Understanding the relationship between cognitive performance and function in daily life after traumatic brain injury

Free

Loading

  1. http://orcid.org/0000-0003-4113-2328Lindsay Wilson1,
  2. Lindsay Horton1,
  3. Kevin Kunzmann2,
  4. Barbara J Sahakian3,
  5. http://orcid.org/0000-0001-6044-9035Virginia FJ Newcombe4,
  6. Emmanuel A Stamatakis4,
  7. Nicole von Steinbuechel5,
  8. Katrin Cunitz5,
  9. Amra Covic5,
  10. Andrew Maas6,
  11. Dominique Van Praag7,
  12. David Menon4
  13. The CENTER-TBI participants and investigators
    1. 1 Division of Psychology, University of Stirling, Stirling, UK
    2. 2 MRC Biostatistics Unit, University of Cambridge, Cambridge, UK
    3. 3 Department of Psychiatry, University of Cambridge, Cambridge, UK
    4. 4 Division of Anaesthesia, University of Cambridge, Addenbrooke's Hospital, Cambridge, UK
    5. 5 Institute of Medical Psychology and Medical Sociology, University Medical Center Goettingen and Georg-August-University, Goettingen, Germany
    6. 6 Department of Neurosurgery, Antwerp University Hospital and University of Antwerp, Edegem, Belgium
    7. 7 Department of Psychology, Antwerp University Hospital and University of Antwerp, Edegem, Belgium
    1. Correspondence to Professor Lindsay Wilson, Division of Psychology, University of Stirling, Stirling FK9 4LA, UK; l.wilson{at}stir.ac.uk

    Abstract

    Objective Cognitive impairment is a key cause of disability after traumatic brain injury (TBI) but relationships with overall functioning in daily life are often modest. The aim is to examine cognition at different levels of function and identify domains associated with disability.

    Methods 1554 patients with mild-to-severe TBI were assessed at 6 months post injury on the Glasgow Outcome Scale—Extended (GOSE), the Short Form-12v2 and a battery of cognitive tests. Outcomes across GOSE categories were compared using analysis of covariance adjusting for age, sex and education.

    Results Overall effect sizes were small to medium, and greatest for tests involving processing speed (ηp 2 0.057–0.067) and learning and memory (ηp 2 0.048–0.052). Deficits in cognitive performance were particularly evident in patients who were dependent (GOSE 3 or 4) or who were unable to participate in one or more major life activities (GOSE 5). At higher levels of function (GOSE 6–8), cognitive performance was surprisingly similar across categories. There were decreases in performance even in patients reporting complete recovery without significant symptoms. Medium to large effect sizes were present for summary measures of cognition (ηp 2 0.111), mental health (ηp 2 0.131) and physical health (ηp 2 0.252).

    Conclusions This large-scale study provides novel insights into cognitive performance at different levels of disability and highlights the importance of processing speed in function in daily life. At upper levels of outcome, any influence of cognition on overall function is markedly attenuated and differences in mental health are salient.

    Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center's RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Traumatic brain injury (TBI) is a leading cause of disability, creating a huge burden on individuals and society.1 Over half of patients presenting with mild TBI report limitations in function at 6 months,2 and disability may persist for many years.3 Despite a high prevalence, much of this disability is unexplained, representing a barrier to effective treatment.4 Studies show that cognitive test performance is associated with aspects of function in daily life after TBI, including independence and return to work.5 6 However, the relationship between cognitive impairment and everyday functioning is incompletely understood.

    It is often assumed that cognitive impairment will have a strong influence on overall functional outcome; however, reported associations are typically modest.6 Chaytor et al 5 found that cognitive test performance accounted for 20%–30% of outcome variance on the Functional Status Exam, which provides a multidomain evaluation of function. A systematic review7 of studies relating cognition to global functional outcome found that multiple dimensions of cognition were associated with the Glasgow Outcome Scale—Extended (GOSE), explaining 31% of the variance in outcome. However, these studies were relatively small, with a median sample size of 135 (range 37–334).7 As a consequence, there is little information concerning cognitive performance at different levels of functional recovery.

    Past work indicates that cognitive impairment is present in individuals who are unable to return to work, and is even greater in those that are dependent. However, the role of cognition in higher levels of functional recovery is unclear. Impaired performance on cognitive tests has been reported in individuals graded as Good Recovery on the GOSE.8 Possible explanations include unrecognised cognitive impairment identified by objective testing, poorly matched normative data for tests or use of a coarse global scale.

