The Commander’s Wellness Program: Assessing the Association Between Health Measures and Physical Fitness Assessment Scores, Fitness Assessment Exemptions, and Duration of Limited Duty

The Commander’s Wellness Program: Assessing the Association Between Health Measures and... Abstract Introduction Air Force Medical Service health promotions staff have identified a set of evidenced-based interventions targeting tobacco use, sleep habits, obesity/healthy weight, and physical activity that could be integrated, packaged, and deployed as a Commander’s Wellness Program. The premise of the program is that improvements in the aforementioned aspects of the health of unit members will directly benefit commanders in terms of members’ fitness assessment scores and the duration of periods of limited duty. The purpose of this study is to validate the Commander’s Wellness Program assumption that body mass index (BMI), physical activity habits, tobacco use, sleep, and nutritional habits are associated with physical fitness assessment scores, fitness assessment exemptions, and aggregate days of limited duty in the population of active duty U.S. Air Force personnel. Methods This study used a cross-sectional analysis of active duty U.S. Air Force personnel with an Air Force Web-based Health Assessment and fitness assessment data during fiscal year 2013. Predictor variables included age, BMI, gender, physical activity level (moderate physical activity, vigorous activity, and muscle activity), tobacco use, sleep, and dietary habits (consumption of a variety of foods, daily servings of fruits and vegetables, consumption of high-fiber foods, and consumption of high-fat foods). Nonparametric methods were used for the exploratory analysis and parametric methods were used for model building and statistical inference. Results The study population comprised 221,239 participants. Increasing BMI and tobacco use were negatively associated with the outcome of composite fitness score. Increasing BMI and tobacco use and decreasing sleep were associated with an increased likelihood for the outcome of fitness assessment exemption status. Increasing BMI and tobacco use and decreasing composite fitness score and sleep were associated with an increased likelihood for the outcome of limited duty status, whereas increasing BMI and decreasing sleep were associated with the outcome of increased aggregate days of limited duty. The observed associations were in the expected direction and the effect sizes were modest. Physical activity habits and nutritional habits were not observed to be associated with any of the outcome measures. Conclusions The Commander’s Wellness Program should be scoped to those interventions targeting BMI, composite fitness score, sleep, and tobacco use. Although neither self-reported physical activity nor nutritional habits were associated with the outcomes, it is still worthwhile to include related interventions in the Commander’s Wellness Program because of the finding in other studies of a consistent association between the overall number of health risks and productivity outcomes. INTRODUCTION Employers increasingly recognize worker health as significantly contributing to organizational productivity and profitability.1 There is a growing body of scientific literature, particularly within the occupational and preventive medicine domains, demonstrating associations between modifiable health risks, health conditions, and worker productivity.2–13 Several studies suggest that the cost of lost productivity secondary to health-related conditions (presenteeism) exceeds the costs of direct medical care.14,15 And another study reported that the minority of companies that do have access to productivity data tend to focus less on direct medical care costs and more on productivity outcomes.16 Based on a literature review, Riedel and colleagues proposed that improvements in worker health could lead to corresponding improvements in the quality of goods and services, greater creativity and innovation, enhanced resiliency, and increased intellectual capacity.17 Not surprising, employers are increasingly focusing on health promotion activities as a means of optimizing worker productivity. Health promotion activities in the military are most effective when they are driven by line officers (i.e., officers in operational/combat and combat support specialties) rather than professional officers in the medical service. Line officers are more apt to take ownership of health promotion activities when doing so positively impacts measures that are valued by unit commanders. Based on discussions with commanders, such valued measures include physical fitness assessment scores, fitness assessment exemption status, and Airman Availability, the latter reflecting the proportion of Airmen without duty limitations because of health conditions. Air Force Medical Service health promotions staff have identified a set of evidenced-based interventions targeting tobacco use, sleep habits, obesity/healthy weight, and physical activity that could be integrated, packaged, and deployed as a Commander’s Wellness Program. The premise of the program is that improvements in the aforementioned aspects of the health of unit members will directly benefit commanders in terms of members’ fitness assessment scores, fitness assessment exemption status, and the duration of periods of limited duty. The purpose of this study is to validate the Commander’s Wellness Program assumption that body mass index (BMI), physical activity habits, tobacco use, sleep, and nutritional habits are associated with physical fitness assessment scores, fitness assessment exemption status, and the duration of periods of limited duty time in the population of active duty U.S. Air Force personnel. The following hypotheses guided this study: H1: Physical fitness assessment scores are associated with BMI, physical activity habits, tobacco use, sleep, and nutritional habits; H2: Likelihood of fitness assessment exemption status is associated with BMI, physical activity habits, tobacco use, sleep, and nutritional habits (An Air Force medical provider may grant an Airman a time-limited exemption from the aerobic and muscle fitness components of fitness assessments based on medical evaluation. Exemptions are granted for medical conditions that limit function or may be aggravated by performance of fitness assessment activities. Typical medical conditions include back, upper extremity, or lower extremity injuries or illnesses.); and H3: Aggregate-limited duty time in the past year is associated with BMI, physical activity habits, fitness assessment score, tobacco use, sleep, and nutritional habits. METHODS Study Design This study was conducted under a human-use protocol approved by the 711th Human Performance Wing Institutional Review Board. A waiver of informed consent of participants was granted due to the impracticality of obtaining written consent from each participant in the study population. This study was a cross-sectional analysis of active duty U.S. Air Force personnel during fiscal year 2013. Archival data were extracted from the following three databases: the Air Force Web-based Health Assessment (AF Web HA), the Air Force Fitness Management System, and the Aeromedical Services Information Management System. Social Security numbers were used to match participant data across the three datasets and were then removed from the study dataset to preserve de-identification. Study Population This study enrolled all active duty U.S. Air Force personnel with an AF Web HA accomplished between October 1, 2012, and September 30, 2013. Participants were excluded from the study if they did not have fitness assessment data or an exemption during the corresponding period. Measurements The AF Web HA is a web-based health questionnaire completed by Airmen during their mandatory annual Periodic Health Assessment. Variables extracted from the AF Web HA included age, gender, physical activity level (moderate physical activity [days per week], vigorous activity [days per week], and muscle activity [days per week]), tobacco use (smoker, chewing tobacco user, both smoker and chewing tobacco user, former tobacco user, or never used tobacco), sleep (days unrested in the past 30 d), and dietary habits (consumption of a variety of foods [yes, no, or unsure], daily servings of fruits and vegetables [0, 1–2, 3–4, and ≥5], consumption of high-fiber foods [rarely, daily, every meal, one to two times per day, three to five times per day, or missing data], and consumption of high-fat foods [rarely, daily, every meal, one to two times per day, three to five times per day, or missing data]). The Air Force Fitness Management System provided data on measured height and weight, which were used to calculate BMI, fitness assessment exemption status, and composite fitness score. Data on annual aggregate days of limited duty, defined in terms of the presence of a duty, mobility, and/or fitness restriction, were obtained from the Aeromedical Services Information Management System. Statistical Analysis The study dataset was randomly partitioned into two samples: a learning sample (70% of the observations) for exploratory analysis and a validation sample (30% of the observations) for model building and statistical inference. Nonparametric methods were used for the exploratory analysis and parametric methods were used for model building and statistical inference given the greater ease of interpretation of the latter (e.g., standard errors and p-values). Separating variable selection and model building ensured that the reported standard errors and p-values were valid. Tree-based gradient boosting machine (GBM)18 modeling was used for exploratory analysis on the learning sample. The GBM variable importance capability was used to select the most influential predictors to include in the parametric analysis; larger variable importance scores suggest greater importance in terms of predicting the response. Parametric models were then used to model the validation samples: an ordinary multiple linear regression (MLR) was used to model composite fitness score, an ordinary binomial generalized linear model (GLM) with a logit link was used to model the probability of a fitness assessment exemption, and a hurdle model was used to model aggregate-limited duty time. A hurdle model is a count model with two components: a zero component for modeling zero versus positive counts and a positive component for modeling the positive counts. For the zero component, a binomial GLM with a logit link was used. The positive component used a (zero-truncated) negative binomial GLM. The hurdle model assumes that a binomial probability model governs the binary outcome of whether limited duty time will be zero or a positive value. If the value is positive, the “hurdle is crossed,” and the conditional distribution of limited duty time is governed by a zero-truncated negative binomial model.19 Goodness of fit for the MLR model and the binomial GLM were assessed using the coefficient of determination statistic (R2) and McFadden’s R2, respectively. No analogous goodness-of-fit statistic exists for the hurdle model. Data were analyzed using R, version 3.2.0.20 The R package gbm21 was used for implementing the GBM models for the exploratory analyses.22 The R package pcsl was used for fitting a hurdle model to limited duty time.23 Statistical significance was a priori defined at the 0.05 level. RESULTS Study Population Summary Data Based on the availability of AF Web HA, the eligible study population comprised 221,239 participants. Table I provides descriptive statistics for the measured variables for the final study population. The median composite fitness score was calculated for the subpopulation comprising participants without a fitness assessment exemption. The median aggregate days of limited duty was calculated for the subpopulation comprising participants with aggregate days of limited duty greater than zero. Overall, the study population was comprised predominately of young males who were normal-to-overweight by BMI. The majority of the population reported eating a variety of foods, consumed one to four servings of fruit and vegetables daily, consumed high-fiber foods on a daily basis, and consumed high-fat foods 1–2 d/wk. The majority were also non-smokers who engaged in moderate-to-vigorous exercise three to four times per week and scored in the satisfactory-to-excellent range on fitness assessments. Approximately one-sixth had a fitness assessment exemption and one-quarter had a period of temporary limited duty for a medical condition. Table I. Descriptive Statistics for the Study Population Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) IQR, interquartile range. Table I. Descriptive Statistics for the Study Population Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) IQR, interquartile range. H1: Composite Fitness Score The GBM model for composite fitness score achieved a pseudo-R2 of 14.357% on the independent test sample; that is, the model included predictors explaining approximately 14% of the variance in composite fitness score. Based on variable importance scores, BMI, age, tobacco use, and gender were (in descending order) the most important predictors of composite fitness score. These four predictors accounted for roughly 95% of the total relative influence from all the available predictors. The GBM model also suggested an age and BMI interaction. The MLR model for composite fitness score included BMI, age, tobacco use, and gender as well as a BMI and age interaction term (F8,55531 = 931.609, p < 0.001). The MLR model obtained an adjusted R2 of 12.06% on the test sample. Residual plots did not indicate any sign of heteroscedasticity (non-constant variance). Normal Q-Q plots showed signs of a negatively skewed distribution for the residuals due to the small cluster of extremely low fitness scores. However, the extremely large sample size still allowed interpretation of the confidence intervals, etc., per usual (due to the central limit theorem). Cook’s distance indicated that only three of the outlying observations were influential, but again the large sample size mitigated this concern. Table II displays the estimated regression coefficients, including approximate standard errors and p-values. There was a significant negative association between BMI and composite fitness score. Chewing tobacco users, non-tobacco users, and former tobacco users had higher composite fitness scores relative to smokers/chewing tobacco users; there was no difference between smokers only and smokers/chewing tobacco users. Accordingly, we partially accept hypothesis 1: BMI and tobacco use were associated with composite fitness score, but physical activity, sleep, and nutritional habits were not associated with composite fitness score. Effect size, as ascertained based on the partial correlation coefficient, was small for tobacco use (r = 0.11) and negligible for BMI (r = 0.02).24 Table II. Regression Output for the MLR Model Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 CI, confidence interval; SE, standard error. Table II. Regression Output for the MLR Model Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 CI, confidence interval; SE, standard error. H2: Fitness Assessment Exemption Status The GBM model for fitness assessment exemption status achieved an area under the receiver operating characteristic curve of 0.676 on the independent test sample. Based on variable importance scores, BMI, sleep, gender, age, and tobacco use were (in descending order) the most important predictors of fitness assessment exemption status. These five predictors accounted for roughly 92% of the total relative influence from all the available predictors. The binomial GLM for fitness assessment exemption status included age, BMI, gender, and sleep (χ2 = 3,533.018, p < 0.001). The model obtained a McFadden’s R2 of 8.67% on the test sample. Table III displays the estimated odds ratios and p-values. Increasing BMI and days unrested (i.e., poor quality sleep) were associated with an increased likelihood of fitness assessment exemptions, although the observed effect size was small.24 Chewing tobacco users, non-tobacco users, former tobacco users, and smokers had a lower likelihood of a fitness assessment exemption relative to smokers/chewing tobacco users. These associations had small effect sizes with the exception of non-tobacco users for which the effect size was medium.24 Accordingly, we partially accept hypothesis 2: BMI, sleep, and tobacco use were associated with fitness assessment exemption status, but physical activity habits and nutritional habits were not associated with fitness assessment exemption status. Table III. Logistic Regression Model Results for Exemption Status Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 CI, confidence