There were an estimated 348,453 hospital admissions due to acquired brain injury (ABI) in 2016–2017 in the United Kingdom (Headway, 2018). ABI can arise through “trauma, vascular accident (e.g., stroke), cerebral anoxia, other toxic or metabolic insult (e.g., hypoglycemia), infection (e.g., encephalitis) or other inflammation (e.g., vasculitis)” (Turner-Stokes, 2003, p. 14) and can result in physical, cognitive, communication and emotional difficulties (Wilson et al., 2009). Those who have sustained a traumatic brain injury often present with difficulties associated with frontal lobe function (McDonald et al., 2002), due to the size, structure and location of the frontal lobes making them particularly vulnerable (Levin et al., 1987; Cicerone et al., 2006). Research on the structure and function of the frontal areas of the brain highlights roles in cognitive, behavioral and emotional processes. These include flexible thinking, planning, monitoring, social behavior, decision making, initiation, inhibition and emotional regulation (Lezak, 1995). The term “dysexecutive problems” has been used to describe difficulties with these functions, which can have a profound impact on a person’s level of independence resulting in challenges in day-to-day life (Hanks et al., 1999).

There have been challenges in the clinical measurement of dysexecutive problems, with various approaches to their assessment. Neuropsychological tests can be time-consuming and when used for assessment of frontal functions can be difficult to interpret and lack ecological validity due to the structure and presence of cues in the testing environment (Eslinger and Damasio, 1985; Damasio et al., 1991; Burgess et al., 1998). As a result, traditional tests often fail to highlight difficulties in this area despite reports of challenges in day-to-day life (Stuss et al., 1983; Shallice and Burgess, 1991). To overcome this, self-report measures have been developed to capture challenges faced in everyday life, which might complement neuropsychological assessments (Isquith et al., 2013). A limitation of using these with people experiencing dysexecutive problems is the issue of reduced self-awareness, meaning they may be more likely to underreport such difficulties (Simblett et al., 2017). Informant versions are available to corroborate or assess the discrepancy compared to the self-report version. Different rating scales have been developed and are available for clinical use, however, there are issues in the standardization and interpretation of scores. This is because the nature of these difficulties, such as decision making, perseveration and flexibility, could lead to issues with the reliability of item responses.

The Dysexecutive Questionnaire (DEX; Burgess et al., 1998) forms part of the Behavioral Assessment of the Dysexecutive Syndrome (BADS: Wilson et al., 1996). The DEX is a self-report measure of dysexecutive problems, designed to predict everyday difficulties. There are 20 items measuring behavioral, cognitive, motivational and emotional changes from pre-morbid functioning generating in a single score.

Simblett and Bateman (2011) assessed the psychometric properties of the DEX using item response theory by deploying Rasch analysis techniques. Their analysis suggested the DEX not to be a unidimensional measure of dysexecutive problems, instead capturing underlying sub-constructs thought to underpin these difficulties. Therefore, a total score on self-report measures may not best capture these challenges. In further research, Simblett et al. (2017) made amendments to the wording of some of the items in the DEX as well as including an additional 14 items to expand its measurement to incorporate Stuss’s (2011) proposed categories of frontal functions. After applying Rasch techniques, data from a clinical sample suggested the revised version of the DEX mapped onto the Stuss model capturing four separate sub-constructs of executive cognitive functions, metacognition, activation/energization and behavioral and emotional self-regulation. This enhances clinical application by enabling specific areas of strength and difficulty to be highlighted which can assist in diagnosis, neuropsychological formulation or become a focus for a person’s individual rehabilitation plan. The DEX-R has received further psychometric validation when applied to healthy aging and mental health samples (Loschiavo-Alvares et al., 2013; Dimitriadou et al., 2018) although factor analyses with these groups did not align with Stuss’s model. Instead, these studies produced a three-factor structure more in line with Fuster’s (2008) theory of frontal lobe functioning (Social Self-Regulation, Motivation and Attention, and Flexibility, Fluency and Working Memory).

