Content uploaded by Mark Sherer
Author content
All content in this area was uploaded by Mark Sherer on Mar 04, 2016
Content may be subject to copyright.
ORIGINAL RESEARCH
Ten-Year Employment Patterns of Working Age Individuals
After Moderate to Severe Traumatic Brain Injury: A
National Institute on Disability and Rehabilitation
Research Traumatic Brain Injury Model Systems Study
Jeffrey P. Cuthbert, PhD, MPH, MS,
a
Christopher R. Pretz, PhD,
a,b
Tamara Bushnik, PhD,
c
Robert T. Fraser, PhD,
d
Tessa Hart, PhD,
e
Stephanie A. Kolakowsky-Hayner, PhD,
f
James F. Malec, PhD,
g
Therese M. O’Neil-Pirozzi, ScD, CCC-SLP,
h,i
Mark Sherer, PhD
j,k
From the
a
Rocky Mountain Regional Brain Injury System, Craig Hospital, Englewood, CO;
b
Traumatic Brain Injury Model Systems National
Statistical and Data Center, Englewood, CO;
c
Rusk Institute for Rehabilitation Medicine, New York University Langone School of Medicine, New
York, NY;
d
University of Washington, Seattle, WA;
e
Moss Rehabilitation Research Institute, Elkins Park, PA;
f
Santa Clara Valley Medical Center,
Rehabilitation Research Center, San Jose, CA;
g
Department of Physical Medicine and Rehabilitation, Indiana University, Indianapolis, IN;
h
Department of Physical Medicine and Rehabilitation, Spaulding Rehabilitation Hospital, Harvard Medical School, Boston, MA;
i
Department of
Speech-Language Pathology and Audiology, Northeastern University, Boston, MA;
j
The Institute for Rehabilitation and Research Memorial
Hermann, Houston, TX; and
k
Baylor College of Medicine, Houston, TX.
Abstract
Objective: To describe the 10-year patterns of employment for individuals of working age discharged from a Traumatic Brain Injury Model
Systems (TBIMS) center between 1989 and 2009.
Design: Secondary data analysis.
Setting: Inpatient rehabilitation centers.
Participants: Patients aged 16 to 55 years (NZ3618) who were not retired at injury, received inpatient rehabilitation at a TBIMS center, were
discharged alive between 1989 and 2009, and had at least 3 completed follow-up interviews at postinjury years 1, 2, 5, and 10.
Interventions: Not applicable.
Main Outcomes Measure: Employment.
Results: Patterns of employment were generated using a generalized linear mixed model, where these patterns were transformed into temporal
trajectories of probability of employment via random effects modeling. Covariates demonstrating significant relations to growth parameters that
govern the trajectory patterns were similar to those noted in previous cross-sectional research and included age, sex, race/ethnicity, education,
preinjury substance misuse, preinjury vocational status, and days of posttraumatic amnesia. The calendar year in which the injury occurred also
greatly influenced trajectories. An interactive tool was developed to provide visualization of all postemployment trajectories, with many showing
decreasing probabilities of employment between 5 and 10 years postinjury.
Conclusions: These results highlight that postinjury employment after moderate to severe traumatic brain injury (TBI) is a dynamic process, with
varied patterns of employment for individuals with specific characteristics. The overall decline in trajectories of probability of employment
An audio podcast accompanies this article. Listen at www.archives-pmr.org.
Supported by the Traumatic Brain Injury Model Systems (TBIMS) National Data and Statistical Center Grant from the National Institute on Disability and Rehabilitation Research (NIDRR) (grant no.
H133A110006); TBIMS Centers grants from NIDRR (grant no. H133A120032); The Institute for Rehabilitation and Research Memorial Hermann (grant no. H133A120020); Moss Rehabilitation Research
Institute (grant no. H133A120037); Spaulding Rehabilitation Hospital, Harvard Medical School (grant no. H133A120085); Rusk Rehabilitation at New York University School of Medicine (grant no.
H133A120100); Indiana University School of Medicine (grant no. H133A120035); and a TBIMS Follow-up Center subcontract via an NIDRR Prime Award (award no. H133A110006).
The TBIMS National Database is supported by the NIDRR and created and maintained by the TBIMS Centers Program.
This article is intended to promote the exchange of ideas among researchers and policymakers. The views expressed in it are part of ongoing research and analysis and do not necessarily reflect the
position of the U.S. Department of Education.
Disclosures: none.
0003-9993/15/$36 - see front matter ª 2015 by the American Congress of Rehabilitation Medicine
http://dx.doi.org/10.1016/j.apmr.2015.07.020
Archives of Physical Medicine and Rehabilitation
journal homepage: www.archives-pmr.org
Archives of Physical Medicine and Rehabilitation 2015;96:2128-36
between 5 and 10 years postinjury suggests that moderate to severe TBI may have unfavorable chronic effects and that employment outcome is
highly influenced by national labor market forces. Additional research targeting the underlying drivers of the decline between 5 and 10 years
postinjury is recommended, as are interventions that target influencing factors.
Archives of Physical Medicine and Rehabilitation 2015;96:2128-36
ª 2015 by the American Congress of Rehabilitation Medicine
The Bureau of Labor Statistics defines employment as any work
completed as a paid employee or work undertaken as part of a
personally or family owned business, profession, or farm.
