Intraindividual variability in sleep duration and college degree attainment

ABSTRACT The objective of the current study was to examine the relationship between sleep characteristics and college degree attainment. Participants were 968 college students (72% female; mean age 19.7 [1.7]). Participants completed a psychosocial and sleep questionnaire battery followed by one week of daily sleep diaries. Academic degree completion data was obtained from the university registrar 10 years later. Logistic regression examined whether mean and variability in sleep duration and sleep efficiency and insomnia symptoms predicted degree attainment, adjusting for age, gender, semester, grade point average (GPA), and perceived stress. The strongest predictors of degree attainment were female gender (OR = 0.67), greater age (OR = 1.32), GPA (OR = 1.97), and lower intraindividual variability in sleep duration (OR = 0.99). Results highlight the importance of examining variability in sleep duration in addition to mean sleep duration in predicting college retention. Future research should use a combination of objective and subjective measures to explore the impact of sleep factors, including variability, on degree completion and other academic metrics.


Introduction
Poor sleep health across numerous dimensions (e.g., short duration, insomnia) is common among college students.A 2019 report from the American College Health Association of over 70 000 college students indicated 62% of students reported they received enough sleep to feel rested on 3 or fewer days per week and 65% felt tired, dragged out, or sleepy on 3 or more days per week (American College Health Association 2019).Approximately 26.9% of college students endorse insomnia symptoms (i.e., difficulty falling asleep, staying asleep, or waking up too early), and approximately 9.5% meet criteria for insomnia disorder (Taylor et al. 2013).Sleep health problems like short sleep duration and insomnia have been related to numerous poor physical and mental health outcomes (Taylor et al. 2003).However, sleep health is an underexplored area for health promotion among college students.
Poor sleep health may be directly or indirectly related to other disruptions of important aspects of students' lives, such as academic performance and college retention.The relationship between poor sleep health and worse academic performance (e.g., grade point average [GPA]) is well-documented (Creswell et al. 2023;Gaultney 2010;Hartmann and Prichard 2018;Hershner 2020;Lund et al. 2010;Okano et al. 2019;Taylor et al. 2013), though some studies have not found a relationship between questionnaire-assessed sleep parameters (e.g., mean sleep duration) and academic performance (Eliasson et al. 2010).Most prior literature has focused on questionnaire-derived sleep parameters.One study (N = 88) examined actigraphy-derived mean sleep parameters and found better sleep quality, longer sleep duration, and lower intraindividual variability in sleep duration were associated with better academic performance in an introductory college chemistry course (Okano et al. 2019).Another study (N = 557) found Fitbit-derived longer mean sleep duration, but no mean of or variability in any other sleep variables, was associated with higher GPA for first year college students (Creswell et al. 2023).
In prior literature, academic performance is typically measured via GPA, which is a limited metric of college success.College retention is a metric which may be of greater interest to colleges and students, and sleep health may be an important intervention point for reducing college dropout.One prior study of college students (N = 4,376) found students at elevated risk for sleep disorders in their first semester of college were more likely to leave the institution over the 3-year period of the study (Gaultney 2016).No studies have examined the association between specific sleep parameters and attainment of a college degree.Further information about which specific domains of sleep are related to degree attainment may give insight into the most salient intervention points for health promotion efforts.
One potentially important yet understudied area of sleep health is intraindividual variability in sleep, or how much a person's sleep fluctuates from night-to-night.Typically, studies examine retrospective or "typical" sleep patterns, which are subject to recall bias and may obscure important within-person variability.Greater intraindividual variability in sleep has been linked to a variety of negative health outcomes (e.g., depression, anxiety, poor physical health; Bei et al. 2016) and worse academic performance (e.g., lower GPA; Taylor et al. 2013), even after accounting for mean sleep patterns.Yet no studies have examined how variability in sleep may be a unique risk factor for college degree completion.
In the current study, we sought to determine whether sleep parameters (i.e., average total sleep time [TST], average sleep efficiency [SE], as well as intraindividual variability in these parameters) and insomnia symptoms predicted attainment of college degree after adjusting for covariates (i.e., stress, GPA, term of study completion, demographics).We hypothesized shorter average sleep duration, lower average sleep efficiency, greater variability in sleep duration and sleep efficiency, and greater insomnia symptom severity would predict lower odds of degree attainment, even after accounting for covariates.

Participants and procedures
Undergraduate participants (N = 968) between the ages of 18-25 were recruited from a large public university in Texas from the department of psychology research participant pool from fall 2006 through spring 2007.Students were compensated for completing the study with course credit.Demographic characteristics are presented in Table 1.
Participants consented to complete measures and allow researchers to access transcript data from the university registrar.At baseline, participants were invited into the lab to complete a paper-and-pencil psychosocial questionnaire battery and then completed one week of daily sleep diaries.The study was approved by the University of North Texas Institutional Review Board (approval #04-351).The parent data was previously published examining baseline relationships between sleep health and academic performance (Taylor et al. 2013).

Baseline demographic questionnaires
Participants completed information on demographic characteristics, including information on age, gender, race/ ethnicity, and academic year.We also noted the semester the study was completed (i.e., Fall 2006 or Spring 2007).

Insomnia Severity Index (ISI)
The ISI (Morin 1993) is a validated measure used to determine an individual's self-reported insomnia severity.The ISI includes seven items on a five-point Likert scale ranging from 0 to 4. Total scores range from 0 to 28, with higher scores indicating greater insomnia severity.The ISI yielded a coefficient alpha of .85 in the current study.

Perceived Stress Survey (PSS)
The PSS (Cohen et al. 1994) includes 10 items that measure how unpredictable, uncontrollable, and/or overloaded an individual has felt over the last month.
Responses range from 0 to 4, with total scores ranging from 0-40 and higher scores indicating greater stress.The PSS yielded a coefficient alpha of .88 in the current study.

