Relationship Between Neuropsychological Test Performance and ...

The Clinical Neuropsychologist 2004, Vol. 18, No. 2, pp. 249265

1385-4046/04/1704-446$16.00 # Taylor & Francis Ltd.

Relationship Between Neuropsychological Test Performance and Productivity at 1-Year Following Traumatic Brain Injury Timothy B. Atchison1,2,3, Angelle M. Sander3,4, Margaret A. Struchen3,4, Walter M. High, Jr.3,4, Tresa M. Roebuck3,4,6, Charles F. Contant5, Jeffrey S. Wefel1,3,7, Thomas A. Novack8, and Mark Sherer9,10 1 Department of Psychology, University of Houston, Houston, TX, USA, Department of Behavioral Sciences, West Texas A&M University, Canyon, TX, USA, 3 Brain Injury Research Center, The Institute for Rehabilitation and Research, Houston, TX, USA, Baylor College of Medicine, Houston, TX, USA, 4Department of Physical Medicine & Rehabilitation and 5 Department of Neurosurgery, 6National Rehabilitation Hospital, Washington, DC, USA, 7 Department of Neuro Oncology, University of Texas M.D. Anderson Cancer Center, USA, 8 Department of Physical Medicine & Rehabilitation, University of Alabama Spain Rehabilitation Center, Birmingham, Alabama, USA, 9Department of Neuropsychology, Methodist Rehabilitation Center, Jackson, Mississippi, USA, and 10The University of Mississippi Medical Center, Jackson, Mississippi, USA 2

ABSTRACT While there has been strong evidence for the ability of neuropsychological performance at resolution of posttraumatic amnesia to predict later productivity, there has been less conclusive evidence for the relationship of neuropsychological test scores to concurrent productivity status. The purpose of the current study was to evaluate the relationship of neuropsychological test performance at 1 year post-injury to productivity assessed at the same time point. Participants were 518 persons with medically documented TBI who were enrolled in the TBI Model Systems Research and Demonstration Project. Stepwise logistic regression was utilized to determine the contributions of neuropsychological test scores to productivity after accounting for demographic characteristics, injury severity, and pre-injury productivity. Missing neuropsychological test scores were accounted for in the model. Variables that remained in the model and accounted for a significant proportion of the variance included age, duration of impaired consciousness, pre-injury productivity, and scores on measures of GOAT, Logical Memory II, and Trail Making Test, part B. The results indicate that neuropsychological test performance provides important information regarding the ability of persons with injury to return to productive activities. The results also indicate that inability to complete neuropsychological tests at 1 year post-injury is associated with non-productive activity.

INTRODUCTION Traumatic brain injury (TBI) is a leading cause of disability. Based on figures yielded from studies conducted at various sites, Kraus

and McArthur (1999) estimated the U.S. incidence rate to be 235 per 100,000 for both fatal and non-fatal TBI combined. The annual cost of TBI to the United States economy has been estimated to be $25 billion (Kibby &

Address correspondence to: Timothy B. Atchison, Ph.D., Department of Behavioral Sciences, WTAMU Box 60296, Canyon, TX 790160001, USA. Tel.: þ 1-806-651-2729. Fax: þ 1-806-651-2728. Email: [email protected] Accepted for publication: January 23, 2003.

250

TIMOTHY B. ATCHISON ET AL.

