학술

Sleep hygiene, caffeine use, and academic performance among university students: A cross-sectional study

PLOS ONE
CC BY
이 매체는 공공·자유 라이선스로 본문을 직접 표시합니다.

Figures
Abstract
Background
Caffeine use is common among university students, but its associations with sleep hygiene and academic performance remain uncertain. This study examined these associations among Palestinian university students.
Methods
A cross-sectional questionnaire-based study was conducted from March 21 to May 28, 2026. Among 379 participants, continuous grade point average (GPA) was available for 354. Measures included estimated caffeine intake, caffeine timing, Sleep Hygiene Index (SHI) score, sleep duration, sleep regularity, and study hours. Associations were assessed using analysis of variance, correlation, and multivariable linear regression. Caffeine intake was entered into the regression model as ln(mg/day + 1).
Results
The mean age was 20.71 ± 1.98 years, and 50.1% were female. Mean Sleep Hygiene Score was 23.32 ± 8.34, with acceptable internal consistency (Cronbach’s α = 0.775). Estimated caffeine intake averaged 253.5 ± 180.9 mg/day, with a median of 223 mg/day. Sleep Hygiene Score was positively correlated with raw caffeine intake (Pearson r = .282, p < .001; Spearman rho = .310, p < .001) and log-transformed intake (Pearson r = .210, p < .001). Mean GPA was 3.03 ± 0.59. No statistically significant association was detected between GPA and caffeine category (p = .584), log-transformed caffeine intake (r = −.084, p = .116), raw caffeine intake (Spearman rho = −.098, p = .066), or Sleep Hygiene Score (r = −.061, p = .249). In multivariable regression (n = 352, adjusted R² = 0.123), study hours, sleep regularity, sleep duration, academic year, academic major, and smoking status were independently associated with GPA; caffeine intake was not (B = −0.026, 95% CI [−0.071, 0.020], p = .270).
Citation: Alshaikh R, Jebriel W, Shahrouri M, Zarura T, Khouli H (2026) Sleep hygiene, caffeine use, and academic performance among university students: A cross-sectional study. PLoS One 21(11): e0363064. https://doi.org/10.1371/journal.pone.0363064
Editor: Ana Cristina Mafla, Universidad Cooperativa De Colombia - Pasto, COLOMBIA
Received: July 20, 2026; Accepted: November 3, 2026; Published: November 20, 2026
Copyright: © 2026 Alshaikh et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The de-identified dataset underlying this study, together with a data dictionary, is publicly available in Zenodo at https://doi.org/10.5281/zenodo.23452065.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: A1, A2A, Adenosine receptor subtypes A1 and A2A; ANOVA, Analysis of variance; B, Unstandardized regression coefficient; CI, Confidence interval; F, F-statistic; GPA, Grade point average; IRB, Institutional Review Board; n, Sample size; n.s., Not significant; p, Probability value; QR, Quick Response (code); r, Pearson correlation coefficient; R², Coefficient of determination; SD, Standard deviation; SPSS, Statistical Package for the Social Sciences; VIF, Variance inflation factor
1. Introduction
Caffeine is the most commonly used psychoactive substance among university and medical students worldwide, obtained primarily from coffee, tea, and other caffeinated beverages [1,2]. Its use is closely tied to long study hours, examination pressure, irregular sleep schedules, and the need to sustain alertness during periods of fatigue [3–5]. Its effects are mediated primarily through antagonism of adenosine A1 and A2A receptors in the central nervous system, which increases the release of neurotransmitters such as dopamine, norepinephrine, and acetylcholine and produces greater alertness, attention, and reaction speed in fatigued individuals [1]. Given the demanding and irregular schedules of university life, students may use caffeine to counter daytime sleepiness and remain awake [6].
Yet the relationship between caffeine and academic outcomes is not straightforward. A recent study among second-year medical students found a dose-dependent, non-linear association between coffee consumption and examination performance, with moderate consumers (one to two cups per day) outperforming both non-consumers and heavier consumers. In the same study, scores also differed by sleep duration and self-reported adverse effects of caffeine, suggesting the effect of caffeine depends on the sleep context in which it is consumed [2]. This is biologically plausible: caffeine has a mean half-life of approximately four hours, and even a single dose close to bedtime can increase sleep latency and reduce total sleep time [7], potentially at the cost of the sleep needed for memory consolidation and learning [8]. In contrast, a recent systematic review found the evidence linking regular caffeine use to subjective sleep quality to be limited and inconsistent [9]. This has led to growing interest in sleep hygiene, which encompasses the behaviors and environmental factors that influence sleep quality and consistency. Unlike sleep duration alone, sleep hygiene provides a broader perspective on sleep-related habits and may better capture their overall relationship with academic performance.
Existing research linking caffeine, sleep, and academic performance has several limitations [6,8,9]. Many studies rely on bivariate comparisons that do not adjust for potential confounders such as study habits, sleep irregularity, and academic year, while even adjusted analyses have often omitted variables such as study aptitude and stress [2]. Additionally, sleep has often been measured only as self-reported duration rather than with validated multidimensional instruments such as the Sleep Hygiene Index [10,11], and much of the evidence comes from single-cohort or single-discipline samples, limiting generalizability [12–14]. This study assessed caffeine exposure and sleep hygiene among Palestinian university students and examined their associations with continuous GPA. Multivariable linear regression was used to estimate adjusted associations with GPA while accounting for demographic characteristics, academic characteristics, residence, smoking status, sleep behaviors, and study hours.