    We address these issues by studying the relationship of the GOSE to cognitive assessment in a cohort of patients who form part of the Collaborative European NeuroTrauma Effectiveness Research (CENTER-TBI) project (www.center-tbi.eu).9 Over 1500 patients had detailed cognitive assessment at 6 months, providing an opportunity for more fine-grained analyses than hitherto possible. The aim of our study was to better characterise the association between cognitive performance and global functional outcome, and to tackle key questions concerning this relationship.

    Methods

    Participants

    CENTER-TBI recruited 4509 patients to core data collection from 65 centres.9 Criteria for inclusion were: a diagnosis of TBI, clinical indication for a CT scan, presentation within 24 hours of injury and consent obtained.9 The study enrolled patients from emergency rooms (ERs), hospital admissions and intensive care units. Participants were only excluded if they had a severe pre-existing neurological disorder that would confound outcome assessments. The sample thus included patients with very mild injuries, as well as those at the most severe end of the spectrum. Analyses here were confined to patients aged 16 or older who had been assessed on the GOSE at the 6-month time point and had completed one or more computerised cognitive tests (online supplementary figure S1). The last criterion was included to select the main group that had contributed cognitive data.

    Supplemental material

    Procedure

    Demographic and clinical data were recorded during the acute stage. A composite baseline Glasgow Coma Scale (GCS) was created using assessment at the time of ER discharge as the preferred measure, and where that was not available, working progressively earlier in time. The first CT scan after injury was used to identify whether imaging abnormalities were present.

    All patients were scheduled for follow-up at the 6-month time point, which is the focus of the current study. When translations of assessments were not available from the publisher, the material originally in English was translated into local languages using a process of linguistic validation based on guidelines.10 Patients agreeing to neuropsychological assessment were seen face-to-face. Assessors recorded completion codes for cognitive assessments, which included reasons for non-completion.11

    Outcome assessments

    Cognitive performance

    The battery consisted of the Rey Auditory Verbal Learning Test (RAVLT), the Trail Making Task (TMT) parts A and B and six subtests from the Cambridge Neuropsychological Test Automated Battery (CANTAB). The CANTAB tasks are language independent, facilitating multinational use. Details of the assessments are provided in table 1 and in online supplementary methods. Cognitive tests covered areas known to be affected by TBI, including memory and learning, processing speed, attention and aspects of executive functions. Procedures for conducting testing were specified in a study manual, and assessors were given face to face training in testing.

    View this table:

    • View inline

    Table 1

    Cognitive tests included in the Collaborative European NeuroTrauma Effectiveness Research traumatic brain injury battery, the domains that they primarily assess, and the specific measures that were used in analyses

    Global functional outcome and health-related quality of life

    GOSE: the GOSE was assessed either as a structured interview12 or a questionnaire completed by the patient or carer.13 A composite GOSE was created by scoring both the interview and the questionnaire centrally, and combining the ratings, with the interview as the preferred source (94% of assessments). Outcome categories on the GOSE are: upper good recovery,8 lower good recovery,7 upper moderate disability,6 lower moderate disability,5 upper severe disability4 and lower severe disability.3 Criteria used to assign the categories are detailed in online supplementary table S1.

    Short-Form-12v2 (SF-12v2). The SF-12v2 is a 12-item health-related quality of life assessment completed by the patient.14 The Physical Component Summary provides an overall measure of global functional outcome, while the Mental Component Summary assesses outcome related to aspects of mental health. When the SF-12v2 had not been completed, but the Short Form-36v2 was available, we used the 12 corresponding items from the latter to derive summary scores. Outcomes are expressed as T-scores (standardised using the normative sample to a mean of 50 and SD of 10).

    Statistical analyses

    A reference group was created by dividing patients who reported complete recovery (GOSE 8) into two groups: GOSE 8a consisted of patients with GCS=15 at recruitment and no abnormality on early CT, while GOSE 8b consisted of remaining patients with upper good recovery. The reference group was GOSE 8a, and our expectation was that individuals in this group would be at the very mildest end of the spectrum of TBI severity, and therefore least likely to have persisting cognitive impairment. These 'ultra-mild' patients are matched to the whole group with respect to the experience of the clinical processes associated with TBI. As a check on the reference group, previously published normative data were used to calculate expected mean scores allowing for the age distribution of the patient group. Demographic and clinical characteristics of the GOSE 8a and GOSE 8b subgroups were compared using χ2 for categorical variables and t-testing for age.

