Convenience sampling through social media, flyers and posts online was used to recruit participants. Some students were also recruited through the Department of Psychology’s research participant pool, made up of all full-time psychology students from Maynooth University in their second year of study. Participants had to be at least 18 years old to partake in the study and to have clearly demarcated “work” and “free” days as part of their usual weekly schedule. Due to the COVID-19 pandemic no responses were collected between March and September 2020 to control for the changing nature of the work environment and the impact of the national lockdowns; after this period had passed and individuals had acclimatised to the new schedule, recruitment recommenced and concluded October 2020. 468 individuals completed an online questionnaire; of these responses, 17 participants were excluded due to work schedules at night, 49 further responses were excluded due to either alarm clock use on free days or stating no free days, two further participants were excluded due to unrealistic values being provided. The final sample with complete data was N = 400; 57% (N = 228) of the final sample data was collected before March 2020, and forty-three% of the sample data (N = 172) was collected in September/October 2020. This study was approved by the Biological Research Ethics Sub-Committee at Maynooth University (BSRESC-2019-009) and all research was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants in order to proceed to the surveys.

All participants reported their age (in years), gender, current job-status (employed/unemployed/retired/student) and whether they worked full-time or part-time. Commute time to work and normal work start and end times were reported. Participants selected an occupation from a drop-down menu guided by the International Standard Classification of Occupations (ISCO, 2008). Participants were also presented with an open-ended box to describe their job so that classifications could be checked by the researcher. This classification system was then used to group participants as full-time workers, students, or non-traditional workers where they may not have had regular hours (i.e., homemaker, carers).

The Munich Chronotype Questionnaire (MCTQ) was used to estimate participants’ underlying circadian phase of entrainment and SJL [15, 16]. This instrument asks about typical sleep behaviour on both workdays and free days over the previous month and has two sections with illustrative diagrams. The key variables of interest derived from MCTQ scores were average sleep duration across the week (SDweek), sleep-corrected mid sleep on free days (MSFsc, an indicator of chronotype), and social jetlag (SJL; the difference between the timing of midsleep on “work” days and “free days” [12, 16]). SJL was calculated as both a relative (with directionality as to whether free day sleep timing was earlier or later than workdays) and absolute values. The Pittsburgh Sleep Quality Index (PSQI; [17]), was used to evaluate subjective sleep quality over the previous month. The main output of the PSQI is a total score indicating subjective sleep quality, with higher scores indicating poorer sleep quality. The Perceived Stress Scale (PSS) measures perception of stress by assessing individual’s feelings and thoughts over the previous month [18]. The PSS-10 has 10 items that are all rated on a 5-point Likert scale from 0 (never) to 4 (very often). Items 4,5,7,8 were positively worded and were reversed before the total score was calculated, and total scores can range from 0 to 40 with higher scores indicate higher psychological stress [19].

When participants opened the survey link they were presented with detailed information on the study covering the purpose of the study, confidentiality, and data handling policies. Anyone under the age of 18 was automatically directed to the end of the survey. After consenting to proceed, participants submitted demographic details, information on work schedule and occupation before proceeding to complete the MCTQ, the PSQI and the PSS. All of these questionnaires had to be completed in one session and the participants required around 15 min to complete the surveys. The online Qualtrics instrument only saved responses that were 100% complete. Data collected and used in this study, as well as the STROBE checklist for observational cross-sectional studies and results of statistical analyses, are available at https://osf.io/ba8gf/.

For descriptive data analyses, the distribution of all variables was visually inspected and Kolmogorov Smirnov tests were used to assess normality of distribution. For inferential analysis, one-way ANOVAs, ANCOVAs or Kruskal Wallis tests were used to examine differences between groups (student, employed, career/homemaker) for demographic, sleep and circadian variables depending on normality of the dependent variable distribution. Following AN(C)OVA testing, for statistically significant results post-hoc testing was applied with Bonferroni tests. When assessing relationships between variables Pearson’s r was used for normally distributed variables while Spearman’s rho was used when the data was non-normally distributed. As the approach of the study was exploratory and not hypothesis testing, adjustments for multiple hypothesis testing were not applied, and P < 0.05 was taken as indicating statistically significant differences. Following correlation analysis False Discovery Rate (FDR) analysis was applied by the Benjamin-Hochberg method with alpha set to 0.05 to determine whether statistically significant results persisted after control for FDR in the 21 associations examined. All analyses were conducted on SPSS version 26. JASP version 0.14.1.0. To inform the study sample size, an online statistics calculator (https://www.danielsoper.com/statcalc/default.aspx) was used to guide the minimum sample size. In order to detect a small effect size (0.02) of SJL as a predictor on PSS scores with an alpha level of 0.05 and a power of 0.8. in a presumptive regression analysis, a suggested minimum sample size was 390.This approach was taken on the a-priori assumption that SJL would have a weaker association with PSS than other factors such as PSQI score, and as such if the study were powered sufficiently to detect associations of SJL with PSS, then it would also be powered to detect other associations of interest.

Path Analysis was conducted in SPSS AMOS (IBM Corporation) to test conceptually driven direct and indirect paths onto PSS from the endogenous variables that were identified in univariate analysis as statistically significantly associating with PSS. Maximum likelihood estimation was used in the model, and since there were no missing data points no imputation was not required. Bootstrapping with 1,000 iterations was applied to allow for calculation of bias-corrected 95% confidence intervals for direct and indirect effects. Model fit indices were interpreted as per Schermelleh-Engel et al. [20]. The pathways specified in the model one were: (1) age influences MSFsc (as described widely [12]); (2) MSFsc influences SJL [21]; (3) both MSFsc and SJL have direct effects on PSS [14]; (4) sleep quality, assessed by the PSQI, has a direct effect on PSS [8] and conversely PSS scores exert effects on PSQI scores [22]; (5) average nightly sleep duration, as determined from the MCTQ, will exert an effect on PSQI score [17] and (6) age and sex have direct effects on PSS [23]. As the model specified is non-recursive due to the reciprocal relationship between PSQI and PSS, the error terms for PSQI and PSS were specified to co-vary. The overall aim of the path analysis was to investigate the inter-dependencies of MSFsc, SJL, sleep duration and sleep quality on PSS, and not to best account for variance in PSS; as such, we were primarily interested in describing the direct and indirect paths of interest rather than optimising model fit or maximising the variance in the stress measures accounted for.