Data source
This cross-sectional study used the Sleman Health and Demographic Surveillance Systems (Sleman HDSS) dataset release version 10-1-0 [33]. The Sleman HDSS longitudinal research aims to monitor social, economic, and health conditions in the Sleman Regency, Daerah Istimewa Yogyakarta (DIY), Indonesia [34]. Sleman HDSS conducted the first survey wave in 2015, followed by annual face-to-face interviews until 2019. Because of the COVID-19 pandemic in September-October 2020, Sleman HDSS managed data collection for the sixth wave using the telephone interview method. The Sleman HDSS research design is described elsewhere [34], and data may be obtained at the Sleman HDSS website [33].
Study Sample
A total of 1,513 Sleman HDSS individual panels participated in the sixth wave of the Sleman HDSS. The Sleman HDSS individual panel was a person aged >24 selected from each family using the Kish Grid Sampling method [35]. Sleman HDSS used the Kish grid sampling method to choose an eligible household member with the same probability by ordering a list of household members by gender and decreasing age. We limit the subject only to the adult population in the productive ages (24-59 years old). Those who lacked supporting data, such as demographic data, physical activity, and economic status, were excluded from this analysis. As a result, only 1,328 respondents were included in the data analysis (Figure 1).
Measurements
Mental health status
Respondents were interviewed using a Self-Reporting Questionnaire (SRQ) designed by the WHO, comprising 20 questions. The SRQ intends to determine the respondent's mental health problems during the past month [36]. We classified respondents as having a mental health problem if they answered "yes" to six or more items (out of 20 questions) [37]. We also divided mental health problems into five domains, including depression, anxiety, somatic, cognitive disturbance, and decreased energy [35].
Respondents categorize feeling depressed if their hands are shaky, unhappy, cry more often, feel incapable of doing their part in this life, lose interest in many things, feel worthless, and have thoughts of ending their life. Respondents who reported not sleeping well, being terrified easily, and feeling anxious, tense, or worried were categorized as having anxiety symptoms. Respondents were classified as having somatic symptoms if they often had headaches, lost their appetite, indigested, and felt uneasiness in their stomach. Then, respondents who found difficulty thinking, making a decision, and neglecting some daily activities/tasks were grouped as having cognitive symptoms. Respondents who felt challenged to think clearly had difficulty enjoying day-to-day activities, making decisions, neglecting activities/tasks, feeling tired all the time, and getting tired easily were classified as having decreased energy symptoms. Respondents can have more than one symptom [35]
Independent variables
Individual-level factors
The individual factors include demographic variables and health status. Demographic data included gender, age, educational background, marital status, and current employment status. Educational background was subdivided into four categories: not attending school (never attended school and no formal schooling), low (elementary and junior high school), medium (senior high school), and high (diploma/undergraduate/postgraduate). The current employment status was separated into three groups: government sectors (military/police, civil servants, state-owned enterprises employees, and retired), private sectors (private employee, entrepreneur, service business, farmer, labor, and others), and unemployed (unemployed, homemaker, and students).
Respondents reported current job types or job changes during the pandemic (full-time work, part-time work, being laid off, and not working). Respondents were assessed about their perceptions of the economic impact of the pandemic, including worries about losing jobs and being unable to fulfill their primary needs (groceries, electrical bills, gas, etc.) or financial obligations (rental housing costs and vehicle credits). Insurance ownership was categorized into four groups: government assistance health insurance (premium, nonpremium assistance beneficiaries, and Jamkesda - regional health insurance), national health self-insured, private insurance, and does not have any insurance [38].
Health status variables comprised smoking status, physical activities, and NCDs. Smoking habits were classified as "yes" (smoking every day, sometimes, and occasionally) or "no" (never smoking). Physical activity was measured by the Global Physical Activity Questionnaire (GPAQ) version two and then categorized into high physical activity (achieving a minimum of 3000 metabolic equivalents/MET-minutes/week) and not having high physical activity [39]. Respondents also asked whether they had chronic diseases such as stroke, hypertension, diabetes mellitus, coronary heart disease, cancer, and chronic pulmonary obstructive diseases. Access to health services during COVID-19 was divided into three groups: feeling healthy, feeling unhealthy without treatment, and feeling unhealthy and receiving treatment.
Family-level factors
At the family level, we explored the wealth index, family size, and economic impact (changes in expenses, income changes, and receiving support). The family wealth index was generated from several questions on family assets and ownership using principal component analysis (PCA) [40]. The wealth index was categorized into five quintiles (Q1-Q5), with the first quintile as the poorest and the fifth quintile as the richest [35]. Family size was separated into two groups: less than or equal to four family members and over four family members.
Respondents were also asked about changes in expenses and whether they received support. The expenditure change aimed to evaluate any shift in the general expenses and several expenditure aspects, including groceries, ready-made food and beverages, health (out-of-pocket medical care bills, over-the-counter medicine, vitamins, and sanitation), electricity bills, gas, communication (phone credits, internet/data), and public transport (including online transportation). The expenditure changes were divided into three categories: not affected, increased, and decreased. Respondents also reported whether they used their savings to meet financial obligations. Furthermore, respondents were asked if they had any positive economic impact during the COVID-19 pandemic, such as increasing their salary, obtaining a new job, or not being financially affected. They also inquired if they received support during the COVID-19 pandemic.
Community-level factors
We added ethnicity, living arrangement, and community support as community-level factors to investigate other determinants of mental health. The respondents’ living arrangements were categorized into rural and urban. Indonesia has more than 300 ethnic groups. However, the Javanese ethnicity makes up over 40% of the Indonesian population [41]. Sleman HDSS is on Java Island. Accordingly, ethnicity was divided into Javanese and non-Javanese. Community support was the support received by respondents during the COVID-19 pandemic from the government, private organizations, families, colleagues, or other parties. We categorized the respondents receiving community support as “yes” or “no”.
Statistical Analysis
The descriptive analysis was provided as a percentage of all variables. The first bivariable analysis was conducted to determine the association between mental problems and all independent variables using chi-square tests (significantly p<0.05). Then, we conducted bivariable analysis stratified by gender with a significance level of p<0.25. Finally, we conducted a multilevel analysis using regression logistics, and then we presented the results using odds ratios (ORs) with 95% confidence intervals (CIs). Three models were generated from the findings of the multilevel study. Model one contained all variable levels (individual, family, and community), model two included family-level factors alone, and model three only included community-level variables. Stata 13 (Stata Corp., College Station, TX) was used for all data analyses [42].