Family Consumer Behaviors, Adolescent Prediabetes and Diabetes

NHANES is a nationwide cross-sectional survey conducted by the Centers for Disease Control and Prevention (CDC), a detailed description of which is available elsewhere. The NHANES administers annual face-to-face health surveys in a nationally representative sample of civilian, non-institutionalized households in the US. In addition to core questions documenting household demographic and health characteristics, survey instruments collect information on current health and disease status, including several questionnaires on health quality indicators and lifestyle practices. Besides survey instruments, NHANES collects information on select clinical examinations and laboratory tests.

The study will include NHANES subjects from 2007-2010 who were 12-19 years old and from households where an eligible person had administered the consumer behavior questionnaire (CBQ). The CBQ was developed in partnership with the Economic Research Service of the US Department of Agriculture and has been a part of NHANES from 2007-2010. A total of 2,520 children and adolescents met the eligibility criteria for the study.

Food Availability and Expenditure The primary variables of interest for the study included household healthy food availability and supermarket spending on food that was assessed in the CBQ. Availability will be examined for five food items in the household including fruits, dark green vegetables, fat-free/low fat milk, salty snacks, and soft drinks. The responses for each item were ordinal, ranging from always, most of the time, sometimes, rarely, to never. Dark green vegetables included fresh dried, canned and frozen vegetables, and interviewers were instructed to exclude lettuce. Salty snacks included chips and crackers, but not nuts. The types of milk considered low-fat were skim or 0% and 1%. Soft drinks included soda, fruit-flavored drink and fruit punch; while 100% juice or sports drinks were excluded.

A composite variable that incorporated answers to all the five food items will be created for food availability. First, a decreasing score ranging from 5 to 1 was assigned from "always" to "never" responses for the healthy food items including fruits, dark green vegetables, and fat-free/low fat milk; whereas an increasing score ranging from 1 to 5 was assigned for "always" to "never" response for unhealthy food items including salty snacks and soft drinks. A cumulative score will be calculated for each participant that ranged from 5 to 25, where a higher value indicated higher availability of healthy food(s) in the household. A cumulative score will only be calculated for participants with a non-missing response on all five questions.

Food spending will be examined by the amount of money (in US Dollars) the household spent in the past 30 days on supermarket store purchases, non-food items, food at other stores, eating out and carryout/delivered foods. Non-food spending was asked as a part of supermarket store spending and it included money used towards cleaning and paper supplies, pet food, alcoholic beverages and cigarettes. This will be first subtracted from the total supermarket spending to give supermarket spending on food items alone. It will then be divided by the total food spending of the household (sum of the above food spending questions) to also calculate the proportion of spending on supermarket store food items.

Prediabetes and Diabetes The primary outcome of interest will be adolescent prediabetes and diabetes. Pre-diabetics will be included with diabetics to determine the impact of family consumers on a wider range of glycemic status. Additionally, they are the most amenable to dietary and lifestyle changes in preventing further progression to overt diabetes. These will be computed by the individual's hemoglobin A1c (A1c) status. A1c values are relatively stable after collection15 and reflect 3 months of glycemic control. It has been shown to have less day-to-day as well as inter-subject and intra-subject variability16,17. The HbA1c concentration is expressed directly as %HbA1c. A participant was considered pre-diabetic or diabetic if he/she had an A1c ≥ 5.7 % (39 mmol/mol)18.

Other Variables Covariates were selected as those designated most important in the association between adolescent obesity and consumer behaviors in existing literature. These included socio-demographic characteristics including age, gender, race, parental educational attainment and income levels, and household size. Physical activity, which was defined as participation in vigorous activity such as sports, fitness and recreational activities that increased the heart rate, was also examined.

Statistical Analysis The prevalence (%) with standard error (SE) of prediabetes and diabetes will be estimated across different levels of study covariates. First, the distribution of responses to the five food availability questions as well as the cumulative healthy food availability score will be examined for the study population. The cumulative score will be dichotomized on the median value of 14.96 and divided into high and low availability of healthy food. Next, a chi square test will be used to look at the relationship between healthy food availability and adolescent prediabetes and diabetes in the population. When examining the monthly household supermarket spending in USD, a log transformation will be used due to a large degree of positive skew. The proportion of supermarket spending will be dichotomized at the median value of 0.67 to divide the study population into high versus low supermarket food spending. The prevalence (%) with SE of high healthy food availability and geometric mean USD with 95% confidence interval (CI) of monthly household supermarket food spending will be summarized across various socio-demographic and physical activity levels of the study population. A chi-square test will be used to demonstrate significant differences in the prevalence of high healthy food availability and high supermarket spending proportion for each demographic characteristic.

An analysis of variance will be carried out to examine the relationship between healthy food availability (from the cumulative score) and supermarket spending. Subsequent correlation analyses were used to examine the inter-relationships between several variables in the consumer behavior questionnaire (data not shown).

The unadjusted and adjusted associations between adolescent prediabetes and diabetes will be first examined with high healthy food availability (versus low availability) and then with log supermarket food expenditure in USD, using separate logistic regression models. Three multivariate models were established for each association: where the first was adjusted for age and gender, the second also adjusted for income and race, and the third further adjusted for education, physical activity and household size. The fully adjusted models will finally be stratified by gender to check for effect modification. The above associations were estimated using prevalence odds ratio (OR) and their corresponding 95% CI.

All statistical analyses will be conducted with SAS version 9.3 (SAS Institute Inc., Cary, North Carolina), and the statistical significance level was 0.05, unless otherwise specified. Since NHANES is a complex probability sample, stratification and clustering of the design as well as sampling weights were incorporated into the analysis for proper estimation of nationally representative effect size and SE.