interval; OR, odds ratio. Table III. Logistic Regression Model Results for Exemption Status Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 CI, confidence interval; OR, odds ratio. H3: Aggregate-Limited Duty Time The GBM model for aggregate-limited duty time achieved a pseudo-R2 of 7.298% on the independent test sample. Based on variable importance scores, age, BMI, composite fitness score, gender, sleep, and tobacco use were (in descending order) the most important predictors of aggregate-limited duty time. These six predictors accounted for roughly 94% of the total relative influence from all the available predictors. The coefficients for the hurdle model, along with approximate odds ratios, standard errors, and p-values, are given in Tables IV and V. The hurdle model zero component addresses the likelihood of having zero or greater-than-zero aggregate days of limited duty based on the predictor variables shown in Table IV. Increasing BMI and days unrested (i.e., poor quality sleep) were associated with an increased likelihood for limited duty, whereas increasing composite fitness score was associated with a decreased likelihood for limited duty. Never using tobacco, former tobacco user, chewing tobacco user, and smoker (in order of decreasing effect size) were associated with a decreased likelihood for limited duty relative to smoker/chewing tobacco user. Table IV. Hurdle Model Zero Component Results for Limited Duty Time Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 CI, confidence interval; OR, odds ratio. Table IV. Hurdle Model Zero Component Results for Limited Duty Time Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 CI, confidence interval; OR, odds ratio. Table V. Hurdle Model Positive Component Results for Limited Duty Time Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 CI, confidence interval; SE, standard error. Table V. Hurdle Model Positive Component Results for Limited Duty Time Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 CI, confidence interval; SE, standard error. The hurdle-model-positive component addresses the log of the expected aggregate days of limited duty as a function of predictor variables shown in Table V. Both BMI and days unrested (i.e., poor quality sleep) were positively associated with aggregate days of limited duty counts. Based on both the zero and positive components of the hurdle model, we partially accept hypothesis 3: BMI, composite fitness score, sleep, and tobacco use were associated with limited duty time, but physical activity and nutritional habits were not associated with limited duty time. DISCUSSION This study observed that BMI, composite fitness score, sleep, and tobacco use were all associated with outcome measures that are valued by unit commanders — that is, fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. BMI and tobacco use were associated with all three outcome measures, whereas sleep was associated with two outcome measures. Composite fitness score was only included as a predictor in the analysis of the outcome measure of aggregate-limited duty time, for which it was significantly associated. Accordingly, interventions targeting BMI, composite fitness score, sleep, and tobacco use should be components of the Commander’s Wellness Program. Physical activity habits and nutritional habits were not observed to be associated with any of the outcome measures, contrary to the study hypotheses. The authors identified two thesis studies,25,26 but no published studies, that examined the association between AF Web HA data and clinically and/or operationally relevant outcomes. In a cross-sectional study of 24,020 Airmen, Bell assessed the association of supplement use as reported on the AF Web HA with fitness scores obtained from the Air Force Fitness Management System and the number of medical visits during deployment as reported on the Post Deployment Health Assessment. Although the majority of the population (80%) reported using dietary supplements, no significant association was observed between supplement use and fitness scores or number of medical visits during deployment.25 In another cross-sectional study, Madrid examined the effectiveness of the AF Web HA mental health screening questions in predicting a mental health disorder diagnosis as derived from electronic health record data. The sensitivity of the mental health screening questions was 3% and their positive predictive value was 9%.26 To the best of the authors knowledge, this study is the first to systematically evaluate and report the association between multiple modifiable risk factors as reported on the AF Web HA and the three outcomes of fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. These outcome measures are relatively unique to the military and lack readily available and comparable correlates in other industry sectors. Perhaps the most appropriate point of comparison is the recent body of occupational medicine research examining the association between modifiable health risks with increased on-the-job productivity loss and presenteeism.1 There are a number of well-designed studies examining the relationship between modifiable health risks and employee productivity.2–13 All but of one of the studies5 reviewed by the authors relied on self-reported survey data as the primary data source for outcome measures. Additionally, these studies were diverse in terms of the populations studied and both the health status measures and outcomes of interest were drawn from a variety of surveys or scales. Although individual studies yielded contradictory findings about individual health risk factors, all the studies yielded the common general finding that healthier employees had more favorable productivity-related outcomes.7 Collectively, the health risk assessments used in the aforementioned studies overlapped with the AF Web HA measures included in this study in terms of the following measures: nutrition, physical activity, tobacco use, and weight. Five studies3,8,9,11,13 reported an association between nutritional habits and productivity, of which three studies8,9,13 found that high-risk nutritional habits were associated with productivity loss. Ten studies2–6,8–11,13 reported results for physical activity habits and productivity, of which six studies3,6,8,10,11,13 found an association between inactivity and productivity loss. Ten studies2–6,8–11,13 also reported on results for tobacco use and productivity with only four studies2,4,6,10 finding an association between tobacco use and productivity loss. Of note, one of these four studies2 found that tobacco users had less productivity loss than non-smokers. Lastly, 11 studies2–6,8–13 reported results for unhealthy weight and productivity with six studies4–6,8,10,12 finding an association between unhealthy weight and productivity loss. Additionally, three studies4,6,10 reported a consistent relationship existed between the number of modifiable health risks and productivity loss regardless of the significance of associations of individual health risks and productivity. It is difficult to put the findings of this study within the context of the aforementioned literature given the variability in the observed associations between individual modifiable health risks and productivity. As previously described, there was considerable variation between studies in productivity research methodology and populations and work domains studied. Also, as the study by Shi and colleagues demonstrates,13 other non-health-related variables likely contribute to productivity loss at the same time. The present study was unlike the vast majority of other studies in that it used objective outcome measures rather than self-reported productivity impairment. Also, this study used a unique instrument (i.e., the AF Web HA) to measure health risks. The AF Web HA adapted tobacco use questions from the National Health and Nutrition Examination Survey III and sleep questions from the Behavioral Risk Factor Surveillance System 2001; otherwise, the lineage of the assessment questions relevant to this study is unknown, and to the best of the authors’ knowledge, the AF Web HA was never validated. Another consideration related to the AF Web HA data is that individuals’ health risk responses are reported to their primary care team, who then have the responsibility to follow up and deliver indicated clinical preventive services. Airmen’s knowledge of this fact may have led to a bias to underreport