Data source
The National Health and Nutrition Examination Survey (NHANES) is a cross-sectional, nationally representative survey of the US Census civilian noninstitutionalized population [13]. The NHANES is a unique survey that uses interviews and physical examinations to assess the health and nutritional status of adults and children in the US. About 5000 people have been surveyed every year since 1999 and data are publicly released in 2-year cycles.
NHANES 2013–2014 and 2015–2016 were used in this study. Overall response rates among interviewed participants were 71% in NHANES 2013–2014 and 61.3% in NHANES 2015–2016; participants were asked about demographics, behavioral and lifestyle characteristics (e.g., smoking, alcohol, physical activities), SDH (e.g., food insecurity, household income), and disease history (e.g., whether they have been told by health professionals they have arthritis, which type of arthritis they have). The physical examination component takes place in a mobile examination center (MEC) and consists of medical, dental, and physiological measurements, as well as laboratory measurements.
Study sample
To estimate the prevalence of depression and food insecurity among adults with RA, participants ≥ 18 years of age with self-reported RA (i.e., had been “told by doctor or other health professional” that they had “RA”) were selected from NHANES 2013–2014 and 2015–2016. To assess the association between depression and SDH as well as patients’ demographics, and behavioral and lifestyle characteristics, participants were required to have complete responses to the Patient Health Questionnaire-9 (PHQ-9) mental health-depression screener.
Study measures
A proposed model showing the interactions between RA, depression and SDH is shown in Fig. 1.
The PHQ-9, a 9-item depression screening instrument, was administered to measure the severity of depression symptoms patients experienced over the past 2 weeks [16, 17]. The psychometric properties of the PHQ-9, including internal consistency, and convergent and discriminant validity, have been examined in several studies of people with RA [18,19,20] and the PHQ-9 has been validated as a reliable diagnostic tool for depression in this population. The PHQ-9 incorporates the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV depression diagnostic criteria [21]. Response categories for the 9-item instrument are given a point value ranging from 0 to 3 based on the following responses: not at all, several days, more than half the days, and nearly every day. The sum of PHQ-9 scores, ranging from 0 to 27, was categorized as no/minimal depression (score of 0–4) or depression (score of ≥ 5). Depression was further stratified into mild depression (score of 5–9) and moderate-to-severe depression (score of 10–27), as used in previous studies, to describe the association between depression severity and other variables [22, 23].
Food insecurity was measured using the 18-item US Household Food Security Survey Module, in which the adult interviewee answered for the entire household [24]. To establish this survey, food security data taken annually from the Current Population Survey (CPS) was used to create a food security scale and a related categorical food-security-status measure to describe the food security of US households over the preceding 12 months. The stability and robustness of these measures were tested and validated in the overall population across years and major population groups. A raw score was created by summing the affirmative responses to the 18 questions, and a categorical variable was created to characterize the overall food security status for the entire household as (1) full food security (no affirmative response to any item); (2) marginal food security (1–2 affirmative responses); (3) low food security (3–7 affirmative responses); and (4) very low food security (8–18 affirmative responses) [12, 25]. As an objective of this study was to examine the prevalence of different levels of food insecurity among adults with RA, food insecurity was defined as any category other than full food security (i.e., marginal, low, or very low food security).
Other SDH assessed in this study included health insurance type, education, marital status, annual household income, number of members in household, and housing characteristics. Key behavioral and lifestyle characteristics were determined using information provided by participants on their general health condition, alcohol use, smoking status, physical disability status (i.e., whether or not to have serious difficulty hearing, seeing, concentrating, walking, dressing, or running errands), body mass index, history of overnight hospitalization, work/recreational activity level, number of physical limitations, sleep disorder status, and attempts to lose weight.
Statistical analyses
To make inference on population parameters and obtain representative estimates of the US noninstitutionalized population, all sample data were weighted to account for different sampling probabilities, non-response rates, stratification, and clustering of observations introduced by the NHANES’ complex survey design [26]. As depression was assessed during the interview at the MEC from 2013 to 2016, the 4-year MEC exam weight was used in this study.
The weighted prevalence of depression and food insecurity in US adults with RA were estimated using SAS survey procedures. Unadjusted prevalence was reported for the 2013–2014 and 2015–2016 waves and odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to assess temporal changes in the prevalence of depression and food insecurity over 4 years (i.e., 2 NHANES waves).
SDH, demographics, and behavioral and lifestyle characteristics were described and compared between RA patients with depression and those with no/minimal depression. Univariable logistic regression models were used to evaluate the association between depression and individual factors. Unadjusted ORs and 95% CIs were calculated. Penalized regression with least absolute shrinkage and selection operator (LASSO) method was performed to select the variables for the final multivariable model [27]. Penalized regression using the LASSO method has become a preferred method for variable selection because it can reduce data dimensionality by shrinking a subset of the coefficients to 0 and reduce the complexity of the model. The association between the selected factors and depression was finally evaluated by fitting a weighted multivariable logistic regression model, and results were presented as adjusted ORs and 95% CIs.
All analyses were performed with SAS (Version 7.1, SAS Institute Inc., Cary, North Carolina) and R (Version 3.6.3, R Foundation for Statistical Computing, Vienna, Austria) software.