Aggregation Across Multiple Indicators Improves the Detection of Malingering: Relationship to Likelihood Ratios

• Published in: Clinical neuropsychologist 2008-07, Vol.22 (4), p.666-679
• Main Author: Larrabee, Glenn J.
• Format: Article
• Language: English
• Subjects: Adult; Aggregation; Biological and medical sciences; Brain Injuries - diagnosis; Brain Injuries - psychology; Chaining; False Negative Reactions; Female; Humans; Likelihood Functions; Likelihood ratios; Male; Malingering; Malingering - diagnosis; Medical sciences; Middle Aged; Multiple indicators; Neuropsychological Tests; Predictive Value of Tests; Psychology. Psychoanalysis. Psychiatry; Psychometrics. Diagnostic aid systems; Psychopathology. Psychiatry; Sensitivity and Specificity; Techniques and methods
• ISSN: 1385-4046; 1744-4144
• DOI: 10.1080/13854040701494987

Recent literature shows that aggregating across multiple symptom validity test (SVT) failures increases the probability of malingering over use of one indicator alone, supporting the criteria proposed by Slick, Sherman, and Iverson (1999) that require multiple sources of evidence for diagnosis of malingering. The present study reanalyzes with likelihood ratios data previously published by Larrabee ( 2003a ) on litigants with definite malingering, contrasted with non-malingering patients with moderate and severe traumatic brain injury. Chaining of likelihood ratios demonstrated an increase in probability of malingering when multiple test scores were positive, with values ranging from. 713 to. 837 for one failed SVT,. 936 to. 973 for two failed SVTs, and. 989 to. 995 for three failed SVTs. Posterior probabilities of malingering derived from chaining of likelihood ratios closely approximated those obtained by direct computation of Positive Predictive Power, particularly when three SVTs were failed. Moreover, the five SVTs employed in these analyses did not intercorrelate significantly, satisfying the assumption that the tests be independent for chaining of likelihood ratios. Posterior probabilities derived from chaining of likelihood ratios, holding sensitivity and specificity constant at. 50 and. 90, respectively, and varying the malingering base rate from. 10 to. 90, showed a wide range of values,. 357 to. 978, for failure of one SVT. Failure of two SVTs yielded probabilities ranging from. 735 to. 996. Failure of three SVTs yielded values ranging from. 933 to. 999, demonstrating high probabilities of malingering irrespective of the base rate. These data support the Slick et al . recommendation that multiple positive findings are necessary for diagnosis of malingering. Suggestions are made for modification of the Slick et al . criteria based on the current results.