Racial Differences in Employment Outcomes After Traumatic Brain Injury

988

SPECIAL SECTION: ORIGINAL ARTICLE

Racial Differences in Employment Outcomes After Traumatic Brain Injury Juan Carlos Arango-Lasprilla, PhD, Jessica M. Ketchum, PhD, Kelli Williams, MPH, MS, OTR/L, Jeffrey S. Kreutzer, PhD, Carlos D. Marquez de la Plata, PhD, Therese M. O’Neil-Pirozzi, ScD, Paul Wehman, PhD ABSTRACT. Arango-Lasprilla JC, Ketchum JM, Williams K, Kreutzer JS, Marquez de la Plata CD, O’Neil-Pirozzi TM, Wehman P. Racial differences in employment outcomes after traumatic brain injury. Arch Phys Med Rehabil 2008;89:988-95. Objective: To examine racial differences in employment status and occupational status 1 year after a traumatic brain injury (TBI). Design: Retrospective study. Setting: Longitudinal dataset of the Traumatic Brain Injury Model Systems national database. Participants: Subjects with primarily moderate to severe TBI (3468 whites vs 1791 minorities) hospitalized between 1989 and 2005. Interventions: Not applicable. Main Outcome Measures: Employment status (competitively employed or unemployed) and occupational status (professional/managerial, skilled, or manual labor) at 1 year postinjury. Results: Race and/or ethnicity has a significant effect on employment status at 1 year postinjury (␹12⫽58.23, P⬍.001), after adjusting for preinjury employment status, sex, Disability Rating Scale at discharge, marital status, cause of injury, age, and education. The adjusted odds of being unemployed versus competitively employed are 2.17 times (95% confidence interval, 1.78 –2.65) greater for minorities than for whites. Race and ethnicity does not have a significant effect on occupational status at 1 year postinjury. Conclusions: With this empirical evidence supporting racial differences in employment outcomes between minorities and whites at 1 year postinjury, priority should be given to tailoring interventions to maximize minority survivors’ work-related productivity. Key Words: Brain injuries; Employment; Outcomes research; Race; Rehabilitation. © 2008 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation

From the Departments of Physical Medicine and Rehabilitation (Arango-Lasprilla, Williams, Kreutzer, Wehman) and Biostatistics (Ketchum), Virginia Commonwealth University, Richmond, VA; Departments of Psychiatry and Neurology, University of Texas Southwestern Medical Center, Dallas, TX (Marquez de la Plata); Department of Physical Medicine and Rehabilitation, Spaulding Rehabilitation Hospital, Harvard Medical School, Boston, MA (O’Neil-Pirozzi); and Department of Speech-Language Pathology and Audiology, Northeastern University, Boston, MA (O’Neil-Pirozzi). Supported by the National Institute on Disability and Rehabilitation Research, U.S. Department of Education (grant nos. H133A020516, HI33B040011). No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors or upon any organization with which the authors are associated. Reprint requests to Juan Carlos Arango-Lasprilla, PhD, Dept of Physical Medicine and Rehabilitation, Virginia Commonwealth University, Rehabilitation Psychology and Neuropsychology, School of Medicine, Theater Row Bldg, 730 E Broad St, Richmond, VA 23219, e-mail: [email protected]. 0003-9993/08/8905-00696$34.00/0 doi:10.1016/j.apmr.2008.02.012

Arch Phys Med Rehabil Vol 89, May 2008

RAUMATIC BRAIN INJURY (TBI) is one of the most T prevalent and debilitating conditions in the United States. Of the estimated 1.4 million people who sustain a TBI annually, about 1.1 million are treated and released from emergency departments, 235,000 are hospitalized, and 80,000 to 90,000 experience permanent disability from injury.1 TBI typically affects people either early in their productive years or once they have established a productive life. Approximately 40% of the 200,000 TBI survivors who are hospitalized each year are between the ages of 15 and 44.2 TBI causes physical disability as well as cognitive, behavioral, and emotional sequelae that affect the ability to return to work, even years after the initial injury. In 2000, direct medical costs and indirect costs, including lost productivity, due to TBI-related injuries in the United States were estimated at $60 billion.3 Besides the economic impact of lost years of work on the person, family, and society, research indicates that employment is one of the most important psychosocial predictors of well-being, quality of life, social integration, and recovery in survivors with TBI.4-6 Wehman et al6 conducted an extensive literature review from the past 20 years in the area of employment and TBI outcomes and reported 3 common sets of factors that predicted successful return-to-work after TBI: (1) preinjury sociodemographic characteristics (eg, age, education, preinjury employment status), (2) injury severity (eg, length of coma or posttraumatic amnesia [PTA]), and (3) levels of disability and functional status preinjury and postinjury. Because of the high incidence of TBI in racial and ethnically diverse communities and an increase in minority survivors with TBI, recent studies have examined the role of race and/or ethnicity on return-towork postinjury. Sherer et al7 examined productivity outcomes in 633 white and 450 minority TBI survivors who received treatment at a Traumatic Brain Injury Model Systems (TBIMS) center. Productivity in this study was defined as being competitively employed at least part-time, being a student at least part-time, or being a full-time homemaker. Univariate analyses indicated that, compared with whites, blacks were 2.76 times and other minorities were 1.92 times more likely to be nonproductive. After controlling for preinjury productivity, education, and cause of injury, blacks and other minorities were still twice as likely as whites to be nonproductive. Rosenthal et al8 examined the relationship between minority status and functional outcomes after TBI in 568 TBI survivors from 4 TBIMS centers. Minorities had lower Community Integration Questionnaire (CIQ) productivity subscale scores compared with whites. At the time of the injury, 60% of whites and 45% of minorities were competitively employed, but at 1 year postinjury, 30% of whites and 13% of minorities had returned to work. In a subanalysis of 626 male veterans with mild head injury, Vanderploeg et al9 found that race, loss of consciousness (LOC), and current region of residence moderately predicted full-time work status. Compared with whites, minorities with a history of LOC had significantly lower rates of employment in