Neurological disorders are known to contribute to reports of dysexecutive problems, however, additional understandings of how individual variation may manifest is useful for clinicians to understand. Research has highlighted individual variations in reported levels of dysexecutive problems in non-clinical populations as measured by the DEX questionnaire (Chan, 2001) due to individual differences or demographics such as age and mood or mental health. Normal aging processes have been associated with a decline in various cognitive functions associated with prefrontal areas (West, 1996; Van Petten et al., 2004). This may be more prominent in cognitive changes related to the dorsolateral regions, with less change from normal aging being found in ventromedial areas thought to underpin the emotional processing aspects of dysexecutive problems (MacPherson et al., 2002). The later maturation of ventromedial areas could have implications in the social and emotional functions being less developed in younger people (Burnett et al., 2009; Pfeifer et al., 2011; Sebastian et al., 2011; Barkley, 2012; Coffman, 2014). Negative affect may mediate the increased reports of dysexecutive problems in younger people (Gerstorf et al., 2008). Correlations have been found between dysexecutive and anxiety and depression symptoms, which may relate to cognitive variability in mood (Shaw et al., 2015). Another factor contributing to variation in reports of dysexecutive problems includes education level (Foss et al., 2013; Faria et al., 2015). The presence of these other factors influencing frontal functions presents a further challenge to the interpretation of clinical neuropsychological assessments.

It would therefore be useful to establish levels of reports of dysexecutive problems in non-clinical populations with questionnaire measures such as the DEX-R, to further explore psychometric properties and contribute toward the establishment of a normative level or cut off for identifying clinical levels of dysexecutive symptoms.

The parallel analysis retained three factors. The principal axis factoring using the oblique rotation resulted in appropriate correlations for this rotation method to be applied (Pedhazur and Schmelkin, 1991; Tabachnick et al., 2007). The factor analysis had adequate sampling (Kaiser-Meyer-Olkin: 0.83) and correlation (Bartlett’s Test of Sphericity: < 0.01). The factor loading matrix is presented in Table 2. The three factors accounted for 42.2% of the total variance, with the first factor accounting for 31.1%. Items with factor loadings below 0.3 were excluded from reporting. The internal consistency of factor one was excellent, good for factor two and questionable for the third factor (Hermans et al., 2011). As there were only three factors, the results, therefore, do not align with the Stuss model of frontal functions. The development of the DEX-R found some items did not map onto any of the subscales, these were retained due to their clinical utility (Simblett et al., 2017). Therefore, despite not all items achieving factor loadings above 0.4, no attempts were made to purify the model to preserve this utility.

Seven of the items cross-loaded onto more than one factor. The 19 items loading onto factor one related to processes associated with the medial/dorsal domain. These items spanned across the proposed Stuss subscales, although mainly encompassed those from the activation subscale. These factors commonly share themes of initiation, maintenance and responsiveness, such as the ability to activate or inhibit a behavior or thought and was therefore labeled as “activation-self regulation.” The 17 items loading onto factor two appeared to relate to dorsolateral domains, typically these items represent cognitive dysexecutive symptoms such as planning, decision-making, abstract thinking, memory and attention. Although it was recognized that both blunted affect items additionally loaded onto this factor, and all seven cross-loadings involved this factor. The higher factor loadings mainly included those in the proposed executive cognition Stuss subscale. This factor was therefore labeled as “cognition.” Items loading onto factor three related to processes associated with the orbitofrontal areas, these also shared the blunted affect items. Additionally, more items corresponded to the Stuss behavioral-emotional self-regulation subscale. However, the items also appear to relate to social-self regulatory dysexecutive symptoms, therefore this factor was labeled “social-emotional.”

The responses on the DEX-R did not show fit to the Rasch model which suggests it measures more than one subconstruct [χ²(74, N = 140) = 205.54, p < 0.001]. Many of the questions showed disordered thresholds, and scale responses were not all endorsed on items, therefore it was not possible to confirm the interval nature of the scale as these have not yet arrived at a stable solution. Due to the small number in the clinical groups, we could not compare the level of endorsement or differential item functioning.

The DEX-R had excellent internal consistency with Cronbach’s α at 0.93 for time one and 0.94 for time two (Hermans et al., 2011). Cronbach’s α were also conducted to establish the level of consistency if each item were removed (see Table 3). Removal of specific items did not yield significant changes to the DEX-R reliability, with α ranging from 0.93 and 0.94. Cronbach’s α was 0.92 for factor one, 0.88 for factor two and 0.68 for the third factor. Split-half reliability was 0.93 for time one, and 0.94 for time two. A high degree of reliability was found between DEX-R scores on two-time points. The average measure ICC was 0.92 with a 95% confidence interval from 0.88 to 0.95 [F(88, 88) = 12.4, p < 0.001]. The median interval between the two phases were 23 days (interquartile range: 21–28 days). Table 3 displays scores given for the first phase of completion of the DEX-R.