    Distributions of scores on individual tests were inspected, and a log 10 transform was applied to reduce unequal variance in measures from the following: Trail Making A and B, Choice Reaction Time (RTI), Paired Associate Learning (PAL), Rapid Visual Processing (RVP), Spatial Working Memeory (SWM) and Attention Switching Task (AST) latency. Scores from the RAVLT and Stockings of Cambridge (SOC) were not transformed. To provide a common metric for tests, cognitive measures were converted to z-scores using the mean and SD of the reference group (ie, group GOSE 8a). Scores were coded so that negative values indicated poorer performance than reference. A composite cognition score was calculated by averaging z-scores across tests, when six or more cognitive outcomes were available. Scores that were one or more SDs below the reference were considered to indicate at least borderline cognitive impairment.

    GOSE categories were compared using one-way analysis of covariance (ANCOVA), adjusting for age, sex and level of education. ANCOVA is a linear model, with test score as the dependent variable. Missing values for level of education were imputed using the mice function in R to generate 10 datasets.15 Outcomes were included in the imputation process, but imputed values of outcomes were not used in subsequent analyses. Pooled estimates for adjusted means and 95% CIs were derived from the ANCOVA. CIs are provided to aid interpretation of the graphed results: differences between means are considered to have p<0.05 when the CIs overlap less than 50%, and p<0.01 when they do not overlap.16 The CIs do not make any adjustment for multiple comparisons. Pooled F-values for the omnibus comparison from the imputed datasets were obtained using the mi.anova function in R. Controlling for the family-wise error rate (FWER), the significance threshold is (0.05/15) 0.0033. Effect sizes from the ANCOVA are partial eta-squared (ηp 2), with a conventional interpretation 0.01=small, 0.06=medium and 0.14=large. We conducted 15 pairwise comparisons (12 cognitive tests and 3 composites) of the adjusted means of outcomes for groups GOSE 8a and GOSE 8b, and controlled the results for a 15% false discovery rate using the Benjamini and Hochberg procedure.17 In contrast to FWER adjustment, this procedure corrects for multiple comparisons where differences on specific individual tests are not critical to the overall conclusion.

    Data were collected on an electronic case report form (Quesgen Systems, USA), hosted on the International Neuroinformatics Facility (INCF) platform and extracted via INCF Neurobot (INCF, Sweden). V.2.1 of the CENTER-TBI database was downloaded on 10 November 2019, and analyses were conducted using SPSS V.25.0 and R V.3.5.1 (R Foundation for Statistical Computing, Vienna, Austria).

    Results

    Assessment of cognitive performance at 6 months was carried out in 1554 patients out of 2604 survivors with a GOSE follow-up. Cognitive test completion rates were influenced by GOSE category (see online supplementary table S2): patients with GOSE 3 had the lowest completion rates (8%–21% across cognitive tests), followed by patients with GOSE 4 (30%–57%), while completion rates for patients with higher GOSE outcomes were substantially greater, and generally around 70% (GOSE 5 59%–74%, GOSE 6 65%–81%, GOSE 7, 62%–79% and GOSE 8, 53%–69%). In the GOSE 3 category, a common reason given for non-completion was the presence of cognitive or neurological deficits (online supplementary table S2). Over all categories, logistical reasons and patient availability were most commonly cited for non-completion. Since the absolute number of individuals tested in the GOSE 3 category was small (N=31), GOSE 3 and 4 were combined into one category of patients with severe disability (GOSE 3/4, N=115).

    Demographic and clinical information concerning the study sample is given in table 2, and corresponding information for the whole sample and the sample alive and eligible for assessment at 6 months is provided in online supplementary table S3. Compared with the non-study group, the study sample was more likely to have progressed to a higher level of education, and was more likely to be working and partnered before injury. The study group was also less severely injured by GCS and Abbreviated Injury Scale (AIS) criteria than the non-study group, and had better outcomes on the GOSE. Patients who had completed some cognitive assessment are therefore a selected subgroup of the eligible sample, particularly in relation to injury severity and outcome.

    View this table:

    • View inline

    Table 2

    Demographic and clinical characteristics of the study sample, and the two subgroups with Glasgow Outcome Scale—Extended (GOSE) 8 (upper good recovery)

    Information concerning the subgroups with upper good recovery (GOSE 8a and GOSE 8b) is given in table 2. There were significant differences between subgroups on variables reflecting severity of injury (baseline GCS, care pathway, imaging abnormality and both head and neck and non-head and neck AIS scores). On demographics, differences for employment history reached significance, but not for age, sex, race, level of education or marital status.

