Greater social jetlag predicts poorer NIH Toolbox crystallized cognitive and academic performance in the Adolescent Brain Cognitive Development (ABCD) study

ABSTRACT Academic performance plays a crucial role in long-term educational attainment and occupational function. Chronotype refers to an individual’s daily tendencies for times for waking, activity, and sleep. Social jetlag reflects the mismatch between an individual’s chronotype and their social schedule. Because school typically starts early in the morning, later chronotype is often associated with daytime sleepiness, insufficient sleep, and poor academic performance. However, the relationship between academic performance, chronotype, and social jetlag has not been extensively examined in large samples like the Adolescent Brain Cognitive Development (ABCD) study. We hypothesized that greater social jetlag would predict poorer cognitive and academic performance. Year 2 (ages 11–14) cross-sectional data from the ABCD cohort (n = 6,890 adolescents) were used to evaluate academic performance (i.e. self-reported past year grades), NIH Toolbox cognitive performance measures, chronotype, and social jetlag from the Munich Chronotype Questionnaire. We found that later chronotype and greater social jetlag predicted poorer cognitive and academic performance with small effect sizes. Our findings emphasize the importance of individual differences in chronotype and social jetlag when designing class schedules, as aligning school activities with student optimal sleep-wake times may contribute to improved academic performance.


