Employment Interventions for Return to Work in Working Aged Adults Following Traumatic Brain Injury (TBI): A Systematic Review

1.1.1 Incidence and Causes of Traumatic Brain Injury

Traumatic brain injury (TBI) occurs when an external force causes an alteration to brain function (e.g., loss of consciousness, memory loss, and neurologic defects). A TBI can occur from a blow to the head, blast waves from an explosion, swift acceleration or deceleration, or the penetration of a foreign object into the brain (Maas, Stocchetti, & Bullock, 2008; Menon, Schwab, Wright, & Maas, 2010). TBI is a major concern worldwide and affects 10 million people annually (Hyder, Wunderlich, Puvanachandra, Gururaj, & Kobsingye, 2007). The two most common worldwide causes of TBI are traffic accidents and falls, with traffic accidents declining in predominance and falls rising due to aging populations (Bražinova et al., 2015; Li et al., 2015; Majdan, Rusnák, Bražinová, & Mauritz, 2015; Peeters et al., 2015; Popescu, Anghelescu, Daia, & Onose, 2015; Scholten, Haagsma, Panneman, Beeck, & Polinder, 2014; Shekhar, Gupta, Premsagar, Sinha, & Kishore, 2015; Shivaji, Lee, Dougall, McMillan, & Stark, 2014).

1.1.2 Severity and Sequelae

Levels of TBI range from mild to severe and are determined by measures such as duration of coma or post-traumatic amnesia (PTA), Glasgow Coma Scale (GCS) scores, and the nature and extent of functional impairments following the injury. Individuals with TBI experience various combinations of physical, cognitive, sensory, and emotional symptoms, which are largely determined by the severity of the impact, the location of the impact, and whether the TBI is a penetrating (open) or non-penetrating (closed) injury (BrainandSpinalCord.org, 2013).

Brain injury severity is typically related to the degree and extent of impairment post-injury. Even individuals experiencing mild brain injuries can experience problems in one or more areas of functioning, though most individuals with a mild TBI will recover lost functioning. In most cases of moderate and severe TBI, individuals' recovery will plateau below pre-injury functioning levels.

1.1.3 Return to Work Following TBI

Individuals with TBI often struggle to either obtain or maintain competitive employment after sustaining a TBI, commonly as a result of the post-injury difficulties they exhibit (Andelic, Stevens, Sigurdardottir, Arango-Lasprilla, & Roe, 2009; Dikmen et al., 1994; Ezrachi, Ben-Yishay, Diller, & Rattock, 1991; Mansfield et al., 2015; Ruttan, Martin, Liu, Colella, & Green, 2008). Reported unemployment rates of people with TBI have ranged from 45% to 78% (Cuthbert et al., 2015; Doctor et al., 2005; Kendall, Muenchberger, & Gee, 2006; Yasuda, Wehman, Targett, Cifu, & West, 2001). This variation in return-to-work (RTW) rates is related to differences in sample characteristics, methodologies, and measures of employment. Nevertheless, unemployment for individuals with TBI is considerably higher than for individuals without disabilities.

TBI and its effects can be overwhelming, not only for individuals with this injury and their families (Cifu, Craig, & Rowland, 1996; Liu, Zhu, Liu, & Guo, 2015; Moriarty et al., 2015; Tarn, McKay, Sloan, & Ponsford, 2015), but also for other sectors of society. While individuals and families experience financial burdens due to the loss of jobs and wages, as well as medical and rehabilitation expenses (Costa, Dagher, Lamoureux, Guise, & Feyz, 2015; Fadyl & McPherson, 2009; Joseph et al., 2015; Ma, Chan, & Carruthers, 2014; Samuelsson, Tropp, & Lundqvist, 2014; Te Ao et al., 2014), communities and employers experience lost workdays and productivity (Te Ao et al., 2014; Andlin-Sobocki, Jonsson, Wittchen, & Olesen, 2005; Samuelsson et al., 2014). TBI can also impose substantial burdens on governments through publicly funded healthcare, rehabilitation service costs, and financial support for injured persons (Te Ao et al., 2014; Hyder et al., 2007; Samuelsson et al., 2014). The Centers for Disease Control and Prevention (CDC) in the United States, estimated that the medical and indirect costs of TBI (e.g., lost productivity) in 2010 were $76.5 billion in the United States (Coronado, McGuire, Faul, Sugerman, & Pearson, 2012; Finkelstein, Corso, & Miller, 2006), while the total annual cost in Europe was approximately ***386 billion (Andlin-Sobocki, Jonsson, Wittchen, & Olesenet al., 2005). Thus, the cost of TBI to both the individual and society is staggering. These costs can be ameliorated through rehabilitation and by returning individuals with TBI to work (Doctor et al., 2005; Kendall et al., 2006; Samuelsson et al., 2014; Yasuda et al., 2001).