The DEX-R had good concurrent validity when compared to responses given on another validated measure of dysexecutive problems, the FrSBe (Grace and Malloy, 2001). Both the total scores on the DEX-R and FrSBe were first transformed to achieve adequate normality, the correlation between DEX-R and FrSBe was r = 0.83, p < 0.01.

Pearson product moment correlation coefficient analysis was conducted to establish any influence on participants reports of dysexecutive problems. Gender (r = −0.02, p = 0.836) and years of education (r = −0.52, p = 0.551) were not significantly correlated to DEX-R total and factor scores (p > 0.05). No significant differences were found between males and females t (123) = −0.208, p = 0.84. A negative correlation was found between responses on the first DEX-R administration and age (r = −0.27, p = 0.002). Age significantly correlated with factor one, r = −0.34, p = <0.01 and factor two r = −0.24, p = <0.01, but not with factor three, r = 0.03, p = 0.76. A cut off of three as specified by the literature for the GAD-2 and PHQ-2 indicated 18% of participants scoring above the anxiety threshold, and 7% scoring above the depression threshold. Due to the uneven group sizes correlation analysis was used to preserve validity. Spearman Rho correlations was applied when analyzing the PHQ-2 and GAD-2. Anxiety scores correlated with dysexecutive problems, r = 0.45, p = <0.01. Depression scores were moderately correlated to dysexecutive problems, r = 0.58, p = <0.01. The scores on the GAD-2 significantly correlated with factor one, r = 0.51, p = <0.01 and factor two r = 0.47, p =< 0.01, but not with factor three, r = 0.16, p = 0.16. The scores on the PHQ-2 significantly correlated with all the factors, factor one, r = 0.57, p = <0.01, factor two r = 0.45, p = <0.01, and factor three, r = 0.25, p <0.01.

The effect of background variables on DEX-R responses was analyzed using a multiple regression, the DEX-R total score was the dependent variable, and age, gender, GAD-2 and PHQ-2 scores were the independent variables. In model one, age and gender were kept constant and explained 9.5% of the variance whereas in model two the additional inclusion of the GAD-2 and PHQ-2 scores explained 31.9%, F(4, 120) = 14.05, p <0.001. The results found gender not to significantly predict DEX-R scores (p >0.05). In the first model, DEX-R scores decreased by 0.44 for every year older a participant was, however, when the model also accounted for mood, this decreased to a reduction of 0.17 for every year older and was no longer a significant effect. Controlling for age, gender, anxiety and depression scores, the regression coefficient [B = 2.51, 95% CI (0.21, 4.81) p <0.05] for the GAD-2 indicates that for each increased score on the GAD-2, the total DEX-R score will increase by 2.51. Furthermore, within the same model, the regression coefficient [B = 5.28, 95% CI (2.67, 7.88) p <0.05] for the PHQ-2 indicates that for each increased score on the PHQ-2, the total DEX-R score will increase by 5.28.

The DEX-R was initially developed to map onto the Stuss model. Both the self and informant report versions have demonstrated validity and reliability in both acquired brain injury and healthy aging samples (Simblett et al., 2017; Dimitriadou et al., 2018). The current study adds to the literature regarding the robustness of the psychometric properties of the DEX-R by evidencing its stability and consistency over time. Additionally, it extends the previous literature by evidencing concurrent validity of the DEX-R with the FrSBe. Furthermore, its application with a non-clinical population provides some consideration for clinicians on how individual differences and mood may contribute to responses.

