Socioeconomic Status Explains Sex Differences on the Trail Making Test: The Case of the Epirus Health Study Cohort Normative Data

Abstract

Objectives

Socioeconomic (SES) and health status (HS) are rarely considered when normative data are calculated. In the present study, normative data for the Trail Making Test (TMT) were developed from a large cohort and the association of sex, age and education, as well as HS and SES, with direct and derived TMT scores was explored.

Methods

Two thousand three hundred sixteen participants [1412 (61%) women; mean age: 47.11 (SD = 11.67) years; mean education: 14.82 (SD = 3.39) years] were drawn from the population-based Epirus Health Study. HS was rated on a self-reported scale and participants’ medical conditions were recorded. SES was estimated from participants’ after-tax income per month. The association of sex, age and education with TMT-A, TMT-B, TMT B-A and TMT B/A was explored with linear regression analyses. Hierarchical regression analyses were applied to control for HS and SES.

Results

Direct TMT scores were associated with sex, age and education (TMT-A: B_sex = 0.060, B_age = 0.322 and B_education = −0.191; ΤΜΤ-Β: B_sex = 0.042, B_age = 0.330 and B_education = −0.208). TMT B-A was associated with age (B = 0.176) and education (B = −0.130), whereas TMT B/A was not associated with any tested variable. SES, but not HS, was associated with TMT-A and TMT-B explaining the association of sex with TMT scores when included simultaneously in the model.

Conclusions

TMT performances are associated with age, education and sex. However, sex differences in direct TMT scores are attributed to underlying socioeconomic disparities in this large well-characterized cohort.

INTRODUCTION

Neuropsychological performances are adjusted for demographic characteristics that account to a certain degree for normal variation in cognitive abilities in neurologically healthy individuals. Thus, performances are usually corrected for age and education and less often for sex or ethnicity. However, other factors may also be related to neuropsychological performances, such as socioeconomic status (SES) and general health condition, but these are rarely considered (Bergman & Almkvist, 2015).

SES is a complex construct that describes various aspects of social class, including income, occupation, educational attainment and financial security (American Psychological Association, 2024a) and has been identified as an important determinant of quality of life (Henriques et al., 2020; Nutakor et al., 2023). Community characteristics and exposure to violence (Evans, 2004), access to medical care and education (Arentoft et al., 2015), working conditions and lifestyle (Landsbergis, 2010) are all related to physical and mental health. Studies including large cohorts of older adults reveal that SES, estimated via participants’ income (Nutakor et al., 2021; Bertola et al., 2021; Zahodne et al., 2021; Zhang et al., 2015) or quality of neighborhood (Lang et al., 2008), has an effect on cognitive performance beyond other demographic characteristics. Early life SES also predicts later cognitive performance through its cumulative effect on health (Luo & Waite, 2005), as well as on the opportunities for proper education (Bertola et al., 2021). In addition, general health status and multimorbidity constitute a source of variance in cognition in large epidemiological studies (Aarts et al., 2011; Fabbri et al., 2016). For these reasons, research on the degree of test bias when individuals from underprivileged socioeconomic groups are being examined (Rivera Mindt et al., 2008) and normative corrections for SES have been suggested (Arentoft et al., 2015).

Trail Making Test (TMT) (Reitan, 1958) is one of the most popular neuropsychological tests in clinical practice, widely used as a measure of executive functions in patients with suspected frontal lobe damage (Lezak, 1995). It consists of two parts. Performance in both conditions is timed and although some studies have suggested the analysis of specific error types (Christidi et al., 2013; Hafiz et al., 2023; Mahurin et al., 2006; Stuss et al., 2001), in most cases completion time is the only variable of interest. The speedy completion of both parts of TMT requires adequacy in various cognitive domains, including visual scanning, attention and speed of motor response and information processing (Bowie & Harvey, 2006). In addition, TMT-B elicits the use of higher-order cognitive skills, because the fast and correct switch between letters and numbers requires ability for working memory and mental flexibility (Sánchez-Cubillo et al., 2009). This assumption is also supported by evidence from neuroimaging studies, where performance on TMT-B was associated with increased activation in comparison to TMT-A in middle and superior frontal regions, precentral gyrus and insular cortex, regions inextricably linked to executive functions (Kopp et al., 2015; Talwar et al., 2020; Varjacic et al., 2018). For that reason, derived scores have also been suggested to isolate the contribution of executive functions to TMT performance controlling for the effect of low-level processes (Muir et al., 2015; Sánchez-Cubillo et al., 2009; Stuss et al., 2001).