Introduction
Academic performance plays a crucial role in determining eligibility and the opportunity to attend college, graduate school, professional school, and other careerdetermining educational programs (Meda et al. 2017;Shaw 2007).While intelligence quotient (IQ) remains widely recognized as a key predictor of academic achievement (Zax and Rees 2002), a multitude of external and internal factors influence academic performance.One such factor that has gained traction in the literature relates to an individual's trait-like tendencies with respect to the timing of sleep and activity, or "chronotype" (Roenneberg et al. 2019).A related construct is circadian preference, which captures one's relative preference for morningness versus eveningness.Individuals high in morningness preferring to wake up and go to bed early, and individuals high in eveningness preferring to wake up and go to bed late (Preckel et al. 2011).Chronotype (and circadian preference) exist along a normally distributed spectrum spanning early to late (or morningness to eveningness), and to correlate with biological measures of circadian phase (Kantermann et al. 2015).Furthermore, numerous studies have shown that chronotypes differ in the timing of physiological and psychological variables (Enright and Refinetti 2017), such as physical and cognitive performance (Facer-Childs et al. 2018), alertness (Harrison et al. 2021), and positive affect (Dagys et al. 2012).Studies have also suggested that individuals with a later chronotype may face challenges in aligning their sleep-wake patterns with traditional morning-oriented school schedules (Roenneberg et al. 2012), resulting in reduced sleep duration, poorer sleep quality, and difficulties in concentration and attention during morning classes (Arbabi et al. 2015;Beebe 2011;Dewald et al. 2010).
The association between chronotype and level of intelligence has been highly mixed.Some groups have found no association between chronotype and overall intelligence, e.g. when comparing Mensa members with matched nonmembers (Ujma et al. 2020).Other studies have found in a convenience sample of adults directly assessed for fluid intelligence in the afternoon that evening chronotypes had higher levels of fluid intelligence than morning chronotypes (Zajenkowski et al. 2019).In a different group of high-achieving graduate students, evening chronotypes scored higher on measures of general intelligence than morning chronotypes (Piffer et al. 2014).Therefore, the association between chronotype and measures of intelligence needs further clarification.
Intelligence has been measured in a variety of ways including crystallized cognitive ability and fluid cognitive functioning.The National Institutes of Health Toolbox for Assessment of Neurological and Behavioral Function Cognition Battery provides seven cognitive tests with two for crystallized cognitive ability (vocabulary and reading) and five for fluid cognitive functioning (working memory, memory, speed of processing, and executive functioning) (Holdnack et al. 2017).The NIH Oral Reading Recognition Test of the Language Construct that assesses reading decoding skills and crystallized abilities has the highest test-retest reliability of 0.90 of those measures in the cognitive battery of the NIH toolbox (Weintraub et al. 2014).Many studies have focused on more traditional measures such as intelligence quotient (Pishghadam et al. 2022;Ujma et al. 2020), but fewer have investigated intelligence using the NIH Toolbox Cognition Battery that offers assessment of both crystallized and fluid cognitive abilities.
The persistent discrepancy between an individual's internal circadian time and their social schedule is commonly referred to as social jetlag, which quantifies the misalignment between an individual's circadian rhythm and their daily schedule on school or work days versus free days, as well as the change in sleep-wake timing between school/work days and free days (Roenneberg et al. 2019;Wittmann et al. 2006).Social jetlag has been associated with a range of negative physical and behavioral outcomes (Beauvalet et al. 2017), such as obesity (Antunes et al. 2010), diabetes (Antunes et al. 2010), cardiovascular disease (Caliandro et al. 2021), depressive symptoms (Levandovski et al. 2011), ADHD (Kivelä et al. 2018), bipolar disorder (Chung et al. 2012) and substance misuse (Nelson et al. 2023;Wittmann et al. 2010).These adverse outcomes, in turn, can have implications for academic performance.
Several cross-sectional studies have found that daytime sleepiness and sleep disturbances have a significant impact on cognitive performance in academic settings (Ferguson et al. 2018;Mirghani 2017;Montaruli et al. 2019;Shimura et al. 2022).These studies have examined students in various educational contexts, including college freshmen, Italian students, high school students, and medical students, demonstrating the widespread impact of sleep disruption on academic performance.However, many of these studies were conducted on students from one or a few schools, which can affect the generalizability of their findings.Reproducibility on a larger scale and across diverse populations is crucial for establishing robust conclusions regarding the impact of sleep disruption on academic performance.
The Adolescent Brain Cognitive Development (ABCD) study is a longitudinal study that follows a large and diverse sample of 11 880 participants aged 9-10 years from 21 research sites across the United States over a 10-year period ("About the Study" 2023).The study collects extensive data on various domains, including academic performance, cognitive assessments, chronotype, and social jetlag (Volkow et al. 2018).It also gathers information on environmental factors such as family, school, socioeconomic status, and exposure to risk factors or adversity (Garavan et al. 2018).Given its expansive data, use of standardized cognitive measures like the NIH Toolbox, and longitudinal design, the ABCD study provides an ideal group of subjects to examine whether greater social jetlag (sleep corrected) predicts poorer cognitive and academic performance.We hypothesize that individuals with higher levels of social jetlag will have poorer cognitive and grade performance.Specifically, we hypothesize that lower levels of crystallized intelligence and fluid intelligence will be associated with higher levels of social jetlag.

Participants
Data on demographic variables, academic performance, NIH Toolbox cognitive performance measures, and circadian rhythm parameters were obtained from the ABCD study ("NIH's Adolescent Brain Cognitive Development ABCD Study" 2018).All together, these data comprise a sample of n = 9,852, after removing participants that declined to disclose or did not know their past year grades or household income.This resulted in a sample of n = 6,890 for performing hierarchical linear regression.Data included was collected prior to the COVID-19 pandemic (2018COVID-19 pandemic ( -2020)), which changed sleep patterns significantly (Pelham et al. 2021).

Academic performance
Academic performance was assessed using the students' grades from the previous academic year.The data on past year grades were obtained from the ABCD Youth School Attendance and Grades data archive (element = sag_grades_last_yr).Grades were self-reported by the participants and were assigned numerical values ranging from 1 to 12, with lower numbers indicating better grades (1 = A+, 2 = A, 3 = A-, 4 = B +, 5 = B, . . ., 12 = F).In our study, we specifically focused on data obtained from the two-year follow up.This crosssectional analysis was conducted as the first part of a long-term line of research to track these factors longitudinally over multiple years of data (e.g.later years of school).

