Introduction
Most people experience a full spontaneous recovery after a mild traumatic brain injury (mTBI; Moore, Terryberry-Spohr & Hope, Reference Moore, Terryberry-Spohr and Hope2006; NIH, 2002); however, a proportion of people go on to develop chronic problems post-mTBI (Giskiewicz, Marshall & Bailes, Reference Giskiewicz, Marshall and Bailes2007; McMahon et al., Reference McMahon, Hricik, Yue, Puccio, Inoue, Lingsma and Investigators2014; Mooney & Speed Reference Mooney and Speed2001). One of the most common areas in which such chronic issues develop is in self-reported cognition, particularly in terms of processing speed and executive functioning (Holsinger et al., Reference Holsinger, Steffens, Hillips, Helms, Havlik, Breitner and Plassman2002; Koponen et al., Reference Koponen, Taiminen, Portin, Himanen, Isoniemi, Heinonen and Tenouvo2002). A systematic review of cognitive impairments at least 6 months after traumatic brain injury (TBI) (Wood, Reference Wood2004) found clear evidence of long-term cognitive deficits after severe or moderate TBI. However, the evidence for whether long-term cognitive deficits were evident after mTBI was insufficient.
In a systematic review of meta-analyses focused on the cognitive sequelae of mTBI (Karr, Arechenkoff & Garcia-Barrera, Reference Karr, Arechenkoff and Garcia-Barrera2014), it was concluded that studies had consistently identified adverse cognitive effects that were acute and rapid to resolve, with executive functioning appearing to be the most impacted by injury. Another report looking at more long-term sequela indicated that up to 39% of individuals complained of cognitive deficits at 6 months post-mTBI (Theadom et al., Reference Theadom, Parag, Dowell, McPherson, Starkey, Barker-Collo and Feigin2016). However, in this relatively small study (n = 79) of emergency department admitted individuals with mTBI, it was concluded by Theadom et al. (Reference Theadom, Parag, Dowell, McPherson, Starkey, Barker-Collo and Feigin2016), that self-reported cognitive complaints may be related to pre-morbid traits and physical and emotional state factors rather than to actual cognitive impairments. Inconsistencies in findings suggest that any assessment of cognition in this population should include objective measures of cognitive abilities, as well as an assessment of emotional functioning.
In our population-based study (Barker-Collo et al., Reference Barker-Collo, Jones, Theadom, Starkey, Dowell, McPherson and Group2015), memory, processing speed, executive function, psychomotor speed/reaction time, complex attention and cognitive flexibility all improved significantly over the first year after mTBI. At baseline, over 20% of individuals produced very low scores on executive ability, complex attention and cognitive flexibility. At 1- and 6-month follow-ups, over 20% of participants were very low for complex attention, with 16.3% remaining so at 12 months. Thus, while significant improvements occurred over time, a significant proportion of individuals still performed poorly on neuropsychological measures at 12 months. While no pre-morbid contrast data were available, the proportion of those within the very low ranges was well in excess of that expected within the population (i.e., <1%). This implies a longer trajectory for recovery than has previously been suggested, which has implications for provision of assessment and rehabilitation services for more extended periods. These findings suggest the need to examine trajectories for recovery beyond 12 months after mTBI.
To summarise, the evidence supports arguments that there are likely to be a greater proportion of people than expected that have poor cognitive outcomes after mTBI, and that recovery of functioning can occur later than 3 months post-mTBI. Further, the above suggests that it may be possible to identify and perhaps treat people at risk of poorer late outcome who exhibit characteristics such as emotional state factors linked to poor outcomes.
In the present study, we examined the trajectories of recovery for cognition over an extended time frame, including data collected 4 years post-injury. In addition to presentation of group means and standard deviation at each time of assessment, a visual representation of the flow of participants overall cognitive functioning across the assessments is presented using a Sankey diagram, as well as growth curve modelling of repeated measures to measure the average trajectory of the individual data. Sankey diagrams provide a visual representation of an individuals’ performances within categories (e.g., average, above average) and how these shift over time between categories. In this study, a key assumption of regression analysis (i.e., that data points measured are independent from each other and not repeated measures from the same individual) is violated. As such, growth curve modelling was selected a priori, as it allows for repeated measurement in the same individual. Multilevel models (hierarchical models) have the ability to capture the correlation between responses in clusters allowing for better estimation of standard errors. When time is used as an explanatory variable to describe the change in the outcome, this is known as a growth model.
