Pain, Malingering, and Performance on the WAIS-III Processing Speed Index

Transcription

1 Journal of Clinical and Experimental Neuropsychology, 28: , 2006 Copyright Taylor & Francis Group, LLC ISSN: DOI: / Pain, Malingering, and Performance on the WAIS-III Processing Speed Index NCEN Journal of Clinical and Experimental Neuropsychology, Vol. 28, No. 07, April 2006: pp. 0 0 PSI J.L. and Etherton Pain et al. JOSEPH L. ETHERTON, 1,3 KEVIN J. BIANCHINI, 2,3 MATTHEW T. HEINLY, 2 AND KEVIN W. GREVE 2,3 1 Loyola University, New Orleans 2 University of New Orleans 3 Jefferson Neurobehavioral Group, Metairie, Louisiana Pain patients often report cognitive symptoms and many will include them in their claims of disability. The Processing Speed Index (PSI) of the WAIS-III was investigated as one aspect of cognitive functioning in six groups. Slight impairment was found for PSI and Digit Symbol subtest performance, but not for Symbol Search, in a Laboratory-induced Pain group and a Clinical Pain group. The lowest scores were found in a Simulator group instructed to fake cognitive impairment and a Clinical Pain group diagnosed as Malingering. Results suggest that PSI scores are only slightly reduced by laboratory-induced pain or chronic pain, and that unexpectedly low scores in the absence of significant/documented brain dysfunction suggest poor effort or deliberate misrepresentation. Introduction Complaints of pain are common in patients involved in disability claims, both as a primary complaint and as a secondary symptom in patients whose presenting problem is brain injury. Psychologists frequently evaluate patients with chronic pain, many of whom complain of cognitive problems such as impairment of memory or concentration. Iverson and McCracken (1997) found that 42% of patients with chronic pain and no history of head injury endorsed at least one cognitive symptom and that by self-report alone 39% of their sample would have met criteria for post-concussion syndrome. Consistent with these findings, pain patients may also demonstrate deficits on cognitive testing (Eccleston, 1994; Kewman, Vaishampayan, Zald, & Han, 1991; Sjogren, Olsen, Thomsen, & Dalberg, 2000). Chronic pain patients are typically more impaired on more complex attentiondemanding tasks; however, tasks that require fewer attentional resources are unaffected even when pain levels were high (Eccleston, 1994). But the relationship between pain and cognitive impairment is complex. For example, although 32% of Kewman et al. s (1991) pain sample demonstrated cognitive impairment, the association between pain level and cognitive performance was not significant after controlling for psychological distress; education was almost as strongly associated with cognitive performance as pain intensity. Received 19 April 2005; accepted 8 September This project is one component of a much larger data collection project. Our research assistants, including Kelly Curtis, Laura Thompson, Raven Brasseux, Thomas Harris, Katie O Flaherty, Lyse Jackson, Thomas Tumage, and Melanie Rabalais have been tireless and their efforts are much appreciated. Address correspondence to Kevin W. Greve, PhD., Department of Psychology, University of New Orleans-Lakefront, New Orleans, LA kgreve@uno.edu 1218

2 PSI and Pain 1219 In addition, other problems that may accompany pain (e.g., sleep disturbance, depression) have the potential to affect cognitive function (Brown, Glass, & Park, 2002; Cohen, Malloy, & Jenkins, 1998; lezzi, Duckworth, Vuong, Archibald, & Klinck, 2004; Menefee et al., 2000). Further hindering a clear understanding of the impact of pain on cognitive function is the fact that the presence of financial incentive influences symptom report and test performance. For example, pain patients involved in worker s compensation claims and with no history of head injury reported more symptoms of cognitive disability than did patients with head injury but without the presence of financial incentive (Iverson, King, Scott, & Adams, 2001). In addition, a number of studies have shown that chronic pain patients involved in disability litigation have much higher failure rates on symptom validity measures than do non-litigating pain patients or even non-litigating TBI patients, indicating likely intentional misrepresentation of their performance on tasks that appear to measure cognitive ability. Meyers and Diep (2000) reported that 29% of litigating chronic pain patients failed two or more of six validity measures in an assessment battery, compared with 0% of non-litigating pain patients. Gervais, Green, Allen, and Iverson (2001) found that litigating pain patients failed two symptom validity measures (the Computerized Assessment of Response Bias [CARB; Conder, Allen, & Cox, 1992] and the Word Memory Test [WMT; Green, Iverson, & Allen, 1999] at high rates (CARB, 41%; WMT, 42%) compared with patients with no financial incentive and with verified neurological impairment (CARB, 2%; WMT, 1%). However, when litigating patients in their study were informed in advance that the CARB was a measure of effort rather than a measure of cognitive ability, the failure rate dropped to 6%, suggesting volitional control of performance on such measures. Thus, current research indicates that some pain patients may intentionally exaggerate or fabricate cognitive symptoms. Purpose Since cognitive problems are reasonably expected in the context of chronic pain, it is important to understand the relationship between pain and cognitive impairment and to be able to discriminate legitimate clinical cognitive complaints from exaggerated cognitive complaints in this population. Accordingly, the present studies addressed two goals. 1. Investigate the effects of pain and factors associated with pain on cognition as measured by the Processing Speed Index (PSI) of the Wechsler Adult Intelligence Scale-3 (WAIS-III; Wechsler, 1997). The PSI was selected because it is sensitive to cognitive impairment in neurologically compromised patients (Axelrod, Fichtenberg, Liethen, Czamota, & Stucky, 2001; Langeluddecke & Lucas, 2003) and therefore may be sensitive to the cognitive effects of pain. 2. Determine whether the PSI could be utilized to identify intentional exaggeration of cognitive deficits in patients with pain. We have observed clinically that some malingering patients attempted to appear disabled by slowing their performance on timed tasks. Thus, it is possible that PSI and its constituent subtests may also be sensitive to intentional efforts to appear cognitively impaired. The present study will examine the extent to which the PSI is reduced by pain-related factors versus intentional slowing. The two studies described in this paper examine the impact of pain on speed of cognitive processing, first through laboratory-induced pain in healthy volunteers (Study 1), then in clinical chronic pain patients (Study 2). The two studies described in this paper used