2. Methods
2.1 Study design, setting, and period
A cross-sectional study was conducted from March 21, 2026 to May 28, 2026 to examine the association between sleep hygiene, caffeine exposure, and academic performance among university students. Participants were recruited from An-Najah National University and other Palestinian universities.
2.2 Study population and eligibility criteria
The target population included undergraduate and postgraduate university students of both sexes. Using convenience sampling, eligible participants were university students aged 18 years or older. A total of 385 responses were collected across all recruitment channels. Of these, 6 were excluded prior to analysis (2 duplicate responses, identified and removed after data collection was complete, and 4 incomplete or invalid questionnaires), yielding a final analytic sample of 379 participants. Fig 1 summarizes participant recruitment and exclusions. Continuous GPA was available for 354 of these participants (93.4%); the remaining 25 (6.6%) had no recorded cumulative GPA, most commonly because they were first-year students who had not yet completed enough coursework to have one. During data cleaning, these 25 missing GPA values were found to have been miscoded as the lowest GPA category rather than as missing; this was identified and corrected prior to analysis (Section 2.7). The final multivariable regression model was based on 352 participants with complete data on all model predictors.
Of 385 responses collected across paper, QR code, online link, and WhatsApp channels, 6 were excluded (2 duplicates, 4 incomplete/invalid), yielding a final analytic sample of 379. Continuous GPA was available for 354 participants (93.4%); the remaining 25 (6.6%) had no recorded GPA, most commonly reflecting first-year students without sufficient completed coursework. The final multivariable regression model included 352 participants with complete data on all model predictors.
2.3 Sampling method and recruitment
A convenience sampling method was used. Students were recruited on campus and invited to complete a paper questionnaire or to access the survey electronically by scanning a QR code or following a direct link. The survey was also distributed through WhatsApp groups and student networks at several Palestinian universities to broaden participation beyond a single institution or discipline.
This exploratory study used convenience sampling, and no a priori sample-size calculation was performed. The analyses included all eligible responses remaining after data cleaning. Estimates, particularly those involving small subgroups, should be interpreted in light of their confidence intervals.
2.4 Data collection tool
Data were collected using a self-administered questionnaire, available in both paper and electronic (Google Forms) formats with identical items. The questionnaire was adapted from previously published and validated instruments, including the Sleep Hygiene Index [10], with additional items developed for this study. It captured demographic and academic characteristics, self-reported GPA (recorded both as a continuous value and as a categorical range), sleep hygiene practices, caffeine intake, sleep duration, sleep schedule regularity, and study habits. The questionnaire underwent forward translation into Arabic by a public health professor and back-translation into English. Discrepancies were reviewed by public health experts in An-Najah National University. The Arabic questionnaire was piloted with 25 students, and minor wording changes were made to improve clarity. The SHI comprised 13 items rated from 0 (never) to 4 (always). Item scores were summed to produce a total score ranging from 0 to 52, with higher scores indicating poorer sleep hygiene. The 25 pilot participants were included in the final analytic sample. Completion was voluntary and anonymous; no personally identifiable information, such as names or email addresses, was collected.
2.5 Study variables
The primary outcome was academic performance, measured as continuous grade point average (GPA); a four-category GPA range (less than 2.00, 2.00 to 2.99, 3.00 to 3.49, 3.50 to 4.00) was also derived and used for descriptive purposes. Explanatory variables included Sleep Hygiene Index score (with higher scores indicating poorer sleep hygiene), estimated daily caffeine intake in milligrams (used as a continuous, log-transformed measure in the regression model, and also categorized into six consumption levels for descriptive purposes), timing of the last caffeinated drink before sleep, average daily sleep duration, sleep schedule regularity, daily study hours, academic year, academic major, residence, age, sex, and smoking status.
2.6 Ethical considerations
2.7 Statistical analysis
Statistical analyses were performed using IBM SPSS Statistics version 27. Participant characteristics and study variables were summarized using descriptive statistics: means and standard deviations for continuous variables, and frequencies and percentages for categorical variables. Internal consistency of the Sleep Hygiene Index was assessed using Cronbach’s alpha.
During data cleaning, a coding error was identified in which participants with a missing continuous GPA had been defaulted to the lowest GPA category rather than coded as system-missing; this was corrected before analysis, and continuous GPA (n = 354) was used as the primary outcome throughout. Responses from all recruitment channels were screened for potential duplicates by comparing response patterns across items. Two records were excluded because they were identical to another record across every questionnaire item, indicating the same respondent had submitted the survey more than once. To reduce duplicate participation during data collection, the questionnaire included an instruction asking respondents not to complete the survey again if they had already done so. Because responses were anonymous, duplicate participation could not be excluded with certainty. Reported GPA values were recorded on the 0–100% scale. Values for some participants required conversion to the 0–4 scale using the university’s letter grading and then converting that letter grading to the 4.00 equivalent (for example, a 95% in Birzeit University is an A, and an A at An-Najah is 4.00). GPA categories were assigned after this conversion. All GPA values were recorded using the 4.00-point scale. GPA analyses included participants with recorded continuous GPA (n = 354). Multivariable regression used complete-case analysis; two additional participants were excluded because of missing values for residence, leaving 352 participants.