1
For
individuals in the United States, particularly people between
the ages of 16 and 65 years, employment is a central indicator
of success. In addition to the monetary gains associated with
paid work, employment has numerous psychological benefits,
including greater happiness,
2
quality of life,
3
life satisfaction,
4
and
social well-being.
5
Conversely, unemployment and loss of
employment are linked to depression,
6,7
hopelessness,
8,9
and
anxiety.
7
Unemployment also has collateral effects, including
increased familial stress and decreased familial functioning.
10-12
Traumatic brain injury (TBI) can significantly impact
employment. People who incur TBI severe enough to require
acute care and inpatient rehabilitation have an increased likeli-
hood of physical, cognitive, emotional, behavio ral, social, and
functional problems after injury, substantially reducing their
ability to assume or resume purposeful work postinjury.
13,14
Previous estimates of unemployment rates for people with
moder ate to severe TBI have ranged wide ly given time post-
injury and severity.
15-17
Research targeting this population
demonstrates that postinjury unemployment has monetary and
psychological costs, diminishingtheabilitytoacquireincome
and reducing life satisfaction,
18
quality of life,
19
and psycho-
social adjustment.
20
Substantial research exists regarding employment and un-
employment post-TBI. Factors shown to be consistently asso-
ciated with postinjury employment include demographic
variables (age,
21-26
sex,
21,22,26,27
race,
21-23
marital status
21,22
),
preinjury status (occupation type,
21-23,28,29
substance misuse
16
),
injury characteristics (injury severity,
7,15,21-24,2 7-31
injury etiol-
ogy
23,29
), and postinjury functioning (pain,
32
neuropsychologi-
cal function, coping strategies
27,30-32
). Despite this expansive
knowledge base, few studies have attempted to examine
employment as a dynamic outcome. Rather, most available
research, including longitudinally focused efforts, report cross-
sectional population (mean/average) based results. Studies that
have attempted to assess employment as a dynamic construct
have relied on analyses of categorical surrogates for change (eg,
employed-employed-employed vs employed-unemployed-un-
employed).
9,33-36
Analysis of longitudinal dichotomous outcomes a t the indi-
vidual level is achieved by combining generalized linear mixed
modeling and random effects modeling.
37
Through this
approach, researchers are able to model probabilities of out -
comes over time (ie, trajectories), where the shape of a trajectory
is influenced by a variety of individual-level characteristics
represented by covariate s; however, 3 temporally spa ced
outcome measures are required to model trajectories.
38
Such an
approach offers a description of the time-dependent probability
of event data. It is the goal of this stu dy to improve under-
standing of how the probability of employment changes over
time as mediated by factors known to be important predictors of
employment status post-TBI. The product of these analyses will
provide an interactive tool with which clinicians may evaluate
possible postinjury employment scenarios for cli ents wi th TBI
and may be used to discuss postinjury employment with clients
and tailor employment- focused rehabilitative approaches for 10
years after injury.
Methods
Settings and participants
Data used in this study originated from the National Institute on
Disability and Rehabilitation Research’s Traumatic Brain Injury
Model Systems (TBIMS) National Database (NDB). For purposes
of the TBIMS NDB data collection, TBI is defined as damage to
brain tissue caused by an external mechanical force as evidenced
by medically documented loss of consciousness or posttraumatic
amnesia (PTA) or by objective neurologic findings on physical or
mental status examination that is attributed to the brain injury.
Data collection for the TBIMS involves retrospective review of
acute care information, prospective record review and interview-
ing during rehabilitation hospitalization, and follow-up inter-
viewing at 1, 2, and 5 years postinjury and every 5 years
thereafter.
39
Table 1 Temporal profile estimates from reduced random
intercept generalized linear mixed model
Covariate F P
Time 11.61 <.0001
Preinjury vocation 56.47 <.0001
Sex 42.43 <.0001
Age 269.24 <.0001
Race/ethnicity 54.60 <.0001
Preinjury drug/alcohol use 18.39 <.0001
Primary payer source 57.29 <.0001
Education 60.20 <.0001
Year of injury 75.23 <.0001
PTA 782.46 <.0001
Timeeducation 2.24 .0172
Timeage at injury 19.49 <.0001
Timepreinjury vocation 2.43 .0015
Timepreinjury drug/alcohol use 8.32 <.0001
TimePTA 7.75 <.0001
List of abbreviations:
NDB National Database
PTA posttraumatic amnesia
TBI traumatic brain injury
TBIMS Traumatic Brain Injury Model Systems
Post-TBI employment patterns 2129
www.archives-pmr.org
For this study, data selected from the TBIMS NDB included a
cohort of individuals enrolled between 1989 and 2009, aged 16 to
55 years, and who were not retired at the time of injury (nZ7204).
Based on common retirement practices in the United States, in-
dividuals who would have reached 65 years of age before the 10-
year follow-up were excluded. Cases missing baseline values
used as covariates (subsequently described) were excluded
(nZ2481), as were cases missing >2 assessments of employ-
ment status during the 1-, 2-, 5-, and 10-year follow-up in-
terviews (nZ1105). Based on these criteria, the study sample
included 3618 cases.