3 1220 J.L. Etherton et al. simulator and known-groups designs (Rogers, 1997) and follow the guidelines proposed by Greve and Bianchini (Bianchini, Greve, & Glynn, 2005; Greve & Bianchini, 2004) for developing malingering indicators. This general method has been used effectively to develop malingering indicators in traumatic brain injury (e.g., Bianchini, Mathias, Greve, Houston, & Crouch, 2001; Greiffenstein, Baker, & Gola, 1994; Larrabee, 2003a, b) and is particularly useful because the results of known-groups studies, unlike those of simulator studies, can be directly generalized to clinical patients. Also, the use of simulator and known-groups designs in concert allows for converging data on the accuracy and validity of the PSI as an indicator of cognitive malingering in pain. Study One: Pain Induction in Healthy Volunteers Method In Study 1 pain was induced in healthy volunteers via the cold-pressor procedure (Peckerman et al, 1991). Although healthy volunteers experiencing acute pain differ from chronic pain patients in important ways, this procedure allows for the examination of the effect of pain in isolation on PSI performance compared with the performance of similar healthy volunteers who did not experience induced pain. This procedure enhances internal validity by eliminating the influence of extraneous variables other than pain in PSI performance. Procedure. Upon arrival, participants were informed that they would be randomly assigned to one of three conditions and that one of these conditions would involve experiencing moderate pain by means of the cold-pressor procedure. They were assigned to one of three conditions (Control, Simulator, Cold-Pain) by drawing a number out of a box. All participants were administered the Word Memory Test (WMT; Green et al., 1999) and the Computerized Assessment of Response Bias (CARB; Conder et al., 1992) as part of a separate study, and were administered in this order the two subtests (Digit Symbol and Symbol Search) that comprise the PSI during the waiting period required for the WMT. A fourth condition, Procedural Distraction, was added after the initial data collection period. This condition was designed to match the distracting elements of the cold-pressor procedure without inducing pain, and is described further below. Processing Speed Index (PSI). The PSI of the WAIS-III consists of two subtests, Digit Symbol and Symbol Search, which involve processing visual information and responding under time constraint. In Digit Symbol participants refer to a key which matches digits with symbols, and then draw the appropriate symbols in blank boxes, each of which is associated with a digit. In Symbol Search participants examine two figures in a left column, and then indicate by marking Yes or No whether either of these figures appears in a group of 5 figures in a right column. Two minutes are allotted for each of these two subtests. Participants. A total of 82 undergraduate women and men were recruited at a liberal arts university in the Southern United States either in fulfillment of course requirements or for extra credit. Participants were asked to self-exclude if they did not wish to undergo the cold pressor procedure or if they had any of the following conditions: diabetes, peripheral neuropathy, myotonia congenita, Reynaud s disease, complex regional pain syndrome, reflex sympathetic dystrophy, or any circulatory conditions. No participants self-excluded on this basis. One participant in the Simulator condition was replaced because of failure to adhere to instructions, and a participant in the Procedural Distraction group was replaced because he