health risks. A strength of the current study was the use of a large cohort of active duty Air Force personnel to better understand the association between modifiable health risks and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. A limitation of this study was its cross-sectional design, which limits the ability to draw inferences of cause and effect. Accordingly, future research using a retrospective prospective or pre–post prospective study design should evaluate the impact of changes on health risks and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. Another limitation of this study was the small effect sizes of the statistically significant associations between health risks and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. Nevertheless, Grossmeier and colleagues make the point that even very small changes in population-level health risks result in substantial impacts when applied across the entire workforce.7 Their study also suggested that the impact of population-level health improvement is realized over subsequent years, so it is likely this cross-sectional analysis underestimates the effect sizes of interventions deployed as part of the Commander’s Wellness Program. However, these small effect sizes also suggest that there are additional variables not considered in this study that also are associated with the outcomes. Therefore, leadership’s expectations need to be appropriately managed when implementing the Commander’s Wellness Program — targeting health risks is not a panacea for improving valued unit metrics. In conclusion, this study identified that there was an association between some of the health risks targeted in the proposed Commander’s Wellness Program and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. The Commander’s Wellness Program should focus on interventions targeting BMI, composite fitness score, sleep, and tobacco use. Although neither self-reported physical activity nor nutritional habits were directly associated with the outcomes, it is still worthwhile to include related interventions in the Commander’s Wellness Program because of the finding in other studies4,6,10 of a consistent association between the overall number of health risks and productivity outcomes. Acknowledgments The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Air Force, the Department of Defense, or the U.S. Government. References 1 Kirsten W : Making the link between health and productivity at the workplace – a global perspective . Ind Health 2010 ; 48 : 251 – 5 . Google Scholar CrossRef Search ADS PubMed 2 Alavinia S , Molenaar D , Burdorf A : Productivity loss in the workforce: associations with health, work demands, and individual characteristics . Am J Ind Med 2009 ; 52 : 49 – 56 . Google Scholar CrossRef Search ADS PubMed 3 Boles M , Pelletier B , Lynch W : The relationship between health risks and productivity . J Occup Environ Med 2004 ; 46 : 737 – 45 . Google Scholar CrossRef Search ADS PubMed 4 Burton W , Chen C , Conti D , Schultz A , Edington D : The association between health risk change and presenteeism change . J Occup Environ Med 2006 ; 48 : 252 – 63 . Google Scholar CrossRef Search ADS PubMed 5 Burton W , Conti D , Chen C , Schultz A , Edington D : The role of health risk factors and disease on worker productivity . J Occup Environ Med 1999 ; 41 : 863 – 77 . Google Scholar CrossRef Search ADS PubMed 6 Burton W , Chen C , Conti D , Schultz A , Pransky G , Edington DW : The association of health risks with on-the-job productivity . J Occup Environ Med 2005 ; 47 : 769 – 77 . Google Scholar CrossRef Search ADS PubMed 7 Grossmeier J , Mangen D , Terry P , Haglund-Howieson L : Health risk change as a predictor of productivity change . J Occup Environ Med 2015 ; 57 ( 4 ): 347 – 54 . Google Scholar CrossRef Search ADS PubMed 8 Merrill R , Aldana S , Pope J , et al. : Self-rated job performance and absenteeism according to employee engagement, health behaviors, and physical health . J Occup Environ Med 2013 ; 55 : 10 – 8 . Google Scholar CrossRef Search ADS PubMed 9 Pelletier B , Boles M , Lynch W : Change in health risks and work productivity over time . J Occup Environ Med 2004 ; 46 : 746 – 54 . Google Scholar CrossRef Search ADS PubMed 10 Riedel J , Grossmeier J , Haglund-Howieson L , Buraglio C , Anderson D , Terry P : Use of a normal impairment factor in quantifying avoidable productivity loss because of poor health . J Occup Environ Med 2009 ; 51 : 283 – 95 . Google Scholar CrossRef Search ADS PubMed 11 Robroek SJW , van Lenthe F , Burdorf A : The role of lifestyle, health, and work in educational inequalities in sick leave and productivity loss at work . Int Arch Occup Environ Health 2013 ; 86 : 619 – 27 . Google Scholar CrossRef Search ADS PubMed 12 Schultz AB , Edington DW : The association between changes in metabolic syndrome and changes in cost in a workplace population . J Occup Environ Med 2009 ; 51 : 771 – 9 . Google Scholar CrossRef Search ADS PubMed 13 Shi Y , Sears LE , Coberley CR , Pope JE : The association between modifiable well-being risks and productivity: a longitudinal study in a pooled employer sample . J Occup Environ Med 2013 ; 55 : 353 – 64 . Google Scholar CrossRef Search ADS PubMed 14 Collins J , Baase C , Sharda C , et al. : The assessment of chronic health conditions on work performance, absence, and total economic impact for employers . J Occup Environ Med 2005 ; 47 : 547 – 57 . Google Scholar CrossRef Search ADS PubMed 15 Goetzel R , Long S , Ozminkowski R , Hawkins K , Wang S , Lynch W : Health absence, disability, and presenteeism cost estimates of certain physical and mental health conditions affecting U.S. employers . J Occup Environ Med 2004 ; 46 : 398 – 412 . Google Scholar CrossRef Search ADS PubMed 16 Lynch W , Riedel J , Hymel P , Loeppke R , Nelson R , Ashenfelter J : Factors affecting the frequency of value-focused health activities and policies by employers . J Occup Environ Med 2004 ; 46 : 1103 – 14 . Google Scholar CrossRef Search ADS PubMed 17 Riedel J , Lynch W , Baase C , Hymel P , Peterson K : The effect of disease prevention and health promotion on workplace productivity: a literature review . Am J Health Promot 2001 ; 15 : 167 – 91 . Google Scholar CrossRef Search ADS PubMed 18 Friedman J : Stochastic gradient boosting . Comput Stat Data Anal 2002 ; 38 ( 4 ): 367 – 8 . Google Scholar CrossRef Search ADS 19 Zeileis A , Kleiber C , Jackman S : Regression models for count data in R . J Stat Softw 2008 ; 27 ( 8 ): 1 – 25 . Google Scholar CrossRef Search ADS 20 R Core Team : R: a language and environment for statistical computing. R Foundation for Statistical Computing. Available at www.R-project.org; accessed March 15, 2017 . 21 Ridgeway G : gbm: generalized boosted regression models, R package version 2.1.1. 2015 . Available at https://CRAN.R-project.org/package=gbm/; accessed March 15, 2017. 22 Cameron A , Trivedi P : Regression Analysis of Count Data . Cambridge , Cambridge University Press , 1998 . 23 Jackman S : pscl: classes and methods for R developed in the Political Science Computational Laboratory, Stanford University, R package version 1.4.9. 2015 . Department of Political Science, Stanford University. Available at https://pscl.stanford.edu/; accessed March 15, 2017. 24 Cohen J : Statistical Power Analysis for the Behavioral Sciences , 2nd ed. , Hillsdale, NJ , Lawrence Erlbaum , 1988 . 25 Bell M : Dietary Supplement Use Associated with Air Force Fitness and Deployment Health [Thesis] . Dayton , Wright State University , 2015 . 26 Madrid M : Air Force Web Preventive Health Assessment Mental Health Screening Effectiveness [Thesis] . San Antonio , University of Texas School of Public Health , 2010 . Published by Oxford University Press on behalf of the Association of Military Surgeons of the United States 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Military Medicine Oxford University Press

The Commander’s Wellness Program: Assessing the Association Between Health Measures and Physical Fitness Assessment Scores, Fitness Assessment Exemptions, and Duration of Limited Duty

Military Medicine , Volume 183 (9) – Sep 1, 2018

Loading next page...
 
/lp/ou_press/the-commander-s-wellness-program-assessing-the-association-between-PohMJFPsQf
Publisher
Oxford University Press
Copyright
Published by Oxford University Press on behalf of the Association of Military Surgeons of the United States 2018.
ISSN
0026-4075
eISSN
1930-613X
D.O.I.
