Population and Data Source
Our study used a serial cross-sectional design using Behavioral Risk Factor Surveillance System (BRFSS) survey data from 2011 through 2015, for adults aged 18 or older with complete covariate data. BRFSS is an annual survey of randomly selected US residents contacted via landline telephone and cellular telephone in all 50 states, the District of Columbia, and 3 US territories, collected in either English or Spanish. Only 1 member of each household is surveyed, and the data are valid, reliable, and generalizable to the US population. The average response rate for the 2011 through 2015 BRFSS was 48%. BRFSS data are publicly available and contain no personal identifiers; for this reason, this study was determined to be exempt from review by the National University Institutional Review Board.
Measures. We used data for respondents who self-reported a diagnosis of diabetes or prediabetes for whom there were full covariant data, based on their answers to 2 questions: 1) "Have you ever been told by a doctor or a health provider you have diabetes?"; and 2) "Ever been told by a doctor or a health provider you have prediabetes or borderline diabetes?" The aggregate 5-year affirmative responses for the questions were 1) n = 215,441 [12.7%; weighted frequency 10.5%] and 2) n = 63,567 (3.7%; weighted frequency 3%). Women who self-reported having diabetes or prediabetes during pregnancy (gestational diabetes) were excluded from this study. Clinically, prediabetes is defined as the condition where glycemic parameters are above normal but below diabetes thresholds. The American Diabetes Association describes prediabetes as fasting plasma glucose (FPG) of 5.6 to 6.9 mmol/L, referred to as an impaired fasting glucose level, and/or postload plasma glucose level of 7.8 to 11.1 mmol/L, referred to as impaired glucose tolerance, or hemoglobin A1c levels of 5.7% to 6.4%. We used data on respondents who self-reported a diagnosis of type 2 diabetes or prediabetes.
Demographic and socioeconomic factors. Self-reported age in years (18–34, 35–49, 50–64, or ≥65), marital status (married, never married, or other), military veteran status (yes or no), education level (≤high school graduate, or some college and above), annual household income (<$15,000, $15,000 to <$25,000, $25,000 to <$35,000, $35,000 to <$50,000, or ≥$50,000), consistent access to health provider (yes or no), routine annual checkup in the past 12 months (yes or no), race/ethnicity (white non-Hispanic, black non-Hispanic, Hispanic, or other), and sex. Data were also analyzed by geographic regions according to the 9 US Census Bureau designations: Northeast (New England division and Middle Atlantic division); Midwest (East North Central division and West North Central division); South (South Atlantic division, East South Central division, and West South Central division); and West (Mountain division and Pacific division) (for detailed list of states included in each division see https://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf).
Health variables. Self-reported body mass index (BMI) was stratified into 4 categories: underweight (BMI <18.5 kg/m2 [weight in kg divided by height in m2]), normal weight (BMI, 18.5–24.9), overweight (BMI 25.0–30.0), or obese (BMI >30.0). Other variables were general health condition (excellent/very good, good/fair, or poor); limited activity because of physical, mental, or emotional health (yes or no); and ever diagnosed with any of the following health conditions (yes or no): arthritis (eg, rheumatoid arthritis, gout, lupus, fibromyalgia), depressive disorder, asthma, chronic obstructive pulmonary disease (COPD) or pulmonary disease, kidney disease, cancer, or cardiovascular disease (CVD) (including chronic heart disease, heart attack, and stroke).
Lifestyle variables. Binary questions (yes or no) included smoking (100 or more cigarettes in lifetime), habitual drinking (men >14 drinks/week, women >7 drinks/week), and habitual exercise (any physical activity or exercise other than daily work-related routine in the past 30 days).
Descriptive and univariate analyses of the study population, prediabetes, and diabetes were conducted for all variables (P < .05 to assess significance). BRFSS weighting was used to adjust for differences in noncoverage and nonresponse in the sample to produce more generalizable estimates. Weighted multivariable logistic regression controlling for demographic, health, and lifestyle variables was used to obtain weighted and adjusted odds ratios (AORs) and 95% confidence intervals (CIs) for each variable with respect to prediabetes and diabetes. A multicollinearity assessment, using a variance inflation factor, was performed, with values 4 and above indicating collinearity. Fisher scoring algorithm was used to calculate maximum likelihood and identify the most influential factors in diabetes and prediabetes. Statistical analysis and data management were performed by using SAS software, version 9.4 (SAS Institute, Inc.).
Prev Chronic Dis. 2018;15(3):e36 © 2018 Centers for Disease Control and Prevention (CDC)