Long, 1996). The cost of TBI is partly attributable to lost wages due to high rates of unemployment in persons with injury (Goran Fabiano, & Crewe, 1997). Researchers have reported rates of unemployment ranging from 22% to 66% (Brooks, McKinlay, Symington, Beattie, & Campsie, 1987; Dikmen et al., 1994; Rao et al., 1990; Rappaport, Herrero-Backe, Rappaport, & Winterfield, 1989; Sander, Kreutzer, Rosenthal, Delmonico, & Young, 1996; Weddell, Oddy, & Jenkins, 1980). Individuals who return to employment following TBI often experience reduced productivity, frequent job changes, and reduced responsibilities (Wehman et al., 1991). Accurate understanding of the factors associated with productivity in persons with TBI could assist in rehabilitation planning and in educating persons with injuries and their family members regarding employment potential. Cognitive and behavioral impairments are commonly believed to be a major factor impacting the productivity of persons with TBI (Brooks, Campsie, Symington, Beattie, & McKinlay, 1986; Sherer, Madison, & Hannay, 2000). Neuropsychological assessment has been an important means of documenting cognitive and behavioral impairments and providing feedback to persons with injury and their family members. However, the ability of neuropsychological measures to predict real-world functioning has been questioned (Hart & Hayden, 1996; Sbordone & Long, 1998). Previous studies have documented that neuropsychological test performance conducted early after injury (at resolution of posttraumatic amnesia or 1 month post-injury) was predictive of employment status at > or ¼ 1 year postinjury (Boake et al., 2001; Cifu et al., 1997; Dikmen et al., 1994; Fleming, Tooth, Hassell, & Chan, 1999; Fraser, Dikmen, McLean, & Miller, 1988;Najenson, Groswasser,Mendelson, & Hackett, 1980; Sherer et al., 2002b). A separate group of studies has provided some evidence that neuropsychological performance at a post-acute period ( > or ¼ 6 months after injury) has predictive ability with regard to

employment at a subsequent period (Ezrachi et al., 1991; Isaki & Turkstra, 2000; Lam, Piddy, & Johnson, 1991; O’Connell, 2000; Prigatano et al., 1984; Ruff et al., 1993). However, there is minimal evidence for the relationship of neuropsychological test performance to employment status assessed at the same point in time. The association between neuropsychological test performance and concurrent employment status is potentially very useful information for clinical neuropsychologists. Many practicing neuropsychologists do not have the advantage of access to an earlier neuropsychological evaluation when consulted on a particular case. In such cases, the clinician is often asked to make recommendations about the client’s potential for functioning based on current assessment. Some examples are determining the capacity to return to work for forensic evaluations and worker’s compensation claims, as well as determining disability for the purposes of obtaining benefits. Knowledge of the relationship between neuropsychological test performance and employment or productivity would also be useful in rehabilitation treatment settings. The information would assist in treatment planning, including setting treatment goals (employment vs. general independence) and targeting specific cognitive domains that are important for employment and productivity. Research documenting the relationship between neuropsychological test performance and productivity assessed at the same time point would afford some empirical basis for neuropsychologists faced with such assessment and treatment decisions. Sherer et al. (2002a) noted that there is no theoretical reason that assessments conducted closer in time to the point that outcome is measured should be less related to outcome than those conducted at an earlier time point. A few studies have attempted to address the relationship between neuropsychological performance and employment outcome assessed concurrently. These studies will be reviewed in detail. Three of the studies that examined the relationship between neuropsychological

NEUROPSYCHOLOGICAL PERFORMANCE AND PRODUCTIVITY

performance and concurrent employment utilized only one neuropsychological test variable. Weddell et al. (1980) investigated a small sample of 44 persons with severe TBI (post-traumatic amnesia > 7 days) who were consecutively admitted to an inpatient rehabilitation facility and were available for follow-up at 2 years post-injury. Only persons age 1639 were included. They found that persons who were unemployed had poorer scores on Raven’s Progressive Matrices compared to persons who were employed. Using the Tinker Toy Test, a purported measure of executive functioning, Bayless, Varney, and Roberts (1989) compared the performance of 25 persons with TBI who had returned to work to that of 25 persons who had not resumed work and 25 controls. The origin of the sample was not described. While the average injury severity was not reported, their inclusion criterion specified at least 30 min combined loss of consciousness and post-traumatic amnesia. Use of this inclusion criterion makes it probable that many subjects had experienced mild TBI. They found that persons who were employed for at least 75% of the time since their injury performed better than those who were not employed. Both employed and unemployed persons with TBI performed worse than the control group. Melamed, Stern, Rahmani, Groswasser, and Najenson (1985) investigated the relationship between employment and performance on a dual task involving simultaneous performance on rotary pursuit and delayed digit recall tasks. Subjects were 51 persons with injury who were seen at 23 years following discharge from inpatient rehabilitation. Persons who were competitively employed performed significantly better on the rotary pursuit task compared to persons who were employed in a ‘‘protected’’ setting. Persons working in the ‘‘protected’’ setting performed better than those who were unemployed. Brooks et al. (1987) reported on a sample of 101 patients who were admitted to a neurosurgical unit and were available for follow-up