4 PSI and Pain 1221 reported pain during the procedure, which was intended to involve no pain; he noted that the pain was not due to the procedure but to postural discomfort. For the final participant pool there were 23 men and 57 women, ranging in age from 18 to 25 (mean age = 19.5). Control group. Participants in the Control group completed the PSI subtests under standard WAIS-III instructions (Wechsler, 1997). Simulator group. In the second condition, participants were asked to simulate painrelated memory impairment during administration of the PSI subtests. At the conclusion of the experiment, Simulators were asked to describe in writing their strategy for feigning memory impairment. Prior to test administration they read the following instructions: Imagine that you have been in an accident and suffered an injury to your neck and shoulder. Initially you experienced pain in that arm and hand but now you are completely healed and experiencing no problems. Nevertheless, you have filed a lawsuit and you stand to gain a very large settlement if you are disabled. In your lawsuit you are claiming that your pain has affected your ability to think, especially your memory. Because of the memory problems you have developed, you cannot do college level schoolwork and now your future employment opportunities are limited. You have been sent to a psychologist to evaluate your claim of memory problems and are now about to take a memory test for that purpose. Your task is to perform on that test as if your memory were impaired because of severe, persistent, chronic pain. However, you must fake your memory impairment in a way that is believable because if you are caught, your lawsuit will be thrown out of court and you will get nothing. Cold Pain condition. Participants assigned to the third condition completed the PSI subtests during the cold-pressor procedure, a procedure that has been used for decades to generate pain in the laboratory setting (e.g., Mikail, Vandeursen & von Baeyer, 1986; Peckerman et al, 1991; Saab, Liabre, Hurwitz & Schneiderman, 1993; Worthington, 1978). Prior to beginning test administration, participants placed one foot into a bucket of ice water and were asked to keep it in place for as long as they could tolerate the discomfort or until completion of the tasks. They were informed that they could remove their foot from the water if the pain became too great, but were asked to return their foot to the water as soon as they were able to do so. No participants removed their foot from the water for more than 30 seconds. An 11-point numerical pain rating scale was used to record pain intensity. Such scales are commonly used in both clinical and research applications (Gracely, 1989; Gracely, McGrath, & Dubnear, 1978; Peckerman et al., 1991). The scale and associated verbal descriptors (No Pain [0], Mild Pain [1 3], Moderate Pain [4 6], Severe Pain [7 9], Very Severe Pain [10]) were visible throughout the procedure. Pain ratings were recorded prior to beginning the Digit Symbol test, between the Digit Symbol and Symbol Search tests, and immediately following completion of Symbol Search. Water temperature was recorded at the beginning and end of the procedure. Procedural Distraction condition. It was noted that participants in the Cold Pain condition appeared to be distracted by non-pain variables related to the cold-pressor procedure, such as needing to maintain a relatively stable posture to keep their foot submerged in the water, being asked to rate their pain, and referring to the pain rating scale to select their response. Accordingly, any differences between the Cold Pain group and the Control group may have been attributable to pain, to the distracting elements of the cold pressor procedure,

5 1222 J.L. Etherton et al. or both. In order to control for these extraneous variables, a fourth condition was added after the first three conditions had been completed. The Procedural Distraction condition followed the same procedures as the Cold Pressor procedure except that non-painful warm water, ranging from degrees Fahrenheit, was substituted for the cold water. This condition allowed for a comparison of PSI performance for two groups that experienced the same procedural distractions, differing only in the presence or absence of pain. Pain ratings for all participants in the Procedural Distraction condition were at 0 throughout. Results Descriptive statistics and the analysis of variance (ANOVA) results for the PSI and subtests are reported in Table 1. Significant condition effects were observed. Bonferroni follow-up comparisons demonstrated that the Simulator group scored significantly lower on all measures than the Control, Cold Pain, and Procedural Distraction Groups. The Cold Pain group also scored significantly lower on Digit Symbol than did the Control and Procedural Distraction groups, which did not differ from each other on any measure. The Cold Pain group did not differ from the Control and Procedural Distraction groups on PSI or Symbol Search. Pain ratings for the Pain condition were evaluated (mean pain rating = 5.77, s.d. = 2.36, range = 1 to 9). Eight of 20 participants reported overall mean ratings of Severe pain (ratings of 7 9/10), 10 reported Moderate pain (4 6/10), and 2 reported Mild pain (1 3/10). Mean pain severity was negatively correlated with performance on Digit Symbol (r =.68, r 2 =.46, p <.01) and for Processing Speed Index (r =.54, r 2 =.29, p <.02), but not with Symbol Search performance (r =.23, p =.33). Thus, pain as an isolated variable produced minor impairment in Digit Symbol performance; however, Symbol Search and the overall PSI were not significantly affected. Performance in the Procedural Distraction group suggests that impaired performance in Table 1 Processing Speed Index and scaled scores for Digit Symbol and Symbol Search for the healthy volunteer group Control Simulator Cold pain Procedural Distraction F p < eta 2 N Processing Speed Index Mean a b a a SD Range Digit Symbol scaled score Mean a 2.80 c 9.15 b a SD Range Symbol Search scaled score Mean a 3.65 b a a SD Range ab Row means with same superscript letter did not differ significantly at alpha <.05. Df = 3, 76.