RACE AND EMPLOYMENT OUTCOMES AFTER TBI, Arango-Lasprilla

the Midwestern, Northeastern, and Southern regions of the United States. Moderating factors in return to work and job stability 1 to 4 years post-TBI were examined by Kreutzer et al10 in 186 adults from 6 TBIMS centers. Job stability was categorized as stably employed (employed at 3 follow-up periods), unstably employed (employed at 1 or 2 of 3 follow-ups), and unemployed (not employed at any of the 3 follow-up periods). Minorities were less likely to be employed and have stable employment compared with whites in this study. Regardless of race and/or ethnicity, TBI survivors tend to be unemployed and/or unproductive for at least 1 year postinjury. Unfortunately, rates of unemployment after TBI are even higher for minorities compared with whites, causing financial difficulties and stress for survivors and their families. In fact, minorities are less likely to be employed postinjury, independent of other factors known to influence return-to-work. In general, the limitations of previous work in the area of race, TBI, and return-to-work include relatively small or homogenous samples, minority groups consisting mainly of blacks, and a lack of control for potential confounders. To our knowledge, the present study will be the first to examine the relationship between race and employment status (competitively employed vs not competitively employed) at 1 year post-TBI in a large, longitudinal sample of diverse patients (whites, blacks, Hispanics, Asians, Native Americans) from across the 16 TBIMS centers. It was hypothesized that minorities would be more likely to lack competitive employment at 1 year postinjury compared with whites, after controlling for factors that significantly affect employment status. Furthermore, of those TBI survivors who were employed, minorities were expected to be more likely to have a manual labor occupation (vs a skilled or professional/managerial occupation) at 1 year postinjury compared with whites, even after controlling for factors that significantly affect occupational status. METHODS Participants The National Institute on Disability and Rehabilitation Research funds 16 comprehensive TBI rehabilitation programs known as the TBIMS. Records of all patients receiving rehabilitation are maintained in a centralized database. Approval for informed consent for each funded TBIMS center was given by the individual institutional review boards. Every patient (or patient’s legal guardian or family member if appropriate) provided informed consent to be enrolled in the study. For the present study, data from patients with primarily moderate to severe TBI from 1989 to 2005 were extracted from the TBIMS national database. Moderate to severe traumatic brain injury is determined in the TBIMS by either PTA for longer than 24 hours, intracranial neuroimaging abnormalities due to trauma, LOC for more than 30 minutes (excluding unconsciousness as a result of drug or alcohol use), or emergency department– documented Glasgow Coma Scale (GCS) score of less than 13 (excluding low scores resulting from drugs or intubations).11 The final sample consisted of 5259 subjects, 3468 whites and 1791 minorities (1238 blacks, 384 Hispanics, 142 Asians, 27 Native Americans). Measures Demographics. The race and/or ethnicity variable was categorized according to those who self-reported as black, Hispanic, Asian, Native American, or white. Age was measured in years and ranged from 16 to 89 years. Sex was a dichotomous

989

variable (male, female). Years of education were dichotomized into less than high school (grades 1–11) or high school degree, general educational diploma (GED), or greater (GED, GED/ high school, high school, trade school, high school diploma, some college, associate’s degree, bachelor’s degree, master’s degree, doctoral-level degree). Marital status was dichotomized as married or not married (including the database categories single, divorced, separated, and widowed). Preinjury employment status was categorized as either competitively employed (only subjects categorized as engaging in paid employment) or unemployed (unemployed, full-time student, part-time student, homemaker, volunteer work, others). Injury and rehabilitation characteristics. Injury severity was determined by the GCS score at admission to the emergency department and modeled as a continuous variable. The GCS scores ranged from 3 to 15. Cause of injury was dichotomized as either violent (gunshot, blunt assaults, other violence) or nonviolent (vehicular, sports-related, fall, pedestrian accident). Disability status at admission and discharge. The Disability Rating Scale (DRS) measures general functioning of a person (from coma to full activities at home or in the community) and has been used with adolescent and adult TBI survivors in inpatient rehabilitation. The scale has 8 items, and raters must evaluate a patient’s arousal and awareness, cognitive ability to handle self-care, physical dependence on others, and psychosocial adaptability. DRS scores were measured at admission (range, 0 –29; median, 11; mean, 12.5) and at discharge (range, 0 –26; median, 7; mean, 6.1). A DRS score of 0 indicates no disability and a DRS of 29 indicates an extreme vegetative state. Interrater reliability has been found to be high (␬ range, .97–.98).12 DRS score at discharge was chosen as a potential covariate rather than DRS score at admissions because this score is a more likely predictor of unemployment 1 year after injury. Employment status at 1 year postinjury. Employment status was dichotomized as either competitively employed or unemployed. Those classified as unemployed are considered not competitively employed, and this includes the employment categories unemployed, full-time student, part-time student, homemaker, volunteer work, and others. Occupational status at 1 year postinjury. The TBIMS census occupational categories at 1 year postinjury were clustered into the 3 groups: professional/managerial (executive, administrative, managerial), skilled (technicians and related support, sales, administrative support, protective service, farming, forestry, fishing, precision production, craft, rapier), and manual labor (machine operators, assemblers, inspectors, private household, transportation, material moving, handlers, equipment cleaners, helpers, laborers). Statistical Analysis Preliminary analyses. All statistical analyses were conducted using SAS.a Because the design was not randomized, to identify potentially confounding factors, the ethnic groups were compared in relation to other demographics at the time of injury (sex, marital status, employment status, education level, cause of injury, age, and injury severity [GCS, DRS]) using chi-square tests for categorical variables and equal-variance t tests for continuous variables. The primary hypotheses address the effects of race and/or ethnicity on unemployment and on occupational status 1 year after TBI. To correctly understand these effects, the effects of other variables (covariates) that may affect unemployment or occupational status must be adjusted for in the final analyses. The covariates considered for adjustment were sex, marital Arch Phys Med Rehabil Vol 89, May 2008

