tombaugh, 1996); WMT = Green's Word Memory Test (Green, 2003); MSVT = Green's Medical Symptom Validity Test (Green, 2004); ROC = receiver-operating characteristic; MMPI-2 = Minnesota Multiphasic Personality Inventory-2; PPP = positive predictive power; NPP = negative predictive power; ADHD = attention-deficit hyperactivity disorder; SVT = Symptom validity test.

*Dionysus et al. (2011) express their cutoff as ≥14, but for consistency with the data reported in Table 5, this cutoff is reported here as >13.

^§These articles used an early prototype of the RBS, not the final version of this scale.

¹This article has a published correction (Gervais, Ben-Porath, Wygant, & Sellbom (2010a).

²Postscreening diagnostic category percentages.

¹Wygant et al. (2010) report RBS cutoffs using T-scores; a T-score of 100 approximates an RBS raw score of 17.

Note. N = 83 (15 poor-effort, 68 genuine responders). In cases where patients had multiple diagnoses, they were categorized according to their most prominent diagnosis.

²The analyses described in this article were rerun without these two participants. These exclusions did not change the interpretation of results. Analyses including these participants are reported.

Note. N = 83 (15 poor-effort, 68 genuine responders); df = 81; RBS scores are T-scores.

*p < .05.

Note. N = 83 (15 poor-effort, 68 genuine responders).Nk = Nagelkerke's pseudo R²; e ^b  = exponent b; Mod. χ² = omnibus chi-square test results for the whole model with 2° of freedom; Block χ² = chi-square test results for the specific block with 2° of freedom; % = overall percentage of group classifications correctly predicted by the model.

*p < .05. **p < .01. ***p < .001.

Note. N = 83 (15 poor-effort, 68 genuine responders). A cutoff of ≥17 was recommended by Gervais et al. (2007), and a cutoff of 19 was recommended by Whitney et al. (2008); a cutoff of ≥14 was recommended by Dionysus et al. (2011). Wygant et al. (2010) recommend cutoffs between T-scores of 90 and 100, which correspond to a raw score of approximately 15 to 17.

³Dionysus et al. (2011) reported a PPP of 1.00 at a cutoff of 17.

⁴Dionysus et al. (2011) reported an NPP value of .57 at a cutoff of 17.

The views expressed in this article are those of the authors and do not reflect the official policy of the Department of Defense or U.S. government.