Lack of adequate sleep quality and duration are major public health concerns in the U.S.; nearly 35% of Americans report sleeping ≤ 6 hours per night (Liu et al., 2016). Numerous chronic diseases, such as cancer, heart disease, and stroke, are linked to poor sleep (Ayas et al., 2003; Gangwisch et al., 2006; Kakizaki et al., 2008; Patel et al., 2004). Recent epidemiologic evidence indicates that there may also be a relationship between sleep quality and duration and dietary intake (Grandner et al., 2013, 2014, 2010; Kant & Graubard, 2014; Kim et al., 2011; Kurotani et al., 2015; Shi et al., 2008). The impact of sleep quality and duration on chronic disease risk may be in part through an influence of sleep on diet, a known risk factor for many chronic diseases (US Department of Health and Human Services, 2015).

There is evidence for an association between poor sleep quality, short sleep duration, and total energy consumption as well as dietary intake of certain macronutrients and alcohol (Grandner et al., 2013, 2010; Kant & Graubard, 2014). Campanini et al. reported a positive association between adherence to a Mediterranean diet and self-reported sleep quality in the Seniors-ENRICA cohort (Campanini et al., 2017). Researchers have also observed higher total energy, fat, sugar, and alcohol intake in people with shorter compared to longer sleep duration (Galli et al., 2013; Grandner et al., 2013, 2010; Kant & Graubard, 2014; Patterson et al., 2014). In a study of 68,832 Chinese women over 45 years of age, Tu et al. observed that lower fruit and meat intake was associated with shorter self-reported sleep duration (Tu et al., 2012). Stamatakis et al. reported similar findings in a population of older Americans in rural communities (Stamatakis & Brownson, 2008). Despite accumulating evidence to suggest an association between sleep quality, sleep duration and specific facets of diet, the current literature is more limited on the association between sleep quality and duration and patterns of dietary intake, which more comprehensively demonstrates the entirety of all beverages and foods consumed by an individual (US Department of Health and Human Services, 2015). Therefore, dietary patterns may also be more indicative of disease risk than individual diet components (e.g., fat intake) (US Department of Health and Human Services, 2015; Hu, 2002). Poor sleep quality and duration is of particular concern among first responders, who often work long hours and rotating shifts and are at greater risk for sleep disorders. Police officers experience a higher prevalence of sleep apnea (34%) and excessive daytime sleepiness (29%) (Rajaratnam et al., 2011) compared to the general U.S. population (24% in men and 9% in women for sleep apnea and 13% overall for excessive daytime sleepiness) (Ford et al., 2015; Young et al., 2009). In a previous Buffalo Cardio-Metabolic Occupational Police Stress (BCOPS) study, 54% of officers self-reported sub-optimal sleep quality using the Pittsburgh Sleep Quality Index (PSQI), comparable to the prevalence reported by Neylan et al. in over 700 police officers across police departments in the U.S. using the same measure (64%) (Fekedulegn et al., 2016; Neylan et al., 2002). Participants in the BCOPS study were also more likely to self-report short sleep duration compared to the general U.S. working population (33.0% vs. 8.0%, respectively) (Hartley et al., 2011). Police officers are also at greater risk for certain health conditions compared to the general population, including high cholesterol, metabolic syndrome, and obesity (Hartley et al., 2011). The potential intermediate role of dietary intake in relation to sleep quality and duration and chronic disease risk may have important implications for workplace interventions.

Using data from the BCOPS cohort study, we assessed the association of sleep quality and duration with dietary patterns derived from a food frequency questionnaire (FFQ). We hypothesized that police officers reporting poor sleep quality or short sleep duration would have a different dietary pattern than officers with a good sleep quality and long sleep duration.

Of the 422 participants included in the analysis, 199 reported optimal sleep quality, 102 reported borderline sleep quality, and 121 reported poor sleep quality (Table 2). Those with poor sleep quality reported greater mean physical activity than those with optimal sleep quality, with a higher proportion of physical activity derived from occupational tasks (data not shown in tables). Those with poor sleep quality were more likely to work a non-day shift and to have worked as a police officer for > 15 years compared to those with optimal sleep quality. Those with short sleep duration (< 6 hours/night) also had a higher mean physical activity than those with long sleep duration (≥ 7 hours/night), while those with moderate sleep duration (6–< 7 hours) had a higher mean percent of their daily caloric intake from fat (data not shown). We did not observe statistically significant differences for other covariates examined according to sleep quality or sleep duration.