    Raw scores for assessments of the reference group were compared with norms predicted from the healthy population (see online supplementary table S4). For some cognitive tests, the reference group scores are almost identical to the predicted scores (Trail Making A and B, RTI Choice reaction time, Stockings of Cambridge (SOC) problems solved), while for others they are better (RTI movement time, RVP latency, SWM Between Errors) or worse (PAL total errors, RAVLT, RVP A prime). Overall, there are no systematic trends that would indicate cognitive impairment in the reference group. The SF-12v2 summary score means for the reference group were close to norms.

    Percentages of scores one SD or more below the reference group mean are given in table 3. The scores are based on z-scores of transformed variables and have not been adjusted for covariates. The expectation from the normal distribution is that 16% of scores will be one SD or more below the mean. As can be seen, tests for the reference group conform closely to expectation (11%–20%), while the percentages of clinically significant scores for the most disabled groups (GOSE 3/4 and GOSE 5) are always greater than 16% (20%–59%).

    View this table:

    • View inline

    Table 3

    Number (percentage ±95% CI) of standardised test scores 1 SD or more below the mean for the reference group (GOSE 8a)

    The results of omnibus comparisons of outcome categories adjusting for age, sex and level of education are given in table 4. Estimated means and 95% CIs for each measure are shown in figure 1, after adjusting for covariates. As can be seen, there are differences in cognitive performance across groups for all measures. Overall effect sizes were small to medium (table 4), and greatest for tests involving processing speed (RTI decision time and Trail Making Test A and B) and learning and memory (PAL and RAVLT), followed by tests of sustained attention (RVP) and attention switching (AST). The smallest effect sizes were observed for two tasks assessing executive functions (SWM and SOC).

    View this table:

    • View inline

    Table 4

    Summary of ANCOVA for the overall difference across Glasgow Outcome Scale—Extended categories adjusted for age, sex, and education level

    Figure 1

    • Download figure
    • Open in new tab
    • Download powerpoint

    Figure 1

    Glasgow Outcome Scale—Extended (GOSE) outcome categories and cognitive performance. Means and 95% CIs of z-scores. PAL, Paired Associate Learning; RAVLT, Rey Auditory Verbal Learning Test; TMT A, Trail Making Part A; TMT B, Trail Making Part B; RTI Dec, Choice Reaction Time decision time; RTI Mov, Choice Reaction Time movement time; RVP A, Rapid Visual Processing A prime; RVP Lat, Rapid Visual Processing latency; AST Cor, Attention Switch Task correct; AST Lat, Attention Switching Task latency; SWM, Spatial Working Memory; SOC, Stockings of Cambridge.

    Figure 1 indicates generally monotonic associations between cognitive test performance and order of outcome categories. Lower cognitive performance was particularly evident in patients who were dependent (GOSE 3 or 4) or who were unable to participate in one or more major life activities (GOSE 5). As can be seen from figure 1, these two groups consistently have poorest performance. On many measures (PAL, RAVLT, TMT A and B, RVP latency, RTI and AST), there are clear differences between GOSE 5 and GOSE 6 (ie, the CIs do not overlap). In contrast, performance for the groups with GOSE 6, 7 and 8 was surprisingly similar across categories (ie, many of the CIs overlap by 50% or more).

    Overall effect sizes from ANCOVA for the three summary measures were medium to large: 0.111 for cognition (F=36.98, df 5, 36 425, p<0.0001), 0.252 for the physical health summary (F=100.74, df 5, 11 0351, p<0.0001) and 0.131 for the mental health summary (F=45.22, df 5, 2 680 243, p<0.0001). Figure 2 displays z-score differences from group GOSE 8a.

    Figure 2

    • Download figure
    • Open in new tab
    • Download powerpoint

    Figure 2

    Contrasting relationships with functional status. Data are differences (z-scores) from reference (dashed line) graphed as means and 95% CIs. The Short-Form-12v2 (SF-12v2) Physical Component Summary (Physical) and the Glasgow Outcome Scale—Extended (GOSE) are in good agreement, as indicated by the near straight line relationship. The cognition composite (Cognition) strongly separates lower levels of outcome, but not upper; although some cognitive impairment is present even among patients reporting complete recovery. The SF-12v2 Mental Component Summary (Mental) does not distinguish between the lowest categories, but strongly differentiates more favourable outcomes. The shaded area covers data points where test completion rates were relatively low and additional caution is required concerning interpretation.