Shames, Treger, Ring, and Giaquinto (2007) conducted a comprehensive review of the literature regarding factors associated with successful RTW following TBI and current rehabilitation strategies. They note that predicting a successful RTW involves a complex interaction of variables, including pre-morbid factors (e.g., age, education, and prior employment history), injury-related factors (type, severity, and location), post-injury impairments, and personal and environment factors (e.g., marital status, alternative income sources, and social support). Because of this complexity and the unique characteristics of each individual with a TBI, prediction of RTW is not feasible. They note, however, that the literature includes numerous examples of high-risk individuals who have been successful when given sufficient rehabilitation and support. In addition, the TBI employment research literature is international (e.g., Cancelliere et al., 2014; Fear et al., 2008; Li et al., 2015; Shekhar et al., 2015; Te Ao et al., 2014), which brings into consideration differences across economies, cultures, health care and rehabilitation systems, and national landscapes of public and social disability policies.

Despite the variety of factors related to poor employment outcomes for individuals with TBI, some factors are more frequently reported to contribute to poor employment outcomes (Shames et al., 2007):

More severe injury (Cuthbert et al., 2015; Kreutzer et al., 2003; Trexler, Trexler, Malec, Klyce, & Parrott, 2010; Wagner, Hammond, Sasser, & Wiercisiewski, 2002), as measured by the GCS, PTA, duration of rehabilitation, or other means

1.2.1 Post-Acute Return-to-Work Interventions

Following acute care, post-acute rehabilitation services focus on helping individuals adjust to ongoing impairments and re-enter their communities, workplaces, schools, etc. Post-acute interventions are broadly classified into the following groups (Shames et al., 2007): (1) Residential community reintegration programs that provide intensive behavioral interventions; (2) Comprehensive day treatment programs that emphasize training in self-awareness, social skills, daily living skills, and coping mechanisms; and (3) Community re-entry programs that focus on vocational and social reintegration.

Within these, RTW interventions are funded and delivered through multiple avenues. In the United States, the primary sources of RTW interventions are employee disability management programs and state vocational rehabilitation (VR) agencies. Many other countries have equivalent organizations (United Nations World Health Organization & The World Bank, 2011).

Brief descriptions of employee disability management and VR programs are presented in this section.

Employee Disability Management Programs, typically provided directly by employers or through their insurance programs, are a combination of many services (e.g., work injury prevention programs, early intervention, transitional work programs, and case management) to better address the physical and psychological needs of the worker (Ongori, 2012; Shrey, 1996). Job accommodations, support services, and determining the use of adaptive devices are all potential components of disability management (Ongori, 2012; Shrey, 1996). For example, gradual RTW would fall into this category. Additionally, some companies may offer on-site training as a disability management component. This may be similar to transitional programs, where the client receives physical, cognitive, and behavioral training while developing job skills in a practice work environment. This allows employers to see how and if they need to make accommodations, and helps the employee develop the skills they need to perform on the job (Ben-Yishay et al., 1985; Jacobs, 1997; Klonoff et al., 2007; Prigatano et al., 1994).

State VR Services are provided to individuals with various types of disabilities with the goal of increasing their independence and employment. Services include physical restoration (prosthetic devices), vocational training, counseling, and job development assistance. Supported employment is an increasingly utilized VR service option for clients with TBI, where a client obtains paid work that is integrated with people without disabilities while receiving ongoing support from an employment specialist or job coach (Wehman et al., 2003). An unique feature of the supported employment model is that individuals with TBI receive training on the job as opposed to pre-training.

An individual placement model of supported employment consists of an employment specialist or job coach who assists the individual with TBI in finding a job. This specialist then helps coordinate pre-employment supports (e.g., benefits counseling or transportation) and may provide on-the-job support and training (Wehman et al., 2003), in addition to other counseling or advocacy services that are needed (Wehman et al., 1993). These supports provided by the employment specialist vary from person to person and change over time. Eventually, the specialist's involvement reduces to periodic checks, as mandated by the Rehabilitation Act (U.S. Department of Education, 2004). These checks are critical to long-term support and the ongoing employment of individuals with TBI. Such long-term support is unique to the supported employment model (Wehman et al., 2003).

Job accommodations are common to the generally differing systems of Employee Disability management programs and state VR services. Title I of the Americans with Disabilities Act (ADA; Americans with Disabilities, 2008) requires employers with 15 or more employees to consider providing reasonable accommodation for employees who meet the ADA's definition of disability. Accommodations are negotiated between the employee and the employer and are determined on a case-by-case basis. Before becoming eligible for a reasonable accommodation, the applicant or employee with TBI must disclose his or her disability status to the employer and make a request for the accommodation.

Accommodations are alterations made for the employee that do not conflict with the responsibilities of the job, such as flexible scheduling to accommodate the person“s disabilities (e.g., epilepsy, chronic fatigue, or psychosocial disabilities). Environmental modifications are a type of job accommodation and include building ramps for wheelchair access, adding elevators, relocating an employee's office, and widening aisles. With the increasing use and availability of technology such as high-speed Internet, some businesses permit telework from home, allowing an employee to accomplish their daily tasks without having to leave their residence (West & Anderson, 2005). In addition, employers may also reassign non-essential job functions that the employee cannot perform due to his or her disability. Employers are not required to make accommodations for essential job functions by eliminating them from the job description. Essential functions are the fundamental, crucial job duties performed in a position.