Estimated daily caffeine intake was calculated from reported beverage consumption using standard reference values for common caffeinated drinks: brewed coffee, approximately 95 mg per 240 mL cup; tea, approximately 47 mg per cup; and energy drinks, approximately 80 mg per 250 mL can [7,15]. Additional caffeine sources reported under an open-ended “other sources” item were assigned fixed per-serving values based on published reference ranges: instant coffee (e.g., Nescafé), 63 mg; cappuccino, 75 mg; iced coffee, 95 mg; cola, 34 mg; pre-workout supplements, 150 mg; caffeine pills, 200 mg; and caffeine-containing analgesic tablets (e.g., Panadol Extra), 65 mg. Total daily caffeine intake was calculated by multiplying the reported frequency of each source by its estimated caffeine content and summing across all sources. Participants reporting no caffeine consumption were retained as 0 mg/day. Because caffeine content varies by serving size, brand, and preparation method, these values should be interpreted as estimated rather than directly measured intake. For descriptive purposes, estimated daily caffeine intake was categorized into six consumption levels: no consumption (0 mg/day), very low (1–49 mg/day), low (50–99 mg/day), moderate (100–199 mg/day), high (200–399 mg/day), and very high (≥400 mg/day), consistent with caffeine intake thresholds referenced by the European Food Safety Authority [7]. The questionnaire items and the beverage-specific conversion rules, including the recall period (past month), response options (0, 1, 2, or more than 3 servings per day), and the handling of open-ended “other sources”, are provided in S1 Appendix.
Estimated daily caffeine intake (mg) was strongly right-skewed (skewness = 1.13, SE = 0.13; Shapiro-Wilk W(379) = 0.924, p < .001) and was therefore log-transformed (ln[mg + 1]) prior to correlation and regression analyses. This continuous, log-transformed measure was used in place of the six-category caffeine consumption variable in the multivariable regression model; the categorical variable was retained for descriptive reporting (Table 2) and as a secondary bivariate comparison (Table 4).
Bivariate associations between continuous GPA and categorical study variables were assessed using one-way ANOVA; associations with continuous variables (age, Sleep Hygiene Score, and estimated caffeine intake) were assessed using Pearson correlation, with Spearman correlation reported alongside Pearson for the untransformed caffeine intake variable given its skewed distribution. Where Levene’s test indicated that the assumption of homogeneity of variance was violated, the Welch-corrected F-test is reported alongside the standard ANOVA.
Multivariable linear regression was used to estimate adjusted associations with GPA. The model included age, sex, academic major, academic year, residence, average daily sleep duration, sleep schedule regularity, daily study hours, log-transformed estimated caffeine intake, SHI score, and smoking status. Categorical variables were represented by indicator variables, with reference categories specified in Table 5. City and third academic year were used as reference categories for residence and academic year, respectively, to facilitate comparisons with larger groups. Multicollinearity was assessed using variance inflation factors (VIFs). Individual dummy-variable VIFs depend on the coding of categorical variables and were interpreted alongside subgroup sizes and confidence intervals. Model assumptions were assessed using a normal probability plot of standardized residuals and a plot of standardized residuals against standardized predicted values. All tests were two-sided, with p < 0.05 considered statistically significant.
3. Results
3.1 Participant characteristics
A total of 379 university students participated in this study; their demographic and academic characteristics are summarized in Table 1. The mean age was 20.71 ± 1.98 years. The sample was almost equally distributed between males (189, 49.9%) and females (190, 50.1%). Engineering and Technology students represented the largest academic group (109, 28.8%), followed by Medicine (95, 25.1%) and Dentistry (64, 16.9%). Most participants were in their third academic year (142, 37.5%).
Continuous GPA was available for 354 participants (93.4%). GPA was missing for the remaining 25 participants (6.6%); a data check found that this missingness was concentrated among first-year students, consistent with many not yet having completed enough coursework to have an official cumulative GPA. A formal comparison confirmed this pattern: participants missing GPA were younger by an average of 1.28 years, (95% CI [0.48, 2.07], p = .002, Cohen’s d = 0.65) and more often first-year students (60.0% vs. 11.6%, χ²(5) = 45.2, p < .001), female (72.0% vs. 48.6%, χ²(1) = 5.12, p = .024), and reporting an irregular sleep schedule most of the time (72.0% vs. 40.4%, χ²(3) = 9.65, p = .022) than those with recorded GPA. They did not differ significantly in estimated caffeine intake, log-transformed caffeine intake, Sleep Hygiene Score, sleep duration, daily study hours, academic major, residence, or smoking status (all p ≥ .07). During data cleaning, a coding error was also identified in which these missing GPA values had originally been defaulted to the lowest GPA category rather than coded as missing; this was corrected prior to analysis. The mean GPA was 3.03 ± 0.59 (median 3.00, range 1.00 to 4.00). Using the corrected categorical variable, the most common GPA category was 3.00 to 3.49 (127, 35.9%), followed by 2.00 to 2.99 (119, 33.6%), 3.50 to 4.00 (94, 26.6%), and less than 2.00 (14, 4.0%).
The Sleep Hygiene Index demonstrated acceptable internal consistency (Cronbach’s alpha = 0.775). The mean Sleep Hygiene Score was 23.32 ± 8.34.
Table 1 summarizes the demographic characteristics and academic performance distribution of the study participants.
3.2 Sleep and caffeine characteristics
The most commonly reported sleep duration was 6–7 hours per night (30.3%), followed by 7–8 hours (24.5%) and 5–6 hours (21.9%). Most students reported irregular sleep schedules, with 42.5% indicating that their sleep schedule was irregular most of the time.