Among these scores, the difference in completion times between TMT-B and TMT-A (TMTB-A) and the ratio of TMT-B to TMT-A (TMT B/A) are the most widely used. In TMT B-A, performance on TMT-A serves as a baseline reference to which performance on TMT-B is compared. This score has been described as an index of the time needed to perform alternations between different mental sets (Sánchez-Cubillo et al., 2009; Senior et al., 2018) and, therefore, it is used as a measure of cognitive flexibility. However, as has been commented by Espenes et al. (2020), TMT B-A focuses only on the difference between the two TMT parts without taking into account the time spent for the completion of the two conditions separately. Ιn other words, two hypothetical examinees, one with a very fast and one with a very slow performance on both conditions, would end up at the same percentile if they produced the same difference scores. On the other hand, ΤΜΤΒ/Α, which has been found to be negatively associated with TMT-A (Espenes et al., 2020; Llinàs-Reglà et al., 2017), addresses that issue by controlling for the speed of information processing, so that the same difference between TMT-A and TMT-B would correspond to better performance for examinees with longer completion times.

Due to its sensitivity to cognitive dysfunction and its quick and easy administration, TMT has been used for the investigation of cognitive impairment in various clinical groups (Chan et al., 2019; Hafiz et al., 2023; Rogers et al., 2023; Stenberg et al., 2020). Its clinical value has led to an increasing number of studies reporting normative data across different national populations (Abi Chahine et al., 2020; Bezdicek et al., 2012; Cangoz et al., 2009; Siciliano et al., 2019), as well as culturally and linguistically diverse groups (Hsieh & Tori, 2007; Suarez et al., 2021). Most studies agree that age and education significantly predict direct TMT scores, so that older age and fewer years of education are associated with longer completion times (Abi Chahine et al., 2020; Espenes et al., 2020; Llinàs-Reglà et al., 2017; Siciliano et al., 2019). Findings about the effect of sex remain inconsistent, with most studies reporting no sex differences (Bezdicek et al., 2012; Suarez et al., 2021), whereas some others observe greater performances by men versus women (Cangoz et al., 2009; Llinàs-Reglà et al., 2017). Derived TMT scores appear less affected by demographic variables. TMT B/A, for example, has been found to be associated only with education and not with age (Arampatzi et al., 2024; Espenes et al., 2020; Llinàs-Reglà et al., 2017).

Specific to the Greek population, to our knowledge, four studies reporting normative data have been published so far (Arampatzi et al., 2024; Christidi et al., 2015; Vlahou & Kosmidis, 2002; Zalonis et al., 2008). In most, the effect of age and education on direct TMT scores was confirmed. Only in Arampatzi et al. (2024), age was positively associated with TMT-A but not with TMT-B presumably due to the restricted range in participants’ age, as this study was conducted in a cohort of older individuals. Regarding derived TMT scores, a significant effect of education on TMT B-A (Christidi et al., 2015) and a significant effect of age (Christidi et al., 2015) and education (Arampatzi et al., 2024; Christidi et al., 2015) on TMT B/A were noted. Sex differences were only reported for TMT B-A by Christidi et al. (2015), in the largest existing normative Greek sample, where women outperformed men. Interestingly, the aforementioned studies either used smaller samples to stratify their normative data [e.g., from 195 individuals in Vlahou & Kosmidis (2002) up to 775 individuals in Christidi et al., 2015] or they targeted participants of specific age [e.g., older adults in Arampatzi et al., 2024]. Additionally, all studies reporting norms for young adults date before 2016 and, most importantly, none of them has taken into account the possible contribution of participants’ SES or their general health state, which may both be associated with cognitive performance (Lyu & Burr, 2016; Schüz et al., 2020), and could possibly affect the reported results.