NIH Toolbox
The NIH Toolbox offers a comprehensive range of over 100 stand-alone measures, available in 30-minute batteries, to evaluate Cognition, Emotion, Motor, and Sensation.We focused on the NIH Oral Reading Recognition Test of the Language Construct that assesses reading decoding skills and crystallized abilities in participants due to it having the highest test-retest reliability of 0.90 of those measures in the cognitive battery of the NIH toolbox (Weintraub et al. 2014).Previous studies have established the reliability and validity of this measure in evaluating cognitive abilities.Weintraub et al. (2014) found a strong (r > 0.7) correlation between the Oral Reading Recognition Test and other cognitive ability measures, such as memory and executive function critical to academic performance.Furthermore, Shields et al. (2020) found that each test in the cognitive battery, as well as the Crystallized and Fluid Composite scores, correlated moderately to strongly (r > 0.5) with IQ; IQ has found to predict academic performance (Pishghadam et al. 2022).
In our study, we utilize scores from the NIH Oral Reading Recognition Test of the Language Construct as an indicator of individual IQ to determine if these scores can predict academic performance from the previous year.We specifically use age-corrected standard scores (M = 100, SD = 15) sourced from the ABCD Youth NIH TB Summary Scores data archive (element= nihtbx_reading_agecorrected).These scores reflect an individual's neurocognitive performance relative to their peers of the same age and are helpful for assessing developmentally appropriate cognitive functioning (Casaletto et al. 2015).Using this well-established benchmark, we expected that individuals with higher NIH Toolbox scores would have higher grades.

Circadian rhythm parameters
Munich chronotype questionnaire.The Munich Chronotype Questionnaire (MCTQ) is a 17-item, selfrated scale that characterizes chronotype by calculating the midpoint between sleep on-and off set on free days (Roenneberg et al. 2015).Total participant chronotype scores ranged from 16 to 40, with higher values representing tendency towards later timing, ostensibly reflecting later circadian phase of entrainment.Units reflect sleep midpoint and are measured in hours, so that, for example, 16 on the scale corresponds to 16:00.A chronotype value of 25.3 would be 1:20 of the following morning.A chronotype value of 40 would be 16:00 the following day.Chronotype data were extracted from the ABCD Youth Munich Chronotype Questionnaire data archive (element = mctq_msfsc_calc).Our chronotype data may be impacted by weekend oversleeping in the sample since we did not utilize any correction of midpoint of sleep on free days compared to scheduled days (Roenneberg et al. 2012).We instead employed participant-reported raw data for the midpoint of sleep on free days.To address this weakness, we calculated sleep corrected relative social jetlag as described below.
Social jetlag.The Munich Chronotype Questionnaire can be used to calculate social jetlag (SJL), defined as the difference between an individual's midpoint of sleep times on workdays (MSW) and free days (MSF).This will provide a measure of the extent to which an individual's social schedule is disrupting their natural sleep pattern.Sleep corrected social jetlag was utilized as the difference between sleep onset on free days and sleep onset on workdays (Jankowski 2017) to more accurately reflect jetlag across a heterogenous population of chronotypes (Roenneberg et al. 2019).

Statistical analyses
Hierarchical linear regression models were used to examine the associations between chronotype or social jetlag (sleep corrected), NIH Toolbox cognitive battery (e.g.oral reading) scores (age corrected), and past year grades.All variables were checked by histogram for normality prior to running models.In Model 1, we regressed chronotype or social jetlag (sleep corrected) onto demographic variables including year 2 age, sex, ethnicity, and household income normalized by number of household members, with family ID included as a random effect (intercept).Next, we added BMI (Model 2) followed by average weekly sleep duration (Model 3).In the subsequent model (Model 4), we added age-corrected NIH Toolbox cognitive battery scores (e.g.oral reading) along with time of testing in decimal form.In Model 5, we added grades in the past year.All analyses were conducted in R (R Core Team 2021 and RStudio v 4.2.2).We also constructed independent models investigating prediction of cognitive and academic performance from BMI, average weekly sleep duration, and normalized household income.The lm.beta package was used to standardize all regression coefficients (Behrendt 2023).Effect sizes were quantified using difference in variance accounted for between models (i.e.ΔR 2 ) with values greater than 0.02, 0.13, and 0.26 interpreted as the cutoffs for small, medium, and large respectively (Cohen 1988).Bivariate relationships between age, SJL (sleep corrected), average weekly sleep duration, BMI, NIH Toolbox oral reading scores, and past year grades were analyzed using Pearson's correlation coefficients.Differences between participants with complete data (n = 6,890) and participants with incomplete data (n = 9,852) were analyzed using student t tests and proportion tests.