In the overall model, we observed a linear association between the PSQI global score for sleep quality and the FV dietary pattern (β [SE] = −0.03 [0.02]; p = .05) (Table 3). Individuals with poor sleep quality had a lower average FV score than those with optimal sleep quality (β [SE] = −0.32 [0.13]; p = .01). After sex stratification, women with poor sleep quality had significantly lower mean scores on the FV pattern compared to women with optimal sleep scores (β [SE] = −0.81 [0.29]; p = .01). Sleep duration was significantly linearly associated with the dairy products dietary pattern among women (β [SE] = 0.24 [0.10]; p = .02) (Table 4). Women with short sleep duration (< 6 hours) had significantly lower mean scores on the dairy products pattern as compared to women with long sleep duration (≥7 hours) (β [SE] = −0.69 [0.29]; p = .02). Sleep quality and sleep duration were not associated with dietary patterns in men.

Women reporting ≥ 10 sleep disturbances per week had lower mean FV pattern scores compared to those who reported none (β [SE] = −1.30 [0.61]; p = .04). Compared to women with optimal subjective sleep quality, those reporting borderline subjective sleep quality had higher mean scores on the meat and eggs pattern (β [SE] = 0.72 [0.34]; p = .04). In men, borderline daytime dysfunction compared to none was associated with a lower mean FV pattern score (β [SE] = −0.38 [0.12]; p = .002). Use of sleep medications ≥ 1/week compared to no use was associated with a lower mean starch and fried foods score (β [SE] = −0.62 [0.17]; p = .0004). We did not observe statistically significant associations for sleep latency or habitual sleep efficiency in men or women.

In this cross-sectional study of urban police officers in the United States, we classified individuals according to four dietary patterns: pattern 1: high consumption of fruits and vegetables, pattern 2: high consumption of dairy products, pattern 3: high consumption of starches and fried foods, and pattern 4: high consumption of meat and eggs. Police officers reporting borderline or poor sleep quality had lower mean FV dietary pattern scores than those reporting optimal sleep quality, suggesting that better sleep quality is positively associated with greater intake of healthy foods, such as dark green and red-orange vegetables, citrus and other fruits, and beans. We observed this association in women, but not in men. Sleep quality was not significantly associated with other dietary patterns.

Our findings are consistent with previous studies conducted in different populations. A study of 2,025 Japanese manufacturing workers, aged 18–70 years, who completed a dietary questionnaire on eating habits in the last month, reported that a dietary pattern high in vegetable consumption was associated with shorter self-reported sleep latency (Kurotani et al., 2015). Katagiri and colleagues reported an association between poor sleep quality using the PSQI and low consumption of vegetables and fish among 3,129 employed Japanese women aged 34 to 65 years (Katagiri et al., 2014). Cao et al. also found that sleep initiation, measured via polysomnography, was positively associated with a dietary pattern high in fruits and vegetables in 784 Australian men participating in the Male Androgen Lifestyle, Environment, and Stress study (Cao et al., 2017). In a cross-sectional study of 23,829 adults from the Swedish EpiHealth cohort, Theorell-Haglöw et al. observed that good sleep quality and long sleep duration were associated with consuming a Mediterranean diet (i.e., high intake of fruits, vegetables, and whole grains and low intake of red meats) (Theorell-Haglöw et al., 2020).

Though our analysis utilized a cross-sectional study design, we propose that additional studies explore the biologically plausibility that sleep quality and duration influence dietary intake. Individuals with poor sleep quality and short sleep duration may be more likely to make poor dietary choices, as sleep influences appetite-related functions in the brain (Greer et al., 2013; Hanlon & Van Cauter, 2011). Poor sleep quality and short sleep duration are associated with decreased levels of the hormone leptin, which regulates appetite suppression, and elevated levels of the hormone ghrelin, which stimulates hunger (Klok et al., 2007). In a small 3-condition crossover trial of nine men aged 20–40 years, Schmid et al. reported that one night of total sleep deprivation led to higher levels of ghrelin compared to one night of sleeping seven hours (Schmid et al., 2008).