Regarding caffeine consumption, high caffeine intake was the most frequently reported category (35.1%), followed by moderate consumption (26.6%). A substantial proportion of participants reported consuming their last caffeinated beverage within three hours before sleep (160, 42.2%). Estimated daily caffeine intake was available for all 379 participants. The mean intake was 253.5 ± 180.9 mg/day, with a median of 223 mg/day and an interquartile range of 128–350 mg/day. Intake ranged from 0 to 944 mg/day, and 20 participants (5.3%) reported no caffeine intake. The distribution was right-skewed (skewness = 1.132, SE = 0.125; Shapiro–Wilk W = 0.924, p < .001), so caffeine intake was log-transformed using ln(mg/day + 1) for correlation and regression analyses.
Table 2 presents the distribution of sleep duration, sleep schedule regularity, timing of caffeine intake, and overall caffeine consumption among the participants.
3.3 Association between sleep hygiene and sleep/caffeine variables
Significant differences in Sleep Hygiene Scores were observed across categories of sleep duration (F(4,374) = 6.565, p < .001), sleep schedule regularity (F(3,375) = 36.610, p < .001), caffeine consumption category (F(5,373) = 7.763, p < .001), and timing of the last caffeine drink before sleep (F(3,375) = 7.073, p < .001).
Descriptively, mean SHI scores were lowest among students with very regular sleep schedules and highest among those with irregular sleep every day. The very-high-caffeine group had the highest mean SHI score, although the category means were not strictly monotonic. Students reporting caffeine consumption within three hours of bedtime had the highest mean SHI score among the caffeine-timing groups. These descriptive comparisons should not be interpreted as statistically significant pairwise differences because post hoc comparisons were not performed. SHI score was positively correlated with raw estimated caffeine intake (Pearson r = 0.282, p < 0.001; Spearman rho = 0.310, p < 0.001) and log-transformed intake (Pearson r = 0.210, p < 0.001).
Table 3 presents SHI scores across sleep-duration, sleep-regularity, caffeine-timing, and caffeine-consumption categories.
3.4 Association between GPA and study variables
GPA did not differ significantly by caffeine consumption category (F(5,348) = 0.754, p = .584) or by average daily sleep duration (F(4,349) = 0.714, p = .583). GPA was also not significantly correlated with age (r = .069, p = .194, n = 354), Sleep Hygiene Score (r = −.061, p = .249, n = 354), log-transformed caffeine intake (Pearson r = −.084, p = .116, n = 354), or raw caffeine intake (Pearson r = −.090, p = .089; Spearman rho = −.098, p = .066; n = 354).
GPA differed by sleep schedule regularity, although the standard ANOVA result was only borderline (F(3,350) = 2.558, p = .055); because Levene’s test suggested unequal variances were plausible for this comparison, the Welch-corrected test was also examined and was significant (Welch F(3, 87.0) = 4.241, p = .008). Students with very regular sleep schedules had the highest mean GPA (3.32 ± 0.40), while those with irregular sleep every day had the lowest (2.92 ± 0.58).
GPA differed significantly by academic year (F(5,348) = 2.843, p = .016; Welch F(5, 80.8) = 3.027, p = .015), with first-year students showing the lowest mean GPA (2.79 ± 0.74) and fifth-year students the highest (3.20 ± 0.37).
GPA differed significantly by smoking status (F(1,352) = 6.819, p = .009), with non-smokers reporting a higher mean GPA (3.09 ± 0.59) than smokers (2.92 ± 0.57).
GPA did not differ significantly by sex (F(1,352) = 0.255, p =.614).
GPA also differed significantly by daily study hours (F(3,350) = 5.315, p = .001; Welch F(3, 157.5) = 5.344, p = .002), increasing from 2.91 ± 0.58 among students studying less than two hours a day to 3.18 ± 0.59 among those studying more than four hours a day.
GPA differed significantly by academic major (F(5,348) = 3.390, p = .005), with the lowest mean GPA among Medicine students (2.84 ± 0.65) and the highest among Dentistry students (3.15 ± 0.50), though this association did not persist for any individual major after adjustment in the multivariable model.
Table 4 summarizes these associations. Fig 2 shows the relationship between Sleep Hygiene Score and GPA directly; consistent with the null correlation reported above, no clear linear pattern is apparent.
Each point represents one participant with complete data on both variables (n = 354). The points form a dense, unstructured cloud with no visible linear trend, consistent with the near-zero, non-significant correlation reported above (r = −.061, p = .249).
3.5 Multivariable linear regression analysis
The multivariable linear regression model included 352 participants with complete data on GPA and all model predictors. Estimated caffeine intake was entered as ln(mg/day + 1). City and third academic year were used as the reference categories for residence and academic year, respectively. The overall model was statistically significant (F(27,324) = 2.816, p < 0.001), with R² = 0.190 and adjusted R² = 0.123.