6 PSI and Pain 1223 the Cold Pain group was not attributable to any distracting elements of the Cold Pressor procedure independent of cold-induced pain. However, factors associated with chronic pain in clinical patients such as sleep deprivation, fatigue, depression, and others, could not be addressed via the induction of time-limited acute pain in a non-clinical volunteer population. Accordingly, Study Two was conducted with clinical patients. Study Two: PSI Performance in Clinical Pain Patients Study Two compared the performance of chronic pain patients who had been identified as definitely malingering with chronic pain patients who had objectively documented spinal pathology and no evidence of malingering and also with nonmalingering neurological patients (moderate-severe brain trauma, memory disorders). In this study, the influence of variables such as medication usage, sleep deprivation, impaired medical status, and emotional distress that frequently accompany chronic pain and have the potential to impair cognition, were not eliminated, allowing for generalization of the results to clinical pain patients. Method Participants. To form the two pain groups, records of approximately 200 patients referred for psychological pain evaluations at a clinical psychology group practice were reviewed. These patients had been referred by physicians, workers compensation companies, and attorneys. Extensive medical records reviewed in the context of these evaluations provided objective medical diagnostic test results, as well as physicians clinical diagnoses and injury descriptions that contributed to study group assignment. Patients were assigned into either malingering or non-malingering groups or were excluded from the study (see below for inclusion and exclusion criteria). Most patients (all patients in the malingering group) had financial incentive in the form of either a workers compensation claim or personal injury suit. The medical diagnoses of the patients in each of the two clinical pain groups are reported in Table 2. Some patients received more than one diagnosis. All patients completed the WAIS, California Verbal Learning Test, and MMPI-2. Almost all completed the Test of Memory Malingering (TOMM; Tombaugh, 1996) and the Portland Digit Recognition Test (PDRT; Binder, 1993); more recently assessed patients also completed the WMT and the CARB. Non-Malingering Clinical Pain Sample. Patients were included in this group if their medical records demonstrated objective clinical abnormalities of the spine as indicated by computerized tomography (CT), magnetic resonance imaging (MRI), x-ray, myelography or electromyography studies, or surgery. They were excluded from this group if there were physician reports of non-physiological findings or behavioral inconsistencies. Patients were also excluded if they demonstrated any psychometric evidence of poor effort or symptom exaggeration as defined by any of the following: scores below published clinical cutoffs on any forced choice symptom validity test (SVT); T-scores above 85 on MMPI-2 F or Fb scales (Greve, Bianchini, Love, Brennan, & Heinly, in press); or Fake Bad Scale (FBS; Lees-Haley, English, & Glenn, 1991) raw scores above 23 for men and above 25 for women (Larrabee, 1998). Of the cases reviewed, 48 patients (30 men and 18 women; see Table 3 for other demographic data) met criteria for inclusion. Definite MND Clinical Pain Patients. Thirty-two clinical pain patients (21 men and 11 women; see Table 3 for other demographic data) met the Bianchini et al. (2005) criteria

7 1224 J.L. Etherton et al. Table 2 Injury characteristics of nonmalingering and definite malingering pain patients Diagnosis/symptom report Nonmalingering Malingering Vertebral disc herniation/rupture (any level) 15 2 Neck or back strain or sprain 5 8 Spinal Stenosis 3 0 Other back pain or injury (any level) Sacroiliac joint dysfunction 1 0 Degenerative disc disease 1 1 Vertebral fracture 2 0 Facet pain/hypertrophy 1 2 Low back pain unspecified 1 1 Spondylolysis 4 1 Degenerative joint 1 Radiculopathy 5 7 Sciatica 1 Neuropathy/Neuropathic pain 1 1 Nerve impingement syndrome 2 2 Failed back syndrome 2 Fibromyalgia 1 Myofascial pain 3 Reflex Sympathetic Dystrophy 1 2 Table 3 Demographic data and estimated premorbid intelligence for the clinical groups Non-mal. pain Definite mal. pain Mod-Severe TBI Memory disorder F p < eta 2 N Age Mean 43.9 a 43.0 a 31.2 b 70.1 c SD Education Mean 12.1 ab 11.5 a 12.4 ab 13.3 b SD eviq Mean 98.8 ab 94.4 bc acd cd SD epiq Mean 98.3 ab 93.9 abc acd cd SD efsiq Mean 99.0 a 94.4 b c c SD abcd Row means with same superscript letter did not differ significantly at alpha <.05. Non-Mal. Pain = non-malingering clinical pain group; Definite Mal. Pain = Definite malingering clinical pain group; Mod-Severe TBI = non-malingering moderate-severe TBI group. Df = 3, 148 for age and education analysis; df = 3, 139 for estimated premorbid IQ analysis because missing data for some patients (Non-Mal. pain = 3; Definite Mal. pain = 1; Mod-Severe TBI = 1; Memory Disorder = 4) prevented the calculations; eviq = estimated premorbid verbal IQ; epiq = estimated premorbid performance IQ; efsiq = estimated premorbid full scale IQ.1