Sleep diary
Participants were asked to complete daily paper-andpencil sleep diaries for one week using a sleep diary similar to the Consensus Sleep Diary (Carney et al. 2012).In the current study, two mean sleep parameters were calculated from the sleep diary across the week: total sleep time (TST) and sleep efficiency (SE; TST/ Time in Bed × 100).Intraindividual variability was calculated as the intraindividual standard deviation (SD) of SE and TST across the week of sleep diaries.

Grade point average
The grade point average (on a 0-4 scale) during the term that the student completed the study was obtained from the registrar.

Degree attainment
Degree attainment information was obtained from the registrar's office in 2018 (i.e., at least 10 years after the final participant's data was collected).Receiving a degree from the institution was coded as 1, and not receiving a degree was coded as 0.

Analyses
Starting from an original sample size of N = 1032, we retained participants for whom we were able to obtain registrar-derived degree attainment data (N = 968).No variables included in the models were missing more than 4% of datapoints.We computed means, counts, and standard deviations for all relevant variables (Table 1).Then, we conducted multiple imputation (25 iterations) for missing data on all variables.Findings for the logistic regression were similar on the multiply imputed and non-imputed datasets; we reported the findings for the pooled imputed datasets.
Bivariate correlations were run to examine all variables' relationships (Supplemental Table S1).Next, logistic regression was conducted to examine whether mean and intraindividual variability in TST and SE and insomnia symptom severity were associated with degree attainment after adjusting for perceived stress, GPA, semester the study was completed, age, and gender.We reported structure coefficients (bivariate correlation between the independent and dependent variables divided by the total model R) to quantify the amount each independent variable contributes to overall model variance per recommendations by Courville and Thompson (2001) and Henson (2002).We used IBM SPSS Version 23 for all analyses.
In contrast to some prior research, mean sleep duration and intraindividual variability in sleep duration were not correlated in our study.This may be due to the fact we had a relatively healthy sample.Mean and intraindividual variability in sleep duration may be more strongly correlated in those with underlying sleep disorders.Insomnia symptoms and perceived stress demonstrated substantial overlap with one another and both accounted for a modest amount of variance in odds of degree attainment.This overlap may explain why neither was statistically significant in the final model, though they may still be conceptually important in this relationship.The findings suggest future research should consider a focus on increasing the regularity of sleep duration on a night-tonight basis in addition to the traditional approach of emphasizing the average duration of sleep as a potential pathway to improve college retention.Intraindividual variability in total sleep time may warrant unique intervention, but more research is needed to determine intervention targets and timing, and to confirm the importance of this construct in predicting academic performance.
In contrast to our hypotheses, neither mean sleep duration nor mean or variability in sleep efficiency significantly predicted odds of degree attainment.These findings are somewhat at odds with prior literature that has found a relationship between actigraphy/Fitbitderived longer sleep duration and GPA (Creswell et al. 2023;Okano et al. 2019).This discrepancy may be related to our use of self-report, as opposed to inferred, measures of sleep duration which produce overlapping yet distinct sleep parameters, or due to the use of a much more distal outcome (i.e., degree attainment vs. GPA a few months later).Variability in sleep efficiency did account for a modest amount of variance in degree attainment.Significantly reduced sleep efficiency (e.g., <85%) is a hallmark of insomnia disorder, which is common in college samples (Taylor et al. 2013).Variability in sleep efficiency may not have been driving a factor contributing to likelihood of degree attainment because of a restriction of range; less than 16% of the sample had sleep efficiency ≤ 85%.
One key strength of this research is the longitudinal measurement strategy.To assess sleep variables, we used validated sleep disorders questionnaires and a sleep diary, the most precise method of subjective sleep assessment (Dietch and Taylor 2021), which is unusual in studies of sleep and academic outcomes.This approach allowed us to examine variability in sleep parameters in addition to mean values, which revealed novel insights about the role of variability in sleep duration in degree attainment.Further, the degree attainment data was derived from an objective source (i.e., the registrar) which increases confidence in data accuracy over self-report.
This research also demonstrated key limitations.Only one week of sleep diary was collected, which limits the extent to which we can assume that these sleep data represent habitual sleep in this sample.Sleep diary was collected in a paper-and-pencil format, meaning that no timestamps for completion were available and thus "backfilling" may have occurred and increased measurement error.Further, the use of a single site for data collection means that, although the demographics of the sample were similar to the university, generalizability beyond the university is limited.The age range of the sample was limited to 18-25, which further limits generalizability and reduces the practical relevance of considering age as a relevant factor in predicting degree completion.Although use of a psychology subject pool as a recruitment source can also offer some limits to generalizability, majors in the sample were quite diverse (with only 31% reporting a psychology major).The use of registrar-reported degree attainment does not capture students who may have transferred universities and attained degrees elsewhere or at a time beyond the study's 10-year time horizon, and limited information was available from the registrar at follow-up.Finally, the study design did not allow us to rule out potential cohort effects.Future research should examine the relationship between sleep parameters and degree attainment across multiple universities with focus on purposefully increasing sample diversity.Prior research has been limited by measurement strategies and scope of both sleep health parameters and metrics of academic performance and success.Examining reasons for dropout and mediators of the relationship between sleep health and degree attainment may provide insight into students' support needs.Future projects should use a combination of objective and self-reported sleep health assessment methods and should continue to examine both mean and variability of sleep health parameters in the prediction of academic outcomes.

Disclosure statement
No potential conflict of interest was reported by the author(s).
*Due to small cell sizes, this variable will only be displayed in aggregate.Data presented in Table1is based on actual values, not imputed data.