Considering these, we estimate that the analysis of TMT scores drawn from a large-scale well-characterized population-based sample would further contribute to the extraction of precise normative data, because the inclusion of a large number of participants could yield findings more safely generalizable to the whole population. Furthermore, it would contribute to the understanding of previously reported discrepant findings (i.e., the association of sex with TMT scores). Most importantly, though, including participants with well-defined lifestyle and health information, a population study would allow for the investigation of the simultaneous effect of self-reported health status (HS) and SES to participants’ performance along with the other demographic variables. The aims, therefore, for undertaking the present study were: (a) to explore the association of sex, age and education with direct and derived TMT scores to suggest appropriate normative data and (b) to investigate whether these results would differ when self-reported HS and SES were considered.

MATERIALS AND METHODS

Participants

The present study constitutes a part of the Epirus Health Study (EHS). EHS is a population-based prospective cohort study initiated in June 2019 to investigate the etiology of complex multifactorial chronic diseases and to promote the improvement of the general health state in the Greek population (Kanellopoulou et al., 2021). The recruitment was completed in October 2023. The EHS cohort consisted of 2,540 participants between 20 and 80 years old, who are permanent residents of Epirus, a northwestern geographical region in Greece. Participants’ recruitment was based on advertisements to the local press and social media, promotions via study’s website (https://ehs.med.uoi.gr/), events organized by local health agencies and invitations to Epirus residents working in the private or public sector. Even though pregnancy and active infections were the only exclusion criteria applied during the initial sampling, especially for the present analyses individuals with self-reported medical conditions that could possibly interfere with their neuropsychological status were excluded. More specifically, out of the 2,540 participants, 59 were excluded due to neurological disorders, 85 due to major mental health disorders, 47 due to severe cardiovascular diseases, 7 due to alcohol abuse and 14 due to other medical conditions. The procedure of participants’ selection and the exclusion criteria are presented in detail in Fig. 1.

The final sample of the present study, therefore, consisted of 2,316 participants [1,412 (61%) women] with a mean age of 47.11 (± 11.67) years and a mean educational attainment of 14.82 (± 3.39) years. All participants were literate having completed at least 6 years of formal schooling. Most participants (66.8%) were married. About 76% were employed, 11.5% were retired, 7% were unemployed and 5% were university students or homemakers. In terms of the occupation, 32% were office workers, 28.7% were scientists or artists, 12.3% were craftsmen, farmers/breeders or unskilled workers, 3.9% worked for the armed forces or in the public sector, and the rest were sellers or entrepreneurs. About 39% self-reported having a chronic health condition, with dyslipidemia (26.6%) and hypertension (14.3%) being the most common.

Study design

The EHS research protocol was approved by the Research Ethics Committee of the University of Ioannina and was conducted in accordance with the Declaration of Helsinki and its later amendments. All participants gave their written informed consent prior to participation. At recruitment, participants were interviewed with a close-ended standard questionnaire, where information about demographic characteristics (i.e., sex, age, place of birth, marital status, level of education, and current employment status), lifestyle factors (i.e., physical activity, smoking habits, and alcohol consumption) and personal and family medical history were collected. Additionally, they participated in a series of clinical exams by medical professionals and in a brief neuropsychological assessment with measures of executive functions and episodic memory by a trained neuropsychologist. More information about the EHS research protocol can be found in previously published papers (Kanellopoulou et al., 2021; Koutsonida et al., 2023).

