Meta-Analysis of the MMPI-2 Fake Bad Scale: Utility in Forensic Practice

ABSTRACT

Some clinical researchers disagree regarding the clinical utility of the MMPI-2 Fake Bad scale (FBS ) within forensic and clinical settings. The present meta-analysis summarizes weighted effect size differences among the FBS and other commonly used validity scales (L, F, K, Fb, Fp, F-K, O-S, Ds2, Dsr2 ) in symptom overreporting and comparison groups. Forty studies that included FBS were identified through exploration of online databases, perusal of published references, and communication with primary authors. Nineteen of the 40 studies met restrictive inclusion criteria, resulting in a pooled sample size of 3664 (1615 overreporting participants and 2049 comparison participants). The largest grand effect sizes were observed for FBS (.96), followed by O-S (.88), Dsr2 (.79), F-K (.69), and the F- scale (.63). Significant within-scale variability was observed for seven validity scales, including FBS (Q = 119.11, p < .001). Several subsequent FBS moderator analyses yielded moderate to large effect sizes and were statistically significant for level of cognitive effort, type of overreporting comparison group, and condition associated with overreporting (e.g., traumatic brain injury, posttraumatic stress, chronic pain). Findings suggest that the FBS performs as well as, if not superior to, other validity scales in discriminating overreporting and comparison groups; the preponderance of the present literature supports the scale's use within forensic settings.

Notes

Elhai et al. (2000) was excluded as it appears to have used the same simulating group as Elhai et al. (2001). The current study did not allow for redundant contributions of the same samples across studies. Elhai et al. (2001) was selected instead of Elhai et al. (2000) as the latter study was said to have taken data from the former.

The Posthuma and Harper (1998) study was excluded as it included comparison groups comprised of child custody litigation groups, a context within which underreporting of psychological symptoms is expected. Excluding these custody litigants left no other group comparisons to be made with the personal injury litigant sample.

Rogers et al. (1995) was excluded because it used a within-groups design; pre-post designs are thought to yield effect sizes that are not comparable to those obtained from independent samples (Lipsey & Wilson, 2001, p. 45).

A. Inclusion Criteria: ¹The study contained appropriate MMPI-2 data (i.e., n's, means, standard deviations) that would allow computation of effect sizes, or availability of other psychometric data (e.g., t, F, p values) that would allow for FBS effect size estimations.

² FBS was reported with adult participants in two or more independent groups. One of these groups was obtained in a forensic context known to increase the likelihood of over-reporting of symptoms (e.g., personal injury, workers' compensation, disability claims, or a simulation condition designed to mimic a forensic context), and there was independent evidence that the level of symptom over-reporting was greater than comparison group(s). Comparison groups were (a) either evaluated within a context in which the likelihood of over-reporting would not reasonably be expected (e.g., a normal control group; a clinical group evaluated outside of a forensic setting), or (b) if in an overreporting context, there was a likelihood that the comparison group(s) did not overreport symptoms to the same extent as the overreporting group.

³If the same data was included in multiple published studies, it was only represented once. Abstracts of conference presentations known to have been subsequently published in peer-reviewed journals were only represented once.

B. Exclusion Criteria: ¹Studies involving litigant groups that probably contained an unknown number of malingerers, which were compared only to other groups of exclusively malingering litigants or simulators (thus, precluding the calculation of an overreporting effect size), and for which there were no means of determining a differential degree of overreporting between groups, were excluded.

²Comparison groups evaluated within a context in which underreporting of symptoms (e.g., personnel selection, custody evaluation) would be expected were excluded.

³Study designs that employed FBS as an a priori measure of symptom validity to assign group membership were not included. Study designs that selected the comparison group on the basis of significantly elevated MMPI-2 scores were excluded.

Note. ^a Every study in the left column: met all inclusion criteria, did not meet any of the exclusion criteria, and contributed an independent mean effect size to the current meta-analysis.

^b Every study in the right column either did not meet one or more of the inclusion criteria or met one or more of the exclusion criteria.

Note. In the above four formulae, s _p  = pooled standard deviation, N = total sample size, n = number, g = group, s = standard deviation, ES _sm  = standardized mean effect size, ES′ _sm  = Hedges' (1981) unbiased effect size estimate, and w _sm  = inverse variance weights. Formulae are from Lipsey and Wilson (2001, pp. 47, 49).

Note. N = the number the number of contributing means summed across overreporting and comparison groups. Studies that did not report these data, or whose report of demographics was unclear, did not contribute to these variables. Due to the nature of comparisons made, some individuals were counted in both overreporting and comparison groups. The total number of participants across all overreporting groups was 1615, and the total number of participants across all comparison groups was 2049.

Note. k = number of effect sizes; Q _T  = within-scale heterogeneity.

*p < .001.

Note. MND = malingered neurocognitive dysfunction (Slick et al., 1999); PTSD = posttraumatic stress disorder; PTS = posttraumatic stress; TBI = traumatic brain injury; MVA = motor vehicle accident.

*Subsequently excluded as a statistical outlier.

Note. k = number of effect sizes; Q _W  = within-category heterogeneity, Q _B  = between-category heterogeneity.

^a Moderator comparisons are orthogonal; each study's effect size(s) contribute(s) to only one moderator category.

^b For the known category, studies either: (1) included a priori cognitive effort measures and/or the Slick et al. (1999) criteria of malingered neurocognitive dysfunction to establish a likelihood of cognitive malingering between groups, or (2) reported post hoc analyses that discriminated over-reporting and comparison groups on the basis of insufficient/sufficient cognitive effort.

^c For the simulator category, subjects were confederates, often students, enrolled in experimental studies and instructed to feign or exaggerate symptoms.

^d This moderator refers to studies comparing over-reporting bias against a number of other comparison groups. OR = a comparison of a group in an over-reporting context with a group that has a lesser degree of overreporting. NOR = a comparison group not in an overreporting context, which did not show evidence of overreporting.

^e Includes individuals who had multiple conditions, and groups in whom individuals had different conditions. These included, mixed psychiatric conditions (e.g., depression, anxiety), traumatic brain injury, simulated “neurosis” and “psychosis,” neurotoxicity, etc.

Note. *p < .05

***p < .001.

*Contributed effect size(s) to the current meta-analysis.