General characteristics
Participants ranged in age from 12 to 14 years old (mean 11.95) at time of cognitive testing, past year grades, and administration of the MCTQ, and were approximately evenly divided between the sexes (47.9% female sex assigned at birth).Approximately three-fifths (58.1%) identified as being White.Participants had a mean past year grade of 3.15 in the last year but with a wide range (1 to 12).Household income normalized by number of household individuals has been provided in Table 1.Demographic characteristics were similar between complete (N = 6890) and incomplete (N = 9852) datasets.NIH Toolbox scores for Picture Vocabulary, Flanker, List, Card Sort, Pattern, Picture, and Oral Reading, along with Chronotype, Social JetLag, BMI, and Average Weekly Sleep Duration have been summarized in Table 2. Sample sizes for NIH Toolbox List and Card sort scores were inadequate to construct meaningful predictive models.

Hierarchical linear regression models
The results for all linear regression models are found in Tables 3-7, in addition to Supplementary Tables S1-S8.Demographic model subexpansion of age and sex results can be found in Supplementary Table S9.Correlation Matrix Among Common Variables is found in Supplementary Table S10.Note that lower numbers for self-reported grades reflected better grades (i.e. an inverse relationship).
Table 6 (Sleep Duration Oral Reading Predictive Hierarchical Regression Model): In Model 3, NIH Toolbox Age-Corrected Oral Reading Scores explained significantly more variance in average weekly sleep duration with a marginal-small effect size (ΔR 2 of 0.006); greater average weekly sleep duration was associated with lesser NIH Toolbox Age-Corrected Oral Reading Scores (B = −0.007,β = −0.074,SE = 0.001, p < 0.001, CI = −0.010 to −0.005) adjusted for time of testing.In Model 5, past year grades explained more variance in average weekly sleep duration with a marginal effect size (ΔR 2 of 0.004); greater average weekly sleep duration was associated with better past year grades (B = −0.053,β = −0.064,SE = 0.010, p < 0.001, CI = −0.073 to −0.033).