Neuropsychological assessment

Regarding the administration of the TMT, the version suggested by Reitan and Wolfson (1993) was used. All participants were given standard instructions following the guidelines presented by Strauss et al. (2006). In part A, participants were presented with encircled numbers from 1 to 25, and in part B with encircled numbers from 1 to 13 and letters of the Greek alphabet from A to M, which were randomly distributed in a white page. In part A, participants were instructed to connect the circles in numerical order drawing a continuous line as fast and as accurately as possible. In part B, they were instructed to switch between numbers and letters drawing a continuous line as fast and as accurately as possible. Before the completion of each part, all participants had a rehearsal trial. Time from the first pen stroke to test completion was recorded manually with a stopwatch. In case of a mistake, the examiner pointed it out immediately and asked the participant to go back and correct the sequence. Number of errors was not recorded. The time limits for TMT-A and TMT-B were 150 and 300 seconds, respectively. Variables of interest were the completion times for TMT-A and TMT-B, as well as derived scores TMT B-A and TMT B/A. In both direct and derived scores, lower scores corresponded to better performance.

Estimation of HS and SES

Participants HS was measured with a self-reported questionnaire, where they were asked to rate on a 4-point Likert scale their general health condition from 1 = poor to 4 = very good. Detailed questions about the presence of specific chronic health conditions (e.g., dyslipidemia, diabetes mellitus, high blood pressure, other cardiovascular diseases, asthma, anxiety disorders, macular degeneration, short-sightedness/myopia, glaucoma, chronic kidney disease, rheumatoid arthritis/osteoarthritis/osteoporosis, and thyroid disease) and about limitations to daily life activities due to health issues followed.

For SES, since to our knowledge no “gold standard” measure exists and with respect to the high heterogeneity observed in the broad literature (Arentoft et al., 2015; Shavers, 2007), we opted to estimate participants’ self-reported current after tax income per month as a proxy, following the example of other studies in Greece (Michou et al., 2019; Pappa et al., 2009). Five SES subgroups were, thus, formed based on participants’ income (<500 euros: very low, 501–900 euros: low, 901–1,400 euros: average, 1,401–2,000 euros: high and >2,000 euros: very high SES).

Data analysis

The association of demographic variables with TMT scores was investigated in two ways. Stepwise multiple regression analysis was used to identify the factors that contribute to TMT performance. First, the association of sex, age, and education with direct and derived TMT scores was explored. In the second phase, hierarchical analyses tested the association of the aforementioned demographic variables after controlling for the self-reported HS and SES. For that reason, HS and SES were entered in the first block and sex, age, and education were entered in the second block, to assess their predictive value beyond and above of that of HS and SES. In both analyses all independent variables were entered in a backward manner. Age, education and TMT scores were treated as continuous variables, whereas sex, self-reported HS and SES as nominal variables with two, four, and five levels, respectively. All statistical analyses were undertaken using STATA (version 14; StataCorp, College Station, TX, USA). A p-value of .05 was considered an indicative threshold of statistical significance.

RESULTS

On direct TMT scores, age was identified as the strongest predictor, followed by education and lastly by sex. A positive association was noted between age and completion times in both TMT-A and TMT-B, so that for every 1-year increase in age completion time would increase by 0.322 s in TMT-A and by 0.330 s in TMT-B controlling for the rest of the variables. Education was negatively associated with TMT-A and TMT-B, so that for every 1-year increase in educational attainment completion time would decrease by 0.191 s in TMT-A and by 0.208 s in TMT-B controlling for the rest of the variables. A significant association with sex was also found, with women performing more poorly than men both on TMT-A and TMT-B. TMT B-A was significantly associated with age and education, so that for every 1-year increase in age and education completion time would increase by 0.176 s and decrease by 0.130 s, respectively. TMT B/A was not predicted by any demographic variable. Compared to direct TMT scores, the association of age and education with TMT B-A was weaker, as expressed by the values of standardized coefficients. Results are summarized in Table 1.

To stratify normative data, we divided participants into subgroups according to their age and the stages of formal education in Greece. In Fig. 2, TMT performance appears to remain relatively stable until the age of 50, when an apparent decline begins. Similarly, a considerable difference is present for examinees with lower educational attainment, up to 13 years of education, when it remains relatively stable for those participants with higher educational attainment. Based on this observation, we present normative data for TMT scores within Tables 2–4. Due to sex differences, for TMT-A and TMT-B separate normative data for male and female participants are suggested. For TMT B/A, where no demographic association was found, unified normative data are presented for all participants.

