In this research, a cross-sectional study was conducted based on The Indonesian Family Life Survey Wave 5 (IFLS-5) database, which was fielded on the full sample in 2014-2015. IFLS5 contains detailed information collected at the individual and household levels, including multiple indicators of socioeconomics and health. It is the only large-scale longitudinal survey available for Indonesia and conducted by multistage random sampling method, which was conducted in 13 Provinces representing 83% of the population in Indonesia(19).
We included respondents aged 40 years and older and had hypertension categorized by measurement systolic> 140 mmhg or diastolic> 90 mmhg on 3 times the measurement of blood pressure. Based on the inclusion criteria, a sample of 6302 people was obtained. The dependent variable in this study is the number of unmet need for health services. The definition of unmet need for health services is if the respondent has hypertension but does not access health services such as primary health care, clinics and hospitals in the last 4 weeks.
We use some variable to investigate unmet need health care based on socio-economic and demographic status at the level individual. Economic status was measured from variable log household expenditure and the number of health post for elderly (Posyandu Lansia) at the community level. Posyandu lansia is a community-organised health promotion centre at village level supervised by staff from the nearest community health centre. Since the mid-1980s, the Indonesia Ministry of Health has launched services to older people through Posyandu Lansia. To deal with the increasing prevalence of hypertension and other chronic conditions, several preventive and health promotion activities are provided by local communities through Posyandu Lansia(20). Older people frequently obtained anti-hypertensive medications (26%) through community health centers performed by its health staff members (midwives or nurses)(21). Other covariates at the individual level are age, sex (female and male as reference), educational attainment (primary school or less as reference; secondary class; and college or higher), marital status (married as reference; single; divorced,and widower), employment status (casual workers as reference; government workers; private workers; self-employed, and not working), health insurance ownership. We also provide the descriptive statistics of age categorized in eight groups (40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, +75),
We conducted data analysis in two steps: bivariate analysis and multivariate analysis. The bivariate analysis assessed the relationship between two variables: 1) the area of residence and 2) each of its determinants (separately). We used Kruskal-Wallis one-way analysis of variance for numerical variables and ordinal chi-square tests for categorical variables. The multivariate analysis identified the association between the healthcare utilization and all of the risk factors together using tree-level hierarchical logistic regression models to take into account of the household and community level information available from the IFLS. The first level comprised individual characteristics, the second level was household characteristics, and community characteristics made up the third level. Considering individual i nested in a household, and community k:
Yijk = γ000 + ∑ γ00kUk + ∑ γ0jkWjk + ∑ βijkXijk + u00j + r0jk + єijk
Yijk = cognitive function as an ordinal variable (normal, CIND and dementia) for the individual in household j in community k.
Uk is a set of community characteristics,
Wjk is a set of household and community characteristics,
Xijkis a set of individual characteristics,
u00j are the random intercept varying over the household
r0jkis the random intercept varying over household and community
єijk is normally distributed with mean zero and variance σє2.
The multivariate analysis used two models. The first model included only the individual-level variables, including socio-demographic variables of age, gender, marital status, education, employment status, and health insurance ownership. We added the household expenditure as the household level determinant, and rural/urban category and the number of Posyandu Lansia as the community level determinants in the second model. We conducted the hierarchical logit regression using xtmelogitcommands in STATA 14.0 software(22).