Another accommodation strategy common to both systems is to provide assistive technology services and devices. These can include memory aids, touch-screen computers, adjustable mounts at desks, specialized computer software, or alternative input devices to allow a person with limited manual dexterity or fine motor skills to use a computer (Inge & Targett, 2007). With the advent of handheld devices such as Smartphones and tablet computers, individuals can use pre-installed programs to create schedules and set alarms (Gentry, Lau, Molinelli, Fallen, & Kriner, 2012). There are Applications (apps) for these devices that may be used to train employees with disabilities such as TBI by providing visual cues for how to complete tasks, ways to deal with interpersonal situations, and individualized task lists (Burke, Andersen, Bowen, Howard, & Allen, 2010; Chang, Chen, Chuang, 2011; Gentry et al., 2012; Tsai, 2012). Since Individuals with TBI often have memory deficits, the use of handheld devices with apps is an example of an accommodation that can facilitate skill acquisition and job retention.

Finally, both service systems can offer ancillary therapeutic services which, though not directly employment-focused, support the RTW process. These include physical therapy, occupational therapy, speech-language therapy, neuro-cognitive therapy, and peer or professional counseling. Utilizing these programs as part of a cohesive rehabilitation strategy emphasizes a holistic approach (Ben-Yishay et al., 1985; Prigatano et al., 1984, 1994).

1.3.1 Supply-Side and Demand-Side Interventions

Employment barriers and interventions may be broadly grouped into either supply-side or demand-side sources (Chan et al., 2010). Supply-side barriers are those related to individuals with disabilities, such as skill deficits, interfering behaviors, functional limitations related to the disability, and accommodation needs. Vocational rehabilitation provides support to individuals with disabilities on the supply-side of the labor market by providing training and placement services to job seekers (Luecking, 2008). The intent of these services is to prepare individuals with disabilities for employment and then supply these trained workers to employers. Supply-side interventions typically include the assessment of skill deficits followed by skill training and the provision of accommodations (including assistive technology) to prepare the job seeker for employment. Supported employment is a supply-side intervention. However, this is a “place and then train” approach to provide a labor supply to employers (Wehman, Inge, Revell, & Brooke et al., 2007). Individuals with TBI are supported in identifying their preferences for employment, and then are presented to potential employers for meeting labor market needs (Wehman & Targett, 2006). An employment specialist or job coach provides training and support on the job until the worker with disabilities performs to the satisfaction of the workplace. VR counselors, employment specialists, therapists, and educators typically deliver supply-side interventions. The duration of these interventions can range from days to years depending on the nature of the services. For example, an individual receiving supported employment services should receive ongoing support throughout his or her work history.

Conversely, demand-side interventions are increasingly seen as important for assisting individuals with disabilities in becoming employed (Chan et al., 2010; Gilbride & Stensrud, 1999; Luecking, 2008; Wehman, Kregel, & Brooke, 2008). Examples of demand-side barriers include employers' limited experience with employees with disabilities, employer attitudes toward hiring individuals with disabilities, high turnover or absentee rates, high production demands, and high costs related to injured workers.

Demand-side interventions may include disability management practices such as gradual return to full duty following injury, employer training and technical assistance related to disability and accommodation, co-worker and supervisor support and assistance, and employer-provided internships, skills training, or support programs.

The Logic Model (see Figure 1) presents the supply-side and demand-side intervention activities, recipients of intervention activities, and short- and long-term outcomes. Both types of interventions for individuals with TBI are included.

OPEN IN VIEWER

1

Logic Model, Employment Interventions for RTW in Working Aged Adults After TBI

3.1.1 Types of Studies

A priori criteria were developed for the selection of studies to be included in this review. The following criteria were used. Each study must have: (1) explored interventions that assist working-aged adults with TBI to return to competitive employment (with competitive employment as the outcome), including self-employment; (2) at least included the primary outcome and employment status, and may also have included secondary outcomes, length of time employed, and mean hours worked; (3) participants who were between the ages of 18 and 65, with a non-penetrating TBI, engaged in either full-time or part-time employment at the time of injury, and unemployed or on medical leave at the time of receipt of intervention (participants could have been injured on or off the job); and (4) presented data separately for competitive (i.e., open or independent) employment and the non-competitively employed participants or students. Studies that included individuals with other disabilities were included only if the results were provided for TBI participants separate from those with other disabilities. The designs of the included studies could be randomized controlled trials (RCTs) or quasi-experimental designs, with a treatment group and a control/comparison group consisting of treatment as usual, other appropriate interventions, or no intervention.

3.2.1 Language and Geographic Origin

The search was restricted to published or unpublished research written in English or Spanish and published between 1973 and 2015. To include reports in all languages would be untenable and fiscally burdensome to include in this report.

3.2.2 Electronic Searches

Two trained research assistants conducted the literature search, sixteen databases (with 125 sub-search engines within them) were divided between the two research assistants. Each research assistant searched one database at a time, using appropriate search terms (see Table 2) for the database, before moving on to the next assigned database. One research assistant was assigned to conduct grey literature searches not related to databases.