Primary outcome
The primary outcome for this study was employment. Employ-
ment included the report of any vocation that involved any
amount of employment for pay within the month before the
follow-up interview. Conversely, unemployment was any voca-
tion that did not involve paid work, including student,
homemaker, unemployed (any reason), retired (any reason),
hospitalized, and other.
Covariates
Covariates selected a priori included those found to demonstrate
relations with employment in previous cross-sectional analyses of
employment after TBI and those determined to be of theoretical
interest to the author group. Sociodemographic covariates included
age at injury (y), sex, race/ethnicity (white, black, Hispanic, other),
education (less than high school graduate, high school graduate,
greater than high school graduate but less than college graduate,
college graduate, greater than college graduate), primary rehabili-
tation payment source (private insurance, Medicare, Medicaid, self/
no payment, workers’ compensation, other), and preinjury voca-
tional status as defined by a combination of employment status and
vocation type (employed, professional; employed, skilled;
employed, manual; unemployed; student; volunteer/other).
40
Injury-
related covariates were days of PTA and calendar year in which the
injury occurred. PTAwas computed as days from hospital admission
to the first day that the participant achieved 2 consecutive scores of
76 on the Galveston Orientation and Amnesia Test,
41
11 on the
Galveston Orientation and Amnesia Test-Revised,
42
25 on the
Orientation Log, or 8 on the nonverbal Orientation Log,
43
signi-
fying that the individual was grossly oriented and was able to retain
new information day to day. Cases remaining in PTA at the time of
inpatient rehabilitation discharge and those missing PTA measure-
ments (nZ998) had their days of PTA imputed via expectation
maximization.
44
Preinjury substance misuse as defined by Corrigan
et al
45
(yes, no, unknown) was also included.
Analyses
All analyses were performed using SAS 9.4.
a
The overall analysis
strategy followed the process outlined in Pretz et al.
37
In the initial
step, a random intercept generalized linear mixed model was used
for the purpose of estimating a logit-based individual-level tem-
poral profile. Once profiles were generated (which are the
outcome data with which the study analyses are completed)
(table 1), profiles were assessed using random effects modeling or
individual growth curve analysis. Individual growth curve analysis
provided a detailed understanding of time-dependent outcomes at
the individual level. Within the individual growth curve analysis,
an unconditional model (a model free of covariates) that optimally
associated the outcome (estimated logits) with time was sought
(for details see Pretz et al
46
). To account for variability across the
profiles over time, covariates were subsequently introduced into
the modeling process where relations between covariates and
growth parameters were estimated.
Table 2 Characteristic comparisons of the study sample with all
excluded cases enrolled between 1989 and 2009, aged 16 to 55
years, and not retired at injury
Variables
Study
Sample
(NZ3618)
Excluded
Cases
(nZ3586)
Sex
Female 27 21
Male 73 79
Race/ethnicity
White 72 58
Black 18 26
Hispanic 7 12
Other 3 5
Education
<High school 26 36
High school graduate 34 34
>High school but <college
graduate
27 21
College graduate 13 9
Preinjury vocational status
Employed, professional/managerial 13 9
Employed, skilled 43 36
Employed, manual labor 20 21
Unemployed 12 21
Student 10 8
Volunteer and other 2 3
Primary rehabilitation payment source
Private insurance 57 47
Medicare 2 2
Medicaid 25 30
Workers’ compensation and other 8 7
Self or no pay 8 13
Preinjury substance abuse
No 47 29
Yes 44 43
Unknown 9 28
Age at injury (y) 3111 3111
Days of PTA 3024 3125
NOTE. Values are mean SD or percentages.
Table 3 Growth parameter estimates of the unconditional model
(NZ3618)
Growth
Parameter Estimate P
95% Confidence
Interval
Intercept 1.4681 <.0001 1.5720 to 1.3642
Linear term
(time)
0.3373 <.0001 0.3263 to 0.3484
Quadratic term
(timetime)
0.03145 <.0001 0.03222 to 0.03067
2130 J.P. Cuthbert et al
www.archives-pmr.org
Results
Table 2 provides comparisons of the analyzed study cohort
(NZ3618) and cases excluded for lacking 2 valid follow-up as-
sessments or for missing covariate data (nZ3586). Compared
with the excluded cases, the study sample had a higher percentage
of individuals who reported being of white race (72% vs 58%,
respectively) and paid for inpatient rehabilitation with private
insurance (57% vs 47%, respectively) and had a lower percentage
of people reporting black race (18% vs 26%, respectively), skilled
preinjury vocations, and less than high school education (26% vs
36%, respectively).
Table 3 contains estimates for the intercept, linear term, and
quadratic term (ie, growth parameters). In addition to the indi-
vidual profiles, figure 1 provides the trajectory generated by the
estimates in the unconditional model.