8 PSI and Pain 1225 for Definite Malingered Pain-Related Disability (MPRD) by virtue of a statistically below-chance (p <.05) performance on a forced-choice SVT. Typically, they failed either the PDRT or the TOMM. However, several patients failed a tactile SVT (Greve, Bianchini, & Ameduri, 2003). The interpretation of a below-chance result as definitive evidence of intentional exaggeration of cognitive deficits even in the context of objective pathology has become well established in the neuropsychological literature (Bianchini, Mathias, & Greve, 2001; Bianchini, Greve, & Love, 2003; Reynolds, 1998). Patients with evidence of symptom or deficit exaggeration who did not meet criteria for Definite MPRD (Bianchini et al., 2005) were excluded from the study. TBI Patients. For comparison purposes, 34 patients (27 men and 7 women; see Table 3 for other demographic data) with documented moderate to severe TBI and no evidence of poor effort, suspect behavior, or malingering were included. All completed at least the PDRT or the TOMM, most completed both, and most also completed the MMPI-2. They were also negative on indicators from the California Verbal Learning Test (CVLT; Millis, Putnam, Adams, & Ricker, 1995) and the Wisconsin Card Sorting Test (WCST; Greve, Bianchini, Mathias, Houston, & Crouch, 2002). These patients were among those included in a study of Reliable Digit Span (RDS) in TBI (Heinly, Greve, Bianchini, Love, & Brennan, under review) and malingering in chronic pain (Etherton, Bianchini, Ciota, Heinly, & Greve, 2006; Etherton, Bianchini, Greve, & Heinly, 2005c). All had documented evidence of acute neurological pathology indicating an injury that was worse than a mild TBI as defined by the Mild Traumatic Brain Injury Committee of the Head Injury Interdisciplinary Special Interest Group of the American Congress of Rehabilitation Medicine (1993). Specifically, they were considered to have suffered a moderate to severe TBI if they met any of the following criteria: 1) posttraumatic amnesia (PTA) greater than 24 hours; 2) an initial Glasgow Coma Scale (GCS) < 13; 3) loss of consciousness > 30 minutes, and/or positive neuro-radiological findings (e.g., skull fracture, hemorrhage, hematoma) or focal neurological signs. Memory Disorder Patients. A sample of 38 patients (14 men and 24 women; see Table 3 for other demographic data) diagnosed with memory disorder was derived from the general neuropsychological cases of the same clinical practice. These were patients who had experienced a decline in memory functioning with no known precipitating event or etiology to account for the memory loss though most were suspected of having Dementia of the Alzheimer s Type, vascular dementia, or both. None of these patients had any incentive to appear impaired. Pain Ratings For most patients in the two clinical pain groups, pain ratings had been recorded during the interview portion of the psychological evaluation using an 11-point numerical scale as described in Study 1. Typically, ratings were collected for current, worst, and least pain. Full pain rating data were not available for all patients. Results The clinical pain malingering and nonmalingering groups did not differ in pain ratings (Nonmalingering group: current pain mean: 6.82, sd = 1.52; least pain mean: 4.87, sd = 1.95; and worst pain mean: 9.37, sd =.94; Definite Malingering group: current pain mean:

9 1226 J.L. Etherton et al. 6.70, sd = 2.14; least pain mean: 4.78, sd = 1.96; worst pain mean: 9.26, sd = 1.58). For the clinical pain patients there was no correlation between pain rating and performance on PSI or any subtest. The two pain groups were significantly older than the TBI patients and all groups were younger than the memory disorder patients. Age was not associated with performance on the PSI or any subtests. There was a significant group effect for education with the memory disorder group completing significantly more education than the TBI group and the two pain groups which did not differ from each other. A significant group effect was seen for all estimated premorbid IQ scores (using the demographic-based estimates from the Wechsler Test of Adult Reading; Psychological Corporation, 2001), with the TBI and Memory Disorder patients having significantly higher scores than the pain groups. The Nonmalingering pain group was significantly higher than the Definite Malingering group on estimated premorbid Performance IQ (epiq) and Full Scale IQ (efsiq) but not Verbal IQ (eviq). Table 3 presents group means and ANOVA results for age, education, and estimated IQ. When the persons with the least education (< 9 years) were removed from the analyses of estimated intelligence, the two pain groups no longer differed. Premorbid intelligence was minimally correlated with PSI score (average r =.32, sd =.02) in the pain groups and the amount of observed variance accounted for was small (average r 2 =.10, sd =.02). However a multiple regression analysis within the pain patients indicated that malingering group membership accounted for over 40% of the variance in PSI scores; including efsiq in the regression model increased the R 2 by about 6%. Thus, it is possible that lower premorbid intelligence contributed somewhat to lower PSI scores, though the effect was minimal. Ultimately it will be important to determine if the inclusion of patients with lower education adversely impacted the observed classification accuracy of the PSI score. This is addressed below. Significant group effects were observed in that the Definite Malingering Pain group performed significantly worse than the other three groups on the PSI and on both subscales (see Table 4 for details). The only significant difference seen among the other clinical groups was that the Memory Disorders group scored lower than the Nonmalingering Pain group on Symbol Search, but not lower than the TBI group, yet the Memory Disorder group still scored significantly higher than the Definite Malingering Pain group. Comparison with Table 1 indicates that the mean scores of the healthy volunteer Simulator and clinical Definite Malingering groups are very similar and that both are substantially lower than the scores of healthy volunteers in acute pain and nonmalingering clinical pain patients, as well as patients with TBI and memory disorder. The PSI scores of the Simulator and Definite Malingering groups did not differ significantly (F(1,50) = 3.87, p >.05). Although not originally developed as a malingering indicator, the group analyses reported for these two studies suggest that scores on the PSI may have utility in detecting intentionally exaggerated cognitive impairment attributed to pain. Greve and Bianchini (2004; Bianchini et al., 2005) have emphasized the necessity of reporting classification accuracy of potential malingering indicators, and specifically recommended providing tables with classification accuracy for a range of scores. Accordingly, Tables 5 through 7 report cumulative percentages by group for scores on the PSI and two subtests, and include the healthy volunteer groups as well as the clinical groups. Percentage values for the Simulator and Clinical Malingering group indicate the sensitivity of each score to intentional exaggeration. Percentage values for the remaining groups indicate the false positive error rate for each score. The distributions of scores indicate clear distinctions among the nonmalingering and nonsimulating groups compared with the Simulator and Malingering groups. For PSI, scores