Considering SES, 14.3% of study participants reported very low SES, 27.6% low, 43.4% average, 10.9% high, and 3.8% reported very high SES. Regarding HS, 0.7% of study participants reported poor HS, 13.1% moderate, 50.5% good, and 35.7% reported very good HS. A series of chi-square analyses revealed that participants who rated their HS as poor or moderate reported higher rates of chronic health conditions and more limitations in daily life activities within the last year than those with self-reported good or very good HS. Detailed information about participants’ medical conditions according to their self-reported HS is presented in Table 5. SES and HS were correlated with each other, so that participants of a higher SES rated their HS as better, but that correlation was found to be weak [χ²(12) = 51.606, P < .001, Cramer’s V = 0.088].

A significant association was noted between SES and direct TMT scores, so that participants of a higher SES performed faster. SES accounted for 3.5% and 3.4% of the total variances of the TMT-A and TMT-B, respectively. On the contrary, HS was not associated with any direct or derived TMT score. Interestingly, when SES was taken into consideration, the association of sex with TMT-A and TMT-B was eliminated (Table 1). Based on these findings, we performed two consecutive Sobel tests to explore the possible mediating role of SES in the relationship between sex and direct TMT scores. Interestingly, Sobel test yielded a Z score of 4.19 (p < .001) for TMT-A and 4.43 (p < .001) for TMT-B confirming that the mediating effect of SES is statistically significant (Abu-Bader & Jones, 2021). Even though women did not differ from men in age and educational attainment, they performed worse than men on both TMT-A and TMT-B, but that discrepancy was found to be weak, as indicated by the value of the effect size. Women also significantly differed from men in SES, where 49.6% of women were placed at the range of very low and low income in comparison to 29.9% of men. Similarly, 8.8% of female participants reported high/very high income in comparison with 23.8% of male participants. The association between sex and SES was strong, as can be inferred by the value of effect sizes. Sex differences in demographic characteristics, TMT scores, and SES are summarized in Table 6.

DISCUSSION

In the present study, we explored the association of sex, age, and education with direct and derived TMT scores using a large sample drawn from a population-based cohort in order to develop normative data for the Greek population. Additionally, we investigated the association of SES and HS with neuropsychological performances.

A significant association of sex, age and education was noted with direct TMT scores (i.e., TMT-A and TMT-B), so that male sex, younger age and higher education were related to shorter completion times and, thus, to better performance. The association of age and education with direct TMT scores is not surprising, as it has been consistently confirmed in the extant literature (Cavaco et al., 2013; Espenes et al., 2020; Málišová et al., 2022; Specka et al., 2022). However, while most studies have reported no sex differences on TMT (Abi Chahine et al., 2020; Bezdicek et al., 2012; Siciliano et al., 2019; Suarez et al., 2021), a few others indicated better performance either by men (Cavaco et al., 2013; Specka et al., 2022) or women (Magnusdottir et al., 2021). Except for Specka et al. (2022), who noticed a large standard error on the effect of sex and opted not to include it in the estimation of normative data, the other two studies did not report possible differences in other sociodemographic characteristics, which could possibly explain sex differences on TMT. Regarding derived TMT scores, age and education predicted TMT B-A, but that association was weaker in comparison with direct TMT scores, whereas TMT B/A was not predicted by any demographic variable. This finding is in line with other studies that suggest TMT B/A as a more sensitive index of executive dysfunction, because it is less affected by demographic factors (Cavaco et al., 2013; Hester et al., 2005; Specka et al., 2022) and less correlated with direct TMT scores (Espenes et al., 2020; Siciliano et al., 2019). Given that TMT B/A was not found to be vulnerable to the effect of demographic characteristics and SES, a greater emphasis should be given on its clinical use among TMT scores when executive functions are being assessed in Greek examinees. Of interest, a significant association of SES was observed explaining a small proportion of the total variance in direct TMT scores. Notably, when SES was controlled, the contribution of sex to direct TMT scores was no longer significant.