990

RACE AND EMPLOYMENT OUTCOMES AFTER TBI, Arango-Lasprilla

status, education level, cause of injury, DRS score at admissions (or discharge), GCS score at admission, PTA, and age. In addition, preinjury employment status was considered for the employment analysis, and preinjury occupational status was considered for the occupational status analysis. Minority groups (blacks, Hispanics, Asians, Native Americans) were initially examined in terms of outcomes (employment and occupational status at 1-year follow-up). For the employment outcome a multiple logistic regression model was fit that included an effect for minority ethnicity and adjusted for the covariates sex, marital status, preinjury employment status, education level, cause of injury, DRS score at admissions (or discharge), GCS score at admission, PTA, and age. If this model indicated that minority ethnicity did not have an effect on employment at 1-year follow-up, then the minority groups were to be collapsed into 1 group, and the primary analysis was to model minorities versus whites. If this model indicated that minority ethnicity does have an effect on employment at 1-year follow-up, then the minority groups were not to be collapsed and the primary analysis was to model each ethnicity separately. For the occupational status outcome, similar methods comparing the minorities were used with a multinomial multiple logistic regression model adjusting for sex, marital status, preinjury occupational status, education level, cause of injury, DRS score at admissions (or discharge), GCS score at admission, PTA, and age.

Primary analyses for employment status and occupational status. Initially, 2 logistic regression models were fit to determine the unadjusted effect of ethnicity on employment and occupational status. After this, the adjusted models for employment and occupational status were obtained using model building strategies as outlined in Hosmer and Lemeshow.13 This process involved the following steps: (1) univariate logistic regression models for each variable were fit, and any variable with a P value less than .25 in the univariate model was considered for the adjusted model; (2) the adjusted models were fit with all potential covariates obtained from step 1, and the adjusted effect of each variable was examined; (3) any variable that no longer contributed to the fit of the model was then removed; (4) the assumptions of linearity in the logit for continuous variables were assessed; (5) interactions among the variables were then examined by adding each interaction term to the adjusted model (independently) and assessing their contribution to the fit of the model; and (6) the final adjusted model was then assessed with respect to goodness of fit. RESULTS Preliminary Analysis Descriptive statistics (means, percentages) at the time of injury for minorities and whites are shown in table 1. At the time of injury, minorities were more likely to have lower levels

Table 1: Demographic and Injury Characteristics Nominal Variables

n

Sex Male Female Marital status Married Not married Preinjury employment status Employed Unemployed Education ⬍High school ⱖHigh school Cause of injury Violent Nonviolent

3884 1374 1624 3627 3276 1878 1725 3372 805 4405

DRS score at admission DRS score at discharge FIM score at admission FIM score at discharge GCS score at admission PTA (d)

Minorities

(n⫽3468) 71.7% 28.3% (n⫽3464) 34.3% 65.7% (n⫽3411) 66.8% 33.2% (n⫽3364) 27.9% 72.1% (n⫽3433) 7.8% 92.2%

(n⫽1790) 78.2% 21.8% (n⫽1787) 24.3% 75.7% (n⫽1743) 57.1% 42.9% (n⫽1733) 45.4% 54.7% (n⫽1777) 30.3% 69.7%

␹2 (df) P

OR (95% CI)

26.37 (1) P⬍.001

1.42 (1.24–1.62)

56.29 (1) P⬍.001

1.62 (1.43–1.85)

46.42 (1) P⬍.001

1.51 (1.34–1.70)

152.74 (1) P⬍.001

2.14 (1.90–2.42)

432.89 (1) P⬍.001

5.18 (4.41–6.09)

Whites Mean ⫾ SD

Minorities Mean ⫾ SD

t (df) P

Difference (SE) (95% CI)

(n⫽3462) 38.81⫾18.22 (n⫽3383) 12.46⫾5.90 (n⫽3394) 6.07⫾4.03 (n⫽1885) 55.39⫾26.27 (n⫽1935) 96.35⫾23.32 (n⫽2624) 9.01⫾4.49 (n⫽2474) 27.70⫾30.26

(n⫽1790) 36.11⫾15.29 (n⫽1753) 12.63⫾5.35 (n⫽1758) 6.37⫾3.73 (n⫽1113) 56.59⫾25.40 (n⫽1146) 96.07⫾21.92 (n⫽1569) 9.08⫾4.26 (n⫽1226) 26.84⫾24.72

5.36 (5250) P⬍.001 ⫺1.02 (5134) P⫽.308 ⫺2.58 (5150) P⫽.010 ⫺1.22 (2996) P⫽.224 0.33 (3079) P⫽.744 ⫺0.54 (4191) P⫽.590 0.87 (3698) P⫽.386

2.70 (0.50) (1.71 to 3.68) ⫺0.17 (0.17) (⫺0.50 to 0.16) ⫺0.30 (0.12) (⫺0.52 to ⫺0.07) ⫺1.19 (0.98) (⫺3.12 to 0.73) 0.28 (0.85) (⫺1.39 to 1.94) ⫺0.08 (0.14) (⫺0.35 to 0.20) 0.86 (1.00) (⫺1.09 to 2.82)

Continuous Variables

Age (y)

Whites

Abbreviations: CI, confidence interval; OR, odds ratio; SD, standard deviation; SE, standard error.

