Community-based interventions for adolescents following traumatic brain injury: A systematic review

1.1 Objectives

The main objective of this systematic review was to examine the effectiveness of existing community-based supports and intervention programmes available to adolescents following TBI. For the purposes of this review, a community-based intervention was defined by setting and geographical location, including those delivered in a home environment via the internet, or using community facilities. For widespread coverage of community-based rehabilitation strategies, all settings outside of clinical inpatient care and school-based rehabilitation were included. This was to ensure that neither the context of intervention development nor where the intervention was initiated could bias the types of interventions included. It also guaranteed full-coverage across disciplines and sectors, including those which integrate multiple services, and those which involve repeated contact with professionals outside of a community setting.

2.1 Search strategy

A comprehensive search of peer-reviewed studies was conducted in ten databases: Medline (Pubmed); Medline (Ovid); PubMed Central; PSYCHinfo; EBSCOhost; JSTOR; Cochrane Library; Web of Science; ASSIA and Scopus. In each search, all combinations of keywords were grouped into those relating to sample age, injury, intervention, intervention delivery setting, and outcomes. Adaptations to search strategies were made as per the requirements of individual databases. The reference lists of included studies were also examined, alongside a keyword search of the grey literature.

2.2 Eligibility criteria

Studies were eligible for inclusion if they involved a clear clinical population group diagnosed with a TBI or concussion. If the sample was inclusive of other forms of acquired brain injury, studies were only included if the TBI data was presented separately. To ensure a comprehensive view of community-based interventions there was no specific diagnostic criteria of TBI required for studies to be included. The searches were limited to a sample age of 11 to 25 years inclusive; studies including children or adults outside of this range were only included if the data for the required age range was presented separately. Included studies had to apply intentional interventions that were delivered in a community setting, such as web-based packages accessed at home, or rehabilitation programmes in the local youth centre. Interventions delivered in schools, hospital or neurorehabilitation inpatient settings were excluded. Research designs had to be experimental, and include at least two points of measurement, including one at baseline and one post-intervention. Qualitative studies and opinion articles were excluded. All outcome measures were considered. No restrictions were placed on the year of publication of the study, with databases searched in their entirety. No language restrictions were applied, and the searches were completed in March 2017.

2.3 Study selection and data extraction

Results from all databases were merged using reference management software and duplicates were removed. All abstracts from the initial search were reviewed by two independent reviewers and categorised into (i) ‘likely include’, (ii) ‘maybe include’, and (iii) ‘exclude’. The same method was followed for secondary abstract screening, and lastly full text review, using a study inclusion form to identify eligible studies. In the cases where reviewers disagreed, a third reviewer resolved this by reaching a consensus. A log of decisions at each strategy was maintained to provide an audit trail.

Two independent reviewers then used a data extraction form to systematically obtain the relevant data items for assessment. The data items extracted from each study included: study characteristics; overall intervention focus; specific characteristics of the intervention; delivery setting; mode and intensity of delivery; intervention provider; economic and replication costs of the intervention; outcome measurements; summary of intervention effectiveness; and conclusions of the study. Due to the clinical heterogeneity and marked variation in the components, modes and delivery of interventions, and intended outcomes, a meta-analysis could not be conducted.

2.4 Risk of bias assessment

The methodological quality of studies was assessed using a strategy based on the guidelines proposed by Hayden and colleagues (2006). This strategy involves assessing risk of bias on 25 separate domains, which are grouped into six key areas including: study participation; study attrition; prognostic factor measurement; outcome measurement; confounding measurement and account; and analysis. On each domain study quality was categorised as either: (i) ‘criteria fulfilled’; (ii) ‘criteria partly fulfilled’; (iii) ‘criteria not fulfilled’; (iv) ‘unclear whether criteria fulfilled’; or (v) ‘criteria not applicable’. The quality of evidence for each of these key areas was summarized according to a template that was consistent with that of the Scottish Intercollegiate Guidelines Network (SIGN) methodology (Mollayeva et al., 2014). For each of the 6 key areas, studies are categorized into a methodological quality hierarchy of: ‘high quality’ (+++) when all or most of the quality criteria in an area are fulfilled; ‘good quality’ (++) when the majority of criteria are fulfilled; ‘fair quality’ (+) when few criteria are fulfilled; and ‘low quality’ (–) when no criteria are fulfilled. Decisions were recorded at every stage of the assessment process.