251

at 27 years after injury. All subjects had been employed prior to injury. Participants were administered a neuropsychological test battery including Progressive Matrices, Mill Hill Vocabulary Scale, word fluency, Token Test, Wechsler Memory Scale Revised (WMSR) Logical Memory, Buschke Selective Reminding Test, learning of three verbal paired associates, Rey-Osterreith Complex Figure, and Paced Auditory Serial Addition Test (PASAT). Serial t-tests indicated that persons who were employed (defined as working at least part-time in the week prior to follow-up) scored significantly higher on Progressive Matrices, Logical Memory, Buschke long-term storage, paired associate learning, Rey Figure Copy, and PASAT. When scores from these tests were entered into a stepwise linear regression model, only Logical Memory and PASAT entered the equation. No other neuropsychological test score significantly added to the prediction of employment. Unfortunately this study did not account for demographic factors or injury severity in the prediction of current employment. Therefore, it is unclear whether completion of the neuropsychological testing contributed information that was not redundant with that known from the more easily obtained demographic and injury-related variables. Hanlon, Demery, Martinovich, and Kelly (1999) reported on 100 individuals with mild TBI who were consecutively admitted to an outpatient clinic. A neuropsychological test battery was administered between 3 and 40 months after injury, and included Trail Making Test, Wechsler Adult Intelligence Scale Revised (WAISR) Digit Span, WMSR Logical Memory, WMSR Visual Reproduction, California Verbal Learning Test, Boston Naming Test, Controlled Oral Word Association, Judgement of Line Orientation, Finger Tapping Test, Grooved Pegboard, Wisconsin Card Sorting Test, and Beck Depression Inventory. Neuropsychological test performance was used to predict employment at approximately 1 year after injury. Employment was coded into three

252


categories: good (returned to work at previous level); modified (returned to work at reduced capacity); and poor (failed to return to work). A logistic regression model was used employing the following predictor variables: (1) age; (2) injury mechanism (acceleration=deceleration with head hitting object, acceleration=deceleration without head hitting object, head struck by falling object); (3) type of injury (motorvehicle accident, fall, assault, etc.); and (4) neuropsychological test scores that showed a univariate relationship to employment status (WMSR Logical Memory Immediate, WMSR Visual Reproduction Delayed, Judgment of Line Orientation, Trail Making Test B, and the Beck Depression Inventory). Scores on Logical Memory and Visual Reproduction distinguished between good and modified=poor outcomes, while age and injury characteristics were not predictive. In contrast, no neuropsychological measure distinguished between good=modified and poor outcomes. Only injury mechanism resulted in an accurate classification of these outcomes. The generalizability of this study was limited by the inclusion of only persons with mild TBI. Furthermore, the fact that employment status was assessed at 1 year while neuropsychological assessment was conducted over an extended period makes it unclear how neuropsychological scores from different time points relate to outcome at 1 year. Sherer et al. (2002a) concluded that these studies provided inconclusive evidence for the ability of neuropsychological test performance to predict concurrently measured employment status. Conclusions based on these studies were limited by methodological concerns, including small sample size; the use of referred rather than consecutively admitted samples; the use of multiple measures and multiple analyses with insufficient sample size for power; and inadequate control for other factors that may impact outcome, such as age and injury severity. Notably, for many of the studies, determining the relationship between neuropsychological performance and outcome was not the primary

purpose of the study but was an incidental finding. The present study investigates neuropsychological test performance’s association with productivity status at the time of evaluation. Information gained from relationship models could be helpful to neuropsychologists who are making recommendations regarding return to work and other productive activities in addition to judgments about forensic and disability questions. The study attempts to address methodological limitations in the existing literature by using a large multicenter sample of persons with TBI consecutively admitted to major rehabilitation centers; by accounting for demographic and injury characteristics that could impact outcome; by using statistical techniques to account for missing data; and by utilizing a homogenous time point for the assessment of neuropsychological performance and productivity status. The current study utilized both employment and academic pursuit to measure productivity, since persons under the age of 21 may not have competitive employment as their primary means of productivity. Exclusion of this age group would leave out a large proportion of persons with TBI. Accounting for educational pursuits as ‘‘productive’’ activities allows for a more accurate representation of productivity status among persons with TBI. It is hypothesized that neuropsychological test performance will add significantly to the variance in productivity status beyond what is accounted for by demographic factors and injury severity.