Arch Phys Med Rehabil Vol 89, May 2008

991

RACE AND EMPLOYMENT OUTCOMES AFTER TBI, Arango-Lasprilla Table 2: Preinjury Employment Versus Follow-Up Employment by Ethnicity One-Year Follow-Up Whites

Minorities

Preinjury

Employed

Unemployed

Employed

Unemployed

Employed Unemployed

793 (44.48) 108 (12.63)

990 (55.52) 747 (87.37)

161 (23.75) 35 (7.07)

517 (76.25) 460 (92.93)

NOTE. Values are number (%).

of education (odds ratio [OR]⫽2.14; 95% confidence interval [CI], 1.90 –2.42), be unmarried (OR⫽1.62; 95% CI, 1.43– 1.85), be male (OR⫽1.42; 95% CI, 1.24 –1.62), be unemployed (OR⫽1.51; 95% CI, 1.34 –1.70), and incur their TBIs as a result of an act of violence (OR⫽5.18; 95% CI, 4.41– 6.09) compared with whites. Furthermore, at the time of injury, minorities were 2.7 years (standard error [SE]⫽.50; 95% CI, 1.71–3.68) younger than whites and had greater DRS scores at discharge (difference, ⫺.30; SE⫽.12; 95% CI, ⫺.52 to ⫺.07) but did not have significantly different injury severity (GCS) scores (difference, ⫺.08; SE⫽.14; 95% CI, ⫺1.09 to 2.82) compared with whites. Employment status at follow-up is summarized according to preinjury employment status in table 2. For those who were employed preinjury, 55.52% of whites were unemployed at 1-year follow-up compared with 76.25% of minorities. Similarly, occupational status at follow-up is summarized according to preinjury occupational status in table 3. Of those who were employed in professional/managerial positions preinjury, 24.06% of whites and 15% of minorities were employed in skilled positions at 1 year postinjury and 1.07% of whites and 5% of minorities were employed in manual labor positions at 1 year postinjury. Of those who were employed in skilled positions preinjury, 7.76% of whites and 7.95% of minorities were employed in professional/managerial positions at 1 year postinjury and 18.35% of whites and 20.45% of minorities were employed in manual labor positions at 1 year postinjury. Of those who were employed in manual labor positions preinjury, 4.92% of whites and 1.96% of minorities were employed in professional/managerial positions at 1 year postinjury and 34.43% of whites and 35.29% of minorities were employed in skilled positions at 1 year postinjury. Next, several univariate simple logistic regression models were fit for each of the predictors, with employment status at 1 year as the outcome variable, to assess which of the covariates were associated with employment status and to consider for inclusion in the adjusted analysis. The results of these analyses are summarized in table 2. The simple logistic regression models for employment status indicate that DRS score at admission to rehabilitations and DRS at discharge, PTA, GCS score at admission to emergency department, cause of injury,

age, education level, sex, marital status, and preinjury employment status each have significant effects on employment status at 1 year postinjury when examined individually. Similarly, several univariate multinomial logistic regression models were fit for each of the predictors, with occupational status at 1 year as the outcome variable, to assess which of the covariates were associated with occupational status and to consider for inclusion in the adjusted analysis. The results of these analyses are summarized in table 4. The simple logistic regression models for occupational status indicate that age, sex, education level, marital status, cause of injury, occupational status preinjury, and DRS score at discharge each have significant effects on occupational status at 1 year postinjury when examined individually. Before fitting the final adjusted models, the minority group was examined more closely with respect to employment and occupational status outcomes at the 1-year follow-up. For the employment outcome, a multiple logistic regression model was fit that adjusted for the covariates DRS score at discharge, PTA, GCS score at admission to emergency room, cause of injury, age, education level, sex, marital status, and preinjury employment status. DRS score at admission was not included, because DRS at discharge is thought to be a more significant predictor of outcomes at the 1-year follow-up. This model also included an effect specifying minority status (black, Hispanic, Asian, Native American). This model indicated that minority status was not a significant predictor of employment at the 1-year follow-up among minority subjects (␹32⫽.24, P⫽.972). Thus, there is no evidence that employment at 1-year follow-up differs among the minority groups. Based on this result, a decision to collapse the 4 minority groups was made. With regard to the occupational status outcomes, there were only 3 Native Americans and 26 Asians. As a result, the minority ethnicity effect was unstable in the occupational status model. Based on blacks and Hispanics, there was not evidence of a significant difference in occupational status outcomes among the minority ethnicity groups (␹22⫽.34, P⫽.846). Thus, the final analyses compared whites with minorities (regardless of minority status).