Table 2 provides a detailed account of individual study bias ratings. Inter-rater agreement in assessing risk of bias was good at 88%. Overall, potential risk of bias in the included studies was highly variable, which is a reflection of the quality and range of study designs. Risk of bias assessment of study participation found that whilst reporting was clear for some included studies, in others it was not, making risk of bias difficult to assess (study participation). It was also more difficult to assess participation bias due to case study designs. The prognostic factor for all studies included was brain injury. Assessment of whether TBI was adequately measured (prognostic measurement) found that comprehensive diagnostic criteria (such as that of the Diagnostic and Statistical Manual of Mental Disorders, 5th edition) were often not followed, with many studies dependent on loss of consciousness measures alone, yet not clearly reporting their interpretation of score when diagnosing severity (e.g. Glasgow Coma Scale). The risk of bias assessment of whether outcomes were adequately measured (outcome measurement) showed that the case studies tended to be of lower quality, due to the use of non-validated scales, self-reported observations dependent on memory, and caregiver- and staff-observation at limited time points. There appeared to be less consideration of confounders in research design (confounding measurement and account), which resulted in an increase in the number of studies considered lower quality.

Table 3 shows a summary of the intervention components used in each included study. All interventions included neurorehabilitation approaches with an aim to circumvent any impairments by developing strengths, either through a compensatory (n = 14), or restorative approach (n = 3) (SIGN, 2013).

Interventions sought to address specific aspects of neurorehabilitation across the range of categories developed by SIGN (2013), which include: physical rehabilitation and management; cognitive rehabilitation; rehabilitation of behavioural and emotional disorders; communication and swallowing; and vocational rehabilitation. Cognitive rehabilitation strategies included an intervention targeting gist reasoning and memory, attention process training, and assistive technology to aid orientation. Behavioural and emotional rehabilitation included verbal and non-verbal communication, social participation, and behaviour management techniques. Rehabilitation of communication problems included providing supportive relationships for service users, and a theatre-based creative arts therapy seemingly focused on intonation and prosody using strategies, such as voicework, breathing, movement, character development, and group dynamics. The vocational rehabilitation interventions provided social and communication support strategies and functional skills training to compensate for any residual impairments.

It should be noted that the same web-based counsellor assisted problem-solving intervention (CAPS) was replicated in five different studies, which use similar study design, and the same sample population. (Study identifiers in table: ‘Kurowski 2013’;- ‘Tlustos 2016’; ‘Wade 2014’;-‘Wade, Kurowski … 2015’; Wade, Taylor ...  2015’) These studies used a comprehensive approach to rehabilitation, focused on family problem-solving and adjustment, as well as emphasizing specific strategies regarding individual behavioural, social, and cognitive rehabilitation. CAPS was designed with the participation of assigned clinical psychologists to deliver counselling, which involved designing a treatment overview, goal planning, and assistance coping with injury-related issues. A further three studies also described a Teen Online Problem Solving intervention (TOPS), which also used a similar comprehensive rehabilitation approach consisting of self-guided web exercises and linked family video-conference therapy sessions (Wade, Walz, Carey, & Williams, 2008; Wade et al., 2010, 2011).

A variety of underlying intervention components were used to deliver these neurorehabilitation strategies, from using a creative arts therapy to a computerised gist reasoning training package. In all of the included studies, interventions were delivered in community-based settings, either solely face-to-face (n = 6), via a combination of face-to-face, web-based, and/or by phone (n = 11).