Discussion
We have shown for the first time using NIH Toolbox measures of crystallized and fluid intelligence that later chronotype and greater SJL predicted poorer cognitive performance, in addition to past year grades in a large national sample of youth.We focused on the impact of social jetlag as a potential driver of cognitive impairment impacting academic performance due to lack of sleep.
As expected, our study found that later chronotype and greater SJL predicted poorer crystallized intelligence (reading, picture vocabulary) from the NIH Toolbox, confirming a negative relationship between chronotype and SJL, and cognitive performance.However, later chronotype and greater SJL did not predict poorer fluid intelligence using 3 battery assessments (picture, flanker, pattern) from the NIH Toolbox and lacked sufficient sample using 2 additional battery assessments (list, card sort) to complete analyses.Our findings focusing on crystallized intelligence differ from those of other groups that have found later chronotypes to have higher levels of crystallized intelligence (Piffer et al. 2014) despite controlling for time of testing, but  are in agreement with other groups in school settings (Arbabi et al. 2015).However, effect sizes were more pronounced with greater SJL.This suggests that despite controlling for time of testing, there may have been an effect present related to sleep phase delay leading to less average weekly sleep duration as measured by our models.Our findings focusing on fluid intelligence are most consistent with no association between chronotype and fluid intelligence (Ujma et al. 2020).The timing of tests may impact performance; a study done by van der Vinne et al. (2015) found that late chronotypes performed just as well as early chronotypes when exams were administered in the afternoon instead of in the morning.However, the influence of test times has been debated, as another study found that early chronotypes outperformed late chronotypes regardless of exam time (Enright and Refinetti 2017).We controlled for test times for all cognitive battery testing under the NIH Toolbox.
In line with our hypothesis regarding SJL and academic performance, we found that higher levels of SJL were associated with poorer grades.This finding supports previous research studies (Ferguson et al. 2018;Haraszti et al. 2014;Mirghani 2017;Montaruli et al. 2019;Shimura et al. 2022) conducted in diverse student populations, which have consistently demonstrated the impact of SJL on academic performance.A noteworthy distinction is that some studies [N = 247 undergraduates at Semmelweis University in Hungary, N = 1101 participants between 11 and 23 in Russia] (Borisenkov et al. 2010;Haraszti et al. 2014) have found higher levels of SJL to be associated with poorer grades, but not reduced sleep duration.Our study agrees in that we found SJL to be associated with poorer grades, but only marginally with reduced sleep duration.The association between SJL and grades can be attributed to the scheduling of classes during hours that may be more favorable for individuals with earlier chronotypes (e.g.08:00h to 17:00h), with numerous studies (Cohen-Zion and Shiloh 2018;Giannotti et al. 2002;Hahn et al. 2011) reporting that early chronotypes tend to perform better academically compared to their late chronotype counterparts because there is greater social jetlag for the majority of late chronotypes.For a minority of late chronotypes, these individuals may have accommodating schedules that reduce social jetlag and contribute to the heterogeneity of our findings.
We were surprised to find that lower BMI only marginally predicted higher average weekly sleep duration, higher NIH Toolbox battery scores (e.g.Oral Reading), and better past year grades.One study in 98 biochemistry college student sample with mean age of 21.7 at the University of Nebraska-Lincoln found normal BMI individuals had higher academic performance (GPA, ACT) than their overweight peers (Franz and Feresu 2013).However, our results are consistent with another group that found no relationship between BMI and academic performance (GPA) in a sample of N = 424 students with mean age of 15.44 in Taif city, Kingdom of Saudi Arabia (Alswat et al. 2017).Possible reasons for differences in our sample compared to the first study include differing age, as well as larger sample size.
Prediction of NIH Toolbox cognitive measures and academic performance in all models was performed while controlling for age and sex, in addition to BMI.Sex differences have been previously to have an impact on the effect of social jetlag on cognitive and academic performance in a sample of N = 796 adolescents aged 12-16 years old in Madrid, Spain (Díaz-Morales and Escribano 2015); social jetlag was found to be negatively associated with poorer cognition and academic performance in girls.However, there did not appear to be a significant link between social jetlag and age or sex in this ABCD cross-sectional sample.This may be related to cultural differences in the United States or the younger age group of this cross-sectional ABCD sample.