METHOD Participants Participants were 805 persons with TBI who were enrolled (as of the Spring of 1999) in the National Institute on Disability and Rehabilitation Research (NIDRR) TBI Model Systems Research and Demonstration Project, a prospective, multicenter longitudinal study on outcome following TBI. Data were collected at 17 centers across the United States. Criteria for enrollment in the


Model Systems study included: diagnosis of TBI resulting in admission to the emergency department of a Model System Level I trauma center within 24 hr of injury; age 16 years at the time of injury; receipt of acute care and inpatient rehabilitation within the Model System; and provision of informed consent by the patient or a relative. As part of the Model Systems protocol, participants are followed up at 1 year after injury. For inclusion in the current study, subjects had to have information available on both neuropsychological test performance and productivity outcome at this 1-year follow-up. Thirty-five subjects were not administered neuropsychological testing because they were too cognitively or behaviorally impaired to complete testing. These subjects were not excluded from analysis because their inability to complete testing was not random, but was related to cognitive factors of interest. Other subjects did not receive the neuropsychological evaluation for reasons other than cognitive impairment or behavioral disturbance and these subjects were excluded from analysis. The reasons for exclusion follow: unable to schedule a testing time due to having returned to work or school (n ¼ 23); primary language other than English or Spanish (n ¼ 5); inability to arrange transportation (n ¼ 22); refused follow-up or did not show up for scheduled appointments (n ¼ 59); and missing information in the data base as to why the battery was not administered or what battery was administered (n ¼ 121). An additional 30 subjects were excluded due to missing data on the Glasgow Coma Scale (GCS) at the time of injury. Because the outcome variable of interest was productivity, persons over the age of 65 were excluded, since they had the potential to be retired for reasons unrelated to injury. Following exclusions, 518 subjects remained and were included in analyses. The individuals excluded from the analysis were not significantly different from those included with regard to pre-injury productivity, injury severity, marital status, gender, or education. As anticipated based on inclusion criteria, those excluded were significantly older than those individuals retained in the analysis ( p < .05). The average age of the 518 research participants was 33.7 years (SD ¼ 12, range ¼ 1664) and the average number of years of education was 11.7 (SD ¼ 2.2, range ¼ 820). The mean of the highest first day GCS score was 9.6 (SD ¼ 3.7, range ¼ 315). It should be noted that, regardless of injury severity, all participants had impairments that were severe enough to warrant

253

an inpatient rehabilitation stay on a specialized brain injury rehabilitation unit. Persons with Glasgow Coma Scale scores of 1315 had positive neuroimaging findings or other abnormal neurological findings that would make their course of recovery similar to that of a person with moderate injury (Williams, Levin, & Eisenberg, 1990). The racial=ethnic composition of the sample was of 53% white, 36% black, 8% Hispanic, and 3% other.

Measures Predictor Measures The variables used as predictor measures are shown in Table 1. The variables selected for analysis were limited to those available in the TBI Model System database. Predictor measures fell into three categories: demographic variables, injury severity indices, and neuropsychological tests. For the demographic variables, age and education were represented as continuous variables. Preinjury productivity was represented as a categorical variable with two levels. The productive group included persons who were competitively employed, attending school full time, or attending

Table 1. Associated Variables. Demographic variables Age Pre-injury productivity (two groups) Education Injury severity indices Days until follows commands Pupillary reactivity (four groups) GCS Highest in the first 24 hours Neuropsychological tests GOAT Token Test COWA Benton Visual Form Discrimination Grooved Pegboard WAIS-R Block Design WMS-R Digit Span RAVLT Total Trials IV RAVLT immediate recall Logical Memory I & II Trail Making A & B Symbol Digit Modalities Test Oral Wisconsin Card Sorting Test (# of categories)

