3.4 million real-world learning management system logins reveal the majority of students experience social jet lag correlated with decreased performance

Abstract

Misalignments between endogenous circadian rhythms and the built environment (i.e., social jet lag, SJL) result in learning and attention deficits. Currently, there is no way to assess the impact of SJL on learning outcomes of large populations as a response to schedule choices, let alone to assess which individuals are most negatively impacted by these choices. We analyzed two years of learning management system login events for 14,894 Northeastern Illinois University (NEIU) students to investigate the capacity of such systems as tools for mapping the impact of SJL over large populations while maintaining the ability to generate insights about individuals. Personal daily activity profiles were validated against known biological timing effects, and revealed a majority of students experience more than 30 minutes of SJL on average, with greater amplitude correlating strongly with a significant decrease in academic performance, especially in people with later apparent chronotypes. Our findings demonstrate that online records can be used to map individual- and population-level SJL, allow deep mining for patterns across demographics, and could guide schedule choices in an effort to minimize SJL's negative impact on learning outcomes.

Figure 1

Student LMS logins showed signs that they contain circadian rhythm information, and social disruptions thereof. A double-plotted actogram (or raster plot) of login times from an example student across one semester (A) showed daily rhythms with an apparent sleep window between 23:00 and 06:00. Each student’s activity generated a unique hourly histogram (B, single row). When arranged by average phase of activity, a range of activity phases were apparent from early (B, top) to late (B, bottom), though all students shared an apparent sleep window in the late night and early morning. A histogram for all students on days with classes (C, red) showed crenulations that aligned to class start times (C, grey area). The same histogram for non-class days (C, blue) showed a delayed phase and no such crenulations. Consistent with the resulting supposition that non-class days are more representative of circadian rhythms, and class days more disruptive to circadian rhythms, histograms of activity on non-class days still showed a range of activity phases when sorted by average non-class day phase of activity (D), whereas the same ordering of individuals for class day activity (E) showed activity on class days to be largely during class times, and devoid of apparent chronotypes.

Figure 2

Non-class day activity distributions changed in ways expected of human circadian rhythms. Non-class day histograms sorted by gender (A) revealed that men (red) are more likely to stay active later than women (blue), with significantly increased activity between midnight and 06:00, while women showed increased activity in the evening (18:00–24:00). Sorted by decade of life (age, B), older students had significantly advanced phases, apparent both in earlier activity onset times (06:00–12:00), and decreased activity in the evening (18:00–24:00) for each additional decade of life. Sorted by season (C), there was a significant difference between fall semesters (blue) and spring semesters (red). Both fall semesters showed a consolidation of activity in the middle of the day, whereas both spring semesters showed a broader distribution of activity into the morning and evening. *p ≤ 0.05.

Figure 3

Social jet lag correlated with decreased academic performance for both advances and delays. Some students delayed from average non-class days (blue) to average class days (red) (A, left), some changed phase less than half an hour, on average, between non-class and class days (A, center), and some students advanced from average non-class days to average class days (A, right) (hourly histograms from example individuals shown for each condition). If SJL is calculated by linear subtraction (B), then amplitude of SJL showed a significant negative correlation with GPA for students who advanced on class days and a non-significant trend of correlation was apparent for students who delayed on class days. If SJL is calculated on a log scale (C), then amplitude of SJL showed a significant negative correlation with GPA for students who advanced on class days, and for students who delayed on class days (for both B and C, 24 groups are used, so that if SJL were random, 1 group would appear per hour of potential SJL).

Figure 4

Owls had an academic disadvantage regardless of class start times. Average distribution of activity across non-class days for larks (blue), finches (purple), and owls (red) (A) revealed significant differences across the three chronotype categories. Owls experienced greater average SJL between non-class days and class days (B, red, linear SJL) than either larks (blue) or finches (purple). Owls also took a significantly higher proportion of morning classes than larks or finches (C) did. Analysis of class grades as a function of both chronotype category and class start time (D) revealed a significant disadvantage for owls across the day and a significant increase in class grade across the day for all chronotype categories. When time of day was normalized across all chronotype categories (E), then there is no longer an effect of time of day, and owls showed a significant disadvantage compared to larks and finches at all times of day. In plots D and E, the lowest bar represents the average of all morning class grades taken by individuals classified as owls. *p ≤ 0.05.