A spreadsheet was constructed and shared with authors and research assistants to maintain a record of searches. Each database had a separate tab in which the name of the searcher, date, database, search string, year, first author, journal/source, title, and DOI were entered. Each day, research assistants reviewed titles of documents located during the search that appeared to describe an employment intervention. If research assistants could not determine relevance from the title, they read the abstract. These relevant abstracts and all search results were saved as PDFs by the research assistants and uploaded to a shared electronic folder, as well as printed out for record-keeping. Two authors (Graham and West) reviewed each relevant abstract and requested the full article for those that appeared appropriate for the systematic review. If documents were not available online, research assistants acquired copies from the university library or submitted an interlibrary loan request. These documents were subsequently added to the shared online folder for review.

Studies for this review were identified by searching 16 databases: Australia Education Index/Australian Council for Educational Research, CIRRIE - Center for International Rehabilitation Research Information and Exchange Database, www.ClinicalTrials.gov, Cochrane Central Registry of Controlled Trials, Ebrary: Academic Complete Collection, EBSCOhost Research Database, First Search, Google Scholar, MEDLINE/PubMed, NARIC REHABDATA, ProQuest, PsycEXTRA, PsycINFO, Science Direct, Web of Knowledge, and Web of Science. Within First Search, nine sub-search engines were searched. Within ProQuest, 47 sub-search engines were searched. Within EBSCOhost, 69 sub-search engines were searched. Further, unpublished papers and grey literature were also searched. Reference lists of papers included in the analysis and previous systematic reviews concerning RTW employment and TBI were also searched.

Grey papers were also searched via 67 national and state websites, which included eight international websites:

Great Britain ○

○

https://www.gov.uk/financial-help-disabled/overview

3.3.1 Selection of Studies

Three team members (Graham, West, & Inge) independently reviewed the amassed abstracts of articles and reports selected by the research assistants as potentially eligible for the systematic review. Articles were then screened by title and abstract. If the title and abstract suggested that the article might meet the inclusion criteria, the article was assigned to two of the three authors for review. Each author was assigned a set of articles to review, which included coding and determining whether the study should be included. Each reviewer used the approved coding template, in which characteristics of the study were coded, such as type and description of intervention, number of subjects, group and subject characteristics, and data needed for analysis. (See Appendix or Graham and West, 2014, for coding template.) reviews stopped if the reviewer determined that the article did not meet the inclusion criteria. After both assigned authors had reviewed the article, the authors met to discuss the article, coding, and inclusion decision. Each article was discussed concerning coding. If there was a difference in coding, the authors discussed the coding and came to a consensus, inclusion required both assigned reviewers to come to a consensus on coding and the inclusion decision. A consensus of included studies for this systematic review was reached without requiring a third reviewer to resolve selection conflicts. None of the authors were involved in any of the studies reviewed, and a third-party reviewer was not needed to minimize bias.

3.3.2 Data Extraction and Management

After coding included studies, inter-rater reliability was established above 85%. Using coded forms for each included study, the authors entered data into Comprehensive Meta-Analysis 2.2 software.

4.1.1 Results of the Search

An extensive literature search was completed on VR and RTW of adults with TBI (see Figure 2).

OPEN IN VIEWER

2

Specific overview of article selection.

The literature search utilized 16 search engines and 125 sub-search engines to search for published, peer-reviewed articles and grey literature. In addition, 67 websites were searched for grey literature. This yielded a total of 705,362 studies, which included duplicates (511,471 possible documents within the search engines 193,874 documents examined on Internet websites with grey literature and 21 additional articles found through hand searching past systematic reviews and primary study reference lists). In order To manage data, a series of Google Drive folders were constructed. The “All Search Documents” folder contained a folder for each search engine. Each of these search engine folders consisted ofcontained the actual search results for each term used research assistants who were assigned search engines kept track of the number of documents found in the “Unduplicated Results” folder. After each search term was run in a search engine, the research assistant would determine the overall number of hits, upload the search hits, screen each hit by article title of the article to determine whether further examination was needed by other team members, and determine whether there were duplicates from past searches. Of those 705,362 documents, 6,941 were unduplicated. The document title and abstract were read to determine whether documents were relevant to the systematic review. Unpublished articles with titles and abstracts that were considered potentially eligible were recorded in the “All GreyDocument” spreadsheet file. The record for a potential grey document consisted contained of name of the research assistant, date found, website address, year the document was published, author, title, type of document (e.g., dissertation, presentation, etc.), search string that located the document, and note (access status). When the document was available, a copy of the document was uploaded into the “Grey PDFs” folder. Once the document had been read by a researcher, the document would be either moved to the “Print Out” folder for further review or to the “Viewed” folder. Researchers discussed which studies should be examined further. The identified documents were then printed out for further review. If it was decided that the study might fit the criteria, researchers coded the document. Documents in the “Viewed” folder were those found not to meet the inclusion criteria. Peer-reviewed articles were processed similarly. From these 6,941 documents, the abstracts of 374 peer-reviewed articles and 40 grey literature documents were reviewed. Of the 414 peer-reviewed and grey documents examined, 67 presented findings on the effectiveness of interventions for RTW for working-aged adults with TBI and were assessed for eligibility. The last search was conducted on November 7, 2015.