    Inspection of table 3 and figure 1 suggests small but systematic deficits in cognitive performance in GOSE group 8b compared with 8a. The following pairwise comparisons between these groups met the Benjamini and Hochberg criteria (raw p values in brackets): Trail Making Part A (p=0.007), cognition composite (p=0.013), RAVLT (p=0.027), AST correct (p=0.042), RTI reaction time (p=0.050), RVP latency (p=0.067), Trail Making Part B (p=0.077) and SWM between errors (p=0.079). There are thus systematic differences in cognitive performance between groups 8a and 8b. In contrast, the groups had similar the physical and mental health summary scores (p=0.417 and 0.956, respectively).

    Discussion

    Cognitive impairment is believed to be a key driver of disability, but important gaps remain in our understanding of this relationship. The current study provides two key pieces of information about this relationship: (1) it identifies cognitive domains most strongly related to function in daily life, and (2) it establishes where cognitive deficits most impact on difference between levels of function.

    Prior to discussing the main findings, it is appropriate to consider the issue of missing data.18 Much past work on cognition in TBI has used test completion as an inclusion criterion (either explicitly or implicitly), and thus provides little information on completion rates. In clinical trials that have included cognitive outcomes as an endpoint, completion by around half of patients recruited has been reported.19 20 We found that completion of cognitive assessments was strongly related to outcome on the GOSE, a result that echoes a report by Clifton et al.21 In their group of 110 cases included in the Traumatic Coma Data Bank, there was 6-month cognitive assessment in 60% of patients with good recovery while only 6% of patients with severe disability were tested. Explanations include cognitive impairment that prevents testing and logistical factors relating to availability of patients who are dependent.19 Particular caution is therefore needed concerning interpretation of findings for the severely disabled category (as indicated in figure 2). The comparison of completers and non-completers suggests other factors, such as level of education, play a role in follow-up but the differences here appear relatively minor.

    Cognitive domain and function in daily life

    The results provide novel insight into cognitive test performance at different levels of disability, and highlight the particular importance of processing speed in function in daily life. TMT part A and RTI decision time are both measures of processing speed and were among tests showing the greatest differences between outcome categories. TMT part B adds task switching demands, but this did not appear to make the task more sensitive. We conclude that processing speed has the strongest overall relationship with functional outcome in current analyses. In addition to processing speed, learning and memory and aspects of attention were related to functional outcome, while two measures of executive function showed the smallest differences across categories of outcome. This cognitive profile replicates that reported by Salmond et al 22 and may reflect cholinergic dysfunction. Impairment of processing speed and on tasks such as SOC implicates other neurochemical systems, including dopaminergic pathways.23

    Our findings are concordant with previous evidence that processing speed particularly influences functional outcome at 12 months post injury.24 Furthermore, Ponsford et al 25 found that slow processing speed was the area of cognition most strongly related to the GOSE 10 years after TBI, implying that this deficit has a long-term impact.26

    It is thought that slowing disrupts timing and synchrony of mental operations and has a general impact on cognitive function.23 Processing speed after TBI is thus a prime target for pharmaceutical interventions,23 27 and the current study supports the potential value of cognitive enhancers for improving function in daily life. The critical impact of processing speed in this context may be explained by the widely distributed neuroanatomical network which underpins it, and hence makes it vulnerable to diffuse pathologies.28–30

    Cognition and level of functional recovery

    The study allowed cognitive performance across levels of disability to be examined in greater detail than previously. The most disabled groups showed clearest evidence of poor cognitive performance. In addition, as already noted, many severely disabled patients did not complete testing, consistent with profound cognitive impairment in this group. There were also clear decreases in performance in patients with GOSE 5. Individuals in this category are unable to participate in one or more major areas of activity, such as work or social and leisure activities. Cognitive assessment in these cases can identify particular barriers to functional recovery, and areas that can be targeted by rehabilitation.31

    A surprising finding is that cognitive performance is similar across higher levels of functional recovery (GOSE 6–8b), explaining why the amount of variance in functional outcome accounted for by cognition can be modest.5 7 In these groups, decreases in performance are small, and any evidence of impairment limited to a minority. It appears to have been assumed in the past that the relationship between the GOSE and cognition is essentially linear. It has been suggested, for example, that cognitive testing could help to improve the granularity of upper categories of functional outcome.32 However, our findings argue against such a conception of the relationship.