10.1093/milmed/usx221
Publisher site
See Article on Publisher Site

Abstract

Abstract Introduction Air Force Medical Service health promotions staff have identified a set of evidenced-based interventions targeting tobacco use, sleep habits, obesity/healthy weight, and physical activity that could be integrated, packaged, and deployed as a Commander’s Wellness Program. The premise of the program is that improvements in the aforementioned aspects of the health of unit members will directly benefit commanders in terms of members’ fitness assessment scores and the duration of periods of limited duty. The purpose of this study is to validate the Commander’s Wellness Program assumption that body mass index (BMI), physical activity habits, tobacco use, sleep, and nutritional habits are associated with physical fitness assessment scores, fitness assessment exemptions, and aggregate days of limited duty in the population of active duty U.S. Air Force personnel. Methods This study used a cross-sectional analysis of active duty U.S. Air Force personnel with an Air Force Web-based Health Assessment and fitness assessment data during fiscal year 2013. Predictor variables included age, BMI, gender, physical activity level (moderate physical activity, vigorous activity, and muscle activity), tobacco use, sleep, and dietary habits (consumption of a variety of foods, daily servings of fruits and vegetables, consumption of high-fiber foods, and consumption of high-fat foods). Nonparametric methods were used for the exploratory analysis and parametric methods were used for model building and statistical inference. Results The study population comprised 221,239 participants. Increasing BMI and tobacco use were negatively associated with the outcome of composite fitness score. Increasing BMI and tobacco use and decreasing sleep were associated with an increased likelihood for the outcome of fitness assessment exemption status. Increasing BMI and tobacco use and decreasing composite fitness score and sleep were associated with an increased likelihood for the outcome of limited duty status, whereas increasing BMI and decreasing sleep were associated with the outcome of increased aggregate days of limited duty. The observed associations were in the expected direction and the effect sizes were modest. Physical activity habits and nutritional habits were not observed to be associated with any of the outcome measures. Conclusions The Commander’s Wellness Program should be scoped to those interventions targeting BMI, composite fitness score, sleep, and tobacco use. Although neither self-reported physical activity nor nutritional habits were associated with the outcomes, it is still worthwhile to include related interventions in the Commander’s Wellness Program because of the finding in other studies of a consistent association between the overall number of health risks and productivity outcomes. INTRODUCTION Employers increasingly recognize worker health as significantly contributing to organizational productivity and profitability.1 There is a growing body of scientific literature, particularly within the occupational and preventive medicine domains, demonstrating associations between modifiable health risks, health conditions, and worker productivity.2–13 Several studies suggest that the cost of lost productivity secondary to health-related conditions (presenteeism) exceeds the costs of direct medical care.14,15 And another study reported that the minority of companies that do have access to productivity data tend to focus less on direct medical care costs and more on productivity outcomes.16 Based on a literature review, Riedel and colleagues proposed that improvements in worker health could lead to corresponding improvements in the quality of goods and services, greater creativity and innovation, enhanced resiliency, and increased intellectual capacity.17 Not surprising, employers are increasingly focusing on health promotion activities as a means of optimizing worker productivity. Health promotion activities in the military are most effective when they are driven by line officers (i.e., officers in operational/combat and combat support specialties) rather than professional officers in the medical service. Line officers are more apt to take ownership of health promotion activities when doing so positively impacts measures that are valued by unit commanders. Based on discussions with commanders, such valued measures include physical fitness assessment scores, fitness assessment exemption status, and Airman Availability, the latter reflecting the proportion of Airmen without duty limitations because of health conditions. Air Force Medical Service health promotions staff have identified a set of evidenced-based interventions targeting tobacco use, sleep habits, obesity/healthy weight, and physical activity that could be integrated, packaged, and deployed as a Commander’s Wellness Program. The premise of the program is that improvements in the aforementioned aspects of the health of unit members will directly benefit commanders in terms of members’ fitness assessment scores, fitness assessment exemption status, and the duration of periods of limited duty. The purpose of this study is to validate the Commander’s Wellness Program assumption that body mass index (BMI), physical activity habits, tobacco use, sleep, and nutritional habits are associated with physical fitness assessment scores, fitness assessment exemption status, and the duration of periods of limited duty time in the population of active duty U.S. Air Force personnel. The following hypotheses guided this study: H1: Physical fitness assessment scores are associated with BMI, physical activity habits, tobacco use, sleep, and nutritional habits; H2: Likelihood of fitness assessment exemption status is associated with BMI, physical activity habits, tobacco use, sleep, and nutritional habits (An Air Force medical provider may grant an Airman a time-limited exemption from the aerobic and muscle fitness components of fitness assessments based on medical evaluation. Exemptions are granted for medical conditions that limit function or may be aggravated by performance of fitness assessment activities. Typical medical conditions include back, upper extremity, or lower extremity injuries or illnesses.); and H3: Aggregate-limited duty time in the past year is associated with BMI, physical activity habits, fitness assessment score, tobacco use, sleep, and nutritional habits. METHODS Study Design This study was conducted under a human-use protocol approved by the 711th Human Performance Wing Institutional Review Board. A waiver of informed consent of participants was granted due to the impracticality of obtaining written consent from each participant in the study population. This study was a cross-sectional analysis of active duty U.S. Air Force personnel during fiscal year 2013. Archival data were extracted from the following three databases: the Air Force Web-based Health Assessment (AF Web HA), the Air Force Fitness Management System, and the Aeromedical Services Information Management System. Social Security numbers were used to match participant data across the three datasets and were then removed from the study dataset to preserve de-identification. Study Population This study enrolled all active duty U.S. Air Force personnel with an AF Web HA accomplished between October 1, 2012, and September 30, 2013. Participants were excluded from the study if they did not have fitness assessment data or an exemption during the corresponding period. Measurements The AF Web HA is a web-based health questionnaire completed by Airmen during their mandatory annual Periodic Health Assessment. Variables extracted from the AF Web HA included age, gender, physical activity level (moderate physical activity [days per week], vigorous activity [days per week], and muscle activity [days per week]), tobacco use (smoker, chewing tobacco user, both smoker and chewing tobacco user, former tobacco user, or never used tobacco), sleep (days unrested in the past 30 d), and dietary habits (consumption of a variety of foods [yes, no, or unsure], daily servings of fruits and vegetables [0, 1–2, 3–4, and ≥5], consumption of high-fiber foods [rarely, daily, every meal, one to two times per day, three to five times per day, or missing data], and consumption of high-fat foods [rarely, daily, every meal, one to two times per day, three to five times per day, or missing data]). The Air Force Fitness Management System provided data on measured height and weight, which were used to calculate BMI, fitness assessment exemption status, and composite fitness score. Data on annual aggregate days of limited duty, defined in terms of the presence of a duty, mobility, and/or fitness restriction, were obtained from the Aeromedical Services Information Management System. Statistical Analysis The study dataset was randomly partitioned into two samples: a learning sample (70% of the observations) for exploratory