Standardized questionnaires were used to gather information about the study participants’ demographics, medical history, and health behaviors. Sociodemographic variables included age, education (< 12 or ≥ 12 years), marital status, and household income. Participants were classified into two groups according to marital status: married, unmarried (never married, divorced, separated, widowed, or unknown). The study population was also classified into three groups according to household income level: lower, middle, or upper (< 30, 30–50, or ≥ 50 million Korean won/year, respectively). Number of comorbidities was calculated as the total number of the following physician-diagnosed diseases, according to a self-report: stroke, transient ischemic attacks, myocardial infarction, angina pectoris, heart failure, chronic kidney disease, hypertension, dyslipidemia, diabetes, thyroid disease, fatty liver disease, chronic hepatitis, liver cirrhosis, asthma or chronic obstructive pulmonary disease, osteoporosis, arthritis, autoimmune disease, or malignant tumor. Health behaviors included smoking status (current smoker, former smoker, or non-smoker), alcohol intake (current drinker, former drinker, or non-drinker), and physical activity. Physical activity was assessed using the International Physical Activity Questionnaire-Short Form. Regular exercise was defined as moderate- to high-intensity physical activity performed at least three times per week.
Standing height was measured to the nearest 0.1 cm using a stadiometer (DS-102, JENIX, Seoul, South Korea), and body weight was measured to the nearest 0.1 kg on a digital scale (DB-150, CAS, Seongnam, South Korea) according to a predetermined protocol. Body mass index (BMI, kg/m2) was calculated as body weight divided by standing height squared.
We measured participants’ egocentric social networks and collected data on each participant’s social network properties. Through face-to-face interviews by trained interviewers, each participant was asked to provide the names or nicknames of up to seven close individuals, including their spouse and up to five others to whom they talked most frequently, and one more person considered to be their closest and most important relationship. Respondents were also asked to provide information on the quality of their relationship with alters by scoring intimacy on a scale from 1 (not intimate) to 4 (most intimate). The total number of names provided was used as an index of each respondent’s social network size (quantity), and mean intimacy score was used as an index of the social network’s quality. Alters’ ages were used to analyze the age diversity and age difference within a social network. Age diversity was represented by the standard deviation of the ages of members of the social network, whereas age difference referred to the difference between the age of the respondent and the average age of those in their social network.
Depressive symptoms was assessed using the BDI . The BDI includes 21 questions evaluating emotional, cognitive, motivational, physiological, and other symptoms. Each item contains four statements describing the intensity of the symptoms; each item is rated on a scale from 0 to 3, reflecting how participants have felt over the past two weeks. Thus, total BDI scores range from 0 to 63, with higher scores representing greater depression. This instrument has demonstrated acceptable sensitivity and specificity in distinguishing between participants with and without depressive symptoms, and is considered a valid and reliable measure of depressive symptoms [22, 23]. For the purpose of this study, depressive symptoms were considered present for individuals who scored ≥ 20 on the BDI or who took antidepressants.
Descriptive analysis was conducted using t-tests for normally distributed continuous variables and chi-square for categorical variables. Multivariable logistic regression analyses were used to assess independent associations between social network factors and depressive symptoms in four adjusted models:  adjusted for age and menopause status;  additionally adjusted for marital status, education, and income status;  additionally adjusted for number of comorbidities, smoking, alcohol intake, and physical activity; and  additionally adjusted for social network quantity or quality.
Stratified analyses were conducted accordingly: social network quantity was stratified as small (social network size < median) and large (size ≥ median) and social network quality was stratified as low (average intimacy with alters < median) and high (average intimacy ≥ median). We divided by four groups based on social network quantity and quality (large & high, large & low, small & high, small & low). All statistical tests were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Statistical significance was defined as a two-sided P value < 0.05.