Several adjusted category contrasts were statistically significant. Students studying fewer than two hours a day had significantly lower GPA than those studying more than four hours a day (B = −0.305, 95% CI [−0.489, −0.121], p = .001), as did those studying two to three hours a day (B = −0.199, 95% CI [−0.395, −0.004], p = .046). A very regular sleep schedule was associated with significantly higher GPA relative to an irregular schedule every day (B = 0.387, 95% CI [0.078, 0.696], p = .014). Sleeping fewer than five hours a night was associated with significantly higher GPA relative to sleeping more than eight hours (B = 0.313, 95% CI [0.056, 0.569], p = .017). Students majoring in Medicine had significantly lower GPA than those majoring in Dentistry (B = −0.224, 95% CI [−0.418, −0.030], p = .024). First-year students had significantly lower GPA than third-year students (B = −0.335, 95% CI [−0.553, −0.117], p = .003); no other academic-year contrast reached significance. Smokers had significantly lower GPA than non-smokers (B = −0.141, 95% CI [−0.273, −0.009], p = .036).
With city as the reference category, neither the village-versus-city contrast (B = −0.065, 95% CI −0.191 to 0.062; p = 0.318) nor the camp-versus-city contrast (B = 0.490, 95% CI −0.017 to 0.998; p = 0.058) reached statistical significance. The camp estimate was imprecise, reflecting the small subgroup. Re-referencing changes the comparisons represented by the coefficients but does not change model fit or the overall association of residence with GPA.
No statistically significant adjusted associations with GPA were detected for age, sex, SHI score, or estimated caffeine intake. The coefficient for log-transformed caffeine intake was negative, but its confidence interval included zero (B = −0.026, 95% CI −0.071 to 0.020; p = 0.270). All individual VIFs in the reported parameterization were below 2.8. Residual plots suggested approximately normally distributed residuals, with modest deviations in the tails and no strong visual evidence of heteroscedasticity. One participant had a standardized residual of −3.75.
A sensitivity analysis including university of enrollment as a binary covariate is reported in the Limitations section.
Table 5 and Fig 3 summarize the full regression model.
Points show unstandardized B; bars show 95% confidence intervals. Red markers indicate p < .05; grey markers indicate p ≥ .05. All 27 predictors are plotted individually. Residence is referenced to city (rather than camp) and academic year to third year (rather than sixth year) following a sensitivity analysis that resolved severe multicollinearity present in the original small-reference-cell parameterization; the camp contrast remains imprecise given its small subgroup size (n = 5).
4. Discussion
This cross-sectional study explored the relationship of sleep hygiene, sleep behaviors, and caffeine use with academic performance among 379 university students from An-Najah National University and other Palestinian universities.
Higher estimated caffeine intake was associated with higher SHI scores, indicating poorer sleep hygiene, but no statistically significant association with GPA was detected in either bivariate analyses or the adjusted model. Selected contrasts involving study hours, sleep regularity, sleep duration, academic year, academic major, and smoking status were associated with GPA after adjustment. These findings describe associations within this sample and do not establish causal effects or the relative importance of the measured factors.
4.1 Caffeine and sleep hygiene
The consumption patterns observed here are consistent with the broader literature describing caffeine as the most widely used stimulant among university and medical students [1,4,5,14,16]. High or very high intake accounted for more than half of the sample, and 42.2% of participants reported consuming their last caffeinated beverage within three hours of bedtime. This timing is important: caffeine has a substantial pharmacokinetic half-life, and intake in the hours immediately preceding sleep is known to delay sleep onset, reduce total sleep time, and degrade sleep quality [17–19]. The present findings are consistent with this relationship. SHI scores differed across caffeine-consumption categories (F(5,373) = 7.763, p < 0.001), although the category means were not strictly monotonic. SHI scores also differed across categories of caffeine timing (F(3,375) = 7.073, p < 0.001). The highest observed means occurred in the very-high-consumption group and the group reporting caffeine intake within three hours of bedtime. However, the omnibus tests do not establish which individual groups differed significantly. Participants reporting very high caffeine intake had the poorest sleep hygiene of any group. This is consistent with experimental work showing that caffeine taken even six hours before bed measurably disrupts sleep [18]. These findings support examining both the amount and timing of caffeine intake, although the present analyses do not establish their independent contributions. This relationship held when caffeine exposure was analyzed as estimated daily intake in milligrams. Sleep Hygiene Score was significantly correlated with caffeine intake using raw milligrams (Pearson r = .282, p < .001; Spearman rho = .310, p < .001) and log-transformed intake (Pearson r = .210, p < .001). The consistency of this association across categorical, raw continuous, log-transformed, and non-parametric analyses supports an association between higher caffeine intake and poorer sleep hygiene, although this study did not formally test for a linear dose-response trend. Interpretation also requires consideration of overlapping questionnaire content. The SHI includes an item concerning substance use, including caffeine, near bedtime. Consequently, the association between caffeine exposure and total SHI score may partly reflect this shared content rather than an entirely independent relationship.
4.2 Caffeine consumption and academic performance
No statistically significant association was detected between caffeine intake and GPA, whether examined as a consumption category, raw milligram estimate, or log-transformed continuous measure. No statistically significant adjusted association was detected in the multivariable regression (B = −0.026, 95% CI [−0.071, 0.020], p = .270), despite its clear relationship with sleep hygiene. On the surface, this appears to differ from a previous study in which moderate coffee consumers achieved higher examination scores than both nonconsumers and heavy consumers [2].
Two considerations may help explain these differing findings. First, the previous study relied on unadjusted comparisons of mean examination scores and did not use multivariable analysis to account simultaneously for sleep duration, study habits, stress, baseline academic ability, or other potential confounders [2]. Its findings may therefore partly reflect differences in these related factors rather than an independent association between coffee consumption and academic performance. Differences in the populations and academic outcomes examined may also have contributed, as that study included only second-year medical students and used scores from a recent preliminary examination, whereas the present study included students from multiple disciplines and assessed cumulative GPA.