The individual variability of dysexecutive problems in a non-clinical population supports previous research (Chan, 2001) with age and mood found to significantly correlate with the DEX-R as has previously been found (Shaw et al., 2015; Dimitriadou et al., 2018). We found age to be negatively correlated with the DEX-R, indicating that older participants reported less dysexecutive problems. This is inconsistent with wider literature that cognitive functions associated with prefrontal areas decline with normal aging (West, 1996; Van Petten et al., 2004) but is consistent with the finding that aspects of emotion regulation improve with age (Phillips et al., 2008). However, this finding is not robust given the sample demographics, with the oldest participant being 69 years old and only 11% being over 56 years of age. The effect of age seemed to be removed when incorporating mood into the model, linking somewhat to previous research where negative affect mediated responses of dysexecutive problems reported by younger people (Gerstorf et al., 2008). These findings could be due to the younger age of the sample and explained by theories of brain maturation, with prefrontal areas developing into people’s early 30’s (Barkley, 2012; Coffman, 2014). The ventromedial areas of the prefrontal cortex are known to mature later than other regions and are implicated in social and emotional functions and may therefore be less developed in this sample (Gerstorf et al., 2008; Burnett et al., 2009; Pfeifer et al., 2011; Sebastian et al., 2011).

The Rasch analysis evidenced the DEX-R as being multidimensional, supporting its development to capture underlying subconstructs of dysexecutive problems (Simblett et al., 2017). A factor analysis found three factors representing activation: self-regulatory, cognitive and social-emotional functions. When compared to the proposed Stuss domains, the factors appeared to share some overlap with “activation,” “executive cognition” and “behavior-emotional self-regulation.” Those items corresponding with the “metacognition” subscale appeared to be equally distributed across the three factors. Despite the results not aligning fully with the Stuss model, the three factors do appear to have some overlap with theoretical conceptualizations of frontal lobe functioning, as found in prior research (Loschiavo-Alvares et al., 2013; Dimitriadou et al., 2018). These authors propose three factors—Social Self-Regulation, Motivation and Attention, and Flexibility, Fluency and Working Memory—in line with Fuster’s (2008) theory. Therefore, whilst the current study supports previous research on the DEX-R being multidimensional, and factor analysis yielded three factors, a confirmatory factor analysis would be required to confirm fit to either the Stuss or Fuster models.

All the subscales correlated with symptoms of depression, albeit only a weaker correlation for the social-emotional factor. Only the activation-self regulatory and cognition factors correlated with symptoms of anxiety. The social-emotional factor may therefore represent reduced emotional reactivity or neutrality, as those items loading onto it included blunted affect, a lack of concern and no concern for social rules. The cognitive factor correlations may be driven by executive cognitive factors associated with depression and anxiety, such as difficulties with problem-solving and decision making, and working memory (Watkins and Brown, 2002; Hofmann et al., 2011). The activation-self regulatory factor may correlate with symptoms of depression and anxiety as they account for components of apathy and a lack of motivation. Therefore, individual differences found in the study, particularly the different components forming factor one, may reflect an overlay with the cognitive symptoms of depression such as apathy. The issue in determining overlap and directionality of mood and dysexecutive problems is further complicated by the lack of diversity in the sample. The demographics of the sample are reported to experience a higher incidence of anxiety (Jenkins et al., 2020). It may be that particular items on the DEX-R are more likely to be rated as occurring more frequently in those experiencing symptoms of anxiety or depression, such as working memory, problem-solving and decision making as previously noted. This might account for why DEX-R scores increased by two for every GAD-2 score, and by five for every PHQ-2 score.

The differences in the underlying subscales and subconstructs found in the current study compared with previous findings on the DEX and DEX-R may relate in part to the sample characteristics as it is well understood that executive or frontal type difficulties are noted in a range of neurological, neurodevelopmental and mental health conditions. A comparison of the endorsement of items by different clinical groups through differential item functioning could provide insights into those items more likely endorsed by specific clinical groups.

The current study supports the psychometric properties of the DEX-R as found in previous research, being both a valid and reliable measure of dysexecutive problems. The number of factors had similarities with two previous studies. There was individual variation in responses, influenced somewhat by mood and age. This may have implications for clinicians in their interpretation of DEX-R scores. Future research could consider comparing the responses of clinical and non-clinical groups.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The study was approved by the Faculty of Medicine and Health Sciences Research Ethics Committee at the University of East Anglia (UEA), reference number 201819–032. The patients/participants provided their written informed consent to participate in this study.

HW completed this research as part of a doctoral thesis, and was supervised by FG and RR. RR also supported part of the analysis phase using R. FG and RR contributed through regular meetings, guidance on analysis, and editing of the manuscript. AB, SS, and JF were involved in the initial discussions of the topic for the manuscript, advice on analysis, and editing. AB supported the Rasch analysis (RUMM). All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.