This form of modelling in the present context allowed us to examine hypotheses that: (1) While the majority of individuals will show recovery in neurocognition over the first 6 months post-injury, others will continue to show recovery long term; and (2) The rate of recovery will be negatively impacted by increased age at injury, female gender, and presence of mood disorder (i.e., depression and anxiety).
Methods
Study sample
The sample was drawn from a larger population-based incidence and outcomes study, the Brain Injury Incidence and Outcomes In the New Zealand Community (BIONIC) study, whose methods are provided in detail elsewhere (Feigin et al., Reference Feigin, Theadom, Barker-Collo, Starkey, McPherson, Kahan and Ametatunga2013; Theadom et al., Reference Theadom, Barker-Collo, Feigin, Starkey, Jones, Jones and Barber2012). TBI was defined according to World Health Organization criteria as an acute injury to the brain due to mechanical energy to the head from external physical forces (Carroll et al., Reference Carroll, Cassidy, Peloso, Bog, van Holst, Holm and Pepin2004). TBI was operationally defined as the presence, at the time of injury, of one or more of: (1) confusion or disorientation; (2) loss of consciousness; (3) post-traumatic amnesia (PTA); and/or (4) other neurological abnormalities (e.g., focal signs, seizure; Carroll, Cassidy, Holm, Kraus & Coronado, Reference Carroll, Cassidy, Holm, Kraus and Coronado2004). These symptoms needed to be primarily related to the TBI and not other causes (e.g., substance use, other injury, psychological/medical conditions). It should be noted that of the final sample (n = 301), 161 had no PTA recorded at the time of injury, only 1 individual had PTA in hours, and the remainder had PTA in minutes, which is expected given the sample had only mild injury. Each TBI diagnosis was confirmed by review of medical records (or clinical details) for each patient by the study Diagnostic Adjudication Group.
TBI severity was defined according to standard definitions using the Glasgow Coma Scale (GCS) and/or Westmead Post-traumatic Amnesia scale (Teasdale & Jennett, Reference Teasdale and Jennett1974), where mild injury was defined as a GCS score of 13–15; and PTA was not greater than 24 h. Classification of mechanisms of TBI was guided by the International Classification of Disease (ICD-10) classification system and included injuries due to falls, motor vehicle accidents, assaults, exposure to external forces (e.g., being hit/struck by another person/animal/ inanimate object) and other causes (including unspecified causes).
Case ascertainment included both prospective and retrospective surveillance systems to ensure registration of all events in residents in the study area. It included all ages and all severities over a 12-month period (from 1 March 2010 through 28 February 2011). Multiple overlapping sources of information on all new hospitalised and non-hospitalised cases were used (see Theadom. et al., Reference Theadom, Barker-Collo, Feigin, Starkey, Jones, Jones and Barber2012 for detail). The study was approved by the Northern Y Regional Ethics Committee of New Zealand.
Present sample
The present analyses included all adults who had experienced a TBI and been identified through the BIONIC study who: (1) were aged 16 years or older at the time of injury; (2) whose injury was classified as mild severity (i.e., Glasgow Coma Score of 13–15; PTA ≤ 24 h); (3) provided written informed consent to participate in follow-up assessments; and (4) provided data on at least two follow-up time points (as this is required for the analysis). Figure 1 shows the flow of cases from the BIONIC population-based study for the present analysis.
Flow chart showing recruitment into the current sample.
Assessment
Assessments included measures of overall recovery following injury (i.e., measures of cognitive and neurobehavioral impairments, PTA, disability, health-related quality of life and community integration). The full assessment took, on average, 2 h to complete and was completed across two or more sessions if required. A trained research assessor conducted all assessments. Assessment data examined in this analysis were collected at baseline (within 2 weeks of injury), 1, 6, 12- and 48 months post-injury. Measures of interest for the present analysis examining neuropsychological functioning after TBI included injury-related characteristics (e.g., severity, mechanism, intentional vs unintentional injury); demographic characteristics (e.g., gender, ethnicity, age at injury); as well as neuropsychological functioning at each assessment.