4.1.2 Included Studies

Of the 67 intervention study reports identified, only three studies (Man, Poon, & Lam, 2013; Salazar et al., 2000; Twamley, Jak, Delis, Bondi, & Lohr 2014, and Twamley et al. 2015) fit all of the inclusion criteria. Note that Twamley reported outcomes in two articles. These two articles are being counted as one article, as both articles concern the same study. All three RCTs included in this review used other interventions that were suspected to be less effective than the “treatment” group. Below, both the primary treatment and alternate treatment groups are described.

The Man et al. (2013) study was a RCT study conducted in China with a civilian sample. This study was conducted with 50 individuals who had mild to moderate TBI and were recruited from Hong Kong hospitals and rehabilitation facilities. Man et al. did not report The cause of TBI. Inclusion criteria required participants to have a loss of consciousness of less than six hours. There was no report of mean length of loss of consciousness for either group. However, the Glasgow Coma Scale was used to assess the severity of TBI, and the mean score on this scale was 10.25 (SD = 1.51; Range = 9-14) for the artificial intelligent virtual reality-based training program (AIVTS; treatment group) and 10.05 (SD = 1.31; Range= 9-13) for the psycho-educational vocational training system (PEVTS; second intervention). No time since injury information was reported.

In the Man et al. study (2013), 50 participants were randomly assigned to either the first intervention group (AIVST) or the second intervention group (PEVTS) (Man et al., 2013). Each group began with 25 participants. Both groups had similar informational content but the training platform differed by group. The AIVST group received 12 sessions using a 3-D artificial intelligence reality-based training software program. The program begins the training by determining the skill level of the participants. As the participant progresses through training, the program can upgrade, maintain, or downgrade the training level to meet the training needs of the participant based on performance. Therefore, when participants performed well, the program would increase difficulty. If the participant's performance declined, the program would revert to previous levels of training.

The PEVTS group also received 12 training sessions with similar problem-solving tasks, content, and structure, given by a vocational trainer. The trainer provided instruction, time to practice skills, and games on problem solving specified by the manual. The fidelity of the administration of this intervention was assessed by four trained rehabilitation professionals. The fidelity agreement intra-class coefficient was 0.645, P = 0.021.

Although the study authors referred readers to a demographic table, no demographic table was published in the article. Forty participants (20 per group) provided three follow-up reports of employment outcome at one, three, and six months. Primary outcomes consisted of Tower of London assessment, Wisconsin Cord Sorting Test, Vocational Cognitive Rating Scale, and employment outcomes. Employment outcomes were measured by four categories: full-time employment, supported employment, sheltered employment, and unemployed/unable to return to work. No other employment measures were reported.

Of the 20 participants in the AIVTS group (treatment group) that were assessed at follow-ups, 17 (85%) were unemployed and three (15%) were employed at one month; 14 (70%) were unemployed and six (30%) were employed at three months; and 12 (60%) were unemployed and eight (40%) were employed at six months. Of the 20 participants in the PEVTS group that were assessed at follow-ups, 20 (100%) were unemployed and zero were employed at one month; 16 (80%) were unemployed and four (40%) were employed at three months; and 16 (80%) were unemployed and four (40%) were employed at six months. The odds ratio was 2.204 (95% confidence interval [CI] = 0.459, 10.573).

The Salazar et al. (2000) study was conducted on a United States military sample and is based on three articles that used the same dataset from one RCT with parallel group interventions: (1) an intensive in-hospital, cognitive treatment group, and (2) an in-home program without a VR group (Braverman et al., 1999; Salazar et al., 2000; Warden et al., 2000). The Salazar et al. (2000) article was the complete report that met all the inclusion criteria for this systematic review and was used in this review.

The Salazar et al. (2000) study was an RCT with parallel treatment groups, conducted with 120 active duty military service persons who had acquired a TBI. There is some discrepancy in the reporting of the severity of injury. Salazar et al. report that the injuries were “moderate to severe” (p. 3035), but state later that the injuries were “relatively severe” based on “the rates of axonal shear injury on MRI (92% and 95%), cerebral contusions (51% and 54%), posttraumatic amnesia of 7 days or more (41% to 42%), and traumatic unconsciousness for 24 hours or more (30% and 38%), respectively, in the hospital and home treatment groups” (p. 3077). Another article from this study reported “moderate to severe TBI” (p. 1100) based on “all patients reaching] the inclusion criteria of Rancho los Amigos level of 7” (Warden et al., 2000, p. 1097). Most participants obtained TBI through motor vehicle/traffic incident, assault, and alcohol-related accidents. The mean loss of consciousness was 35 hours. All participants were within 90 days of injury. Participants in the study were randomly assigned to one of the two treatment groups. Each group had a separate treatment team from the other, and the treatment teams worked independently. There were periodic reviews of the treatment teams in order to maintain fidelity to each treatment protocol.