10 PSI and Pain 1227 Table 4 Means by clinical group for PSI, Digit Symbol, and Symbol Search scale scores Non-mal. pain Definite mal. pain Mod-severe TBI Memory disorder F p < eta 2 N Processing Speed Index Mean a b a a SD Range Digit Symbol scaled score Mean 7.73 a 4.25 b 6.88 a 7.26 a SD Range Symbol Search scaled score Mean 8.52 a 4.28 c 7.62 ab 6.50 b SD Range abc Row means with same superscript letter did not differ significantly at alpha <.05. Df = 3, 148; Non-Mal. Pain = non-malingering clinical pain group; Definite Mal. Pain = Definite malingering clinical pain group; Mod-Severe TBI = non-malingering moderate-severe TBI group. at or below 70 are associated with a false positive error rate of 4% (specificity of 96%) for the Clinical Pain group with a sensitivity of 63% for the Clinical Malingering group. False positive rates increase when the TBI and memory disorder groups are included. It is important to note that when the six nonmalingering clinical pain patients and four malingering clinical pain patients with less than nine years of education were removed from the data set, the false positive error rate associated with PSI scores of 75 and 70 increased one percentage point (to 9 and 5, respectively) while sensitivity dropped about five percentage points (to 65 and 58, respectively). At lower cutoffs sensitivity increased four-fold while the false positive error rate remained unchanged at 0%. Thus, the minimal education/premorbid IQ effect observed in the aggregate group data appears to have had no meaningful effect on the classification accuracy of PSI, especially at extremely low score levels. Table 8 provides positive predictive power values for a range of cutoffs and base rates using the observed sensitivity and specificity data. The methodology used in the current studies is consistent with recommendations published by Sackett and Haynes (2002) for conducting diagnostic research; accordingly, these findings can be applied clinically. To accurately use PSI as an index of malingering, it is important to characterize the individual cases classified as nonmalingering who had PSI scores worse than 95% of the nonmalingering sample. Six memory disorder and five TBI patients scored in this range. None had incentive. The memory disorder patients were older than 60 and four were older than 70; all but one had at least completed high school. All presented with evidence of progressive cognitive decline. The TBI patients were all aged 18 to 21 and had completed high school, except one who was injured in the 11th grade. All had objective evidence of significant brain trauma. Two pain patients classified as nonmalingering earned scores of 68 and 69 on the PSI. The first was a 36 year old man with 10 years of education. He had suffered injuries to his

11 1228 J.L. Etherton et al. Table 5 Cumulative percentage for Processing Speed Index (PSI) scores all groups GROUP Control Simulate Cold pain Procedural distraction Non-mal pain Def. mal pain M-S TBI Memory disorder n PSI The tabled value is the percent of patients performing at or below the given score. For non-malingering patients this is the False-Positive error rate (100 Specificity); for malingering patients this is the Sensitivity rate. Control, Simulator, Cold pain, and Procedural Distraction = healthy volunteers assigned to control, simulator, cold-pressor pain-induced, and procedural distraction control conditions. Non-Mal Pain = Non-malingering clinical pain patients; Def Mal Pain = Definite malingering clinical pain patients; M-S TBI = non-malingering moderate-severe TBI patients; Memory disorder = Non-malingering memory disorder patients. neck and right shoulder with clear evidence of neurological compromise in the C7 distribution and he had undergone previous neck and shoulder surgery. As a result, he had right upper extremity weakness and restricted movement which could have affected his performance on the PSI subtests, both of which have a significant motor component. His WAIS Index scores, except PSI, were in the 80s, supporting this explanation. The second patient had evidence of neurological compromise and degenerative spine disease in the lumbosacral region. All his WAIS scores were in the 60s. Since he completed high school and was working as a plant manager at the time of his injury, those scores are suspicious. While he had no psychometric findings consistent with malingering, it is possible that this is a malingerer who went undetected (i.e., was a false negative case). Because all individual malingering indicators have imperfect sensitivity at levels associated with acceptable false positive error rates, it is possible that some malingerers

12 PSI and Pain 1229 Table 6 Cumulative percentage of Digit Symbol scale scores (DS ss) by group GROUP Control Simulate Cold pain Procedural distraction Non-mal pain Def. mal pain M-S TBI Memory disorder N DS ss The tabled value is the percent of patients performing at or below the given score. For non-malingering patients this is the False-Positive error rate (100 Specificity); for malingering patients this is the Sensitivity rate. Control, Simulator, Cold pain, and Procedural Distraction = healthy volunteers assigned to control, simulator, cold-pressor pain-induced, and procedural distraction control conditions. Non-Mal Pain = Non-malingering clinical pain patients; Def Mal Pain = Definite malingering clinical pain patients; M-S TBI = non-malingering moderate-severe TBI patients; Memory disorder = Non-malingering memory disorder patients. went undetected using current methods and were incorrectly placed in the non-malingering group. While the use of multiple psychometric indicators (cognitive, emotional, somatic) as well as suspicious clinical findings to exclude patients from the nonmalingering group reduces the likelihood that malingerers were included, this possibility needs to be considered. In malingering research, because the indicator under study (in this case, the entire WAIS) cannot be used for classification purposes, false negative errors are a risk. Thus, the observed false positive error rate for the specific measure under study is likely an over-estimate. Discussion The two studies presented were conducted to examine the potential impact of pain on cognitive performance as measured by the Processing Speed Index of the WAIS-III (Wechsler, 1997) and to examine the capacity of this index to detect negative response bias in patients with pain. Previous studies have reported mixed results regarding the impact of pain on

