Self-reported health behaviors, including sleep, correlate with doctor-informed medical conditions: data from the 2011 Health Related Behaviors Survey of U.S. Active Duty Military Personnel

The U.S. Military maintains extensive records on the health of US Military personnel and their families via its healthcare system. However, information regarding health behaviors is not usually present in these records. Therefore, to systematically assess health behaviors among active duty personnel, the U.S. Department of Defense conducts a triennial Health Related Behaviors Survey (HRBS) of active duty military personnel [1], which assesses a range of self-reported health behaviors and risks, such as physical activity, diet, drug and alcohol use, posttraumatic stress, and doctor-informed medical conditions. The representative survey includes responses from participants in all branches of the Department of Defense [DoD] (Army, Navy, Marine Corps, and Air Force), as well as U.S. Coast Guard [USCG] members. The survey is voluntary and anonymous, and is used to guide the health-related policies in the Armed Forces. The HRBS provides data that are otherwise unavailable in this large subset of the U.S. population, approximately 1.3 million service members, since active duty military are excluded from national or state health-related surveys, such as the National Health and Nutrition Examination Survey (NHANES) or the Behavioral Risk Factor Surveillance System (BRFSS).

HRBS data have been used in a number of ways to study active duty and USCG personnel, for example, to identify whether they meet the Healthy People 2010 objectives [2], to describe mental health patterns and associations [3‐5], and assess changes over time in self-reported overweight/obesity [6, 7]. In depth analysis of demographic characteristics and health-related behaviors, and their association with self-reported health outcomes, may alter the manner in which military personnel are assessed and counseled for potential modifiable and non-modifiable demographic and behavioral risks. Furthermore, associations between health behaviors and disease risk factors may be different in military versus non-military populations due to differences in occupational demands and lifestyle characteristics. For example, deployment absences and frequent moves are sources of family stress that are more common in military than civilian families [8, 9]. Military personnel are required to meet standards of physical fitness and weight (or body composition) [10] and engage in regular physical training [11, 12]. Military personnel are part of a workforce with institutionalized standards and a unique warrior culture and ethos not present in the civilian workforce. [13] The unique occupational demands frequently involve high risk activities (e.g., combat), which are an inherent part of military service [14]. Also in contrast to civilians, and of specific relevance to the present research, military personnel and their dependents are eligible to receive healthcare at little or no cost [9].

Therefore, the aims of this study were to examine relationships between generally unmodifiable sociodemographic factors (e.g., age, sex, race/ethnicity) and self-reported health behaviors (e.g., exercise and diet), as well as relationships between these sociodemographic factors and self-reported, doctor-informed medical conditions (i.e., hypertension, dyslipidemia, and hyperglycemia) and overweight/obesity. We also examined the relationships between these self-reported health behaviors and the medical conditions.

Of the 27,034 respondents included in the analysis, 29, 17, 21, 19, and 14%, were in the Air Force, Army, Marines, Navy, and USCG, respectively (Table 1). The weighted mean (SEM) age of respondents was 29.9 (0.1) years; weighted percentages showed 14.7% were female, 69.8% identified as non-Hispanic white, 8.6% as non-Hispanic black, 25.5% had at least a high school education, 37.2% had completed some college, 58.9% were married, and 39.9% had one or more children residing with them. Just over 10% reported having to lose some weight to join a service, but only 3% reported being currently enrolled in a weight control program. A majority of respondents (61.3%) reported having been deployed at some point since September 11, 2001.

Weighted responses indicated low weekly intake of fried food, sweets, and snacks (~ 3–4 ser/wk), moderate intake of starches and sugary drinks (~ 5–6 ser/wk), and at least daily intake of fruit, vegetables, whole grains, dairy, lean meat, and caffeinated drinks (~ 8–10 ser/wk). Respondents reported 175, 116, and 104 min/wk. of moderate exercise, vigorous exercise, and strength training, respectively. Nearly 60% of the weighted sample reported never smoking. Average intake of alcohol in the prior month was a third of drink per day, with > 20% of the weighted sample reporting no alcohol intake. Average sleep duration was 6.3 h/night, with 32% reporting 6–7 h/night—the largest sleep category.

