This cross-sectional, multi-country study used publicly available data from the GSHS among 35 countries. The GSHS is a representative and extensive health survey of risk factors and behaviors of students. US Centers for Disease Control and Prevention (CDC) and the World Health Organization (WHO), and other United Nations (UN) allies and country-specific institutions developed this survey (1, 26). Content from this survey was drawn from the CDC Youth Risk Behaviors Survey (YRBS) to establish its reliability and validity (27). The information included in the survey is the objective, methodology used in the survey, and the procedure for sampling the GSHS is available at http://www.cdc.gov/gshs/.

In summary, though the participants for the GSHS are primarily school-going adolescents aged 13–17 years, those below 13 and above 17 years were also included in the survey. The sampling procedure included a two-stage standardized probability sampling approach for selecting participants within each country in this survey. In the first stage, the schools were selected through probability proportional to size sampling design. In the second stage, students between the ages of 13 and 17 were randomly selected from their various classrooms in each selected school. Irrespective of the student age in the selected classrooms, all students were eligible to participate in the survey. In this study, we employed the most current data available in GSHS participating nations and territories in this investigation. Countries and territories with data released before 2015 were removed from the analysis. These criteria ensured that the analysis reflects and represents recent or current trends. Also, we excluded data that did not have the variables of interest for the present study. By applying the inclusion and exclusion criteria, data from 35 countries and territories were considered recent and eligible to be included in the analysis. Thus, our analysis included data from 35 countries and territories released between 2015 and 2018. To account for non-response and probability selection, the data was weighted to enable the generalization of results to the targeted population (28, 29).

Trained enumerators collected data from local social work agencies and other research institutions during a regular class period. The GSHS questionnaire was designed to be close-ended and was written in English before being translated into each country’s native language. Multiple-choice questions were included in the questionnaire. The questionnaires were administered after informed consent was obtained from children, parents, and school authorities. Before the questionnaires were given, the trained enumerators thoroughly described the purpose of the study and the instructions to the students.

Furthermore, through anonymous and voluntary participation, the survey ensured that students’ privacy was respected. The study followed the Helsinki Declaration, and all GSHS surveys were evaluated and approved by institutional review boards or ethics committees in each nation. Details of the systematic techniques used to gather student data can also be obtained at http://www.cdc.gov/gshs/.

We assessed food insecurity (hunger) by asking, “During the past 30 days, how often did you go hungry because there was not enough food in your home?” The item had 5-point response options: never, rarely, sometimes, most of the time, and always. It has been argued that single-item measures can be easier to capture and more cost-effective in large-scale surveys like the GSHS than multiple-item measures (30). A single item should be sufficient for the variable, and the attribute of the variable measured is “concrete” to the participants (31). We believe the single item assessing food insecurity in the GSHS was an example of such a concrete measure. Given the exploratory nature of the present study, we believed that the single-item question was a sufficient measure of food insecurity, consistent with previous studies (4, 17, 32, 25). For analytic purposes and in line with other previous studies, we categorized never as no food insecurity, rarely and sometimes as moderate food insecurity and most of the time and always as severe food insecurity. These categories were named as such because evidence suggests that moderate food insecurity is often considered to indicate that the quality/quantity of food intake has been compromised. In contrast, severe food insecurity refers to reduced food intake and disturbed eating patterns (7).

Sleep disturbance was assessed based on the question, “During the past 12 months, how often have you been so worried about something that you could not sleep at night?” Responses were never, sometimes, most of the time, and always. This item has been used in previous global studies involving adolescents using the GSHS (11, 33). In this study, never was categorized as no sleep disturbance, sometimes as moderate sleep disturbance, and most of the time and always as severe sleep disturbance for analytic purposes.

We included empirically and theoretically defined variables as confounders. These covariates were selected because they both have direct and indirect relationship with food insecurity and sleep quality in previous studies (4, 17, 32, 34, 25). Sociodemographic variables included age (11–14 years; 15–17 years; 18+ years) and sex (male; female). Health-related and behavior variables included bullying victimization (no; yes), loneliness (no; yes), suicidal ideation (no; yes), cigarette smoking (no; yes), alcohol use (no; yes), marijuana use (no; yes), physical activity (no; yes), close friends (no; 1–2; 3+), amphetamine use (no; yes), and parents understand adolescent problems (no; yes).