analysis and a validation sample (30% of the observations) for model building and statistical inference. Nonparametric methods were used for the exploratory analysis and parametric methods were used for model building and statistical inference given the greater ease of interpretation of the latter (e.g., standard errors and p-values). Separating variable selection and model building ensured that the reported standard errors and p-values were valid. Tree-based gradient boosting machine (GBM)18 modeling was used for exploratory analysis on the learning sample. The GBM variable importance capability was used to select the most influential predictors to include in the parametric analysis; larger variable importance scores suggest greater importance in terms of predicting the response. Parametric models were then used to model the validation samples: an ordinary multiple linear regression (MLR) was used to model composite fitness score, an ordinary binomial generalized linear model (GLM) with a logit link was used to model the probability of a fitness assessment exemption, and a hurdle model was used to model aggregate-limited duty time. A hurdle model is a count model with two components: a zero component for modeling zero versus positive counts and a positive component for modeling the positive counts. For the zero component, a binomial GLM with a logit link was used. The positive component used a (zero-truncated) negative binomial GLM. The hurdle model assumes that a binomial probability model governs the binary outcome of whether limited duty time will be zero or a positive value. If the value is positive, the “hurdle is crossed,” and the conditional distribution of limited duty time is governed by a zero-truncated negative binomial model.19 Goodness of fit for the MLR model and the binomial GLM were assessed using the coefficient of determination statistic (R2) and McFadden’s R2, respectively. No analogous goodness-of-fit statistic exists for the hurdle model. Data were analyzed using R, version 3.2.0.20 The R package gbm21 was used for implementing the GBM models for the exploratory analyses.22 The R package pcsl was used for fitting a hurdle model to limited duty time.23 Statistical significance was a priori defined at the 0.05 level. RESULTS Study Population Summary Data Based on the availability of AF Web HA, the eligible study population comprised 221,239 participants. Table I provides descriptive statistics for the measured variables for the final study population. The median composite fitness score was calculated for the subpopulation comprising participants without a fitness assessment exemption. The median aggregate days of limited duty was calculated for the subpopulation comprising participants with aggregate days of limited duty greater than zero. Overall, the study population was comprised predominately of young males who were normal-to-overweight by BMI. The majority of the population reported eating a variety of foods, consumed one to four servings of fruit and vegetables daily, consumed high-fiber foods on a daily basis, and consumed high-fat foods 1–2 d/wk. The majority were also non-smokers who engaged in moderate-to-vigorous exercise three to four times per week and scored in the satisfactory-to-excellent range on fitness assessments. Approximately one-sixth had a fitness assessment exemption and one-quarter had a period of temporary limited duty for a medical condition. Table I. Descriptive Statistics for the Study Population Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) IQR, interquartile range. Table I. Descriptive Statistics for the Study Population Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) Variable Descriptive Statistic N 221,239 Age, yr, median (IQR) 28 (10) Body mass index, kg/m2, median (IQR) 25.63 (4.54) Dietary habits, no. (%)  Food variety   Yes 210,919 (95.34)   No 9,224 (4.17)   Unsure 1,089 (0.49)   Missing 7 (0.00)  Daily servings of fruits and vegetables   0 3,393 (1.53)   1–2 102,506 (46.33)   3–4 93,503 (42.46)   ≥5 21,825 (9.86)  Consumption of high-fiber foods   Rarely 2,879 (1.30)   Daily 127,325 (57.55)   Every meal 21,643 (9.78)   1–2 d per week 18,392 (8.31)   3–5 d per week 50,998 (23.05)   Missing 2 (0.00)  Consumption of high-fat foods   Rarely 47,670 (21.55)   Daily 14,203 (6.42)   Every meal 806 (0.36)   1–2 d per week 119,566 (54.05)   3–5 d per week 38,991 (17.62)   Missing 3 (0.00) Gender, male, no. (%) 178,484 (80.67) Physical activity, d/wk, median (IQR)  Moderate activity 3 (3)  Vigorous activity 4 (2)  Muscle activity 3 (2) Sleep, days unrested, median (IQR) 0 (5) Tobacco use, no. (%)  Smoker 27,746 (12.54)  Chewing tobacco user 11,269 (5.09)  Smoker and chewing tobacco user 2,772 (1.25)  Former tobacco user 50,341 (22.75)  Never used tobacco 129,077 (58.34)  Missing 23 (0.01) Fitness  Composite fitness score, median (IQR) 91.7 (10)  On exemption status, no. (%) 36,073 (16.30) Limited duty  On limited duty, no. (%) 51,706 (23.37)  Aggregate days, median (IQR) 45 (69) IQR, interquartile range. H1: Composite Fitness Score The GBM model for composite fitness score achieved a pseudo-R2 of 14.357% on the independent test sample; that is, the model included predictors explaining approximately 14% of the variance in composite fitness score. Based on variable importance scores, BMI, age, tobacco use, and gender were (in descending order) the most important predictors of composite fitness score. These four predictors accounted for roughly 95% of the total relative influence from all the available predictors. The GBM model also suggested an age and BMI interaction. The MLR model for composite fitness score included BMI, age, tobacco use, and gender as well as a BMI and age interaction term (F8,55531 = 931.609, p < 0.001). The MLR model obtained an adjusted R2 of 12.06% on the test sample. Residual plots did not indicate any sign of heteroscedasticity (non-constant variance). Normal Q-Q plots showed signs of a negatively skewed distribution for the residuals due to the small cluster of extremely low fitness scores. However, the extremely large sample size still allowed interpretation of the confidence intervals, etc., per usual (due to the central limit theorem). Cook’s distance indicated that only three of the outlying observations were influential, but again the large sample size mitigated this concern. Table II displays the estimated regression coefficients, including approximate standard errors and p-values. There was a significant negative association between BMI and composite fitness score. Chewing tobacco users, non-tobacco users, and former tobacco users had higher composite fitness scores relative to smokers/chewing tobacco users; there was no difference between smokers only and smokers/chewing tobacco users. Accordingly, we partially accept hypothesis 1: BMI and tobacco use were associated with composite fitness score, but physical activity, sleep, and nutritional habits were not associated with composite fitness score. Effect size, as ascertained based on the partial correlation coefficient, was small for tobacco use (r = 0.11) and negligible for BMI (r = 0.02).24 Table II. Regression Output for the MLR Model Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 CI, confidence interval; SE, standard error. Table II. Regression Output for the MLR Model Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 88.007 1.378 85.307 90.707 <0.001 Age 0.801 0.046 0.712 0.891 <0.001 BMI −0.228 0.052 −0.33 −0.125 <0.001 Gender (ref. female) 1.266 0.104 1.062 1.47 <0.001 Tobacco use  Chewing tobacco user 2.514 0.386 1.757 3.27 <0.001  Former tobacco user 3.603 0.357 2.904 4.302 <0.001  Never used tobacco 4.055 0.351 3.367 4.743 <0.001  Smoker 1.292 0.364 0.58 2.005 <0.001 BMI × age −0.027 0.002 −0.03 −0.023 <0.001 CI, confidence interval; SE, standard error. H2: Fitness Assessment Exemption Status The GBM model for fitness assessment exemption status achieved an area under the receiver operating characteristic curve of 0.676 on the independent test sample. Based on variable importance scores, BMI, sleep, gender, age, and tobacco use were (in descending order) the most important predictors of fitness assessment exemption status. These five predictors accounted for roughly 92% of the total relative influence from all the available predictors. The binomial GLM for fitness assessment exemption status included age, BMI, gender, and sleep (χ2 = 3,533.018, p < 0.001). The model obtained a McFadden’s R2 of 8.67% on the test sample. Table III displays the estimated odds ratios and p-values. Increasing BMI and days unrested (i.e., poor quality sleep) were associated with an increased likelihood of fitness assessment exemptions, although the observed effect size was small.24 Chewing tobacco users, non-tobacco users, former tobacco users, and smokers had a lower likelihood of a fitness assessment exemption relative to smokers/chewing tobacco users. These associations had small effect sizes with the exception of non-tobacco users for which