**2. Materials and Methods**

#### *2.1. Sample*

Data came from the six continuous National Health and Nutrition Examination Survey (NHANES) cycles from 2005–2016 (https://wwwn.cdc.gov/nchs/nhanes/ContinuousNhanes/ Default.aspx?BeginYear=2005). NHANES is a nationally representative, population-based survey for assessing adult and child health and nutritional status in the US. This survey combined health interviews conducted in respondents' homes with health measurements (e.g., DPQ\_I, objective physical measures) performed at mobile exam centers (MECs). The examination components consisted of medical, dental, and physiological measurements, and laboratory tests supervised by trained medical personnel. Furthermore, the adoption of various modern equipment enabled the NHANES to collect reliable, high-quality data. Moreover, compensation and a report of medical findings were given to each participant, which increased the compliance of participants [18]. The total sample size of adults from the 2005–2016 assessments was N = 10,349. Additional details of the study design, sampling, and exclusion criteria are described in Figure 1. Only publicly available data was used in the analysis, and no ethical approval was needed in this study.

#### *2.2. Measures*

Diet quality: The Healthy Eating Index (HEI) is a measure for assessing dietary quality, precisely, the degree to which a set of foods aligns with the Dietary Guidelines for Americans [19]. The HEI-2015 components were the same as in the HEI-2010, except saturated fat and added sugars replaced empty calories, with the result being 13 components [20]. HEI-2015 scores ranged from 0–100, with higher HEI scores reflecting better diet quality. We utilized the total nutrient intakes on the first day (DR1TOT) to calculate the 13 components of HEI-2015. For further weighted Scott–Rao chi-square test and weighted logistic regressions, an HEI-2015 score less than 50, between 50 and 70, and more than 70 was categorized as inadequate, average and optimal, respectively [21].

Depression: Current depressive symptoms were measured by the Patient Health Questionnaire-9 (PHQ-9). The PHQ-9 is a well-validated (Cronbach's α = 0.89) self-report instrument that assesses depression symptoms (i.e., sadness, trouble sleeping, fatigue, problems concentrating) in the past two weeks, and has moderate concordance with clinical psychiatric interviews. The PHQ-9 questionnaire contains nine items, with each item being assessed on a four-point Likert scale, ranging from 0 = not at all to 3 = nearly every day, and summing up a total scale range of 0 to 27. A dichotomous variable indicating no depression (PHQ-9 score <10) or elevated depressive symptoms (PHQ-9 score ≥10) was created using a threshold score of 10 [22].

**Figure 1.** Flowchart of the study population.

### Covariates

Each covariant was categorized into a reference group, and other groups. When analyzing, all other groups were compared to this reference group to estimate the relative odds ratio.

Sex: Sex was categorized as male (reference group) and female.

Age: Age was categorized as 20 to 25 years (reference group), 26 to 49 years, and 50+ years [4].

Race: Race was categorized as non-Hispanic white (reference group), non-Hispanic black, Mexican American, and other races [23].

Education level: Education level was categorized as less than a high school diploma (reference group), high school graduate/GED, some college/AA degree, and college graduate or more [24].

Household income: Household income was categorized as ≤130% (reference group), >130% to 350%, and >350% by the ratio of family income to poverty (FPL) [25].

BMI status: body mass index was calculated from measured height and weight as weight/height<sup>2</sup> (kg/m2), and then categorized into ≤25 (reference group), >25 to 30, and >30 [26].

Smoking status: smoking behavior was measured in the "smoking: cigarette use" questionnaire. In the "smoking: cigarette use" questionnaire, respondents were asked if s/he had smoked at least 100 cigarettes in their life, and smoked cigarettes when being questioned. If the respondent had smoked less than 100 cigarettes in their life, s/he was classified as a never smoker. If the respondent had smoked at least 100 cigarettes in his/her life and still smoked when s/he answered the questionnaire, s/he was classified as a current smoker. The respondent was classified as a former smoker if s/he had smoked at least 100 cigarettes in his/her life, and had quit smoking when s/he answered the questionnaire. Smoking status was categorized into never smoker (reference group), former smoker, and current smoker [27].

Drinking status: drinking behavior was measured in the "alcohol use" questionnaire. In the "alcohol use" questionnaire, each respondent was asked how often s/he had drunk alcoholic drinks in the past 12 months, and the average drinks on those days that s/he drank alcoholic beverages. According to these questions, the average number of alcoholic drinks consumed per week in the past 12 months could be calculated. Then it was categorized into four strata (0, <1, 1–<8, and ≥8 drinks per week) and defined as none (reference group), light, moderate, and heavy alcohol consumption, respectively. A "drink" was defined as a 12-ounce beer, a 5-ounce glass of wine, or one-and-half ounces of liquor [28].

Diabetes: plasma glucose data were obtained from the plasma fasting glucose laboratory data. Respondents whose fasting plasma glucose was ≥6.0 mmol/L were thought to be a diabetic, consistent with American Diabetes Association guidelines [29]. Thus, respondents were categorized into adults with normoglycemia (reference group), and adults with diabetes.