Evaluation of Akaike information criteria values indicated
that a quadratic model best related outcome to time for the
unconditional model. From inspection of figure 1, the trajec-
tory for the unconditional model (white curve) did well to
represent the general sample trend, but it did not account for
variability across individuals. Covariates that demonstrat ed
significant associations to the growth parameters were sex,
age at injury, race, preinjury substance use, preinjury voca-
tion, primary payment source, education, year of injury, and
PTA. Estimated relations between growth parameters and
covariates are presented in table 4 and define the conditional
model (a model that contains covariates). The narrow width of
the 95% confidence intervals about the growth parameters
attests to the precision of the estimates. Ultimately, the re-
lations between the growth parameters and covariates play a
Fig 1 Individual profiles based on logits and unconditional model trajectory.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
12345678910
Probability of Employment
Years
Individual Level Trajectory (Probability)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
12345678910
Probability of Employment
Years
Individual Level Trajectory (Probability)
A
B
Fig 2 Individual trajectories of (A) case 1 and (B) case 2.
Post-TBI employment patterns 2131
www.archives-pmr.org
Table 4 Estimates and confidence intervals regarding the relation between tested growth parameters and covariates (NZ3168)
Parameter Estimate* 95% Confidence Interval
Intercept 2.7776 2.9866 to 2.5685
Continuous variables
Time 0.6847 0.6589 to 0.7106
Age at injury 0.01921 0.02467 to 0.01374
Year of injury 0.1788 0.1979 to0.1597
Days in PTA 0.07496 0.07728 to 0.07263
Categorical variables
Preinjury vocation
Unemployed Reference
Manual Labor 2.1212 1.9117 to 2.3307
Professional/managerial 3.2819 3.0227 to 3.5411
Skilled 2.3290 2.1387 to 2.5193
Student 0.3771 0.1209 to 0.6332
Volunteer/other 1.4474 1.8534 to 1.0415
Sex
Male Reference
Female 0.6272 0.7573 to 0.4971
Race/ethnicity
White Reference
Black 1.4448 1.5959 to 1.2937
Hispanic 0.01543 0.2112 to 0.2420
Other 0.5982 0.8983 to 0.2980
Preinjury substance misuse
Yes Reference
No 0.3195 0.1981 to 0.4409
Unknown 2.1428 1.9305 to 2.3552
Primary rehabilitation payer source
Private insurance Reference
Medicaid 1.3100 1.4538 to 1.1662
Medicare 1.8054
2.3208 to 1.2900
Self or no pay 0.8348 1.0458 to 0.6238
Workers’ compensation and other 1.2336 1.4536 to 1.0137
Education
High school graduate or equivalent Reference
<High school graduate 0.6895 0.8397 to 0.5392
>High school graduate but <college graduate 0.2599 0.1151 to 0.4046
College graduate 0.4842 0.2692 to 0.6991
Interacted variables
Timepreinjury vocation (manual labor) 0.1638 0.1923 to 0.1353
Timepreinjury vocation (professional/managerial) 0.4173 0.4530 to 0.3815
Timepreinjury vocation (skilled) 0.2215 0.2471 to 0.1960
Timepreinjury vocation (student) 0.02319 0.01061 to 0.05698
Timepreinjury vocation (volunteer/other) 0.2664 0.2136 to 0.3193
Timepreinjury vocation (unemployed) 0 NA
Timeage at injury 0.02151 0.02227 to 0.02075
Timepreinjury drug/alcohol use (no) 0.1286 0.1451 to 0.1121
Timepreinjury drug/alcohol use (unknown) 0.5173 0.5423 to 0.4923
Timepreinjury drug/alcohol use (yes) 0 NA
Timepreinjury education (college graduate) 0.4059 0.3763 to 0.4355
Timepreinjury education (>high school but <college) 0.1081 0.08815 to 0.1280
Timepreinjury education (<high school) 0.1324 0.1526 to 0.1122
Timepreinjury education (high school graduate or equivalent) 0 NA
Timeyear of injury 0.03113 0.02813 to 0.03412
Timedays in PTA 0.008086 0.007754 to 0.008418
Timetime 0.06065 0.07076 to 0.05053
Timetimeage at injury 0.001776 0.001312 to 0.002241
Timetimepreinjury drug/alcohol use (no) 0.01733 0.006692 to 0.02798
(continued on next page)
2132 J.P. Cuthbert et al
www.archives-pmr.org
major role in obtaining an individual-level interpretation of
the data.
Estimates are given in terms of logits; however, subsequently,
trajectories are transformed into probabilities to enhance inter-
pretability. In addition to reporting the covariate/growth parameter
relations, it is also important to account for the relations between
growth parameters as displayed in table 5 because these relations
also contribute to the individual-level interpretation of the data.
Although this method presents countless individual-level re-
sults, presentation of the patterns associated with all combinations
of covariates is beyond the scope of this publication. Readers
interested in the contribution of specific covariates or combina-
tions of covariates are directed to use an online interactive tool,
developed as part of this research to consolidate and condense
information from both tables 4 and 5 into a visual format depicting
individual-level change. Specifically, this tool generates
individual-level trajectories based on both logits and probability of
employment for specified covariate values, providing a vast
amount of information regarding individual-level change. The
interactive tool can be retrieved via the following link: https://
www.tbindsc.org/Researchers.aspx. To provide an understanding
of the iterative tool’s capability, the following cases highlight 2
trajectories (figs 2A and B).
Case 1
The first case provides an employment trajectory for an individual
who was unemployed before injury, is a white man whose age at
injury is 30 years old, has a payer source of self- or no pay, has a
high school education, has a history of preinjury alcohol misuse,
was injured in 1998, and was in PTA for 20 days.