A further consideration is the mismatch between the time periods represented by the exposure and outcome measures. Cumulative GPA reflects academic performance over an extended period, whereas caffeine intake and sleep hygiene were assessed at a single survey occasion. Current habits may therefore be imperfect indicators of exposure during the period contributing to GPA. Although sleep could plausibly connect caffeine use with academic performance [8,18], mediation was not tested in this study. The nonsignificant caffeine coefficient should not be interpreted as evidence of no association or as demonstrating that any effect is small or indirect.
4.3 Study habits, sleep patterns, and academic performance
Several study and sleep behaviors were associated with GPA in the adjusted model. Selected study-hour categories differed in adjusted GPA, with students studying less than two hours a day showing significantly lower GPA than those studying more than four hours (B = −0.305, p = .001), and this pattern also held, though more modestly, for two to three hours of study (B = −0.199, p = .046). Previous research on study time and academic achievement has produced mixed findings. Nonis and Hudson found no direct effect of study time on performance, but reported that study time interacted with ability in influencing academic outcomes [20], and other studies suggest that study quantity alone does not predict GPA once the quality and structure of study are taken into account, with deliberate, well-organized practice mattering more than raw hours [21]. The present finding, in which study hours retained significance across two of three contrasts even after adjustment for sleep and demographic variables, is best read as supporting the general relevance of study time in this population, while leaving open the likelihood that how students study is at least as important as how long, consistent with Plant et al. [21].
The very-regular-sleep category was associated with higher adjusted GPA than the irregular-every-day category (B = 0.387, p = .014). This is in line with a growing number of studies showing that the regularity of the sleep-wake cycle, rather than sleep duration alone, is important for academic performance. Phillips and colleagues used a sleep regularity index derived from daily sleep diaries and found that students with more regular sleep timing achieved better grades [11]. Bono and Hill similarly found that first-year students with more stable sleep patterns had higher GPA than those with more variable patterns, after adjusting for baseline ability and well-being [22]. The present finding, from a different cultural and educational context, adds to this body of evidence.
Sleep duration showed a less consistent pattern. In the adjusted model, students reporting fewer than five hours of sleep had higher GPA than those reporting more than eight hours (B = 0.313, p = .017). This finding should not be interpreted as evidence that short sleep improves academic performance. The shortest-sleep subgroup was small (n = 30), the difference was limited to one contrast with the more-than-eight-hours reference group, and none of the intermediate sleep duration categories differed significantly. The finding also differs from studies using objective or repeated sleep measurements, which have generally associated longer and more consistent sleep with better academic performance [8,23]. One possible explanation is reverse causation, where academically driven students reduce their sleep to meet study demands while maintaining relatively high grades. Alternatively, unmeasured characteristics of the more-than-eight-hours group, such as irregular schedules, fatigue, or health-related factors, may have contributed to the observed difference. Given the small subgroup and the absence of a graded pattern across sleep-duration categories, this isolated result should be interpreted cautiously and examined in future longitudinal studies.
In the comparison between academic majors, Medicine students showing significantly lower GPA than Dentistry students (B = −0.224, p = .024), even after adjusting for study hours, sleep variables, and academic year. This may reflect differences in grading standards, curriculum difficulty, or workload between programs, none of which were measured directly in this study and which represent a useful direction for future research.
The Sleep Hygiene Score itself showed no statistically significant association with GPA, either as a bivariate correlation (r = −.061, p = .249) or as an independent predictor in the adjusted model. This may indicate that specific sleep behaviors, particularly sleep schedule regularity, were more closely related to GPA than the overall frequency of poor sleep-hygiene practices captured by the composite index. This is consistent with prior work in which discrete measures of sleep timing and regularity were stronger predictors of grades than composite sleep quality indices [6,11,22]. Nevertheless, the SHI demonstrated acceptable internal consistency in this sample (Cronbach’s α = 0.775) and showed a significant association with caffeine exposure, though the categorical pattern was not strictly monotonic, supporting its usefulness for characterizing sleep-related behavior even though no statistically significant association with GPA was detected.
4.4 Additional predictors and model considerations
Smoking status was significantly associated with GPA in bivariate testing, with non-smokers reporting a higher mean GPA than smokers (3.09 versus 2.92, p = .009), and remained an independent predictor after adjustment for sleep, study habits, and demographic variables (B = −0.141, 95% CI [−0.273, −0.009], p = .036). This is consistent with previous work linking smoking to poorer academic outcomes among health sciences students, including lower GPA, higher absenteeism, and more academic warnings, with a dose-response relationship between cigarette consumption and academic performance indicators [24]. The caffeine coefficient was closer to zero after smoking status was included in the model. This attenuation is compatible with confounding but does not, by itself, establish smoking as a confounder.
With third-year students as the reference group, first-year students had lower adjusted GPA (B = −0.335, 95% CI −0.553 to −0.117; p = 0.003), consistent with the unadjusted comparison reported in section 3.4. This may reflect academic and social adjustment challenges typical of the first year of university, although longitudinal data would be needed to confirm this interpretation.
University of enrollment was not included in the primary model. Most participants attended An-Najah National University (325/379, 85.8%), while each of the other ten institutions contributed only 1–13 participants. These small institutional subgroups limited the precision of institution-specific comparisons. Nevertheless, differences in grading practices across institutions may have contributed to residual confounding, even after GPA values were placed on a common numerical scale.