However, it is important to acknowledge that there are additional factors beyond the scope of our study that may influence these results.A number of groups have found that school start time has an important impact on predicting school performance for later chronotypes, especially in scientific disciplines such as physics (Goldin et al. 2020;Zerbini et al. 2017).Unfortunately, school start time was not available for participants in the ABCD study, so this important factor could not be controlled for so results must be cautiously interpreted with this caveat.In addition, light environment and light timing (e.g.absence or presence of light therapy) have been found to impact wellbeing and work performance (Brown et al. 2022;Mills et al. 2007).However, illuminance information was unavailable for participants in ABCD though this effect may well have varied from individual to individual.Geographical latitude, which affects daylight exposure and may vary across our sample's study sites, also plays a role, with individuals living at higher latitudes generally display increased social jetlag than do individuals at lower latitudes (Borisenkov et al. 2010;Chang and Jang 2019;Leocadio-Miguel et al. 2017).While participant locations were not linked to their data due to privacy restrictions, the location of the ABCD studies across the United States consisted of a variety of locations with both higher and lower latitudes, including the ( 1 2019).found that seasonal differences have been found to be associated with variation of grades with time of the year.The past year grades assessment in ABCD lacks the granularity to be able to detect seasonal variation of academic performance that others have found.
Effect sizes relating SJL, and cognitive and academic performance were small, and thus our findings may not replicate in smaller samples using analogous measures.Nonetheless, these small effects should not be ignored, particularly when considering the large number of children with late chronotype (Funder and Ozer 2019).Our effect sizes might also be in part explained by the fact that our sample consisted of younger participants.Previous research suggests that adolescents' chronotype shifts later during puberty, most notably around the age of 12-14 years (Díaz-Morales and Sorroche 2008).Therefore, it is reasonable to expect that the correlation between SJL and academic performance would be stronger in a study of older adolescents.Furthermore, our findings are consistent with previous research that reported a small overall effect size of 0.14 between SJL and school/university achievement in a meta-analysis of 31 papers (Tonetti et al. 2015).
Our study has the main strength that the ABCD sample is large, diverse, and well-characterized in terms of demographics, cognitive performance measures, and circadian rhythm parameters.The major limitation of our work is that our analysis is based on cross-sectional data that restricts our ability to establish causal relationships.Most similar studies on academic performance have been conducted with cross-sectional designs, and so it is not clear if chronotype plays a directional role in academic performance.In future studies, we plan to extend the analyses and re-assess these same factors in a longitudinal manner in the same participants as they transition into later school years.Time of administration for the MCTQ (e.g.summer versus school year) may impact scores.Additionally, considering the unprecedented impact of the COVID-19 pandemic on education, it would be valuable to explore how it has affected student sleep patterns and whether these changes are reflected in academic performance.
We found for the first-time using NIH Toolbox measures of crystallized and fluid intelligence that later chronotype and greater SJL predicted poorer cognitive performance, along with past year grades in a large national sample of youth.Our results showed no association between chronotype and fluid intelligence on the cognitive batteries completed.No relationship between BMI and academic performance was identified.There did not appear to be a significant link between social jetlag and age or sex.The impact of social jetlag as a potential driver of cognitive impairment influencing academic performance due to lack of sleep may play an explanatory role.The finding that chronotype and SJL predict poorer cognitive performance and grades in the past year is consistent with the fact that individuals are less likely to cognitively perform well in the context of ongoing sleep disruption and daytime sleepiness.Recognizing the importance of tailoring class schedules to accommodate the varying chronotypes of students could potentially improve academic performance, especially among individuals with later chronotypes; better understanding ideal sleep hygiene for different chronotypes is also very crucial for promoting overall wellbeing and academic success.

Table 4 (
Social Jetlag Oral Reading Predictive Hierarchical Regression Model): In Model 4, NIH Toolbox Age-Corrected Oral Reading Scores explained
*Incomplete data refers to the dataset with missing data pertaining to participant's predictor variables (past year grades, NIH Toolbox Oral Reading Scores, Social Jetlag) in Table2.more variance in social jetlag (sleep corrected) with a small effect size (ΔR 2 of 0.013); greater social jetlag was associated with lesser NIH Toolbox Age-Corrected Oral Reading Scores (B = −0.007,β =-0.085,SE = 0.001, p < 0.001, CI = −0.009 to −0.005) adjusted for time of testing.In Model 5, past year grades explained significantly more variance in social jetlag (sleep corrected) with a marginal-small effect size (ΔR 2 of 0.007); greater

Table 4 .
Social jetlag (sleep corrected) oral reading predictive hierarchical regression model.

Table 5 .
Household income (corrected for household members) oral reading predictive hierarchical regression model.

Table 6 .
Sleep duration oral reading predictive hierarchical regression model.

Table 7 .
BMI oral reading predictive hierarchical regression model.