The 67 active duty military participants assigned to the in-hospital, cognitive treatment group received eight weeks of intense cognitive rehabilitation program. Their mean age was 24.7, 93% were male, 69% were White, 24% were African American, and 7% were “other” or not specified. Participants had either moderate to severe TBI or severe TBI with a Glasgow Coma score of 9.4.

This in-hospital, cognitive treatment program consisted of a team of interdisciplinary practitioners, including a physiatrist, neuropsychologist, occupational therapist, speech pathologist, and rehabilitation assistants. The physical therapist and psychiatric and neurological experts were consulted when needed. The participants in the group were expected to follow military standards and a structured daily routine. The structured daily routine consisted of physical fitness training, and group and individual therapies. Therapies included cognitive, psychotherapy, pragmatic speech, community re-entry, and integrated work programs.

The 53 active duty military participants assigned to the in-home without VR intervention treatment group received a limited home rehabilitation program. Their mean age was 26, 96% were male, 70% were White, 17% were African American, 11% were Hispanic, and 2% were Native American. All had moderate to severe TBI or severe TBI with a mean Glasgow Coma score of 9.5.

The in-home rehabilitation program included TBI education and individual counseling by a psychiatric nurse, who made weekly 30-minute phone calls to each participant. Educational materials and strategies to improve cognitive and organizational skills were given to participants. Participants were also trained in card game exercises and were encouraged to watch the news and read. They were also expected to resume daily physical fitness training. No participants in this group received vocational or work rehabilitation.

The Salazar et al. study (2000) reported on the 12-month outcome evaluation. Sixty participants (90%) in the in-hospital, cognitive treatment group returned to work, whereas 50 of 53 participants (94%) in the in-home treatment group returned to work with an odds ratio of 0.514 (95% CI = 0.126, 2.093). See Table 3 for the characteristics of the included study.

Twamley et al. (2014, 2015) also conducted a RCT with parallel treatment groups. Fifty United States veterans with mild to moderate TBI from recent conflicts participated. Twenty-five were randomly assigned to the supported employment with CogSMART group (treatment group) and 25 were randomly assigned to the enhanced supported employment group. Only 42 participants completed the study, with 21 participants in each of the two groups. Time since most recent TBI was 4.08 years (SD=2.78) for the CogSMART group (treatment one group) and 5.04 years (SD=4.61) for the enhanced supported employment group (treatment two group) (p=0.377). The majority of participants in each group were men (96% in each group). The two groups were also similar in composition in terms of the number of participants who identified as Hispanic (36%). Twenty-four percent of the treatment one group and 48% of the treatment two group identified as White. The mean length of loss of consciousness across the last four traumatic brain injuries was six hours for the treatment one group and one hour for the treatment two group (p=0.274). There was a statistically significant difference between the two groups in terms of participants' age. The treatment one group had a mean age of 29.68 (SD=6.05), and the treatment two group had a mean age of 33.84 (SD=7.89) (p = 0.042). A statistically significant difference was also reported in education. The treatment one group had a mean of 12.88 years (SD=1.2.7), versus 14.00 years (SD=1.80) for the treatment two group (p=0.014). The odds ratio for this comparison was 0.881 (95% CI = 0.235, 2.844).

All participants in both groups received supported employment for one year. For the first three months of the intervention, the treatment one group received supported employment and CogSMART, while the treatment two group received additional supported employment sessions. CogSMART addressed post-concussive symptoms, prospective memory, attention and vigilance, learning and memory, and executive functioning (see cogsmart.com for more information). The treatment one group received one hour of CogSMART and one hour of supported employment per week for the first three months, and then one hour of supported employment for the rest of the year. The treatment two group received two hours of supported employment for the first three months, and then one hour of supported employment for the rest of the year. Separate employment specialists worked with each group.

Each group lost four participants during the course of the study or at follow-up, which left each group with 21 participants each. There was no difference in competitive employment at 12 months, with both groups having 13 participants (52%) in employment. Thus, CogSMART did not improve the odds of employment for the treatment one group. The supported employment sessions that both groups received may have contributed to this finding. Further research with three groups—CogSMART only, CogSMART and supported employment, and supported employment only—needs to be conducted.

Missing Data. All three studies had attrition. Man et al. (2013) lost five participants in each group during the study. Salazar et al. (2000) lost seven participants in the in-hospital cognitive intervention group and nine in the at-home intervention group. Twamley et al. (2014, 2015) lost four participants in each group. Participants who dropped out were not included in this analysis. Salazar et al. did not provide attrition information. Data on the Glasgow Coma Scale was incomplete for the Salazar et al. study, with scores for only 44 of the participants in the study available. The Twamley articles did not report Glasgow Coma Scale scores but instead provided length of total loss of consciousness injuries (U.S. military commonly report multiple traumatic brain injuries during deployment).

Man et al. (2013) did not report on any demographic data such as gender, race/ethnicity, or education level. Twamley presented the demographic data previously mentioned, but the reporting across the two articles published on this study varied (Twamley et al., 2014, 2015).