13 1230 J.L. Etherton et al. Table 7 Cumulative percentage of Symbol Search scaled score (Sy ss) by group GROUP Control Simulate Cold pain Procedural distraction Non-mal pain Def. mal pain M-S TBI Memory disorder N Sy ss The tabled value is the percent of patients performing at or below the given score. For non-malingering patients this is the False-Positive error rate (100 Specificity); for malingering patients this is the Sensitivity rate. Control, Simulator, Cold pain, and Procedural Distraction = healthy volunteers assigned to control, simulator, cold-pressor pain-induced, and procedural distraction control conditions. Non-Mal Pain = Non-malingering clinical pain patients; Def Mal Pain = Definite malingering clinical pain patients; M-S TBI = non-malingering moderate-severe TBI patients; Memory disorder = Non-malingering memory disorder patients. cognition, with some finding cognitive impairment only on complex tasks and only at high levels of pain (Eccleston, 1994) and others reporting cognitive impairment as relatively common in chronic pain populations (Kewman et al., 1991). Because relatively high rates of malingering have been reported in situations involving pain-related disability claims (Mittenberg, Patton, Canyock, & Condit, 2002), it is important to understand more completely the relationship between pain and cognitive impairment on instruments such as the WAIS-III that are frequently used in disability evaluations and to do so in a manner that takes into account extraneous variables, including intentional exaggeration, that may also influence cognitive performance. In the current pair of studies neither acute laboratory-induced pain nor chronic pain in clinical patients led to statistically significant impairment, on average, on the PSI or any of its constituent subtests (Digit Symbol and Symbol Search). However, examination of Tables 4, 5, and 6 clearly indicates that some patients in all three nonmalingering clinical groups scored at a level that would likely be considered impaired (e.g., PSI 80). It is important to note, however, that half as many clinical pain patients scored at this level compared to moderate-severe TBI and memory disorder patients. Thus, consistent with

14 PSI and Pain 1231 Table 8 False positive error rate, sensitivity, and positive predictive power for different Processing Speed Index values and hypothetical base-rates of malingering a. Clinical pain patients only 1 Positive predictive power for hypothetical base rates score FP Rate Sensitivity BR: b. Mixed clinical patients 2 score FP Rate Sensitivity BR: False positive error rate (FP Rate) is based only on the performance of the non-malingering clinical pain patients (n = 48). 2 False positive error rate (FP Rate) is based on the performance of all non-malingering clinical patients (pain, TBI, memory disorder) (n = 121). BR = hypothetical base-rate of malingering. past research, the findings of this study do suggest the presence of cognitive processing speed deficits in some patients with pain. However, the vast majority (>95%) of clinical pain patients, even those reporting moderate to severe pain and/or the secondary effects of clinical pain such as sleep deprivation, emotional distress, fatigue, or medication side effects, did not demonstrate extreme PSI impairment (i.e., scores 70). Rather, this level of performance on the PSI in this study identifies deliberate misrepresentation with a high degree of specificity. The most extreme scores were seen in the volunteers instructed to fake cognitive disability and the clinical pain patients identified as definitely malingering, both of which performed worse on all scales than patients with moderate-severe TBI or Memory Disorder. More than 60% of the Malingering clinical pain patients and 80% of the Simulators scored lower than all but about 5% of the nonmalingering clinical pain patients. Interestingly, both the mean scores and overall score distributions of the Simulator group were almost identical to those of the Definite Malingering group on all measures and the two groups were more similar to each other than to any of the nonmalingering groups. Given these results, it appears that while mild impairment on PSI may be seen in some nonmalingering clinical patients, significant impairment in PSI performance (e.g., 65) in the absence of objective and significant brain dysfunction cannot reasonably be attributed to the effects of either acute or chronic pain, even at moderate to severe levels, and that the PSI can be used to accurately identify malingered cognitive problems. Moreover, the finding that the