Case 2
The second case provides an employment trajectory for an indi-
vidual who assumed a managerial position before injury, is a black
woman whose age at injury was 25 years old, has no history of
preinjury substance misuse, has private insurance, is a college
graduate, was injured in 2000, and was in PTA for 5 days.
Discussion
This research represents a unique analysis of employment for
individuals with moderate to severe TBI as a complex dynamic
construct. Results of the study demonstrate the fluidity of
employment over time, with extensive individual variation in the
likelihood of both becoming employed postinjury and continuing
employment in the years after injury.
Covariates demonstrating a significant relation with employ-
ment were consistent with previous research. Age, sex, race/
ethnicity, education, primary rehabilitation payer source, days of
PTA, and preinjury substance misuse were all found to signifi-
cantly predict the trajectories of postinjury employment. Although
both previous and current analyses demonstrated statistical sig-
nificance, comparisons of these new results with previous results
are difficult given the variability of the dynamic trajectories and
the need to consider all covariates simultaneously as opposed to
singularly. In general, however, these findings fit well with
existing literature. Older age, being a woman, nonwhite race,
educational attainment below high school graduation, preinjury
substance misuse, use of government payer sources, and increased
days of PTA all negatively influence the trajectory of probability
of employment. Conversely, all types of preinjury employment
improved the trajectory of probability of postinjury employment
compared with nonemployment-based vocations, as did levels of
education beyond high school graduation compared with lower
levels of education.
One new variable, year of injury, also significantly predicted
employment patterns. Of all the included covariates, year of injury
appeared to have the greatest influence on employment probability
trajectories, with individuals injured at earlier years showing
greater probability of postinjury employment at almost all follow-
up years. Some of this variation may be caused by the small
number of cases analyzed that were enrolled during the early years
of the TBIMS program (1989e1999 average annual enrollment,
29; 2000e2009 average annual enrollment, 329). Overall labor
forces also likely influenced these results, with cases having injury
years of 1995 and later subject to the national recession that
occurred between 2001 and 2003
47
and the later global recession
between 2007 and 2009.
48
In addition to overall increased un-
employment in the United States during these periods, individuals
eligible for disability benefits who had previously selected to seek
Table 4 (continued )
Parameter Estimate* 95% Confidence Interval
Timetimepreinjury drug/alcohol use (unknown) 0.03005 0.01094 to 0.04916
Timetimepreinjury drug/alcohol use (yes) 0 NA
Timetimepreinjury education (college graduate) 0.02847 0.04549 to 0.01145
TimeTimepreinjury education (>high school but <college) 0.00066 0.01354 to 0.01221
Timetimepreinjury education (<high school) 0.01518 0.001903 to 0.02845
Timetimepreinjury education (high school graduate or equivalent) 0 NA
Timetimeyear of injury 0.00484 0.00656 to 0.00311
Timetimedays in PTA 0.00068 0.00089 to 0.00047
Abbreviation: NA, not applicable.
* Estimates are in logits.
Table 5 Covariance between growth parameters
Growth Parameters Estimate P
Intercept/linear term 0.03604 <.0001
Linear term/quadratic term 0.001185 <.0001
Quadratic term/linear term 0.00356 <.0001
Post-TBI employment patterns 2133
www.archives-pmr.org
gainful employment may have chosen to apply for these benefits
because of reduced job availability.
49
Furthermore, as years have
advanced, so has the demand for high skill level technologically
aware employees, with a parallel reduction of less skilled posi-
tions, which may have reduced the overall employment opportu-
nities for individuals with residual deficits associated with TBI.
50
The interactive tool developed as part of this project provides
users access to information regarding probability of employment
over time at a level of detail that until now was unavailable. Re-
searchers and clinicians alike are well served to use this device to
better understand not only which factors influence the probability of
employment postinjury over time, but also how these factors in-
fluence the probability of employment over time. Furthermore, use
of this tool allows for the investigation of specific subpopulations of
individuals with specific characteristics and provides insights into
the patients and subpopulations that are in greatest need of post-
injury employment-related therapies and the times at which these
interventions may be most needed. Although it is impossible to
discuss all potential results, exploration using the interactive tool
revealed numerous trends. For instance, individuals with preinjury
vocations involving professional/managerial positions had the
highest probabilities of early postinjury employment (years 1e5)
when compared with other vocations; however, those with the
highest level of later postinjury employment (years 6e10) reported
preinjury vocations of manual labor, skilled professions, or students
and volunteers. Although this result may be driven by age (those
most likely to be professionals and managers at injury are likely
older than students), it may also represent an inability of those who
incur moderate to severe TBI to keep pace with the increased
complexity and technologic demands of the workplace. The trends
for probability of employment based on payment source appear to
be similar across nonprivate insurance payers.
One concerning trend that appears consistently regardless of
the combinations entered into the interactive tool is the reduction
in the probability of employment for all individuals that occurs
between 5 and 10 years postinjury. These trajectories suggest that
in the early years postinjury, individuals who experience TBI
demonstrate improved recovery and community reintegration.