Neuropsychological functioning was assessed using the CNS-Vital Signs (CNS-VS) (Gualtieri & Johnson, Reference Gualtieri and Johnson2006) computerised test of cognitive functioning. It is a short (30 min), well-validated (Gualtieri & Johnson, Reference Gualtieri and Johnson2006; Gualtieri, Johnson & Benedict, Reference Gualtieri, Johnson and Benedict2004) and self-administered. Scoring is automated, eliminating variability and rater bias. It was used to assess verbal and visual memory, psychomotor speed, reaction time, processing speed, executive functioning, complex attention and cognitive flexibility. Raw test scores were converted to domain scores using age-adjusted norms; these are normally distributed and have a mean of 100 and standard deviation of 10. An overall Neurocognitive Index score was calculated as the average of five domain scores: composite memory, psychomotor speed, reaction time, complex attention and cognitive flexibility. The CNS-VS is designed for serial administration, with stimuli changing at each administration, and with test–retest reliability suggesting relative stablility over time (Gualtieri & Johnson, Reference Gualtieri and Johnson2006). Domain scores on the CNS-VS which fall below 70 are considered ‘very low’ and occur in 0.13% of the normative population, whilst scores of 70–79 are ‘low’ (2.14% of the normative population), 80–89 is considered ‘low average’ (13.59%), 90–109 is considered ‘average’ (68.26%) and, finally, scores above 109 are ‘above average’ (15.86% of the normal population). The CNS-VS test has been used in many clinical populations, including TBI, where it has been found to be related to both severity of brain injury and degree of recovery, with tests of psychomotor speed and cognitive flexibility being most related to TBI status. The subtests have good psychometric properties, with very similar reliability and validity to the conventional neuropsychological tests upon which they are based (Gualtieri & Johnson, Reference Gualtieri and Johnson2006). Patients who had recovered from mild brain injuries scored almost as well as normal controls (Gualtieri & Johnson, Reference Gualtieri and Johnson2008).
Hospital Anxiety and Depression Scale (HADS; Zigmond & Snaith, Reference Zigmond and Snaith1983).The HADS is a common test used to predict the likelihood of underlying anxiety and depression. It contains 14 statements (e.g., ‘I have lost interest in things’), 7 relating to anxiety and 7 to depression. For each statement, the participant chooses one of four responses (e.g., ‘yes, definitely’, ‘yes, sometimes’, ‘no, not much’ or ‘no, not at all’) considering the last week. A number of researchers have examined the HADS to establish cut-off points for caseness of anxiety and depression. In this study, a cut-off of 7 was used, as in a systematic review of a large number of studies, a cut-off of 7/21 was identified with specificity and sensitivity for depression of 0.79 and 0.83 and for anxiety of 0.78 and 0.90 (Bjelland, Dahl, Haug & Neckelmann, Reference Bjelland, Dahl, Haug and Neckelmann2002).
Statistical analyses
Prior to conducting the main analyses, the representativeness of the 301 individuals included in the analysis was examined through contrasting them to the 569 individuals from the initial sample (n = 870) who were not included in terms of demographic and injury characteristics reported in Table 1. This involved independent t-tests of continuous variables (e.g., age) and chi-square analyses for categorical variables (e.g., ethnicity, gender, mechanism of injury).
Characteristics of the sample with CNS data at each assessment and the overall sample of adults with mTBI
MVA, motor vehicle accident.
a Note mechanism of injury was missing for 31 individuals not included in the analyses.
The first stage of data analysis then examined cognitive performance in the sample at each assessment (within 2 weeks of injury, and at 1-, 6- 12- and 48-month follow-up). To better visualise the trajectories for overall cognitive performance over time after injury, a Sankey diagram was generated showing the proportion of individuals falling within each qualitative category (i.e., average, low average, or whose data were missing) at each time of measure as well as the flow between labels over time. Available data on recurrent injury and mood (anxiety/depression) were then examined to reflect on potential contributors to negative recovery trajectories.
A basic growth model assumes that the effect of time (slope effect) is linear (i.e., one unit increase in time always predicts the same influence on the outcome), but this may be too restrictive. For instance, change may be evident at first but slow down later in time. In this study, we included a polynomial to account for the change in slope effect, using a piecewise polynomial function called a spline. Splines allow us to identify separate time intervals where separate linear slope effects can be calculated. To be included in this analysis, all adult patients had to have responded in at least two follow-up interviews. In relation to neurocognitive recovery over time, each outcome measured was modelled using a growth curve model. We modelled the effect of time in a mixed effects model, with random intercept and slopes using an unstructured covariance matrix, to take into account the correlated results of repeated measure data. We allowed for nonlinear trends over time by estimating a spline model, identifying three time intervals (i.e., baseline 6 months, 6–12 months and 12–48 months). This allowed us to estimate three different linear trends, one for each time period. Interactions were also estimated to investigate whether people with certain characteristics recovered faster than others. A significance level of p < 0.05 was used. Stata version 14.0 was used to conduct analyses.