Table 3: Preinjury Occupational Status Versus Follow-Up Occupational Status by Ethnicity One-Year Follow-Up Whites Preinjury

Prof/Mang Skilled Manual

Minorities

Prof/Mang

Skilled

Manual

Prof/Mang

Skilled

Manual

140 (74.87) 33 (7.76) 9 (4.92)

45 (24.06) 314 (73.88) 63 (34.43)

2 (1.07) 78 (18.35) 111 (60.66)

16 (80.00) 7 (7.95) 1 (1.96)

3 (15.00) 63 (71.59) 18 (35.29)

1 (5.00) 18 (20.45) 32 (62.75)

NOTE. Values are count (%). Abbreviations: Mang, managerial; Prof, professional.

Arch Phys Med Rehabil Vol 89, May 2008

992

RACE AND EMPLOYMENT OUTCOMES AFTER TBI, Arango-Lasprilla Table 4: Preliminary Univariate Analyses of Race and/or Ethnicity and Covariates Employment

Occupational Status

Variable

n

Wald ␹2

Unadjusted P

n

Wald ␹2

Unadjusted P

Race/ethnicity DRS score at admission DRS score at discharge GCS score at admission Age Education Sex Cause Marital status PTA Employment status preinjury Occupational status preinjury

3868 3793 3806 3027 3865 3773 3868 3833 3864 2766 3811 NA

116.39 267.62 251.38 12.07 63.97 89.14 10.69 41.05 14.88 65.57 294.59 NA

⬍.001 ⬍.001 ⬍.001 .005 ⬍.001 ⬍.001 .011 ⬍.001 .001 ⬍.001 ⬍.001 NA

1109 1096 1097 819 1109 1091 1109 1100 1108 918 NA 954

9.03 0.54 7.35 4.15 72.65 39.56 39.57 14.89 81.01 0.04 NA 396.39

.011 .764 .025 .126 ⬍.001 ⬍.001 ⬍.001 .006 ⬍.001 .980 NA ⬍.001

Abbreviation: NA, not applicable.

Primary Unadjusted Analysis for Employment Status A simple logistic regression model indicated that race and/or ethnicity was a significant predictor of employment status at 1 year postinjury without adjusting for any of the covariates (n⫽3868, Wald ␹12⫽116.39, P⬍.001). The unadjusted odds for minorities being unemployed at 1 year postinjury were 2.59 times greater than the unadjusted odds for whites being unemployed at 1 year postinjury (95% CI, 2.18 –3.07). Primary Adjusted Analysis for Employment Status A multiple logistic regression model was built to determine the effect of ethnicity on employment status at 1 year postinjury, after adjusting for covariates as previously described. The model initially included an effect for ethnicity and adjusted for the covariates DRS score at discharge, PTA, GCS score at admission to emergency department, cause of injury, age, education level, sex, marital status, and preinjury employment status. After adjustment, GCS score at admission was no longer significant (P⫽.636). The model was refit with the GCS removed, and all remaining effects (all P⬍.009) maintained significance. The sample size for this model, however, was 2645. Although the effect of PTA was significant (P⬍.001) there was a large degree of missing data attributed to this measure. It was noted that the sample size increased to 3636 with removal of the PTA effect. Thus a decision to remove this effect from the model was made to gain sample size. The assumptions for linearity in the logit of the continuous variables (age, DRS score at discharge) were assessed and found to be appropriate. Significant 2-way interactions were identified between age and sex and between age and preinjury employment status. The final model (table 5) included main effects for ethnicity, age, DRS score at discharge, marital status at admissions, sex, education level at admissions, cause of injury, and preinjury employment status, as well as 2-way interaction effects between age and sex and between age and preinjury employment status. A Hosmer and Lemeshow goodness-of-fit test indicated that the final model fits the data well (␹82⫽11.30, P⫽.185) with approximately 12 of 3636 (0.3%) subjects poorly accounted for by the model (Pearson residuals, ⬍⫺4.5; deviance residuals, ⬍⫺2.69). These 12 subjects were all unemployed at follow-up; 4 (33%) were minorities and 8 (67%) were white. After adjusting for covariates, there was a significant effect of ethnicity on employment status at 1 year postinjury (Wald ␹12⫽58.23, P⬍.001). The odds of minorities being unemployed Arch Phys Med Rehabil Vol 89, May 2008

at 1 year postinjury were 2.17 times greater than the odds of whites being unemployed (95% CI, 1.78 –2.65). Adding PTA back into the model increased the odds ratio (OR⫽2.36; 95% CI, 1.89 –2.96). There was evidence of significant effects of age, marital status, sex, education, preinjury employment status, DRS score at discharge, and cause of injury on employment status at 1 year postinjury; however, the effects of preinjury employment status and sex depend on age. Specifically, the odds of being unemployed versus the odds of being employed at 1 year postinjury were 1.57 times greater for unmarried versus married persons (95% CI, 1.28 –1.92), 1.99 times greater for those with less than a high school education versus those with at least a high school education (95% CI, 1.83–2.43), and 1.57 times greater for those with violent versus nonviolent injuries (95% CI, 1.18 –2.09). Increases in DRS discharge scores were also associated with increases in the odds of unemployment (unit OR⫽1.31; 95% CI, 1.27–1.36). The interaction effects between age and sex and age and preinjury employment status were such that (1) women and those previously unemployed were more likely to be unemployed at follow-up and (2) the odds of unemployment for women and those previously unemployed significantly increased with age. Primary Unadjusted Analyses for Occupational Status A multiple logistic regression model indicated that race and/or ethnicity was a significant predictor of occupational status at 1 year postinjury without adjusting for any of the covariates (n⫽1109, Wald ␹22⫽9.03, P⫽.011). The unadjusted odds of obtaining manual labor versus a professional/manageTable 5: Final Model for Unemployment (nⴝ3636) Variables