However, this recovery appears to peak by 5 years postinjury, at
which time these individuals appear to regress, and in some
instance even experience worse outcomes than immediately after
postinjury. Because this study included only individuals who
would not be expected to retire within 10 years postinjury based
on age-related retirement, it is not expected that this trend is
caused by removal from the workforce because of natural aging;
rather, this phenomenon represents an interactive effect between
TBI and aging. One explanation for this decline is that moderate to
severe TBI can have negative effects that are both acute and
chronic. Recent shifts in the field of TBI rehabilitation have
emphasized the reconceptualization of TBI care from a single
medical event requiring intensive focus to one of long-term
management.
17,51
The aforementioned labor market forces and
effects of technologic advancement on the labor market may also
explain this decline. Additional research that further explores all
of the underlying reasons for this decline is recommended.
Study limitations
A number of limitations should be considered when using the
iterative tool in interpreting the results of this study. First, tra-
jectories modeled are estimates based on participants that meet
previously established inclusion/exclusion criteria as dictated by
both entry into the TBIMS and the analytic approach. Although
this database has been shown to largely be representative of all
individuals in the United States who receive inpatient rehabilita-
tion for moderate or severe TBI, the selection criteria applied here
may have resulted in findings that are not fully representative of
this entire population. Further, those in the TBIMS database
with <3 follow-up time points, with missing covariate data, who
died before 3 measures could be collected, and more recent in-
juries did not contribute to these models. As a result, use of the
interactive tool should be executed with caution because results
describe only those who have enough temporal information to be
included in an individual growth curve analysis approachda
limitation not unique to this study, but one of any study that le-
verages this type of longitudinal analytic approach. Consequently,
these results are not meant to apply to the entire population of in-
dividuals who incur TBI in the United States; rather, the focus of this
study was to comprehensively describe participants in the TBIMS
NDB that met specified limitations of the analytic approach.
Additionally, because of the observational nature of the data, results
of this study should not be used for inferential purposes. Because of
the descriptive focus of the study, comparison between trajectories
should be made for the purpose of clinical investigation (ie, based on
clinical relevancy) and not to predetermine a specific individual’s
likelihood for returning to work. The trajectories described and
provided within the interactive tool are mathematical projections
based on relations between the available identified covariates and
growth parameters; however, these parameters are not comprehen-
sive. Additional factors unavailable within the TBIMS collection
system that relate to employment, including concomitant medical
conditions, family support, resilience, and motivation, may have
greatly influenced the results.
Caution should be exercised when extrapolating beyond the
range of data (ie, entering sets of covariate values that our unlikely
to represent a real individual). As an example, someone who is 18
years of age likely does not have a college degree. Additionally,
extrapolation of these data beyond the time at which they were
analyzed will lead to spurious results (ie, examining 10-year data
of an individual injured in 2005). Readers should be mindful that
the analyses here require a double estimation process that includes
the following: an estimation of a set of temporal logits per indi-
vidual using a generalized linear mixed model, and a description
of the patterns of these logits via individual growth curve analysis.
Such an estimation process introduces additional error into the
modeling; however, error remains relatively small because of the
large sample size. Finally, the transformation from logits to
probabilities requires a transformation from an infinite scale
(logits) to a bounded scale (0e1 for probabilities). Therefore,
trajectories on the logit scale will not directly mirror trajectories
transformed into probabilities.
Conclusions
This study presents an analysis of postinjury employment as a
dynamic construct for working aged individuals who incurred
moderate to severe TBI, demonstrating temporal changes associ-
ated with individual-level factors. An interactive tool has been
developed from these results allowing for investigation in the
change of the probability of employment of an individual using an
extensive set of covariates. This tool provides an extremely rich
2134 J.P. Cuthbert et al
www.archives-pmr.org
level of detail, from which researchers and clinicians can address
questions by adjusting the combinations of covariates. Regardless of
the covariate patterns entered into this tool, the declining trajectory
of employment between 5 and 10 years postinjury is concerning and
may be related to chronic effects of TBI magnified by aging labor
market forces during the decades included in the research, or both.
Research into factors responsible for this decline and interventions
to target these long-term outcomes is recommended.
Supplier
a. SAS 9.4; SAS Institute.
Keywords
Brain injuries; Employment; Rehabilitation
Corresponding author
Jeffrey P. Cuthbert, PhD, MPH, MS, Craig Hospital, 3425 S
Clarkson St, Englewood, CO 80132. E-mail address: jcuthbert@
craighospital.org.
References
1. U.S. Department of Labor, Bureau of Labor Statistics. Glossary
2014. Available at: http://www .bls.gov/bls/glossary .htm. Accessed
September 1, 2014.
2. Argyle M. Causes and correlates of happiness. In: Diener E,
Schwarz N, editors. Well-being: the foundations of hedonic psy-
chology. New York: The Russell Sage Foundation; 1999.
3. Carlier BE, Schuring M, Lo
¨
tters FJ, Bakker B, Borgers N, Burdorf A.
The influence of re-employment on quality of life and self-rated
health, a longitudinal study among unemployed persons in the
Netherlands. BMC Public Health 2013;13:503.
4. Lucas RE, Clark AE, Georgellis Y, Diener E. Unemployment alters
the set point for life satisfaction. Psychol Sci 2004;15:8-13.