254


school part-time and working part-time. All other persons were classified as non-productive. For the injury severity variables, days to follows commands was coded as a continuous variable, calculated as the time from date of injury to date when simple motor commands were followed two out of two times within a 24-hr period. Pupillary reactivity, coded from medical records, was categorized into four groups: normal; at least one pupil reacting sluggishly; at least one pupil fixed; or unknown. The individual neuropsychological tests shown in Table 1 are described below. While it is acknowledged that these tests are not the only tests that would be useful for assessing cognitive impairment in persons with TBI, the choice of tests was limited by those that were available as part of the national Model Systems data base. The raw score was used when available, as age and education are entered in the analysis as demographic variables and use of corrected scores may lead to over correction. In order to reduce redundant information in the model, a single score was chosen from each neuropsychological test when more than one score was available. The score was chosen by consensus of 3 neuropsychologists, 1 neuropsychology fellow, and a neuropsychology graduate student. For example, for Symbol Digit Modalities Test, both Written and Oral scores were available, but only the Oral score was included in the final analyses. Similarly, for the Logical Memory Test, only the Delayed Recall score was included. A decision was made to include two measures from the RAVLT, total words recalled on trials IV (i.e., as a measure of learning) as well as recall following the immediate delay (i.e., measure of memory).

Galveston Orientation and Amnesia Test ðGOATÞ ðLevin, O’Donnell, & Grossman, 1979Þ. The GOAT was developed as a prospective measure of emergence from post-traumatic amnesia (PTA). The items include assessment of orientation to person, place, time, and situation. There are also items to assess duration of retrograde and anterograde amnesia. The total number of error points was used as the score for the current analyses.

Token Test from the Multilingual Aphasia Examination ðBenton & Hamsher, 1989Þ. The Token Test assesses comprehension and execution of simple and complex verbal commands. Examinees are asked to carry out commands (e.g. pick up or point to) related to tokens of various sizes, shapes, and colors. This test was scored in the standard

manner (i.e., two points for each correct response on the first trial), and raw scores were used in the analyses.

Controlled Oral Word Association Test from the Multilingual Aphasia Examination ðCOWAÞ ðBenton & Hamsher, 1989Þ. The COWA assesses verbal fluency and word generation. Examinees are asked to generate as many words as possible that begin with a specified letter. The age- and education-adjusted score was used in the analyses as the raw score was not available in the database.

Benton Visual Form Discrimination ðBenton, Hamsher, Varney, & Spreen, 1983Þ. This test assesses visual recognition and discrimination skills. Examinees are shown a target stimulus set consisting of three designs. While still viewing the target stimulus, they are asked to choose the matching stimulus set from among a choice of 4. This test was scored in the standardized manner and raw scores were used in the analyses.

Block Design from the Wechsler Adult Intelligence Scale Revised ðWAISRÞ ðWechsler, 1981Þ. Block Design is a measure of visuoconstruction skills. Examinees are required to construct block designs to match those shown on a stimulus card. This test was scored in the standardized manner and raw scores were used in the analyses.

Grooved Pegboard Test ðKlove, 1963; Matthews & Klove, 1964Þ. This is a measure of manual dexterity that requires speeded placement of grooved pegs into a slotted board. The score used for the current analyses was the time to placement of all pegs, using the dominant hand.

Logical Memory from the Wechsler Memory Scale Revised ðWMSR; Wechsler, 1987Þ. This is a story recall task that requires examinees to recall 2 verbally presented stories immediately following presentation (Logical Memory I) and after a 30-min delay (Logical Memory II). For the current study, the Delayed Recall score was used. The total units of information recalled for the two stories (i.e., raw score) was used in the analyses.

Rey Auditory Verbal Learning Test ðRAVLTÞ ðTaylor, 1959Þ. The RAVLT assesses verbal learning and recall. The test assesses a person’s ability to learn a list of 15 unrelated words over


5 learning trials. The scores used for the current analyses were the total number of words recalled across the five trials (RAVLT Total) and the number of words recalled after administration of a second interference list (RAVLT VI). This test was scored in the standardized manner and raw scores were used in the analyses.