The residence coefficients should be interpreted as specific comparisons with city residents. Neither reported contrast reached statistical significance, and the camp estimate was particularly imprecise. Selecting a different reference category changes the comparisons displayed but does not remove the uncertainty associated with a small subgroup.
The final regression model explained a modest proportion of the variance in GPA, indicating that academic performance is influenced by a wide range of factors beyond those measured in this study. Variables such as psychological stress, motivation, learning strategies, prior academic achievement, and socioeconomic circumstances may also contribute to students’ academic outcomes and should be considered in future research [25].
5. Conclusion
Higher estimated caffeine intake was associated with poorer sleep hygiene in this sample of Palestinian university students. No statistically significant association between caffeine intake and GPA was detected in either bivariate analyses or the adjusted model. Selected contrasts involving study hours, sleep regularity, sleep duration, academic year, academic major, and smoking status were associated with GPA. These findings should be interpreted cautiously because of the cross-sectional design, self-reported measures, and uncertainty in exposure assessment. Longitudinal studies are needed to clarify the temporal relationships among caffeine use, sleep behaviors, and academic performance.
6. Limitations
Several limitations should be considered. The cross-sectional design prevents determination of temporal order or causal effects. Caffeine use and sleep behaviors may influence academic functioning, but academic demands may also influence these behaviors. Current caffeine and sleep measures may not adequately represent habits during the longer period reflected by cumulative GPA.
The study relied on self-reported measures, including GPA, and is therefore susceptible to recall and reporting error [26]. Estimated caffeine intake may also be affected by variation in beverage type, serving size, brand, and preparation method. Convenience sampling and the predominance of participants from one university limit generalizability. Differences in grading practices across institutions and disciplines may remain relevant despite the use of a common numerical GPA scale. The Arabic version of the Sleep Hygiene Index was translated, back-translated, and piloted, but it was not formally psychometrically validated; internal consistency in this sample was acceptable (Cronbach’s α = 0.775).
GPA was unavailable for 25 participants, who differed from participants with recorded GPA in age, academic year, sex, and sleep regularity. Consequently, the GPA analyses may be affected by selection related to GPA availability. Nonsignificant differences in caffeine intake and SHI score between these groups do not establish the absence of selection bias. Missing GPA did not exclude these participants from the caffeine–SHI analyses.
Some subgroups were small, particularly the camp-residence group, resulting in imprecise estimates. Changing the reference category does not resolve this limitation. The study was exploratory, no a priori sample-size calculation was performed, and numerous associations were examined; isolated statistically significant findings should therefore be interpreted cautiously. Finally, potentially important factors such as psychological stress, prior academic achievement, and learning strategies were not measured and may have contributed to residual confounding [25].
As a sensitivity check, a binary university indicator (An-Najah National University as reference vs. all other institutions combined) was added to the final regression model. Model fit improved modestly (R² = 0.207, adjusted R² = 0.138, versus R² = 0.190, adjusted R² = 0.123 without it; F(28,323) = 3.010, p < .001), and the binary indicator itself was a significant predictor (B = 0.237, 95% CI [0.059, 0.414], p = .009, VIF = 1.165), with students at other institutions showing higher GPA than An-Najah students on average. Most predictors were materially unchanged by this addition, including the null association between caffeine intake and GPA (B = −0.027, p = .234). The association between smoking status and GPA was attenuated and no longer reached statistical significance at the conventional threshold (B = −0.128, p = .055, versus B = −0.141, p = .036 in the main model), suggesting this finding should be interpreted with some caution and may be partly related to differences between institutions rather than smoking status alone.
7. Recommendations
University health promotion programs should include sleep hygiene education, encourage regular sleep schedules, and address the possible adverse effects of late caffeine intake. Students should be advised to avoid caffeinated drinks close to bedtime, particularly during examination periods, and to adopt healthier strategies for managing fatigue and academic workload [27].
Academic advisors and student support services should also emphasize balanced study routines. Rather than relying on caffeine or all-night studying, students should be encouraged to plan regular daily study periods, obtain adequate sleep, and maintain a consistent sleep schedule [12,27].
Future studies should measure caffeine exposure using more detailed methods, such as beverage diaries, brand-specific caffeine content, serving size, and timing of intake, to reduce exposure misclassification. Longitudinal studies following students across a semester, particularly before and during examination periods, would help determine whether changes in caffeine intake and sleep patterns predict later changes in GPA. In addition, psychological and lifestyle factors such as stress, depression, physical activity, and other health behaviors should be included to provide a more complete understanding of the factors influencing sleep hygiene and academic performance.