A summary of key findings related to the effectiveness of the interventions is reported in Table 4. All studies concluded that their community-based intervention seemed to have some level of improvement for persons with TBI. This is despite six of the studies providing only descriptive statistics, with one reporting such data graphically (although unclear labelling made it difficult to assess the impact of the intervention). These studies reported descriptive improvements in social skills (O’Reilly, Lancioni, & O’Kane, 2000), social participation (Johnson & Davis, 1998), caregiver-burden (Palmisano & Arco, 2007), productivity (Feeney & Achilich, 2014), behavioural difficulties (Feeney & Achilich, 2014), orientation (Kirsch et al., 2004), and communication (Agnihotri et al., 2014). One of the CAPS studies reported results approaching significance, which suggested possible improvements in social skills (Tlustos et al., 2016), and an Attention Process Training intervention reported no meaningful functional change in attentional capacity however discussed the possibility of some improvement having occurred.

Nine studies reported statistically significant results regarding treatment effect (Cook, Chapman, Elliott, Evenson, & Vinton, 2014; Kurowski et al., 2013; Wade et al., 2008, 2010, 2011, 2014; Wade, Kurowski et al., 2015; and Wade, Taylor et al., 2015; Babcock et al., 2017). Of these, Cohen’s d effect sizes were reported in three (Cook, Chapman, Elliott, Evenson, & Vinton, 2014; Kurowski et al., 2013; and Wade et al., 2008). All effect sizes were moderate to strong and ranged from 0.580 to 1.868. The cognitive training intervention, self-management intervention, and the comprehensive approach of CAPS and TOPS were the only strategies which led to statistically significant improvements. However, this is in part due to the creative arts and social skills interventions being case studies with limited outcome data collected.

Four studies reported statistically significant similar improvement in executive functioning. One of these studies used cognitive training through a gist-reasoning intervention and found an increase in performance on a working memory and inhibition task (p = 0.02, d = 0.939), alongside statistically significant improvement in three gist reasoning tasks (p = <0.01, d = 1.868; p = 0.04, d = 0.750; p = 0.04, d = 0.766) (Cook, Chapman, Elliott, Evenson, & Vinton, 2014). The second study used self-management training strategies and found an increase in parent-reported executive functioning (p = 0.03) (Babcock et al., 2017). The third study used the TOPS multi-component intervention and reported an increase in executive functioning only for more severely injured participants (p <  0.05) (Wade et al., 2010). The final study used the CAPS multi-component intervention and found an increase in executive functioning for the older subgroup (p = 0.01, d = 0.630), although this was not the case for the sample as a whole, nor the younger subgroup (Kurowski et al., 2013).

In addition to executive functioning, both the CAPS and the TOPS interventions resulted in statistically significant improvement in behavioural functioning, depression, and/or community functioning. One of the CAPS studies reported a significant decrease in problematic behavioural functioning post-intervention, including decreases in internalizing (p <  0.05), aggressive (p <  0.05), attention problems (p <  0.05), ADHD (p <  0.05), and conduct disorder symptoms (p <  0.05) (Wade et al., 2014). A decrease in externalizing behaviours was reported consistently at multiple follow up points compared to an internet resource comparison (p <  0.01; p = 0.01; p <  0.01) (Wade, Taylor, et al., 2015). One of the TOPS studies also found a decrease in internalizing behaviours for the intervention group (p = 0.03, d = 0.580) (Wade et al., 2008). However, another TOPS study found the decrease in parent reported internalizing behaviour was limited to those with a higher injury severity (p = 0.01). Similarly, improvements in parent reported externalizing behaviours were also moderated, though by socio-economic status; results showed a decrease for those from a lower SES following the TOPS intervention (p = 0.09). Conversely in the same study, adolescent reported externalizing behaviours decreased significantly only for participants of higher income families (p = 0.01) (Wade et al., 2011).

Additionally post-TOPS a significant reduction in depressive symptoms for both adolescent (p = 0.02, d = 0.750) and parent (p = 0.01, d = 0.800) was reported, as well as significant improvement in a number of adolescent-parent family dynamics. The study also reported decreases in parent-adolescent conflict (p = 0.04, d = 0.740), number of parent-adolescent problem issues (p = 0.01, d = 0.920) and in ratings of severity of family problems (p = 0.01, d = 1.450) (Wade et al., 2008). A further TOPS study also reported a significant decrease in parent-adolescent conflict (p <  0.01) (Wade et al., 2011). Another study reported a significant increase in community functioning post-CAPS (p = 0.04) (Wade, Kurowski, et al., 2015). A vocational rehabilitative skills study also reported significant reductions in perceived adolescent-related symptom burden for both parents and adolescents (p <  0.01; p <  0.01), and for parent-reported functional disability (p <  0.01) (Babcock et al., 2017).