15 1232 J.L. Etherton et al. Simulators (i.e., persons instructed to act cognitively impaired) and the clinically defined Definite Malingerers score at similar levels that, together, are very different from the nonmalingering participants, supports the validity of the Bianchini et al. (2005) criteria with pain patients. In summary, the results of the two studies reported here indicate that some pain patients demonstrate impairments in cognitive processing speed, often at a level consistent with that seen in traumatic brain injury. This is an important finding because, unlike previous studies of cognition in pain, we assessed or controlled for poor effort, a methodological consideration which is considered essential to an accurate interpretation of research findings (Green, 2003). It is also important to note that the cumulative frequency distributions for PSI and its constituent subtests indicate that while some pain patients do evidence reduced processing speed, extremely poor performance on the PSI (i.e., 70) was rarely observed except in patients who were malingering their cognitive deficits or who had significant objectively documented neuropathology. This means that scores at or below 70 in a pain patient cannot reasonably be explained by pain and pain-related problems (e.g., fatigue, medication effects) and likely indicate incomplete effort which may be due to malingering. The more extreme the score, the more likely it reflects malingering. At the same time, it must be remembered that malingering detection techniques are not perfect and should not be used in isolation for the clinical diagnosis of malingering. A formal diagnosis of malingering should be based on the integration of diverse clinical information (Bianchini et al., 2005; Slick, Sherman, & Iverson, 1999). Psychometric indicators of negative response bias, of which PSI appears to be one, are one important source of evidence. Clinical Implications The detection of exaggeration and malingering in chronic pain has important clinical implications. The accurate assessment of the validity of subjective complaints and the behavioral manifestation of pain-related disability (including cognitive impairment) during pain-related clinical evaluations has very real practical clinical relevance. As just one example, the disability guidelines of the American Medical Association (Cocchiarella, & Andersson, 2001) quantify disability based on criteria related to objective pathophysiology and clear causal links between the pathophysiology and observed signs and symptoms. Unfortunately, objective physical findings do not fully explain the breadth and magnitude of disability seen in many patients with pain (Boden, Davis, Dina, Patronas, & Wiesel, 1990; Boden et al., 1990). Therefore, examination of painful conditions and decisions regarding estimates of disability require substantial reliance upon subjective symptom report of pain and pain-related disability (Cocchiarella, Lord, Turk, & Robinson, 2001; Robinson, Turk, & Loeser, 2004). Appreciating the potential problems associated with subjective self-report, the AMA guidelines regarding the rating of pain-related disability state examiners must be careful to provide ratings only for those who provide information that appears to be reasonable and accurate (p. 571). How is the examiner to determine what information is reasonable and accurate? Pain often occurs in the context of a legally compensable event such as a work-related injury or incident in which some other party is potentially liable. The presence and magnitude of compensation are related to increased reports of pain and decreased treatment efficacy (Rohling, Binder, & Langhinrichsen-Rohling, 1995) and compensation status has been linked to outcome in surgery and outpatient rehabilitation for back pain (Rainville, Sobel, Hartigan, & Wright, 1997; Vaccaro, Ring, Scuderi, Cohen, & Garfin, 1997). Interestingly, more than 20% of Americans surveyed felt that purposeful misrepresentation of

16 PSI and Pain 1233 claims in the compensation system was acceptable (Public Attitude Monitor, 1992, 1993). Consistent with this, Mittenberg et al. (2002) reported base rates of malingering as high as 30% in personal injury and workers compensation cases. Thus, exaggeration and malingering are going to be present in a non-trivial proportion of pain patients so the question of reasonable and accurate symptomology is very important. Unfortunately, there are no published studies showing that clinicians can reliably determine whether someone is malingering or not based on interview. To the contrary, Ekman, O Sullivan, and Frank (1999) demonstrated an average false positive error rate of 40% when psychologists were asked to detect deception in video clips and opined that it is unlikely that judging deception from demeanor will ever be sufficiently accurate to be admissible in the courtroom (p. 265). Even the physical examination may not be able to accurately differentiate psychologically-involved cases from those without psychological involvement (Grevitt, Pande, O Dowd, & Webb, 1998). The presence of objective physical findings does not ensure that a patient s report of their subjective limitations is valid; Greve et al. (2003) reported Definite MPRD in patients with objective physical findings. Nor does the willingness to undergo invasive procedures such as spinal surgery rule out intentional exaggeration (Bianchini, Heinly, & Greve, 2004). It is also well understood that the results of psychological tests can be invalidated by biased responding and/or poor effort (Beetar & Williams, 1995; Heubrock & Petermann, 1998; Lees-Haley, 1989, 1990) but psychologists are also not very good at detecting malingered performance in their standard clinical tests without specific tests and indicators (Faust, Hart, Guilmette, & Arkes, 1988; Heaton, Smith, Lehman, & Vogt, 1978). Thus, given the importance of subjective report in determining disability in patients with pain and the fact that subjective report can be biased in clinically undetectable ways, well-validated indicators of exaggeration and malingering are essential in the clinical management of patients with pain. Well-validated means that the indicator has been empirically demonstrated to be capable of separating psychological versus physical mechanisms of symptom complaints and/or separating exaggerated versus non-exaggerated symptom complaints, including intentionally exaggerated versus non-intentionally exaggerated complaints and deficits. Though not as advanced as in neuropsychology, the development of indicators of malingering validated in pain conditions is moving forward (e.g., Etherton et al., 2005a,b,c, 2006; Larrabee, 2003c; Meyers, Millis, & Volkert, 2002). The present paper demonstrates the strongest methodology for developing and validating malingering indicators, an integration of simulator and known-groups designs. Because of the importance of accurate subjective report and behavior to the decision-making of clinicians working with pain patients, positive findings on well-validated indicators of exaggeration related to any of a patient s complaints need to be taken very seriously. Conclusion Chronic pain may reasonably be expected to contribute to cognitive deficits in some patients so it is important to evaluate cognitive function in the overall assessment of disability in pain patients. Therefore it is important to include measures of cognitive effort and malingering in psychological test batteries designed for the assessment of pain patients. This paper suggests that the PSI is sensitive to cognitive processing deficits in some patients with pain. At the same time, because evidence of cognitive impairment may lead to a larger workers compensation or personal injury award, incentive exists for pain patients to perform poorly on cognitive measures (Bianchini et al, 2004, 2005). Thus, it is important to include cognitive malingering measures in the psychological evaluation of