Wald ␹2

Adjusted P

Race/ethnicity (minority vs white) Preinjury employment status DRS score at discharge (unit increase) Age Marital status Sex Education Cause Age ⫻ employment status preinjury Age ⫻ sex

58.23 3.67 203.49 16.62 19.15 7.84 45.66 9.36 50.72 4.15

⬍.001 .056 ⬍.001 ⬍.001 ⬍.001 .005 ⬍.001 .002 ⬍.001 .042

RACE AND EMPLOYMENT OUTCOMES AFTER TBI, Arango-Lasprilla Table 6: Final Model for Occupational Status (nⴝ940) Variables

Wald ␹2

Adjusted P

Race/ethnicity Occupational status at admission Sex Marital status Education

0.58 327.77 16.41 9.78 11.09

.748 ⬍.001 .003 .008 .004

rial position 1 year postinjury were 2.108 times greater for minorities than for whites (95% CI, 1.29 –3.45). The unadjusted odds of obtaining a skilled versus a professional/managerial position 1 year postinjury were 1.55 times greater for minorities than for whites (95% CI, 0.98 –2.46). The odds of obtaining manual labor versus a skilled position 1 year postinjury were 1.36 times greater for minorities than for whites (95% CI, .96 –.19). Primary Adjusted Analysis for Occupational Status A multiple logistic regression model was built to determine the effect of ethnicity on occupational status at 1 year postinjury, after adjusting for covariates previously described. The model initially included an effect for ethnicity and adjusted for the covariates age, sex, education level, marital status, cause of injury, occupational status preinjury, and DRS score at discharge. After adjustment, age (P⫽.640), DRS score at discharge (P⫽.168), and cause of injury (P⫽.058) were no longer significant and removed from the model. The assumptions for linearity in the logit of the continuous variables were irrelevant because there were no longer any continuous variables remaining in the model. There were no significant 2-way interaction effects identified. The final model included main effects for ethnicity, marital status at admissions, sex, education level at admissions, and preinjury occupational status (table 6). There was not significant evidence of lack of fit; however, the model indicated that the proportional odds assumption was not met, hence separate logit slopes were necessary. After adjusting for covariates, there was not evidence that ethnicity was a significant predictor of occupational status 1 year postinjury (n⫽940, Wald ␹22⫽.58, P⫽.748). This suggests that the odds of obtaining manual labor versus a skilled or professional/managerial position and the odds of obtaining a skilled position versus a professional/managerial position 1 year postinjury are not significantly different for whites and minorities. There was evidence of significant effects of marital status, sex, education, and preinjury occupational status on occupational status at 1 year postinjury. Specifically, men were significantly more likely to take on manual labor occupations compared with professional/managerial positions (OR⫽2.88; 95% CI, 1.48 –5.63) and compared with skilled positions (OR⫽2.9; 95% CI, 1.73– 4.87) at 1 year postinjury. Those with less than a high school education were significantly more likely to take on manual labor compared with professional/managerial occupations (OR⫽7.08; 95% CI, 2.18 –23.04) 1 year postinjury and significantly more likely to take on skilled compared with professional/managerial occupations (OR⫽6.74; 95% CI, 2.17–20.96) 1 year postinjury. Those not married were significantly more likely to take on manual labor compared with professional/managerial occupations (OR⫽2.33; 95% CI, 1.36 –3.98) and significantly more likely to take on skilled compared with professional/managerial occupations (OR⫽1.79; 95% CI, 1.15–2.79). There was also highly significant evidence

993

that subjects were more likely to find similar occupations postinjury as they had preinjury. DISCUSSION The purpose of the present study was to examine racial differences in employment status 1 year after a TBI using a large multicenter database. The findings of the present study indicate that race and/or ethnicity has a significant effect on employment status 1 year postinjury (␹12⫽58.23, P⬍.001), after adjusting for employment status at admission, DRS score at discharge, age, marital status, sex, education level, and cause of injury. The odds of minorities being unemployed are 2.17 times (95% CI, 1.78 –2.65) greater than the odds of whites being unemployed. These findings of minorities being less likely than their white counterparts to be employed after TBI coincides with previous literature, despite a paucity of studies.7-10 Although the findings of the present study generally support the conclusions of previous studies examining TBI, race and/or ethnicity, and employment outcomes, there are several key differences in methodology. The first important distinction is the definition of employment outcomes in previous work. Although our results concur with those of Sherer et al,7 those researchers were interested in studying productivity, 1 aspect of which included employment. Ethnic minorities were approximately twice as likely to have worse productivity 1 year postinjury than nonminorities, even after statistically controlling for the effects of preinjury productivity, education, and violent versus nonviolent cause of injury. Productivity was defined as competitively employed, in school at least part-time, or a full-time homemaker. In the present study, employment was defined as competitive employment. Rosenthal et al8 found that, controlling for sociodemographic characteristics, minorities had lower CIQ productivity subscale scores compared with whites. The present study provides similar results regarding less productivity for minorities with a specific focus on objective paid employment outcomes. A study by Kreutzer et al10 examined predictors of job stability over the 4 years after TBI and found that nonminority group members were significantly more likely to be stably employed than minorities. Stable employment was operationalized as competitively employed at the year 1, year 2, and years 3 or 4 follow-ups. The sample is a second distinction between the methodology of the present study and prior investigations. As in the present study, many previous studies also used data from the national TBIMS database. The sample of the present study included 5259 subjects from all 16 TBIMS centers across the United States. The sample in Sherer7 was approximately 1000 TBI survivors from all TBIMS centers. Kreutzer10 initially began the study with 2682 TBIMS participants, but because of the longitudinal design, data from only 186 were used in the final analyses. The sample of Rosenthal8 was made up of 586 TBIMS participants from the 4 existing TBIMS centers at the time of the study. One strength of the present study is its large, representative sample, which may reduce variance and/or bias associated with local idiosyncrasies in referral patterns and standard of care. The second aim of the present study was to examine racial differences in occupational status at 1 year post-TBI. As expected, ethnic minorities had significantly greater rates of manual labor jobs and lower rates of professional jobs than whites 1 year postinjury without adjustment. However, race and/or ethnicity does not have a significant effect on occupational status at 1 year postinjury (␹22⫽.58, P⫽.748), after adjusting for occupational status at preinjury, sex, marital status, and education level. That is, minorities were not more likely than Arch Phys Med Rehabil Vol 89, May 2008