5. Bornstein MH, Davidson L, Keyes CL, Moore KA. Well-being:
positive development across the life course. Mahwah: Lawrence
Erlbaum Associates; 2003.
6. McKee-Ryan F, Song Z, Wanberg CR, Kinicki AJ. Psychological and
physical well-being during unemployment: a meta-analytic study. J
Appl Psychol 2005;90:53-76.
7. Montgomery SM, Cook DG, Bartley MJ, Wadsworth M. Unem-
ployment pre-dates symptoms of depression and anxiety resulting in
medical consultation in young men. Int J Epidemiol 1999;28:95-100.
8. Soares JJ, Macassa G, Grossi G, Viitasara E. Psychosocial correlates
of hopelessness among men. Cogn Behav Ther 2008;37:50-61.
9. Winefield AH, Tiggemann M, Winefield HR. The psychological
impact of unemployment and unsatisfactory employment in young
men and women: longitudinal and cross-sectional data. Br J Psychol
1991;82:473-86.
10. Walsh F. Family resilience: a collaborative approach in response to
stressful life challenges. In: Southwick SM, Litz BT, Charney D,
Friedman MJ, editors. Resilience and mental health: challenges
across the lifespan. New York: Cambridge Univ Pr; 2011. p 149-61.
11. Westman M, Etzion D, Horovitz S. The toll of unemployment does
not stop with the unemployed. Hum Relat 2004;57:823-44.
12. Broman CL, Hamilton VL, Hoffman WS. Unemployment and its
effects on families: evidence from a plant closing study. Am J
Community Psychol 1990;18:643-59.
13. National Center for Injury Prevention and Control, Centers for Dis-
ease Control and Prevention. 2013. What are the potential effects of
TBI? Available at: http://www.cdc.gov/traumaticbraininjury/
outcomes.html. Accessed December 1, 2013.
14. National Institute of Neurological Disorders and Stroke, National
Institutes of Health. 2013. NINDS traumatic brain injury information
page. Available at: http://www.ninds.nih.gov/disorders/tbi/tbi.htm#What_
is_the_prognosis. Accessed September 1, 2014.
15. Keyser-Marcus LA, Bricout JC, Wehman P, et al. Acute predictors of
return to employment after traumatic brain injury: a longitudinal
follow up. Arch Phys Med Rehabil 2002;83:635-41.
16. Wagner A, Hammond F, Sasser H, Wiercisewski D. Return to pro-
ductive activity after traumatic brain injury: relationship with mea-
sures of disability, handicap, and community integration. Arch Phys
Med Rehabil 2002;83:107-14.
17. Sander A, Krentzer J, Rosenthal M, Delmonico R, Young M. A
multicenter longitudinal investigation of return to work and com-
munity integration following traumatic brain injury. Arch Phys Med
Rehabil 1996;11:70-84.
18. Corrigan JD, Bogner JA, Mysiw WJ, Clinchot D, Fugate L. Life
satisfaction after traumatic brain injury. J Head Trauma Rehabil
2001;16:546-55.
19. Tsaousides T, Warshowsky A, Ashma TA, Cantor JB, Spielman L,
Gordon WA. The relationship bewteen employment-related self-ef-
ficacy and quality of life following traumatic brain injury. Rehabil
Psychol 2009;54:299-305.
20. Franulic A, Carbonell CG, Pinto P, Sepulveda I. Psychosocial
adjustment and employment outcome 2, 5 and 10 years after TBI.
Brain Inj 2004;18:119-29
.
21. Corrigan JD, Lineberry LA, Komaroff E, Langlois JA, Selassie AW,
Wood KD. Employment after traumatic brain injury: differences
between men and women. Arch Phys Med Rehabil 2007;88:1400-9.
22. Gary KW, Arango-Lasprilla JC, Ketchum JM, et al. Racial differ-
ences in employment outcome after traumatic brain injury at 1, 2 and
5 years postinjury. Arch Phys Med Rehabil 2009;90:1699-707.
23. Gary KW, Ketchum JM, Arango-Lasprilla JC, et al. Differences in
employment outcomes 10 years after traumatic brain injury among
racial and ethnic minority groups. J Vocat Rehabil 2010;33:65-75.
24. Grauwmeijer E, Heijenbrok-Kal MH, Haitsma IK, Ribbers GM. A
prospective study on employment outcome 3 years after moderate
to severe traumatic brain injury. Arch Phys Med Rehabil 2012;93:
993-9.
25. Shigaki CL, Johnstone B, Schopp LH. Financial and vocational
outcomes 2 years after traumatic brain injury. Disabil Rehabil 2009;
31:484-9.
26. Cuthbert JP, Harrison-Felix C, Corrigan JD, Bell JM, Haarbauer-
Krupa JK, Miller AC. Unemployment in the United States after
traumatic brain injury for working-age individuals: prevalence and
associated factors 2 years postinjury. J Head Trauma Rehabil 2015;
30:160-74.
27. Fraser R, Machamer J, Temkin N, Dikmen S, Doctor J. Return to
work in traumatic brain injury (TBI): a perspective on capacity for
job complexity. J Vocat Rehabil 2006;25:141-8.