For the Salazar et al. (2000) study, the primary outcome was employment status, and it was reported in binary form at twelve-month follow-up. None of the other articles concerning this study reported other time points (Braverman, 1999; Warden et al., 2000). The secondary outcome data—mean hourly wage and total days employed—were not reported in this study.

Man et al. (2013) used four categories to measure employment: (1) full-time employment, (2) full-time supported employment, (3) sheltered employment, and (4) unemployed or unable to work. For this review, full-time employment and full-time supported employment were summed to determine the number of competitively employed participants. Sheltered employment and unemployed/unable to be employed were summed for the unemployed group. Employment data were collected and reported at three time periods: one month, three months, and six months (See Table 1).

Twamley et al. (2014, 2015) reported that employment status was collected at three, six, and twelve months. Twamley et al. reported three-month employment data in one study (2014). The second article states that three-month, six-month, and twelve-month outcome data would be presented, but Twamley et al. (2015) only reported twelve-month results in this article. Six-month data were not reported in either article, and further literature searches did not yield any documents reporting the six-month data. Only the binary data of whether participants were employed or not was collected. No other employment-oriented outcomes were measured, such as those of interest to this study—total days employed and mean hourly wage.

studies reported employment status at different time periods ranging from one month to six months. Two studies (Man et al., 2013; Twamley et al., 2014) reported employment outcomes at three months. Salazar et al. (2000) and Twamley et al. (2015) reported employment outcomes at 12 months. Twamley et al. (2015) stated that three-month, six-month, and twelve-month outcome data would be reported, but only twelvemonth employment findings were reported in this article. The overall analysis was performed on the latest time period for each study, which was 12 months for the Salazar et al. (2000) and Twamley et al. (2015) studies, and six months for the Man et al. study (2013). Sensitivity analysis was conducted to determine whether this was problematic. (See Table 1.)

Lastly, none of the studies reported to be intent-to-treat studies. Two studies had attrition: Man et al. (2013) lost three participants in the artificial intelligence group and none of the participants in the psychoeducational group. Twamley et al. (2014, 2015) lost four in each of the two groups (CogSMART Plus group and enhanced supported employment group). Salazar et al. (2000) did not report attrition. This was an active military sample and the type of participants might have contributed to no attrition being reported. Data were used as reported. In the two studies where participants dropped out, the total number remaining in the sample was used. Thus, findings for the participants who remained are included in the Results section.

4.1.3 Excluded Studies

Sixty-four of the 67 study reports did not meet the inclusion criteria. The excluded studies included eight randomized controlled trials (RCT), 13 non-randomized pre-post two-group studies, one non-randomized pre-post three-group study, 22 pre-post one-group studies, 11 retrospective studies, five case studies, and one description of an intervention. Below, the RCT and non-randomized pre-post designs are summarized. The RCTs included cognitive rehabilitation (Cicerone et al., 2008; Vanderploeg et al., 2008), multidisciplinary rehabilitation programs (Powell, Heslin, & Greenwood, 2002; Zhu, Poon, Chan, & Chan, 2007), early intervention (Wade, Crawford, Wenden, King, & Moss, 1997), motivational interviewing by way of the telephone (Bell et al., 2005), resource facilitation (Trexler, Trexler, Malec, Klyce, & Parrott, 2010), and individualized tailored treatment (Andersson, Emanuelson, Björklund, & Stålhammar, 2007). Five used control groups (Andersson et al., 2007; Bell et al., 2005; Powell et al., 2002; Trexler et al., 2010; Wade et al., 1997), and three used comparison treatment groups (Cicerone et al., 2008; Vanderploeg et al., 2008; Zhu et al., 2007). All but one study used civilians in their study (Andersson et al., 2007; Bell et al., 2005; Cicerone et al., 2008; Powell et al., 2002; Wade et al., 1997; Trexler et al., 2010; Zhu et al., 2007). The other study used veterans (Vanderploeg et al., 2008). TBI severity ranged from mild to severe, with two studies not reporting severity. The sample size ranged from 68 to 483. These studies were conducted in China, Sweden, the United Kingdom, and the United States. Most RCTs were rejected because some of the participants were under 18 years of age (Andersson et al., 2007; Bell et al., 2005; Powell et al., 2002; Wade et al., 1997; Trexler et al., 2010; Zhu et al., 2007), or because some participants did not work at the time of injury (Andersson et al., 2007; Bell et al., 2005; Cicerone et al., 2008; Trexler et al., 2010; Wade et al., 1997; Zhu et al., 2007). One RCT was rejected because it combined RTW with school and did not present separate findings (Vanderploeg et al., 2008).