Digit Span from the Wechsler Memory Scale Revised ðWMSR; Wechsler, 1987Þ. The Digit Span test consists of two parts. Digits Forward (DSf) assesses immediate memory span for verbally presented numbers. Digits Backward (DSb) is a measure of short-term concentration that requires examinees to repeat verbally presented numbers in reverse order. This test was scored in the standardized manner and raw scores were used in the analyses for both digits forward and backwards.

Trail Making Test ðReitan, 1955Þ. The Trail Making Test consists of two parts that assess visuomotor tracking and information processing speed. Trail Making A requires persons to draw lines connecting circled numbers in consecutive order. For Trail Making B, examinees must draw lines to connect letters and numbers in an alternating ascending sequence (e.g. 1, A, 2, B. . .). Time in seconds to complete each of these two trials was used in the analyses. Symbol Digit Modalities Test ðSmith, 1982Þ. This is a test of information processing speed that contains both oral and written portions. Each portion contains a key that matches the numbers 1 through 9 with symbols. Below the key is a series of symbols with an empty box below each. Examinees must write in (written version) or verbalize (oral version) the number that matches each symbol, and performance is timed. For the current study, the score for the oral version of the test was used. The score included the total number of items completed correctly within the time limit.

Wisconsin Card Sorting Test ðWCSTÞ ðGrant & Berg, 1948Þ. The WCST is a measure of problem-solving=reasoning. Examinees must match cards to 1 of 4 stimulus cards. Matches must be made on the basis of color, shape, or number, but the concept for matching is changed at intervals throughout the test. The examinee receives verbal feedback regarding the accuracy of each choice, and must make use of the feedback to determine the appropriate match. The total number of categories achieved out of the six possible was the score used in the current analyses. The variable

255

of categories was used rather than perseverative responses due to the fact that number of perseverative responses was not available in the database.

Outcome Measures The Productive Activity Scale from the Community Integration Questionnaire (CIQ) (Willer, Rosenthal, Kreutzer, Gordon, & Rempel, 1993) was used as the outcome measure. The CIQ is a 15-item questionnaire designed to quantify an individual’s integration into home and family life, social activity, and productive activity. The CIQ Productive Activity Scale was used in the current study. Test-retest reliability for this subscale ranges from 0.83 to 0.97 (Willer, Ottenbacher, & Coad, 1994). This subscale has also been shown to have sensitivity to the types of problems experienced by persons with TBI at 1 year after injury (Hall et al., 1996). In the current study, scoring was based on the recent factor analysis conducted by Sander et al. (1999). In that analysis, the Productive Activity Scale consisted of three items assessing participation in work, school, and volunteer work. Participation in these individual activities are weighted and combined to yield a total score ranging from 0 to 5. The highest score is given to persons who are working full time and attending school part-time or attending school full time and working part-time. The next highest score is given to persons who are either working full time or attending school full time. The next highest group consists of persons who are working part-time or going to school part-time. The next category represents persons who are actively looking for work and=or engaging in volunteer work five or more times a month. The second to lowest score is given to persons who are volunteering between one and four times per month, but are not working, attending school, or actively seeking work. The group with the lowest scores includes those who are not working or looking for work, not attending school, and not volunteering. Due to low cell sizes for some of the groups, the top two groups were collapsed into one category, and the bottom two groups were collapsed into one category, yielding four outcome groups. These four outcome groups served as the dependent variable for the analysis.

Procedure As part of the Model Systems protocol, the tests listed above were administered at 1-year postinjury, plus or minus 1 month. Tests were

256


administered by a trained examiner according to standardized instructions. Attempts were made to administer the entire battery, but for some participants individual tests were not administered due to cognitive or behavioral impairments, patient refusal, time constraints, and so forth. Of the 518 participants, 191 (37%) had data missing on at least one of the neuropsychological tests. A smaller subset of participants (n ¼ 32) had missing scores on all neuropsychological tests, because they were cognitively unable to complete the battery. It was believed that the inability to complete a test might be important in predicting productivity outcome, since factors that could contribute to missing data, such as cognitive impairments, fatigue, and behavioral issues, may also impact productivity. Furthermore, there was a positive relationship between injury severity (as assessed by number of days until commands were followed) and the number of missing tests, r ¼ .28, p < .01, indicating that ability to complete tests was related to the injury. In order to model the missing test data, raw test scores from each neuropsychological test were coded into four groups. The first group was defined as all subjects who had a missing score on that test. For those remaining subjects with

completed neuropsychological data, three groups were formed by dividing the test scores as evenly as possible into a top third, middle third, and bottom third performance group. Table 2 provides the cut off scores for each group on each test and the group mean and standard deviation for each test.