Stata 14.0 was used to perform the statistical analysis. The ‘svyset’ command in Stata was used to adjust for the complex sampling design employed by the GSHS survey using the weight, primary sampling unit (psu) and stratum variables. A p-value ≤0.05 was used to assess the statistical significance. Frequencies and percentages were calculated to describe the characteristics of respondents. Country-specific prevalence of food insecurity and sleep disturbance were summarized using proportions and 95% confidence intervals by sex. A series of stepwise multinomial logit regression analysis was performed to assess the independent effect of food insecurity on sleep disturbance. Three separate models were fitted to estimate the association. Model 1 calculated the crude regression estimates by predicting the effect or influence of food insecurity on sleep quality. In Model 2, we added sociodemographic variables that could predict food insecurity or sleep disturbance. In Model 3 (full model), we included covariates related to lifestyle, support, and health-related variables. Further, moderation analyses of age and sex were performed to examine their modifying roles in the association between food insecurity and sleep disturbance.

A total of 189,619 adolescents from 35 countries and territories were involved in the study. Females (50.4%) were more than the males. The majority, 53.8% of adolescents, were between 15 and 17. The majority (66.4%) of adolescents were not bullied; 65.4% were lonely; 89.6% had no suicidal ideas; 74.3% had more than three close friends; 88.8% had not used cigarettes; 83.8% had not used alcohol; 96.2% had not used marijuana; 96.8% had not used amphetamine; 68.6% were physically active, and 77% had parents who understood their problems (Table 1).

Tables 2, 3 present the overall and country-specific prevalence of food insecurity and sleep disturbance, respectively. The overall pooled prevalence of food insecurity (moderate and severe) was 51%. Food insecurity was more common among males (52%) than females. Overall pooled prevalence of sleep disturbance (moderate and severe) was 61.2%. Unlike food insecurity, sleep disturbance was more common among female adolescents (64.9%) than male adolescents (see Table 2). Regarding gender, the highest prevalence of food insecurity among male adolescents was reported in Samoa (75.7%), and the lowest prevalence was found in Curacao (25%). Among females, the highest prevalence of food insecurity was reported in Samoa (72.3%), and the lowest was found in Curacao (25%). Regarding sleep disturbance, the highest among males was reported in the Philippines (73.3%), and the lowest was observed in Suriname (44.9%). Among females, the highest was reported in the Philippines (79.7%), and the lowest was seen in Myanmar (47.2%; see Table 3).

Table 4 shows a series of multinomial regressions predicting sleep disturbance. Model 1 estimated the crude effect of the association, whilst Models 2 and 3 controlled for sociodemographic, health and lifestyle-related variables. In Model 1, adolescents who were moderately food insecure relative to those who were food secure were 2.09 times more likely to have a moderate sleep (AOR = 2.09 CI = 1.98–2.20) and severe (AOR = 1.90 CI = 1.71–2.12) sleep disturbances. Similarly, adolescents who were severely food insecure were 2.28 and 5.38 more likely to experience moderate (AOR = 2.28 CI = 2.10–2.48) and severe (AOR = 5.38 CI = 4.74–6.11) sleep disturbances, respectively, relative to their food secure counterparts.

In Model 2, there was a slight change for each outcome, indicating a marginal role of the sociodemographic factors in explaining the outcome variables. Model 3 was adjusted for lifestyle, health, and health-related variables (Full model). The results showed that adolescents who were moderately food insecure relative to those who were food secure were 1.70 and 1.63 more likely to experience moderate (AOR = 1.70 CI = 1.59–1.81) and severe (AOR = 1.63 CI = 1.42–1.87) sleep disturbances, respectively. Likewise, adolescents who were severely food insecure were 1.88 and 4.07 more likely to experience moderate (AOR = 1.88 CI = 1.68–2.11) and severe (AOR = 4.07 CI = 3.40–4.87) sleep disturbances, respectively, relative to their food secure counterparts (see Table 4).