Nearly 24% of the sample reported having at least one medical condition of five possible doctor-informed conditions: 13.0% reported high blood pressure; 12.7% high cholesterol; 5.5% high triglycerides; 6.1% low HDL cholesterol; and 1.5% high blood glucose. Nearly 63% of the sample categorized themselves as overweight or obese, and mean BMI was 26 kg/m².

We used a large anonymous self-reported survey of the Armed Forces to assess relationships between sociodemographic characteristics, health behaviors, and doctor-informed medical conditions and overweight/obesity. When we examined individual health behaviors and medical conditions, by far the most consistent and largest relationship was between low sleep duration and higher odds of every medical condition. Otherwise, other behaviors, including dietary and exercise behaviors, were inconsistently associated with medical conditions. Age, sex, marital status, and presence of children appeared to be the only relatively consistent sociodemographic factors related to health behaviors. Older individuals, females, and those residing with children tended to exercise less, but eat healthier foods and not smoke. Older individuals also slept less, as did those with children present. Those who were married appeared to have consumed a better diet, exercised more, and slept more than their non-married counterparts. In terms of medical conditions, older age and female sex appeared to be the only consistent drivers of higher and lower odds, respectively, of the presence of a given medical condition. Relationships of race/ethnicity, education, service branch, and marital status with medical conditions were inconsistent. Finally, in exploratory factor analyses of health behaviors, associations of behavioral patterns—two patterns in particular—with medical conditions were associated with medical conditions as expected: the pattern characterized by a “healthy” diet was associated with lower odds of all medical conditions except high blood pressure, while the pattern characterized by “bad” habits (including lack of sleep) was associated with higher odds of every medical condition.

We observed robust associations between short sleep and each medical condition in the present study, ranging from 36% higher odds of overweight/obesity to over double the odds of hypertension, in contrast to inconsistent findings for diet and exercise, and smoking. There is increasing evidence of the importance of sleep to health [17‐23], for example, a 2015 meta-analysis of 10 prospective cohort studies investigating sleep in relation to type 2 diabetes risk reported a U-shaped, dose-response relationship between sleep duration and diabetes risk [22]. The lowest risk was observed at sleep durations of 7–8 h/day, with incrementally higher risk of diabetes below or above that duration. Several reports have linked short sleep durations with higher risk of overweight/obesity [17, 20, 21], including an analysis of the representative U.S. population, in which authors reported that, relative to those who slept 7–8 h/day, very short sleepers (< 5 h) had 30% higher odds of being overweight and were twice as likely to be obese, while (5–6 h) short sleepers had 20% higher odds of being overweight and 57% higher odds of being obese [20]. In addition, a 2013 meta-analysis of 17 cohorts investigating the relationship between sleep duration and incident hypertension reported that short sleep duration conferred 21% higher risk of developing the condition [18]. Because recommendations to change diet, exercise, and smoking behaviors, but not sleep behaviors, are a part of many primary and secondary prevention recommendations [24‐26], it is possible that individuals are more likely to make changes to these behaviors, rather than sleep, in an effort to prevent or as a result of receiving a diagnosis. In fact, sleep is not mentioned as a modifiable risk factor in disease prevention/management guidelines from the American Diabetes Association or from the American College of Cardiology/American Heart Association [24‐26], and thus may not be a part of lifestyle changes individuals undertake to modify their perceived disease risk.

In military populations, the critical importance of sleep has only recently become part of the “Performance Triad”, an educational initiative introduced in 2013 on health readiness and wellness including 1) physical activity; 2) nutrition; and 3) sleep [27, 28]. Our data therefore confirm the importance of the inclusion of sleep in the “Triad” initiative. In our study, poor sleep may also be a complex product of poor health, and occupational or personal stress [29, 30]. Other studies have indicated that sleep disorders, including insomnia and sleep apnea, appear to be more common in service members than in civilians [31‐33]. Our observations using self-reported survey data, on the strong relationship between poor sleep and behaviors and medical conditions, agree with other studies usingself-reported surveys of military personnel [34].