28. Andelic N, Stevens LF, Sigurdardottir S, Arango-Lasprilla JC,
Roe C. Assocations between disability and employment 1 year after
traumatic brain injury in a working age population. Brain Inj 2013;
26:261-9.
29. Ketchum JM, Getachew MA, Krch D, et al. Early predictors of
employment outcomes 1 year post traumatic brain injury in a pop-
ulation of hispanic individuals. NeuroRehabilitation 2012;30:13-22.
30. Forslund MV, Roe C, Arango-Lasprilla JC, Sigurdardottir S,
Andelic N. Impact of personal and environmental factors on
employment ouctome two years after moderate-to-severe traumatic
brain injury. J Rehabil Med 2013;45:801-7.
Post-TBI employment patterns 2135
www.archives-pmr.org
31. Wehman P, Targett P, West M, Kregel J. Productive work and
employment for persons with traumatic brain injury: what have
have we leared after 20 years? J Head Trauma Rehabil 2005;20:
115-27.
32. Dawson DR, Schwartz ML, Winocur G, Stuss DT. Return to pro-
ductivity following traumatic brain injury: cognitive, psychological,
physical, spiritual, and environmental correlates. Disabil Rehabil
2007;29:301-13.
33. Forslund MV, Arango-Lasprilla JC, Roe C, Perrin P, Sigurdardottir S,
Andelic N. Multi-level modelling of employment probability tra-
jectories and employment stability at 1, 2 and 5 years after traumatic
brain injury. Brain Inj 2014;28:980-6.
34. Ponsford J, Spitz G. Stability of employment over the first 3 years
following traumatic brain injury. J Head Trauma Rehabil 2015;30:
E1-11.
35. Kreutzer JS, Marwitz JH, Walker WC, et al. Moderating factors in
return to work and job stability after traumatic brain injury. J Head
Trauma Rehabil 2003;18:128-38.
36. Machamer J, Temkin N, Fraser R, Doctor J, Dikmen S. Stability of
employment after traumatic brain injury. J Int Neuropsychol Soc
2005;11:807-16.
37. Pretz C, Ketchum J, Cuthbert J. An introduction to analyzing dichotomous
outcomes in a longitudinal setting: a NIDRR Traumatic Brain Injury
Model Systems communication. J Head Trauma Rehabil 2014;29:E65-71.
38. Kozlowski A, P retz C, Dams-O’Connor K, Kreider S,
Whiteneck GG. Applying individual growth curve models to
evaluate change in rehabilitation. Arch Ph ys Med Rehabil 2013;
94:589-96.
39. The Traumatic Brain Injury Model Systems National Data and Sta-
tistical Center. TBI Model Systems presentation 2014. Available at:
https://www.tbindsc.org/. Accessed September 1, 2014.
40. Walker WC, Marwitz JH, Kreutzer JS, Hart T, Novack TA. Occu-
pational categories and return to work after traumatic brain injury: a
multicenter study. Arch Phys Med Rehabil 2006;87:1576-82.
41. Levin HS, O’Donnell VM, Grossman RG. The Galveston Orientation
and Amnesia Test. A practical scale to assess cognition after head
injury. J Nerv Ment Dis 1979;167:675-84.
42. Bode RK, Heinemann AW, Semik P. Measurement properties of the
Galveston Orientation and Amnesia Test (GOAT) and improvement
patterns during inpatient rehabilitation. J Head Trauma Rehabil 2000;
15:637-55.
43. Jackson WT, Novack TA, Dowler RN. Effective serial measurement
of cognitive orientation in rehabilitation: the Orientation Log. Arch
Phys Med Rehabil 1998;79:718-20.
44. Dempster AP, Laird NM, Rubin DB. Maximum likelihood from
incomplete data via the EM algorithm. J R Stat Soc 1977;39:1-38.
45. Corrigan JD, Bogner J, Lamb-Hart G. Technical report on prob-
lematic substance use variables. The Center for Outcome Measure-
ment in Brain Injury; 2003. Available at: http://www.tbims.org/
combi/subst. Accessed September 1, 2014.
46. Pretz C, Kozlowski A, Dams-O’Connor K, et al. Descriptive
modeling of longitudinal outcome measures in traumatic brain
injury: a National Institute on Disability and Rehabilitation Research
Traumatic Brain Injury Model Systems study. Arch Phys Med
Rehabil 2013;94:579-88.
47. Langdon D, McMenamin T, Krolik T. US labor market in 2001:
economy enters a recession. Mon Labor Rev 2002;125:3-33.
48. U.S. Department of Labor , Bureau of Labor Statistics. The recession of
2007-2009. 2012. A v ailable at: http://www .bls.go v/spotlight/2012/recession/.
Accessed September 1, 2014.
49. Thompkins A, Honeycutt T, Gill C, Mastrianni J, Bailey M. To apply
or not apply: the employment and program participation of social
security disability insurance applicants and non-applicants. Prince-
ton: Mathematica Policy Research; 2014.
50. Rotman D. How technology is destroying jobs. Cambridge: MIT
Technology Review; 2013.
51.
Masel BE, DeWitt DS. Traumatic brain injury: a disease process, not
an event. J Neurotrauma 2010;27:1529-40.
2136 J.P. Cuthbert et al
www.archives-pmr.org