Results
Sample description
Table 1 presents demographic and injury characteristics of the sample at each assessment time point, as well as that for the overall sample (n = 870) and the included sample (n = 301). As seen in the table, characteristics of the sample at each time of assessment generally reflected the wider sample of adults with mild injury. The majority of participants were male and of European ancestry, with the majority of injuries caused by falls except at 6 months where this was mechanical forces. Alcohol was involved in 25–29% of injuries and drugs 5.9–8.8% of injuries. Between 22 and 26% of injuries were intentional at any given time of measure.
In examining representativeness of the included sample, those included in the study (n = 301) did not differ from those from the initial sample of all adults who had experienced mTBI who were not included (n = 569) in terms of age, intentional injury, gender distribution, distribution of ethnicities or in the distribution of the mechanisms of injury (p > 0.05).
An additional factor which may impact changes in cognition within the sample is presence of additional injuries during the course of the study. Of the 25 individuals who recorded another head injury after the incident injury, 13 (52%) had experienced a single additional head injury over the course of the study, 24% (n = 6) had experienced two such injuries and 6 (24.0%) had three or more such injuries. Data were also collected for depression and anxiety at each time point, and this is presented in Table 2.
Performance on the HADS for those with CNS data at each time of assessment
a Not all those with CNS data at a particular time point had HADS data for that time point. At baseline, 1 month and 48 months HADS data were not available for one person each, whilst at 6 months, two individuals did not have HADS data.
Change over time
Table 3 presents the means and standard deviations across the CNS-VS domains for each time of measurement. Examining mean performances might mask individual trajectories for recovery. Thus, to look further at trajectories for their presence and fluctuations in the proportion of individuals falling within qualitative ranges of functioning (e.g., above average, average) over time in overall cognitive functioning (Neurocognitive Index Score), growth curve modelling was conducted followed by a Sankey diagram to show the movement of participants across groups of those whose scores fell into each qualitative range, and those who did not provide data at a given time of assessment.
Distribution of performance of the sample across CNS-VS domains at each time of measurement
Using growth curve modelling at baseline, overall neurocognition was estimated to be within the low-average level at a mean of 85.60 (95% CI 83.0–88.20) (Fig. 2). We found significant improvements in overall neurocognition from baseline to 6 months, on average adults improved monthly by 0.90 (95% CI 0.42–1.39) p < 0.001 (Table 4), resulting in the mean at 6 months to lie within the average level. No change in overall neurocognition was found within the time periods 6–12 months and 12–48 months.
Visual representation of linear trends across CNS-Vital Signs categories for the neurocognitive index over time.
Growth curve model for monthly change in neurocognition over 48 months
In relation to the CNS subcategories, most improvements occurred within the first 6 months. Five of the nine subcategories had significant improvements from baseline to 6 months: processing speed, executive, psychomotor speed, reaction time and mental flexibility. Among these subcategories, mean monthly improvement ranged from 0.61 to 1.57. From 6 to 12 months, processing speed was the only subcategory where we found continual improvements. From a monthly rate of 0.61 (95% CI 0.27–0.96) between baseline and 6 months, to 0.47 (95% CI 0.14–0.80) between 6 and 12 months. No significant change occurred in processing beyond 12 months. There was no interaction between time and age at injury, gender, depression or anxiety suggesting that based on these characteristics, people of these different characteristics tended to recover at the same rate.
As seen in the Sankey diagram in Fig. 3, a large proportion of individuals performed within the low or very low range at each time point (22.1% at 1 month, 16.9% at 6 months, 19.7 at 12 months and 15.0% at 4 years). In terms of trajectories across the baseline to 1-month period, the majority of individuals remained within the same performance category or improved. From 1 to 6 months, we see again a large proportion of individuals remaining within the same performance category or improving, though with a large proportion also moving out of the study, into the ‘missing’ category.
Sankey diagram showing trajectories for recovery of overall cognitive score over time, with tabulated data on frequencies within each node.