994

RACE AND EMPLOYMENT OUTCOMES AFTER TBI, Arango-Lasprilla

whites to find occupations in manual labor versus skilled or professional/managerial positions. Thus, it seems that although minorities are less likely to find employment 1 year postinjury, those who do are not more likely to obtain manual labor positions rather than skilled or professional/managerial positions. This is consistent with Rosenthal et al,8 who found that the rate of managerial/professional positions among minorities dropped significantly from preinjury to the 1-year follow-up (from 5%– 0%) and increased slightly among whites (from 18%–20%). However, our data show that race does not significantly influence occupational status after accounting for preinjury occupational status, sex, marital status, and education using a multivariate model. The results of this article provide strong evidence arguing for improvements in vocational planning, employer outreach, and patient counseling. For example, given the results of the present study, rehabilitation counselors will need to dedicate more time to job placements and employers allocate more dollars for additional support. As noted earlier, it does not appear that placement difficulties are due to the TBI itself but more likely the race of the TBI survivor, which is a larger overarching societal problem. Hence, these results should be used by rehabilitation counselors and others involved in program planning to promote career development and educational opportunities that help minorities with TBI achieve better longterm employment outcomes. These findings also raise other questions such as: What is the relationship between race, TBI, and long-term job retention and job stability? What is the relationship between race, TBI, and career advancement? Although answers to these questions are beyond the scope of the present study and the data available via the TBIMS database, it is very possible these long-term employment outcomes are also not likely to be favorable for minorities. There are numerous strategies that can be implemented to potentially circumvent these disparities in employment outcomes. First, increased worksite internships would allow employers to begin to connect more with survivors of TBI before offering a position. The greater amount of direct contact and communication employers and coworkers have with minority employees who have sustained a TBI, the greater the likelihood for hiring and long-term retention.14 It is also possible that minority survivors with TBI are experiencing a type of double discrimination regarding both their race and history of TBI. Second, there should be outreach to local chambers of commerce, some of which have Hispanic and black societies, to promote networking and job placement for minority individuals. Many of these groups are aware of employers who are interested in recruiting a diverse workforce. Third, engaging family, friends, and churches in the job-finding process early on permits use of employment resources in the tightly knit family and church networks in many Hispanic and black communities. Fourth, intensive training of rehabilitation counselors, therapists, and psychologists regarding the barriers that race may pose in work reentry allows discussions on how to overcome these barriers. This training needs to be practical and directly focused on the problem of race issues, not global and overly academic. Moreover, this training should emphasize building personal, community, and regional networks in which to use job marketing skills. To our knowledge, the present study is the first to examine the relationship between race and employment status (competitively employed vs not competitively employed) at 1 year post-TBI in a large, longitudinal sample of diverse subjects (whites, blacks, Hispanics, Asians, Native Americans) from across the 16 TBIMS centers. The results of the present study Arch Phys Med Rehabil Vol 89, May 2008

show that, even after adjusting for factors that may influence postinjury employment status such as employment status at admission, sex, DRS score at discharge, marital status, cause of injury, age, and education, minorities are 2.17 times more likely to be unemployed at 1 year postinjury than whites. Study Limitations The results of this study must be interpreted with caution because of the following limitations. First, confounding factors including concomitant medical disorders, medication usage, site-specific insurance limitations, postdischarge therapy and medical care, social support, and neurobehavioral problems are not extensively measured and therefore could not be controlled. Second, the decision to combine blacks, Hispanics, Asians, and Native Americans in a single minority group limited the ability of examining similarities and differences in employment outcomes at 1 year postinjury between individual minority ethnicities. Third, all TBI survivors in the present study received inpatient rehabilitation; thus the results may not generalize to the U.S. population at large of TBI survivors. In fact, all participants, regardless of ethnicity, received treatment from standardized TBIMS centers. It is possible that the average minority TBI survivor does not receive such state-of-the-art rehabilitation or top-quality care; thus one might expect even worse outcomes for this group compared with whites when examining employment status in the population as a whole. CONCLUSIONS Vocational outcomes post-TBI greatly affects survivors, their families, and society at large. There is an increased incidence of TBI in minority groups.15,16 Previous research has shown that, compared with whites, minorities with a TBI tend to receive less medical care17 and poorer quality care,18 report having significantly worse health and less social support,19 have worse functional outcomes in general after TBI,20,21 have lower levels of social functioning and higher rates of alcohol abuse after TBI,22 and have lower levels of social integration and community productivity at 1 year postinjury.8,20,21,23 Empirical evidence from the present study indicates that racial differences exist in competitive employment outcomes 1 year after TBI. This study contributes to the growing body of evidence of health care disparities affecting TBI survivors and highlights the importance of exploring the reasons underlying racial and/or ethnic minorities being more likely to be unemployed at 1 year postinjury. Once specific factors have been determined, priority should then be given to tailoring interventions to address these issues, thereby maximizing employment outcomes in all TBI survivors, regardless of race and/or ethnicity. As the evaluation of rehabilitation effectiveness continues to focus on functional outcomes and life participation, greater emphasis needs to be given to maximizing minority survivors’ work-related productivity. References 1. Langlois JA, Rutland-Brown W, Thomas KE. Traumatic brain injury in the United States: emergency department visits, hospitalizations, and deaths. Atlanta: Centers for Disease Control and Prevention, National Center for Injury Prevention and Control; 2006. 2. Rutland-Brown W, Langlois JA, Thomas KE, Xi YL. Incidence of traumatic brain injury in the United States, 2003. J Head Trauma Rehabil 2006;2:544-8. 3. Finkelstein E, Corso P, Miller T, et al. The incidence and economic burden of injuries in the United States. New York: Oxford Univ Pr; 2006.