As noted, we observed a direct association between sleep quality and the FV dietary pattern only in women and not men. Haghighathoost et al. found that among 410 Iranian women under 30 years of age, those who slept < 6 hours/day had higher intake of total carbohydrates, but lower intake of fruits, compared to those who slept ≥ 6 hours/day (Haghighatdoost et al., 2012). A late midpoint of sleep (midpoint between falling asleep and waking up) was also inversely associated with vegetable, milk, and egg consumption in a cross-sectional study of 3,304 young Japanese women (Sato-Mito et al., 2011).

Results from studies on sleep behaviors by sex are inconsistent. Some of this inconsistency between studies may be due to differences in participants’ ages and in the measurement of sleep (self-report vs. actigraphy) from study to study (Lauderdale et al., 2006; Mezick et al., 2009; Roane et al., 2015). Mezick et al. found that among middle-aged adults, women had greater variability in sleep duration and disturbances than men (Mezick et al., 2009). In a previous study of the BCOPS cohort, women were slightly more likely to report daytime dysfunction compared to men (Bond et al., 2013). Menopausal symptoms are associated with sleep disturbances and difficulty falling asleep (Brown et al., 2009). Women may also be more likely to change their eating behaviors as a result of stress or depression (Allgöwer et al., 2001; Weinstein et al., 1997). While we observed a significant association between better sleep quality and higher fruit and vegetable consumption in women, we also observed an association of fewer sleep disturbances with higher consumption of these same foods in men, indicating a need for additional research on how various dimensions of sleep may differentially influence dietary patterns in women and men.

Our study included several limitations. Due to the cross-sectional design of the study, we cannot determine if the observed association between sleep quality and dietary intake is causal. Recall bias is possible if those with poor sleep quality or short duration reported their dietary intake over the past year differently than those with optimal sleep characteristics. The use of self-reported sleep duration may also be particularly prone to misclassification bias. Residual confounding may exist if we did not adjust for factors related to both sleep characteristics and diet. For example, no data were available for adjustment of our associations in women by menopausal status and a history of mental health disorders was not evaluated. Further, while the PSQI may broadly indicate whether an exposure under study may be associated with sleep quality and duration, it does not capture specific information on factors that may contribute to greater versus lower scores on the PSQI. In a previous BCOPS study, poor sleep quality and duration were significantly associated with frequent exposure to traumatic workplace events in police officers and these findings varied by sex (Bond et al., 2013). We were also unable to example potential latency effects that may influence the association between sleep quality and dietary patterns, such as retirement duration, though exclusion of retired participants did not change the reported effect estimates.

Our study had numerous strengths. A validated sleep quality questionnaire was used addressing multiple aspects of sleep behavior and was administered using a standardized protocol (compared to clinical and polysomnographic data, the PSQI has a sensitivity of 89.6% and a specificity of 86.5%) (Buysse et al., 1989). Data were available on the intake of over 100 food items and for a wide range of potential confounders with minimal missing data. The BCOPS study also provided a unique opportunity to explore the relationship between sleep quality and duration and diet in a U.S. worker population that is also at high risk for sleep disorders and other chronic health conditions.

Overall, findings from this study are suggestive of an association between poor sleep quality and low fruit and vegetable intake in police officers; and this association differed by sex. This study adds to the limited literature on the role of sleep quality on food-based dietary patterns and may provide direction for future studies exploring methods to reduce obesity and cardiovascular disease among police officers. Consideration of work schedule policies, fatigue management training, and nutrition and wellness counseling for police officers in interventional studies may help mitigate the impact of poor sleep quality and short sleep duration on unhealthy eating habits. Currently, studies examining the association between sleep quality and food-based dietary consumption are largely limited to cross-sectional studies which primarily examine intake of specific nutrients or food groups over comprehensive dietary patterns. Prospective studies are warranted, with a larger sample size and further consideration of factors, such as sex, which may modify the association between sleep quality and dietary patterns.