the effect size was medium.24 Accordingly, we partially accept hypothesis 2: BMI, sleep, and tobacco use were associated with fitness assessment exemption status, but physical activity habits and nutritional habits were not associated with fitness assessment exemption status. Table III. Logistic Regression Model Results for Exemption Status Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 CI, confidence interval; OR, odds ratio. Table III. Logistic Regression Model Results for Exemption Status Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 Variable OR 95% CI p-Value Lower Upper Intercept 0.013 0.010 0.016 <0.001 Age 1.023 1.020 1.026 <0.001 BMI 1.120 1.113 1.128 <0.001 Gender (ref. female) 0.404 0.384 0.426 <0.001 Sleep 1.039 1.036 1.042 <0.001 Tobacco use  Chewing tobacco user 0.809 0.660 0.995 0.042  Former tobacco user 0.701 0.583 0.848 <0.001  Never used tobacco 0.607 0.506 0.732 <0.001  Smoker 0.865 0.716 1.049 0.135 CI, confidence interval; OR, odds ratio. H3: Aggregate-Limited Duty Time The GBM model for aggregate-limited duty time achieved a pseudo-R2 of 7.298% on the independent test sample. Based on variable importance scores, age, BMI, composite fitness score, gender, sleep, and tobacco use were (in descending order) the most important predictors of aggregate-limited duty time. These six predictors accounted for roughly 94% of the total relative influence from all the available predictors. The coefficients for the hurdle model, along with approximate odds ratios, standard errors, and p-values, are given in Tables IV and V. The hurdle model zero component addresses the likelihood of having zero or greater-than-zero aggregate days of limited duty based on the predictor variables shown in Table IV. Increasing BMI and days unrested (i.e., poor quality sleep) were associated with an increased likelihood for limited duty, whereas increasing composite fitness score was associated with a decreased likelihood for limited duty. Never using tobacco, former tobacco user, chewing tobacco user, and smoker (in order of decreasing effect size) were associated with a decreased likelihood for limited duty relative to smoker/chewing tobacco user. Table IV. Hurdle Model Zero Component Results for Limited Duty Time Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 CI, confidence interval; OR, odds ratio. Table IV. Hurdle Model Zero Component Results for Limited Duty Time Fit to the Test Sample Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 Variable OR 95% CI p-Value Lower Upper Intercept 1.381 1.011 1.886 0.043 Age 0.992 0.989 0.995 <0.001 BMI 1.046 1.040 1.053 <0.001 Composite fitness score 0.981 0.979 0.983 <0.001 Gender (ref. female) 0.667 0.635 0.701 <0.001 Sleep 1.025 1.022 1.028 <0.001 Tobacco use  Chewing tobacco user 0.861 0.717 1.034 0.109  Former tobacco user 0.818 0.691 0.969 0.020  Never used tobacco 0.786 0.666 0.929 0.005  Smoker 0.915 0.770 1.086 0.310 CI, confidence interval; OR, odds ratio. Table V. Hurdle Model Positive Component Results for Limited Duty Time Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 CI, confidence interval; SE, standard error. Table V. Hurdle Model Positive Component Results for Limited Duty Time Fit to the Test Sample Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 Variable B SE(B) 95% CI p-Value Lower Upper Intercept 3.025 0.100 2.830 3.221 <0.001 Age 0.034 0.001 0.032 0.036 <0.001 BMI 0.012 0.002 0.007 0.016 <0.001 Composite fitness score −0.001 0.001 −0.002 0.000 0.173 Sleep 0.012 0.001 0.010 0.014 <0.001 CI, confidence interval; SE, standard error. The hurdle-model-positive component addresses the log of the expected aggregate days of limited duty as a function of predictor variables shown in Table V. Both BMI and days unrested (i.e., poor quality sleep) were positively associated with aggregate days of limited duty counts. Based on both the zero and positive components of the hurdle model, we partially accept hypothesis 3: BMI, composite fitness score, sleep, and tobacco use were associated with limited duty time, but physical activity and nutritional habits were not associated with limited duty time. DISCUSSION This study observed that BMI, composite fitness score, sleep, and tobacco use were all associated with outcome measures that are valued by unit commanders — that is, fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. BMI and tobacco use were associated with all three outcome measures, whereas sleep was associated with two outcome measures. Composite fitness score was only included as a predictor in the analysis of the outcome measure of aggregate-limited duty time, for which it was significantly associated. Accordingly, interventions targeting BMI, composite fitness score, sleep, and tobacco use should be components of the Commander’s Wellness Program. Physical activity habits and nutritional habits were not observed to be associated with any of the outcome measures, contrary to the study hypotheses. The authors identified two thesis studies,25,26 but no published studies, that examined the association between AF Web HA data and clinically and/or operationally relevant outcomes. In a cross-sectional study of 24,020 Airmen, Bell assessed the association of supplement use as reported on the AF Web HA with fitness scores obtained from the Air Force Fitness Management System and the number of medical visits during deployment as reported on the Post Deployment Health Assessment. Although the majority of the population (80%) reported using dietary supplements, no significant association was observed between supplement use and fitness scores or number of medical visits during deployment.25 In another cross-sectional study, Madrid examined the effectiveness of the AF Web HA mental health screening questions in predicting a mental health disorder diagnosis as derived from electronic health record data. The sensitivity of the mental health screening questions was 3% and their positive predictive value was 9%.26 To the best of the authors knowledge, this study is the first to systematically evaluate and report the association between multiple modifiable risk factors as reported on the AF Web HA and the three outcomes of fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. These outcome measures are relatively unique to the military and lack readily available and comparable correlates in other industry sectors. Perhaps the most appropriate point of comparison is the recent body of occupational medicine research examining the association between modifiable health risks with increased on-the-job productivity loss and presenteeism.1 There are a number of well-designed studies examining the relationship between modifiable health risks and employee productivity.2–13 All but of one of the studies5 reviewed by the authors relied on self-reported survey data as the primary data source for outcome measures. Additionally, these studies were diverse in terms of the populations studied and both the health status measures and outcomes of interest were drawn from a variety of surveys or scales. Although individual studies yielded contradictory findings about individual health risk factors, all the studies yielded the common general finding that healthier employees had more favorable productivity-related outcomes.7 Collectively, the health risk assessments used in the aforementioned studies overlapped with the AF Web HA measures included in this study in terms of the following measures: nutrition, physical activity, tobacco use, and weight. Five studies3,8,9,11,13 reported an association between nutritional habits and productivity, of which three studies8,9,13 found that high-risk nutritional habits were associated with productivity loss. Ten studies2–6,8–11,13 reported results for physical activity habits and productivity, of which six studies3,6,8,10,11,13 found an association between inactivity and productivity loss. Ten studies2–6,8–11,13 also reported on results for tobacco use and productivity with only four studies2,4,6,10 finding an association between tobacco use and productivity loss. Of note, one of these four studies2 found that tobacco users had less productivity loss than non-smokers. Lastly, 11 studies2–6,8–13 reported results for unhealthy weight and productivity with six studies4–6,8,10,12 finding an association between unhealthy weight and productivity loss. Additionally, three studies4,6,10 reported a consistent relationship existed between the number of modifiable health risks and productivity loss regardless of the significance of associations of individual health risks and productivity. It is difficult to put the findings of this study within the context of the aforementioned literature given the variability in the observed associations between individual modifiable health risks and productivity. As previously described, there was considerable variation between studies in productivity research methodology and populations and work domains studied. Also, as the study by Shi and colleagues demonstrates,13 other