Supporting information
S1 Appendix. Caffeine intake estimation: questionnaire items and conversion rules.
https://doi.org/10.1371/journal.pone.0363064.s001
(DOCX)
References
- 1. Fredholm BB, Bättig K, Holmén J, Nehlig A, Zvartau EE. Actions of caffeine in the brain with special reference to factors that contribute to its widespread use. Pharmacol Rev. 1999;51(1):83–133. pmid:10049999
- 2. Petkar SS, Dabhade S. Association between coffee consumption and academic performance among second-year medical students. Cureus. 2026;18(2):e103534. pmid:41841082
- 3. Lund HG, Reider BD, Whiting AB, Prichard JR. Sleep patterns and predictors of disturbed sleep in a large population of college students. J Adolesc Health. 2010;46(2):124–32. pmid:20113918
- 4. Mahoney CR, Giles GE, Marriott BP, Judelson DA, Glickman EL, Geiselman PJ, et al. Intake of caffeine from all sources and reasons for use by college students. Clin Nutr. 2019;38(2):668–75. pmid:29680166
- 5. Riera-Sampol A, Rodas L, Martínez S, Moir HJ, Tauler P. Caffeine intake among undergraduate students: Sex differences, sources, motivations, and associations with smoking status and self-reported sleep quality. Nutrients. 2022;14(8):1661. pmid:35458223
- 6. Hershner SD, Chervin RD. Causes and consequences of sleepiness among college students. Nat Sci Sleep. 2014;6:73–84. pmid:25018659
- 7. EFSA Panel on Dietetic Products, Nutrition and Allergies (NDA). Scientific Opinion on the safety of caffeine. EFSA J. 2015;13:4102.
- 8. Curcio G, Ferrara M, De Gennaro L. Sleep loss, learning capacity and academic performance. Sleep Med Rev. 2006;10(5):323–37. pmid:16564189
- 9. Phan DM, Lam MY, Trang MN. Regular caffeine consumption & subjective sleep quality: A systematic review. JAR Life. 2025;14:100005. pmid:41424818
- 10. Mastin DF, Bryson J, Corwyn R. Assessment of sleep hygiene using the Sleep Hygiene Index. J Behav Med. 2006;29(3):223–7. pmid:16557353
- 11. Phillips AJK, Clerx WM, O’Brien CS, Sano A, Barger LK, Picard RW, et al. Irregular sleep/wake patterns are associated with poorer academic performance and delayed circadian and sleep/wake timing. Sci Rep. 2017;7(1):3216. pmid:28607474
- 12. Mirghani HO, Mohammed OS, Almurtadha YM, Ahmed MS. Good sleep quality is associated with better academic performance among Sudanese medical students. BMC Res Notes. 2015;8:706. pmid:26597849
- 13. Onyper SV, Thacher PV, Gilbert JW, Gradess SG. Class start times, sleep, and academic performance in college: A path analysis. Chronobiol Int. 2012;29(3):318–35. pmid:22390245
- 14. Kharaba Z, Sammani N, Ashour S, Ghemrawi R, Al Meslamani AZ, Al-Azayzih A, et al. Caffeine consumption among various university students in the UAE, exploring the frequencies, different sources and reporting adverse effects and withdrawal symptoms. J Nutr Metab. 2022;2022:5762299.
- 15.
Staff MC. Caffeine content for coffee, tea, soda and more. Mayo Clinic [Internet]. [cited 2026 Jul 19]. Available from: https://www.mayoclinic.org/healthy-lifestyle/nutrition-and-healthy-eating/in-depth/caffeine/art-20049372
- 16. Ali Haddad R. Caffeine usage among medical students and their knowledge of its benefits and side effects. AIMCJ. 2024;1(1):45–55.
- 17. O’Callaghan F, Muurlink O, Reid N. Effects of caffeine on sleep quality and daytime functioning. Risk Manag Healthc Policy. 2018;11:263–71. pmid:30573997
- 18. Drake C, Roehrs T, Shambroom J, Roth T. Caffeine effects on sleep taken 0, 3, or 6 hours before going to bed. J Clin Sleep Med. 2013;9(11):1195–200. pmid:24235903
- 19. Gardiner C, Weakley J, Burke LM, Roach GD, Sargent C, Maniar N, et al. The effect of caffeine on subsequent sleep: A systematic review and meta-analysis. Sleep Med Rev. 2023;69:101764. pmid:36870101
- 20. Nonis SA, Hudson GI. Academic performance of college students: Influence of time spent studying and working. J Educ Bus. 2006;81(3):151–9.
- 21. Plant EA, Ericsson KA, Hill L, Asberg K. Why study time does not predict grade point average across college students: Implications of deliberate practice for academic performance. Contemp Educ Psychol. 2005;30:96–116.
- 22. Bono TJ, Hill PL. Sleep quantity and variability during the first semester at university: Implications for well-being and academic performance. Psychol Health Med. 2022;27(4):931–6. pmid:34465270
- 23. Okano K, Kaczmarzyk JR, Dave N, Gabrieli JDE, Grossman JC. Sleep quality, duration, and consistency are associated with better academic performance in college students. NPJ Sci Learn. 2019;4:16. pmid:31583118
- 24. Alqahtani JS, Aldhahir AM, Alanazi Z, Alsulami EZ, Alsulaimani MA, Alqarni AA, et al. Impact of smoking status and nicotine dependence on academic performance of health sciences students. Subst Abuse: Res Treat. 2023;14:13–24.
- 25. Almojali AI, Almalki SA, Alothman AS, Masuadi EM, Alaqeel MK. The prevalence and association of stress with sleep quality among medical students. J Epidemiol Glob Health. 2017;7(3):169–74. pmid:28756825
- 26. Kuncel NR, Credé M, Thomas LL. The validity of self-reported grade point averages, class ranks, and test scores: A meta-analysis and review of the literature. Rev Educ Res. 2005;75:63–82.
- 27. Hershner S, O’Brien LM. The impact of a randomized sleep education intervention for college students. J Clin Sleep Med. 2018;14(3):337–47. pmid:29510791

관련 뉴스