Our observations regarding select sociodemographic characteristics, notably age, sex, and marital status, as being associated with select health behaviors generally agree with existing evidence [2, 7]. With the exception of odds of high blood glucose, which was higher in females versus males, other associations of age and sex—well-known biological drivers of health risks—were as generally expected. Married individuals reported better health behaviors than never- or previously married individuals; however this did not necessarily translate into lower odds of having medical conditions for married individuals. Given the cross-sectional nature of the study design, we can only speculate as to why: first, married individuals may be more likely to seek routine medical care at the urging of their spouse [35‐37], and thus have been told by a physician of a prevalent condition that would have otherwise gone undetected. Second, cross-sectional analyses preclude assessment of sequential behaviors; the reported behaviors may be in response to a diagnosis, a form of reverse causality. In addition, although analyses were adjusted for age and presence of children, as well as other sociodemographic characteristics, relationships between marital status and medical conditions are nuanced and complicated by time-related factors, including secular trends [37‐39]. For example, a recent study using data from a panel of nationally representative households observed that while a protective association of long marriage on women’s self-reported health was evident in the earlier birth cohort (born 1955–1964), it had reversed in the most recent birth cohort (born 1975–1984) [39].

There were other unexpected associations. For example, we observed intake of snacks, sweets, and sugary drinks (in addition to the “Unhealthy diet” factor which included these foods) were associated with lower odds of overweight/obesity, which is generally the opposite of what would be expected given evidence regarding the long-term negative consequences of consuming these foods on health [40‐43]. These associations held despite adjusting for history of weight loss and current enrollment in a weight control program. It is possible this finding reflects the generally young age of the respondents in this military population, or that some respondents were simply not at risk of overweight. This observation, in fact, appears to be consistent with cross-sectional nationally representative NHANES data on snacking and sugary beverages in both adolescents [44] and adults [45] which have shown no clearly discernible associations with overweight or obesity. In addition, state-based BRFSS 2013 data indicate that sugar-sweetened beverage intake declines with higher age. Intake was most prevalent among younger adults (age 18–24 y), at 43% reporting consuming these beverages one or more times per day, while those in older age groups reported less consumption (38% of those age 25–34 y, 30% in those age 35–54 y, and 19% in those age ≥ 55 y) [46]. On the other hand, our exploratory observations regarding the “healthy diet” pattern and favorable odds of medical conditions were consistent with dietary pattern literature [47‐50].

As noted above, our analyses are limited by the cross-sectional nature of the survey, which precludes causal inferences. Particularly when assessing health behaviors and medical conditions concurrently, a behavioral pattern may reflect a subset of respondents who have changed their behaviors as result of having been alerted to a health risk [51‐56]. Furthermore, we were limited to gross rather than granular assessments of dietary components. We were also likely limited in the validity of our inferences by social desirability response bias of self-reported data. In addition, non-response bias, length of the survey, and fear of loss of anonymity among either non-respondents or due to selective non-responses in those excluded from the present analysis may also limit our ability to make inferences or generalize our findings to the broader Armed Services, despite using revised sampling weights to accommodate exclusions.

This study presents relationships between sociodemographic characteristics, health behaviors, and doctor-informed medical conditions and overweight/obesity in the U.S. Armed Forces, which are typically not included, or not separately reported, in national or state health surveys.

Despite aforementioned limitations and our inability to deduce causal relationships, we observed pronounced associations between health behaviors—especially sleep duration—and medical conditions, thus adding to evidence that sleep is a critical, potentially modifiable behavior within this population. It is recognized that there are times when sleep restriction is unavoidable in this population (e.g., during periodic training and mission scenarios), however, adequate sleep should continue to be promoted as an important part of overall health and wellness throughout the military community.