non-health-related variables likely contribute to productivity loss at the same time. The present study was unlike the vast majority of other studies in that it used objective outcome measures rather than self-reported productivity impairment. Also, this study used a unique instrument (i.e., the AF Web HA) to measure health risks. The AF Web HA adapted tobacco use questions from the National Health and Nutrition Examination Survey III and sleep questions from the Behavioral Risk Factor Surveillance System 2001; otherwise, the lineage of the assessment questions relevant to this study is unknown, and to the best of the authors’ knowledge, the AF Web HA was never validated. Another consideration related to the AF Web HA data is that individuals’ health risk responses are reported to their primary care team, who then have the responsibility to follow up and deliver indicated clinical preventive services. Airmen’s knowledge of this fact may have led to a bias to underreport health risks. A strength of the current study was the use of a large cohort of active duty Air Force personnel to better understand the association between modifiable health risks and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. A limitation of this study was its cross-sectional design, which limits the ability to draw inferences of cause and effect. Accordingly, future research using a retrospective prospective or pre–post prospective study design should evaluate the impact of changes on health risks and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. Another limitation of this study was the small effect sizes of the statistically significant associations between health risks and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. Nevertheless, Grossmeier and colleagues make the point that even very small changes in population-level health risks result in substantial impacts when applied across the entire workforce.7 Their study also suggested that the impact of population-level health improvement is realized over subsequent years, so it is likely this cross-sectional analysis underestimates the effect sizes of interventions deployed as part of the Commander’s Wellness Program. However, these small effect sizes also suggest that there are additional variables not considered in this study that also are associated with the outcomes. Therefore, leadership’s expectations need to be appropriately managed when implementing the Commander’s Wellness Program — targeting health risks is not a panacea for improving valued unit metrics. In conclusion, this study identified that there was an association between some of the health risks targeted in the proposed Commander’s Wellness Program and fitness assessment scores, fitness assessment exemptions, and aggregate-limited duty time. The Commander’s Wellness Program should focus on interventions targeting BMI, composite fitness score, sleep, and tobacco use. Although neither self-reported physical activity nor nutritional habits were directly associated with the outcomes, it is still worthwhile to include related interventions in the Commander’s Wellness Program because of the finding in other studies4,6,10 of a consistent association between the overall number of health risks and productivity outcomes. Acknowledgments The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Air Force, the Department of Defense, or the U.S. Government. References 1 Kirsten W : Making the link between health and productivity at the workplace – a global perspective . Ind Health 2010 ; 48 : 251 – 5 . Google Scholar CrossRef Search ADS PubMed 2 Alavinia S , Molenaar D , Burdorf A : Productivity loss in the workforce: associations with health, work demands, and individual characteristics . Am J Ind Med 2009 ; 52 : 49 – 56 . Google Scholar CrossRef Search ADS PubMed 3 Boles M , Pelletier B , Lynch W : The relationship between health risks and productivity . J Occup Environ Med 2004 ; 46 : 737 – 45 . Google Scholar CrossRef Search ADS PubMed 4 Burton W , Chen C , Conti D , Schultz A , Edington D : The association between health risk change and presenteeism change . J Occup Environ Med 2006 ; 48 : 252 – 63 . Google Scholar CrossRef Search ADS PubMed 5 Burton W , Conti D , Chen C , Schultz A , Edington D : The role of health risk factors and disease on worker productivity . J Occup Environ Med 1999 ; 41 : 863 – 77 . Google Scholar CrossRef Search ADS PubMed 6 Burton W , Chen C , Conti D , Schultz A , Pransky G , Edington DW : The association of health risks with on-the-job productivity . J Occup Environ Med 2005 ; 47 : 769 – 77 . Google Scholar CrossRef Search ADS PubMed 7 Grossmeier J , Mangen D , Terry P , Haglund-Howieson L : Health risk change as a predictor of productivity change . J Occup Environ Med 2015 ; 57 ( 4 ): 347 – 54 . Google Scholar CrossRef Search ADS PubMed 8 Merrill R , Aldana S , Pope J , et al. : Self-rated job performance and absenteeism according to employee engagement, health behaviors, and physical health . J Occup Environ Med 2013 ; 55 : 10 – 8 . Google Scholar CrossRef Search ADS PubMed 9 Pelletier B , Boles M , Lynch W : Change in health risks and work productivity over time . J Occup Environ Med 2004 ; 46 : 746 – 54 . Google Scholar CrossRef Search ADS PubMed 10 Riedel J , Grossmeier J , Haglund-Howieson L , Buraglio C , Anderson D , Terry P : Use of a normal impairment factor in quantifying avoidable productivity loss because of poor health . J Occup Environ Med 2009 ; 51 : 283 – 95 . Google Scholar CrossRef Search ADS PubMed 11 Robroek SJW , van Lenthe F , Burdorf A : The role of lifestyle, health, and work in educational inequalities in sick leave and productivity loss at work . Int Arch Occup Environ Health 2013 ; 86 : 619 – 27 . Google Scholar CrossRef Search ADS PubMed 12 Schultz AB , Edington DW : The association between changes in metabolic syndrome and changes in cost in a workplace population . J Occup Environ Med 2009 ; 51 : 771 – 9 . Google Scholar CrossRef Search ADS PubMed 13 Shi Y , Sears LE , Coberley CR , Pope JE : The association between modifiable well-being risks and productivity: a longitudinal study in a pooled employer sample . J Occup Environ Med 2013 ; 55 : 353 – 64 . Google Scholar CrossRef Search ADS PubMed 14 Collins J , Baase C , Sharda C , et al. : The assessment of chronic health conditions on work performance, absence, and total economic impact for employers . J Occup Environ Med 2005 ; 47 : 547 – 57 . Google Scholar CrossRef Search ADS PubMed 15 Goetzel R , Long S , Ozminkowski R , Hawkins K , Wang S , Lynch W : Health absence, disability, and presenteeism cost estimates of certain physical and mental health conditions affecting U.S. employers . J Occup Environ Med 2004 ; 46 : 398 – 412 . Google Scholar CrossRef Search ADS PubMed 16 Lynch W , Riedel J , Hymel P , Loeppke R , Nelson R , Ashenfelter J : Factors affecting the frequency of value-focused health activities and policies by employers . J Occup Environ Med 2004 ; 46 : 1103 – 14 . Google Scholar CrossRef Search ADS PubMed 17 Riedel J , Lynch W , Baase C , Hymel P , Peterson K : The effect of disease prevention and health promotion on workplace productivity: a literature review . Am J Health Promot 2001 ; 15 : 167 – 91 . Google Scholar CrossRef Search ADS PubMed 18 Friedman J : Stochastic gradient boosting . Comput Stat Data Anal 2002 ; 38 ( 4 ): 367 – 8 . Google Scholar CrossRef Search ADS 19 Zeileis A , Kleiber C , Jackman S : Regression models for count data in R . J Stat Softw 2008 ; 27 ( 8 ): 1 – 25 . Google Scholar CrossRef Search ADS 20 R Core Team : R: a language and environment for statistical computing. R Foundation for Statistical Computing. Available at www.R-project.org; accessed March 15, 2017 . 21 Ridgeway G : gbm: generalized boosted regression models, R package version 2.1.1. 2015 . Available at https://CRAN.R-project.org/package=gbm/; accessed March 15, 2017. 22 Cameron A , Trivedi P : Regression Analysis of Count Data . Cambridge , Cambridge University Press , 1998 . 23 Jackman S : pscl: classes and methods for R developed in the Political Science Computational Laboratory, Stanford University, R package version 1.4.9. 2015 . Department of Political Science, Stanford University. Available at https://pscl.stanford.edu/; accessed March 15, 2017. 24 Cohen J : Statistical Power Analysis for the Behavioral Sciences , 2nd ed. , Hillsdale, NJ , Lawrence Erlbaum , 1988 . 25 Bell M : Dietary Supplement Use Associated with Air Force Fitness and Deployment Health [Thesis] . Dayton , Wright State University , 2015 . 26 Madrid M : Air Force Web Preventive Health Assessment Mental Health Screening Effectiveness [Thesis] . San Antonio , University of Texas School of Public Health , 2010 . Published by Oxford University Press on behalf of the Association of Military Surgeons of the United States 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US.

Journal

Military MedicineOxford University Press

Published: Sep 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off