As displayed in Table 5, moderation analysis was performed to determine the modifying effect of age and sex on the association of food insecurity and sleep disturbance. The age-wise analysis used the interaction effect of age to determine the association between food insecurity and sleep disturbance. There was a significant association between moderate food insecurity and moderate sleep disturbance among adolescents aged 15–17 (AOR = 3.39 CI = 3.39–3.93) and a similar association between moderate food insecurity and severe sleep disturbance among adolescents aged 18 and older (AOR = 3.24 CI = 2.40–4.38). Moreover, there was a significant association between severe food insecurity and severe sleep disturbance among adolescents aged 15–17 (AOR = 1.15 CI = 0.97–1.33), severe food insecurity and severe sleep disturbance among adolescents aged 15–17 (AOR = 1.75 CI = 1.45–2.06), moderate food insecurity and severe sleep disturbance among adolescents aged 18 or older (AOR = −1.60 CI = 0.81–1.24). With sex-wise analysis, the interaction effect of sex was used to assess the association between severe food insecurity and severe sleep disturbance. There was an association between moderate food insecurity and severe sleep disturbance among females (AOR = 2.77 CI = 2.30–3.33), severe food insecurity and severe sleep disturbance among females (AOR = 8.58 CI = 6.95–10.57; see Table 5).

In this multi-country analysis, 45 and 5.3% were reported to be moderately and severely food insecure, respectively. Similarly, 52.5 and 8.6% of moderate and severe sleep disturbances were found. Adolescents with moderate and severe food insecurity were at higher risk of sleep disturbance than food-secure adolescents. The association between food insecurity and sleep disturbance was evident after adjustments for multiple potential variables. In addition, the effect of food insecurity on sleep disturbance was found in female adolescents. Also, adolescents aged 15–17 years and 18 years or older were found to have an increased effect of food insecurity on sleep than those aged 11–14 years. The study indicates that interventions for quality sleep among adolescents should include ensuring adequate food security.

The study contributes to a growing area of research that broadens understanding of food insecurity and adolescent sleep disorders. Overall, the prevalence of moderate (45.2%) and severe (5.8%) food insecurity were found. Thus, our study’s food insecurity prevalence was within the range of previously documented estimates in multi-country studies (4, 35). The threats of global environmental change in food production systems, sky-rocketing food prices, and natural (and human-induced) disasters could be impacting food insecurity (hunger) prevalence (36). Also, about 52.6% moderate and 8.6% severe sleep disturbances were found, consistent with the rates reported in one multi-country analysis (11). The high prevalence of sleep disturbance among adolescents is troubling because studies have associated poor sleep quality with anxiety disorders, behavioral challenges and psychiatric ailments (37, 38).

Consistent with our first hypothesis, adolescents with moderate and severe food insecurity levels were at higher risks of severe sleep disturbance than those who were food secure. Our findings resonate with prior literature. In a longitudinal study assessing food insecurity among economically challenged households, adolescents in low and moderate-food-insecure households/groups had more sleep disturbances than their counterparts in food-secured households (9). Conterminously, among 223,561 adolescents sampled for a multi-country study on the association between food insecurity and sleep disorders, it was reported that severe food insecurity was significantly associated with a higher risk of sleep disturbance in 48 of the 68 countries studied (11). Transcending the adolescent groups, other studies have reported similar findings among older adults (2) and the general population (10, 13).

Even though the exact mechanisms linking food insecurity (hunger) and sleep disturbances are unclear, some biological and socioeconomic mechanisms may offer plausible explanations. First, considering biological mechanisms, an individual’s metabolic activity may disturb his/her state of vigilance due to lower nutrition, leading to insomnia (39). In addition, chronic stress or psychological distress because of one’s financial ability to access safe and nutritious foods may influence stress related to physiological arousal mechanisms, leading to the instigation of the sympathoadrenal medullary and hypothalamic–pituitary–adrenal-cortical systems that are noted to cause sleep disturbances (10). Besides, adolescents with food insecurity tend to encounter nutritional deficiencies and obesity, which can affect their sleep (40). Similarly, several studies suggest that food insecurity is associated with adverse physical and mental health outcomes that affect sleep (41, 42).

For socioeconomic mechanisms, poverty is a significant risk factor for poor sleep quality (4, 43). While poverty typically does not occur in isolation, with a tight household budget, the family may be unable to pay for food, housing, clothing, health care, and other living expenses simultaneously (43). Consequently, food insecurity is always associated with housing insecurity, and poor housing conditions may link to sleep disturbance due to concerns about personal safety, exposure to more noise from neighbours, and environmental situations within the house (44). This study has a policy implication because addressing food insecurity may be crucial to improve sleep quality and other phycological problems. However, further longitudinal and clinical research is necessary to confirm these hypotheses and explore more complex mechanisms that may underlie adolescent food insecurity and quality sleep.