Analyses As a first step, the relationship of performance on each of the neuropsychological tests to productivity status, after accounting for age and injury severity (as assessed by number of days until commands were followed), was determined using logistic regression. Next, the association of demographic, injury severity, and neuropsychological variables to CIQ productivity levels was determined using a logistic regression backward elimination procedure. The CIQ Productivity Scale was reclassified into four levels as described in the Measures section. All variables shown in Table 1 were entered as predictor variables in the model. Listwise deletion of cases with missing data on the demographic or injury severity variables yielded 467 cases in the final sample.

Table 2. Cut Points for Top, Middle and Bottom Thirds of Sample for Neuropsychological Variables With Means and Standard Deviations. Top 1=3

Middle 1=3

Bottom 1=3

Mean

SD

Variables in model GOAT (error points)a Logical Memory II Trail Making B

n ¼ 495 n ¼ 468 n ¼ 434

0 19 71

19 1018 72119

10 9 120

8.3 14.6 113.8

12.9 9.3 69.5

Variables not in model Token Testa COWA (adjusted) Digits Forward (points) Digits Backward (points) Logical Memory II RAVLT Total RAVLT VI Grooved Pegboard Symbol Digit Oral Trail Making Part A Benton Form Discriminationa Block Design (raw) WCST (categories)a

n ¼ 459 n ¼ 466 n ¼ 487 n ¼ 487 n ¼ 473 n ¼ 465 n ¼ 463 n ¼ 457 n ¼ 450 n ¼ 454 n ¼ 476 n ¼ 427 n ¼ 402

44 35 9 7 24 46 10 75 49 30 32 31 6

4143 2634 78 56 1523 3345 69 7695 3648 3147 2931 2030 35

40 25 6 4 14 32 5 96 35 48 28 19 2

40.5 30.3 7.3 5.6 19.4 39.7 7.2 94.1 42.5 46.3 28.4 25.0 4.3

5.9 11.2 2.3 2.2 8.7 13.5 4.2 39.1 16.3 31.1 4.4 11.5 2.2

Note. aDue to ceiling effects the top group was defined as a perfect score and the rest of the sample was split as evenly as possible to create the other groups. 33% scored 0 on the GOAT, 36% scored 44 on the Token test, 28% scored 32 (40% scored 31 or above) on the Benton Form Discrimination test, 55% completed six categories on the WCST.


Analysis indicated no significant differences on the variables age, education, previous productivity, or GCS scores for those delete compared to those individuals retained.

RESULTS Table 3 shows the results of the individual logistic regression equations assessing the contribution of each neuropsychological test score to productivity status after accounting for age and number of days until commands were followed. As can be seen in the table, each of the neuropsychological test scores contributed significantly to the variance in productivity status. All variables shown in Table 1 were next entered into a regression model to predict productivity. Using a backward elimination regression procedure, variables were removed until a resolution was reached consisting of the following seven variables: days to follow commands; age; pupillary reactivity; pre-injury productivity status; GOAT; Logical Memory II; and Trail Making B. A Wald’s criterion of 0.05 was used to determine which variables remained in the equation. Table 3. Test of Significance Results for Each Individual dual Neuropsychological Variable With Age and Injury Severity Controlled. Variable namea GOAT Logical Memory II Trail Making Part B Token Test COWA Digits Forward Digits Backward RAVLT Total RAVLT VI Grooved Pegboard Symbol Digit Oral Trail Making Part A Benton Form Discrimination Block Design WCST Categories

Wald v2

p value

48.3 65.4 76.9 35.2 23.4 34.1 30.8 41.8 28.3 42.6 64.6 53.8 31.2 36.5 47.0