The 13 non-randomized pre-post two-group studies and the one non-randomized pre-post three-group study (High, Roebuck-Spencer, Sander, Struchen, & Sherer, 2006) will also be described. The interventions used in these studies included neuropsychologic interventions (Prigatano et al., 1984; Prigatano et al., 1994; Sarajuuri et al., 2005; Wilier, Button, & Rempel, 1999), community-based interventions (Heinemann, Corrigan, & Moore, 2004; High et al., 2006; O'Neill, Zugger, Fields, Fraser, & Pruce, 2004; Ponsford, Harrington, Olver, & Roper, 2006), cognitive interventions (Cicerone, Mott, Azulay, & Friel, 2004; Dawson, Binns, Hunt, Lemsky, & Polatajko, 2013), occupational interventions (Radford et al., 2013), work entry interventions (Haffey & Abrams, 1991), comprehensive day interventions (Hashimoto, Okamoto, Watanabe, & Ohashi, 2006), and an employment enhancing program with club houses (Niemeier et al., 2010). There was one non-randomized pre-post three-group study (High et al., 2006), which will also be described. Ten studies used control groups (Cicerone et al., 2004; Dawson et al., 2013; Hashimoto et al., 2006; Niemeier et al., 2010; O'Neill et al., 2004; Prigatano et al., 1984; Prigatano et al., 1994; Ponsford et al., 2006; Radford et al., 2013; Sarajuuri et al., 2005; Willler et al., 1999), and four compared treatments (Haffey & Abrams, 1991; Heinemann et al., 2004; High et al., 2006). All used civilians. TBI severity ranged from mild to severe. The sample size ranged from 13 to 319. These studies were conducted in Australia, Canada, Finland, Japan, the United Kingdom, and the United States. Most of these studies were rejected because not all participants were employed at the time of injury (Cicerone et al., 2004; Haffey & Abrams, 1991; Hashimoto et al., 2006; Heinemann et al., 2004; Ponsford et al., 2006; Prigatano et al., 1984; Prigatano et al., 1994; Radford et al., 2013; Sarajuuri et al., 2005). Other reasons for exclusion included the following: not all participants were over the age of 18 years (Radford et al., 2013; Sarajuuri et al., 2005); employment category was unclear, which made it difficult to determine whether competitive employment was assessed or not (O'Neill et al., 2004); some participants were employed at the beginning of the treatment (Dawson et al., 2013); and separate findings were not provided for TBI (Niemeier et al., 2010; Wilier et al., 1999).

Studies had to have a control or comparison group to be included in this systematic review. Thus, all of the pre-post only-group studies were excluded from the review. The interventions found among the 22 pre-post one-group studies included neurorehabilitation (Bjorkdahl, 2010; Ezrachi, Ben-Yishay, Kay, Diller, & Rattok, 1991; Geurtsen et al., 2011; Geurtsen et al., 2012; Klonoff, Lamb, & Henderson, 2000; Klonoff, Lamb, & Henderson, 2001; Perino, Zappala, Verne, & Rago, 2000); supported employment (Curl, Frasier, Cook, & Clemmons, 1996; Ellerd & Moore, 1992; Malec & Moessner, 2006; Wehman et al., 2003); club house (Vandiver & Christofero-Snider, 2000); cognitive therapy (Braverman et al., 1999; Walker, Onus, Doyle, Clare, & McCarthy, 2005); transitional job coaching (Preston & Ulicny, 1992); in-home rehabilitation (Warden et al., 2000); comprehensive outpatient rehabilitation (Ben-Yishay, Silver, Piasetsky, & Rattok, 1987); multidisciplinary interventions (Christensen, 1992); community-based interventions (Kolakowsky-Hayner, Wright, Shem, Medel, & Duong, 2012); interpersonal communication training; attention training; community training; and personal counseling (Rattok et al., 1992). The severity of TBI ranged from mild to severe, and sample sizes ranged from 9 to 603. Two studies were military (Braverman et al., 1999; Warden et al., 2000) and were one-group analyses of the Salazar et al. (2000) study included in this review. The other 20 studies were civilian. These studies were conducted in Australia, Denmark, Italy, the Netherlands, Sweden, the United Kingdom, and the United States. See Table 4 for more information concerning the characteristics of the excluded studies.

4.1.4 Location of Study

Two of the three included studies were conducted in the United States (Salazar et al., 2000; Twamley et al., 2014, 2015). The other study (Man et al., 2013) was conducted in China.

4.1.5 Participants

Sample sizes ranged from 40 to 120 participants, with a total sample size across the three included studies of 146. Two studies provided the proportion of males and mean age (Salazar et al., 2000; Twamley et al., 2014, 2015). The unweighted mean age for study participants was 28.56. The Salazar et al. and Twamley et al. samples were predominantly male, ranging from 93% to 96% male, with a mean of 95.25. Salazar and colleagues (2000) used active military samples from the United States, and Twamley et al. (2014, 2015) used U.S. military veterans. Man and colleagues (2013) used civilians in China. The Salazar et al. (2000) sample was predominantly African American and White, and there were more Hispanic/Latino and White participants in the Twamley et al. (2014, 2015) study. Man and colleagues (2013) did not report the ethnicity or race of their sample from China. The Twamley et al. (2014, 2015) and Man et al. (2013) studies reported severity of injury as mild to moderate. There were inconsistencies in the reporting of severity in the Salazar et al. (2000) study, with most places in the document reporting moderate to severe and one location in the Salazar et al. document reporting only severe injuries.