    The comparison of group GOSE 8b with the reference group shows that some cognitive impairment is present even in patients reporting complete recovery. The influence of cognitive impairment on function may be attenuated by active compensation.33 For example, the person with cognitive slowing may adapt by taking greater time and effort to complete tasks, show increased compensatory recruitment (eg, of the prefrontal cortex) in processing speed tasks,34 and report more fatigue,35 in keeping with greater effort. Subtle changes in cognition are thus still likely to be consequential, and could, for example, be a source of stress in demanding work settings. In the long-term cognitive decline will reduce reserve, and may make the individual vulnerable to the effects of degenerative illness in later years.36 Formal cognitive assessment is valuable to establish whether impairment is present, even in individuals who have apparently recovered well.

    Diffuse white matter changes are a key neuropathological substrate for both cognitive deficits and poor daily life outcomes.37 Newcombe et al 38 found that diffusion tensor imaging abnormalities were related to the GOSE, and report findings qualitatively very similar to the trajectory shown for cognitive impairment in figure 2. We therefore hypothesise that cognitive impairment is a major mediator of the relationship between diffuse white matter damage and poor functional outcome after TBI.

    Comparison of cognitive, physical and mental health outcomes

    In contrast to the cognitive outcomes, there are clear differences in mental and physical health-related outcomes at the upper levels of recovery. Comparison with the SF-12v2 physical health summary argues against the idea that the relative coarseness of the GOSE explains absence of cognitive differences at upper levels of recovery: participants with GOSE 8a and GOSE 8b are close to the healthy norm, and thus unlikely to include a substantial subgroup with functional limitations. Furthermore, there are clear differences in SF-12v2 mental and physical health outcomes across GOSE categories 6–8b, and no overlap that would explain the similarity in cognitive performance between these groups.

    Overall, the findings indicate contrasting relationships of cognition and mental health with functional outcome after TBI. Cognitive impairment appears to be a key influence on disability at lower levels of outcome. On the other hand, differences in aspects of mental health are prominent at upper levels of recovery,8 and seem likely to play a role in whether the person achieves a complete recovery. Although relationships between function in daily life and mental health will be bi-directional, understanding mental health problems provides a clear focus for treatment which may also improve functional outcome.

    The physical health component of the SF-12v2 tracks outcome across GOSE categories, confirming a close relationship between these assessments of global functional outcome.39 It is also likely that aspects of physical health, such as continuing effects of extracranial injury, are drivers of overall functional outcomes, particularly in individuals with mild injuries.40

    These findings are important in the context of developing multidimensional outcome measures which characterise outcome beyond the GOSE.41 Our data indicate that assessments of cognitive performance and psychological health are likely to have distinct contributions across different parts of the outcome spectrum. The findings thus strongly support separate evaluation of cognition, mental health, and function in daily life after TBI as multidimensional descriptors of outcome.

    Limitations and future directions

    The reference group was at the very mildest end of the spectrum but may nonetheless have included individuals with impaired cognitive performance, and this would result in underestimation of the extent of impairment. Use of additional markers of brain injury, such as post-traumatic amnesia, MRI, or biomarkers could help to define a reference group. The use of an internal comparison group provides a strong control for factors associated with TBI, including variables not explicitly measured such as accident trauma and aspects of TBI care. While acknowledging that it is a conservative approach, the comparison provides even greater confidence in the differences found.

    As already discussed, a further limitation of the study concerns missing outcomes. Since this inevitably includes some individuals with the greatest cognitive impairment, the effect will be to lead to underestimation of cognitive impairment in the most disabled categories of outcome.

    To our knowledge, this is the first study of cognition in TBI large enough to describe in detail performance at different levels of functional outcome. Unexpectedly, we found little difference in cognitive performance at upper levels of outcome. This raises two key issues for future research. The first concerns the drivers of outcome and whether emotional and mental health factors play a key role at the upper end of functional recovery.42 The second relates to the nature and significance of cognitive impairment in these patients. The instruments we used address cognition without the contextual stresses of everyday life. Further, they do not account for increased cognitive fatigability reported by many patients, and provide only limited measures of the effortfullness of these tasks,43 which have identifiable neuroanatomic correlates.44 Tests of 'hot cognition'45 and measures of effortfullness and fatigue may bring additional value to assessments, and identify potential targets for therapy.

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      • Data supplement 1