RACE AND EMPLOYMENT OUTCOMES AFTER TBI, Arango-Lasprilla

4. O’Neill J, Hibbard MR, Brown M, et al. The effect of employment on quality of life and community integration after traumatic brain injury. J Head Trauma Rehabil 1998;13(4):68-79. 5. Abrams D, Barker LT, Haffey W, Nelson H. The economics of return to work for survivors of traumatic brain injury: vocational services are worth the investment. J Head Trauma Rehabil 1993; 8(4):59-76. 6. Wehman P, Targett P, West M, Kregel J. Productive work and employment for persons with traumatic brain injury: what have we learned after 20 years? J Head Trauma Rehabil 2005;20:115-27. 7. Sherer M, Nick TG, Sander AM, et al. Race and productivity outcome after traumatic brain injury: influence of confounding factors. J Head Trauma Rehabil 2003;18:408-24. 8. Rosenthal M, Dijkers M, Harrison-Felix C, et al. Impact of minority status on functional outcome and community integration after traumatic brain injury. J Head Trauma Rehabil 1996;11(4): 69-79. 9. Vanderploeg RD, Curtiss G, Duchnick JJ, Luis CA. Demographic, medical, and psychiatric factors in work and marital status after mild head injury. J Head Trauma Rehabil 2003;18:148-63. 10. Kreutzer J, Marwitz J, Walker W, et al. Moderating factors in return to work and job stability after traumatic brain injury. J Head Trauma Rehabil 2003;2:128-38. 11. Harrison-Felix C, Newton CN, Hall KM, Kreutzer JS. Descriptive findings from the Traumatic Brain Injury Model Systems National Database. J Head Trauma Rehabil 1996;11(5):1-14. 12. Rappaport M, Hall K, Hopkins K, Belleza T. Disability Rating Scale for severe head trauma: coma to community. Arch Phys Med Rehabil 1982;63:118-23. 13. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: John Wiley & Sons; 2000. 14. Inge KJ, Wehman P, Revell WG Jr, Brooke VA. Supported employment and workplace supports. In: Wehman P, Inge KJ, Revell WG Jr, Brooke VA, editors. Real work for real play. Baltimore: PH Brookes; 2007. p 117-38.

995

15. Jager TE, Weiss HB, Coben JH, Pepe PE. Traumatic brain injuries evaluated in U.S. emergency departments, 1992-1994. Acad Emerg Med 2000;7:134-40. 16. Frankowski RF, Annegers JF, Whitman S. The descriptive epidemiology of head injury in the United States. In: Becker DP, Povlishock JT, editors. Central nervous system trauma status report. Bethesda: NINCDS, National Institutes of Health; 1985. 17. Burnett DM, Kolakowsky-Hayner SA, Slater D, et al. Ethnographic analysis of traumatic brain injury patients in the national Model Systems database. Arch Phys Med Rehabil 2003; 84:263-7. 18. Bazarian JJ, Pope C, McClung J, Cheng YT, Flesher W. Ethnic and racial disparities in emergency department care for mild traumatic brain injury. Acad Emerg Med 2003;10:1209-17. 19. Brown SA, McCauley SR, Levin HS, Contant C, Boake C. Perception of health and quality of life in minorities after mild-tomoderate traumatic brain injury. Appl Neuropsychol 2004;11: 54-64. 20. Arango-Lasprilla JC, Rosenthal M, Deluca J, et al. Traumatic brain injury and functional outcomes: does minority status matter? Brain Inj 2007;21:701-8. 21. Arango-Lasprilla JC, Rosenthal M, Deluca J, Cifu DX, Hanks R, Komaroff E. Functional outcomes from inpatient rehabilitation after traumatic brain injury: how do Hispanics fare? Arch Phys Med Rehabil 2007;88:11-8. 22. Jorge RE, Robinson RG, Starkstein SE, Arndt SV. Influence of major depression on 1-year outcome in patients with traumatic brain injury. J Neurosurg 1994;81:726-33. 23. Hart T, Whyte J, Polansky M, Kersey-Matusiak G, FidlerSheppard R. Community outcomes following traumatic brain injury: impact of race and preinjury status. J Head Trauma Rehabil 2005;20:158-72. Supplier a. Version 9.1; SAS Institute, 100 SAS Campus Dr, Cary, NC 275132414.

Arch Phys Med Rehabil Vol 89, May 2008

View publication stats