The majority of downward trajectories over the baseline to 6-month time frame represent individuals who moved into the study after having been initially ‘missing’ from the data set. Indeed, these constitute the majority of the downwards trajectories within the figure, and it must be remembered that they do not represent a decline in function. In regard to declines in function, at each time gap, a small proportion of those who were average lowered in performance to the low average, low and very low ranges; and from the 1-month assessment onwards, we saw a small number from the above-average category drop to the average range. It is not until the 12- to 48-month time gap where a much smaller proportion of individuals came into the study from missing, while a large proportion left the study. We also saw many remain within existing categories; whilst of those in the average range some decline to low average, low and very low, and those in the above range drop to the average range. Proportionally, few showed any improvement – with only a small proportion of those who were very low at 1 year improving.
Discussion
The findings of the present study were in line with the existing literature, which concluded that: (1) a greater proportion of people than previously thought have poor outcomes (15.0–22% in the low or very low ranges across assessments), with a small number of individuals showing declines in function more than 12 months post-injury; (2) the majority of individuals showed stable or improved function with the potential for recovery to occur later than 3 months post-injury; and (3) that it may be possible to identify and perhaps treat people at risk of poorer late outcome. Inconsistent with the literature and hypothesis, there was no interaction between time and age at injury, gender, depression or anxiety, suggesting that these characteristics had not impacted rate of recovery.
The results suggested that while cognitive recovery from mTBI is greatest within the first 6 months, recovery is quite variable across individuals. The findings from growth curve modelling replicate that reported in the literature, which suggests that the majority of significant change in cognition was improvement which occurred within the first 3–6 months post-injury (Barker-Collo et al., Reference Barker-Collo, Jones, Theadom, Starkey, Dowell, McPherson and Group2015; McInnes, Friesen, MacKenzie, Westwood & Boe, Reference McInnes, Friesen, MacKenzie, Westwood and Boe2017; Stulemeijer, van der Werf, Borm & Vos, Reference Stulemeijer, van der Werf, Borm and Vos2008). That only processing speed continued to improve until 12 months is not unexpected. Diffuse axonal injury (DAI) is a mechanism thought to underlie cognitive deficits after mTBI (Heilman & Valenstein, Reference Heilman and Valenstein2003). DAI can be extensive even in mTBI, and is linked to cognitive deficits (Heilman & Valenstein, Reference Heilman and Valenstein2003; Suh, Kolster, Sarkar, McCandliss & Ghajar, Reference Suh, Kolster, Sarkar, McCandliss and Ghajar2006). Both radiological and pathological studies have found diffuse brain damage in the frontal and temporal lobes, corpus callosum and fornices in those with mTBI and have linked these to deficits in information processing speed (Mathias, Beall & Bigler, Reference Mathias, Beall and Bigler2004). Deficits in information processing speed after mTBI are thought to cause other cognitive difficulties, either through reducing an individual’s speed of thinking or in reducing the amount of information that the individual can process concurrently (Bogdanova & Verfaellie, Reference Bogdanova and Verfaellie2012). Executive functioning is the area of function most implicated due to the abundance of white matter connections between the frontal lobes and other areas of the brain (Arfanakisa et al., Reference Arfanakisa, Haughtonb, Carewc, Rogersa, Dempseyd and Meyeranda2003). Executive functioning is complex, including many and varied abilities. As such its relationship to processing speed is also likely to be complex. It was observed here that while executive function did improve alongside processing speed, it is noted that scales for executive function and its components (mental flexibility and complex attention) are the only scaled scores whose means decreased over the course of the study.
Results of a review of the literature on mTBI (McInnes et al., Reference McInnes, Friesen, MacKenzie, Westwood and Boe2017) indicate that, in contrast to the prevailing view that most symptoms of concussion are resolved within 3 months post-injury, approximately half of individuals with a single mTBI demonstrate long-term cognitive impairment, highlighting the need to carefully examine the long-term implications of a single mTBI. The findings presented here suggest that significant improvement can occur after mTBI for up to 6 months post-injury, with continued significant improvement up to 12 months in the area of processing speed. The Sankey diagram further suggested that, while the numbers impacted are small, improvements can still be attained by those in the very low ranges of functioning from 12 to 48 months post-injury.
Additional statistical techniques such as Latent Class Analysis or Latent Transition Analysis would provide further statistical tests of whether or not there are subgroups of individuals who differ in their temporal patterns of cognitive function, but the present sample was too small to support these methods. Obtaining samples large enough to support use of these techniques will be an important challenge for future research.
