Data source
The NHANES is a series of independent and nationally representative health surveys of the US civilian, noninstitutionalized population conducted by the National Center for Health Statistics (NCHS). Since 1999, NHANES collected data continuously and released datasets in 2-year cycles. In each survey, participants are selected with the use of a complex, stratified, multistage probability-cluster sampling design for in-home interviews and visits to a mobile examination center(12). We included data from NHANES 2005-2014 as the NHANES used a different rating scale for depression before 2005. The NCHS approved the study protocols and written informed consent was obtained from each participant. This study followed the STROBE checklist for observational studies.
Definition of Covariates
During the in-home interview, demographic and health-related information was collected by standardized questionnaires(12). Current smoking was self-reported based on questions about whether participants were currently smoking or not. Current excessive alcohol use was defined as drinking, on average, ≥ 5 drinks/day in the 12 months preceding the survey interview. For the antidepressant medication, participants were asked if they had taken any prescription drug in the past 30 days and antidepressant medication was determined using the second level of drug ingredient categorical codes. A combination of bupropion and naltrexone, which is primarily used for obesity, overweight, and weight-related medical problems, was excluded(13). History of CVD and cancer was ascertained by self-report and CVD included stroke, congestive heart failure, angina, and myocardial infarction.
Weight and height were measured during the examination and the body mass index (BMI) was defined as weight in kilograms divided by height in meters squared. Blood pressure (BP) was measured and mean BP was calculated as the mean of three readings according to the standardized protocol. Blood samples were collected at the mobile examination center, stored at −20 °C, and sent to the central laboratories for the measurement of hemoglobin A1c, total cholesterol, and serum creatinine. Urine albumin and creatinine were measured in the same laboratory during the surveys.
Obesity was defined as BMI of 30 or higher; hypertension as systolic BP of 130 mm Hg or higher, diastolic BP of 80 mm Hg or higher, or the use of antihypertensive medications; diabetes as hemoglobin A1c of 6.5% or higher, or the use of antidiabetic medications; dyslipidemia as total cholesterol of 240 mg/dL or higher, or the use of lipid-lowering medications.
We calculated the estimated glomerular filtration rate (eGFR) based on serum creatinine by using the Chronic Kidney Disease Epidemiology Collaboration equation(14). Calibrations for the measurements of serum creatinine in different surveys were conducted according to the recommendation of NHANES. Moderate CKD (stage 3-4) was defined as an eGFR of 15-59 mL/min/1.73 m2 or a one-time urine albumin-to-creatinine ratio ≥30 mg/g. Patients with an eGFR <15 ml/min/1.73 m2, corresponding to CKD stage 5, were excluded owing to the lack of information on participants’ dialysis information.
Depression was assessed using the 9-item Patient Health Questionnaire (PHQ-9) according to the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). The PHQ-9 is a screening instrument that asks about the frequency of symptoms of depression over the past 2 weeks. We defined the major depression as the summed score on the PHQ-9 of 10 points or greater. The use of a PHQ-9 score of 10 points has been proven to provide a favorable balance between sensitivity and specificity(15).
Ascertain of mortality
Mortality data were ascertained by linkage to the National Death Index through December 31, 2015. The International Classification of Diseases-10 was used to determine disease-specific mortality. According to the Codebook for the 2015 Public-Use Linked Mortality File, CVD mortality included heart disease mortality and stroke mortality(16).
Statistical analysis
To account for oversampled population, appropriate 10-year sampling weights were constructed according to the NHANES recommendation. All statistical analysis was conducted in R version 4.1.0 with the “Survey” package after accounting for the complex sampling design. All statistical tests were 2-sided, and P <0.05 was considered statistically significant. We restricted our participants to non-pregnant adults aged 18 years or older. Continuous variables were expressed as means (standard error) and categorical variables as percentages (standard error) in the four different groups. Group 1 was defined as participants without depression and without moderate CKD, Group 2 as participants with depression and without moderate CKD. Group 3 as participants without depression and with moderate CKD, and Group 4 as participants with both depression and moderate CKD. Mortality was expressed as events per 1000 person-years of follow-up. The number of person-years of follow-up time was calculated from the time of entry into the study until the date of death or the termination day of follow-up. Life expectancy and survival times for four groups were estimated by the Kaplan-Meier product-limit method.
Cox proportional hazards regression models were used to calculate multivariate-adjusted hazard ratios (HRs) of death across different groups. Three different models were developed. Model 1 was adjusted for age, gender, race, and education. Model 2 was further adjusted for obesity, diabetes, hypertension, dyslipidemia, history of cardiovascular disease, and history of cancers based on Model 1. Model 3 was further adjusted for current smoke, current excessive alcohol use, antidepressant medication, eGFR, and albumin-to-creatinine ratio based on model 2. Three pairs of post hoc comparisons with Bonferroni correction were conducted after the main analysis to test the difference of death in Group 4 with death in the other three groups(17). A P-value less than 0.017 was considered statistically significant in the post hoc analysis.
We conducted several sensitivity analyses to test the robustness of the results. First, to account for the difference in the definition of CKD, we redefined the CKD as an eGFR of 15 to 59 mL/min/1.73 m2 alone without considering the urine albumin-to-creatinine ratio. Second, to account for the potential effects of age on the outcomes, we excluded young adults in the analysis and restricted our participants to those aged 45 years or older. Third, to account for the interaction between depressive symptoms and functional impairment, we compared mortality rates among four groups of participants based on the presence or absence of functional impairment and moderate CKD. In these sensitivity analyses, we only compared the all-cause mortality in different groups owing to the small number of CVD death events.