Eight of the studies reported on outcomes at least 6 months post intervention (Agnihotri et al., 2014; Kurowski et al., 2013; Tlustos et al., 2016; Wade et al., 2010, 2011, 2014), with a follow up period of 18 months for two of the CAPS studies (Wade, Kurowski et al., 2015; Wade, Taylor, et al., 2015). In the remaining studies, the follow up period was 8 weeks or less (Babcock, et al., 2007; Cook et al., 2014; Johnson & Davis, 1998; Kirsch et al., 2004; O’Reilly et al., 2000; Palmisano et al., 2007; Youse et al., 2009), with two studies capturing outcomes only at the end of the intervention (Feeney et al., 2014; Wade et al., 2008).

Costs and resources associated with the interventions are presented in Table 1. The resources required to replicate the CAPS intervention, whilst systematically incorporating it into clinical practice, are high; the continued involvement of a trained clinical psychologist and the therapy provided is a core component of the intervention (Kurowski et al., 2013; Tlustos et al., 2016; Wade et al., 2014; Wade, Kurowski et al., 2015; and Wade, Taylor et al., 2015). Whilst the mode of delivery was generally web-based, clinicians attended the service user’s home address for an initial 90-minute meeting, and scheduled biweekly meetings with service users over the duration of three months. Considering the potential demand, a large time and resource commitment would be required to help ensure that service provision is adequate. Similar comments can be made regarding the TOPS intervention and the other problem-solving intervention, which both also involved continued clinical participation (O’Reilly et al., 2000; Wade et al., 2008, 2010, 2011).

The creative arts theatre-based intervention also required professional involvement, from professional theatre artists or equivalent; however as a group based intervention it has the potential to be scaled up (Agnihotri et al., 2014). The cognitive training gist-reasoning intervention involved an ‘interventionist’ having to complete approximately 30 hours of in-person training to be considered qualified to conduct the intervention, which then also had to be delivered in-person, individually, increasing the resources needed (Cook et al., 2014). Similarly, although the assistive technology intervention was largely web-based, this was sometimes supplemented by a one-to-one therapist retracing the route alongside the participant. Furthermore the lack of sustainability of treatment gains suggest that an even longer duration of intervention is required for consolidation, further increasing the resources needed (Kirsch et al., 2004).

Other interventions, such as the comprehensive daily goal planning, routine management, communication, and behavioural consequence interventions, did not require staff training to implement the intervention, thus reducing resources needed. However each intervention was individualised and required significant preplanning and development (Feeney & Achilich, 2014; Palmisano & Arco, 2007). The supported relationships intervention required active engagement and participation from an interest-matched nondisabled community participant, which was not always possible, perhaps suggesting that the commitment was too great (Johnson & Davis, 1998). This was also a component of the Attention Process Training, alongside active participation from a therapist (Youse et al., 2009).

4.1 Review limitations

Due to heterogeneity in study design there was some difficulty making valid comparisons in this review; definition of injury severity, time since injury, and age-related group comparisons are all important predictors of outcome, and when they were reported there was large variation. Due to these factors, only a limited evaluation of the interventions could be conducted, and this is exacerbated by incomplete reporting, particularly regarding details of the intervention. Despite the use of a comprehensive search strategy, devised in consultation with an experienced librarian, limiting the search to studies to articles published in peer-reviewed journals may leave some publication bias. This is due to the likelihood of some service evaluations appearing as reports in the grey literature, although attempts were made to highlight such literature. Data synthesis was challenging; the large heterogeneity of included studies meant that conducting a meta-analysis would have been meaningless, misleading, and possibly harmful. However, even descriptive synthesis was difficult making comprehensive internal quality appraisals more problematic.