The current study has several advantages over previous ones. First, it included one of the largest Greek community-dwelling study samples for normative data so far with well-defined characteristics allowing for the generalizability of the results. This is important as many neuropsychological norms are based on inadequate sample sizes, which produce confidence intervals that are deemed too large to be of clinical utility (Bridges & Holler, 2007; Piovesana & Senior, 2018). Second, only a few studies have revealed the importance of HS and SES when considering neuropsychological performances and normative data (Bergman et al., 2016; Bergman & Almkvist, 2015; Smerbeck et al., 2012). Health and socioeconomic factors are of particular importance when sex differences in cognitive performances are interpreted, because female participants have been found to report worse physical (Pappa et al., 2009) and mental health (Eikemo et al., 2018) and they also face economic inequity (American Psychological Association, 2024b), as was noted in our study as well.

The association of SES with neuropsychological performances has been an area of interest. Even though it is not clear whether specific cognitive domains are more vulnerable than others to the effect of SES, correlations have been reported for measures of memory (Arentoft et al., 2015; Uddin et al., 2023), attention (Arentoft et al., 2015), processing speed (Arentoft et al., 2015; Zhang et al., 2015), verbal fluency (Arentoft et al., 2015; Uddin et al., 2023) and executive functions (Shaked et al., 2018; Wong & Yang, 2023). Interestingly, early-life SES has been suggested to predict future neuropsychological performance (Greenfield et al., 2021), presumably through its effect on physical and mental health (Kim et al., 2021; Zahodne et al., 2018). Additionally, improvement in early-life SES has been found to confer benefits on later cognitive status (Turrell et al., 2002). That being said, the current findings do not support the inclusion of SES as a covariate in the regression equations, in order to avoid “normalizing” lower performance in underprivileged women and health-burdened populations (Brandt, 2007) prolonging, thus, socioeconomic inequality. Instead, they highlight the necessity of research focused on understanding the mechanisms that relate SES with cognition, in other words why SES is related to cognition and through which processes. Understanding this phenomenon could contribute to appropriate policies to diminish mental disparities (i.e., cognitive functioning) associated with SES in the long term. Until then, raising awareness of the effect of socioeconomic disparities on mental health, especially among neuropsychologists and mental health providers is critical. Certainly, neuropsychologists in clinical practice could apply more lenient criteria to detect neuropsychological impairment (i.e., apply cut-offs corresponding to 2 SDs below the mean instead of 1.5), when individuals of a lower SES are being examined.

This study has certain limitations. First, SES was determined via participants’ current self-reported after-tax monthly income. Income is considered a common but less stable measure of SES than occupation, for example (Shavers, 2007). A more complex construct encompassing more aspects of SES could, therefore, yield more accurate results. Moreover, participants’ health status was self-reported, and thus, the possibility of including individuals with an undiagnosed condition affecting their neuropsychological status cannot be safely ruled out despite the conservative inclusion criteria applied in our analyses. Finally, despite the recruitment of a large sample of young and middle-aged adults, older adults with low educational attainment are underrepresented in the EHS cohort.

CONCLUSION

The present study provides normative data for the TMT derived from a well-defined large study sample of healthy Greek adults. Most importantly, it highlights the importance of considering SES, which in our study was found to explain sex differences in TMT performances.

FUNDING

This work was funded by the projects: 1) “Understanding pathways of healthy ageing (in health and disease) through integration of high resolution omics data - pathAGE (MIS 5047228) which is implemented under the action “Regional Excellence in R & D Infrastructures” funded by the Operational Programme “Competitiveness, Entrepreneurship and Innovasion (NSRF 2014-2020) and co-financed by Greece and the European Union (European Regional Development Fund) and 2) the Operational Programme Epirus 2014-2020 of the Prefecture of Epirus (MIS H⊓1AB-0028180).

CONFLICT OF INTEREST

None declared.

AUTHOR CONTRIBUTIONS

Yiannis Tsiaras (Formal analysis), Kassiani Styliani Tsantzalou (Data curation, Formal analysis), Myrto Koutsonida (Data curation, Formal analysis, Methodology), and Eleni Aretouli (Conceptualization, Supervision)

References