In line with the differential vulnerability concept, the analysis found significant age-wise differences in the relationship between food insecurity and sleep disturbance. The effect of food insecurity on sleep disturbance was found in adolescents aged 15–17 years and 18 years or older than those between 12 and 14 years. A possible explanation may be due to the socioeconomic conditions of the households, such as poverty and other socioeconomic deprivations, which might cause older adolescents to sacrifice their food intake for younger ones in times of limited household food access. The age differences may be addressed further by decreased brain capacity and functionality as we age, as measured by hippocampal sizes. In the setting of food insecurity, this may raise the likelihood of sleep problems.

Further mixed-method studies would be required to clarify and provide reasons for this result, as this is the first study coming forth with this finding. We found statistically significant sex-wise differences in the relationship between food insecurity and sleep disturbance. This finding agrees with previous propositions that females are less likely than males to achieve their dietary requirements due to general socioeconomic deprivations among females, especially in developing countries (24). Another study on the hidden penalties of gender inequality also found that females likelihood of achieving all the dietary requirements is lower than their male counterparts (45). Our finding, therefore, suggests that boys and girls in the study are disproportionally affected by food insecurity compared to food security. The finding further supports the male–female health survival paradox suggesting that women generally present higher lifetime health problems (46), such as poor sleep and their associated social determinants of health, including food insecurity, than men. Thus, future studies would benefit from further examination of sex differences in the association between food insecurity and sleep disturbances.

Our study findings suggest that policymakers, guardians, and stakeholders should know that food insecurity (hunger) could be an underlying factor for adolescents’ sleep disturbances. Our results can help facilitate more empirical research to explore sleep disorders within the context of poverty and social inequality, with hunger being a relevant daily stressor that can impact poor sleep outcomes. Evidence exists on interventions addressing the negative impacts of food insecurity in children and adolescents (47). However, such interventions are mostly not multi-faceted to address this area of need (3). For instance, the World Food Programme (WFP) was implemented to support school meal programs in 71 countries in 2017 (48). However, one study reports that the benefits of school-based programs do not address food insecurity holistically as adolescents also require to be fed outside the school environment [for instance, dinner, weekends, holidays, and breaks (3)]. Again, these programs would not benefit those who are not in school. Thus, a long-term solution would be alleviating poverty and acquiring agricultural inputs so that households can potentially meet their dietary needs to address sleep problems. Again, one solution may be free or subsidized nutrition programs that promote access to adequate food for adolescents in food-insecure households. Future interventions and programs to address food insecurity and sleep disturbances should consider demographic factors such as age and sex.

There are particular strengths and limitations to the present study. Drawing evidence from 35 countries and territories, this study contributes to the limited knowledge base by assessing the association of food insecurity (hunger) with sleep disturbance among school-going adolescents. Our sample used nationally representative datasets with large sample sizes and across several countries and territories, increasing the study findings’ generalizability. The GSHS followed best standard practices regarding technique and employed professional and well-trained interviewers. Notwithstanding these advantages, our findings should be evaluated and viewed in light of the following limitations. Although we controlled for most known potential confounders, residual confounding may exist and affect or explain the results.

Furthermore, reliance on self-reports implies that recollection and social desirability biases cannot be ruled out. In addition, food insecurity is limited to indicating only food insecurity experiences in the last 30 days, while worry-induced sleep disturbance in the previous 12 months indicates a potential timeframe imbalance between the measures. Again, the use of a single-item measure approach is an important limitation in this study because it does not capture all the different dimensions of food insecurity. Future studies are needed to further validate single-item food insecurity measures among adolescents across regions of the world.

Previous studies have established that of those who are currently food insecure, most were food insecure in the past year (4, 11, 32). In addition, notable studies using the same GSHS studies used the past 30 days and 12 months timeframe for the assessment of key variables (4, 49–53). Moreover, given that this was a cross-sectional observational study, causal relationships and implications from the association between food insecurity and sleep disturbance cannot be inferred. Thus, future research should investigate the associations of interest utilizing longitudinal cohort designs that allow for examining reasonable causal inferences.