Displaying individual trajectories visually across the major time points shows that the majority of individuals remain stable or improve over time, with only a small proportion of individuals declining, and these more evident in the 1- to 4-year time gap. However, what the Sankey diagram does add is the consideration of the flow of individuals in and out of the study. Thus, we propose using Sankey diagrams as a novel way to examine not only trajectories for recovery over time, but retention in such samples over time. In reflecting on the Sankey diagram, examining the patterns for the missing data is quite informative. For example, the greatest proportion of those who withdrew during the course of the study had been functioning within the average range. This was most noticeable from 6 to 12 months post-injury and then again from 12 to 48 months post-injury. That those in the average range were more likely to withdraw contrasts with common perceptions that individuals who are experiencing the greatest difficulties or no difficulties are most likely to withdraw from participation, with studies associating loss to follow-up with negative indicators such as unemployment, low socioeconomic status and injuries due to assault (Langley, Johnson, Slatyer, Skilback & Thomas, Reference Langley, Johnson, Slatyer, Skilback and Thomas2010). While these additional factors should be included in future examinations which use growth curve modelling due to their potential impact on trajectories, the Sankey visualisation suggests other factors may be at play. One additional factor not yet mentioned is performance validity, which has been defined by Larabee (Reference Larabee2012) as ‘the validity of actual ability task performance’ (p. 2). It is possible that this could explain those whose recovery is slow or who decline in ability over time, particularly in contexts where litigation is present. While this was not assessed in this study, it is a factor that could contribute to discrepancies in the literature about the trajectory of recovery (i.e., occurring acutely rather than over a longer trajectory). Performance validity issues are not likely to have impacted the present findings as these are not common in New Zealand’s non-litigious context (Barker-Collo & Fernando, Reference Barker-Collo and Fernando2015).
The greatest limitation of this study was the reduced sample size at some time points. This may have contributed to the presence of the null effects found when investigating demographic and health interactions. However, the growth curve benefits from not needing complete cases at all time points. Therefore, individuals who contribute fewer than the maximum number of observations can be retained in the analysis without further adjustment, assuming data are missing at random (i.e., missingness is associated with earlier outcomes). Furthermore, to test the model assumptions of normality and homoscedasticity, we plotted standardised residuals against normalised scores and standardised residuals against fitted values. There was evidence to suggest approximate normal distribution of residuals and equal variances. Having a better understanding of who is most likely to withdraw from a study could assist in future research through implementation of retention strategies targeting these individuals and time points where this is most likely to be an issue. For example, as emotional factors are thought to impact recovery and retention, readers interested in the 4-year trajectories for anxiety and depression from this sample are directed to Barker-Collo et al. (Reference Barker-Collo, Theadom, Jones, Starkey, Kahan and Feigin2018).
Conclusion
In the present study, half of the sample had some cognitive difficulty at 6 months post-injury, a direct contrast to the prevailing view that most symptoms of concussion resolved within 3 months of injury. In addition, significant improvement in cognitive functioning was found to occur after mTBI for up to 6 months post-injury, with continued significant improvement up to 12 months in the area of processing speed. This has implications for clinical practice, expanding the window of time during which spontaneous recovery might be bolstered with rehabilitative input. Inclusion of a Sankey diagram shows that negative recovery trajectories are also present, though for only a small proportion of individuals. Having noted this, it is important for researchers to determine what factors might place an individual at risk for a negative trajectory and further, to determine if therapeutic measures could be put in place to prevent declines in functioning.
The findings contribute to thinking about new approaches and methods that may be more effective in approaching the problem of identifying people with mTBI who have fluctuating and poor late outcomes. For example, we recommend that regular assessments of cognition take place to identify any declines in functioning immediately. We further recommend the tracking of additional factors, such as additional injuries, which could be useful in explaining the outcomes of those participants who decline. In planning and conducting future research, it will similarly be important to identify those whose outcomes fluctuate overtime, and to collect information that explains these changes, whether these be on the cognitive or other aspects of performance.
Financial Support
This work was supported by the New Zealand Health Research Council Project grants 09/063A; 11/192; and 13/408.
Conflict of Interest
Suzanne Barker-Collo has no conflicts of interest to disclose. Alice Theadom has no conflicts of interest to disclose. Kelly Jones has no conflicts of interest to disclose. Nicola Starkey has no conflicts of interest to disclose. Kris Fernando has no conflicts of interest to disclose. Michael Kahan has no conflicts of interest to disclose. Philip Prah has no conflicts of interest to disclose. Valery Feigin has no conflicts of interest to disclose.
Ethical Standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
