Page 73 - IJPS-11-2

International Journal of
Population Studies Health-care access for the elderly living alone

2.2. Variable selection and measurement (primary school or lower versus secondary school or

The dependent variable (UHN) was assessed as a binary higher), chronic disease (yes versus no), and place of
variable, where the response options were “yes” and “no.” residence (urban versus rural).
The 2019 HWS dataset posed two questions to assess the 2.3. Statistical analysis
UHN of respondents: (i) “during the past 1 month, was
there ever a time when you felt that you needed health A descriptive analysis was performed to summarize the
care?: yes or no” and (ii) “If yes, did you receive it?: received study samples and variables. In the analysis, the bivariate
or not received”. Among older people who answered “yes” association between the dependent variable (UHN) and
to Question 1 (i.e., the study sample), those who answered each independent and control variable was investigated
“not received” to Question 2 were categorized into the using the Chi-squared test. In addition, since UHN was a
“yes” group, and those who answered “received” to the binary variable, a BLR analysis was used to examine how
question were categorized into the “no” group. Living the independent and control variables were associated with
arrangement (independent variable) was also measured UHN.
as a binary variable, with options of “living alone” or “not We developed two separate BLR models: one using the
living alone.” matched sample after the PSM method and another using
In addition, following Aday & Andersen’s (1974) access the entire sample before the PSM method. The BLR model
to the medical care model as well as variable availability in using the matched sample aimed to accomplish the first
the data, this study’s analysis included six sociodemographic research objective (i.e., to investigate the effect of living
variables as control variables. According to Aday and alone on UHN after controlling for potential confounding
Andersen’s model, the factors affecting health-care access effects). The BLR model using the entire sample aimed
are divided into predisposing, enabling, and need-for-care to accomplish the second research objective of this study
factors. Predisposing factors are the demographic and (i.e., to investigate the sociodemographic determinants of
sociocultural characteristics of individuals before the onset of UHN). The performance of the BLR models was examined
illness. Enabling factors are individual- and community-level using Hosmer–Lemeshow goodness-of-fit test (Hosmer &
resources that facilitate access to health care. Need-for-care Lemeshow, 2000).
factors are perceived (subjective) and evaluated (objective) In addition, multicollinearity among the independent
health problems or illness levels that generate the need for and control variables was examined using the variance
health care. In the present study, three sociodemographic inflation factor and cross-comparison between the crude
variables (age, sex, and education), two socioeconomic and adjusted odds ratios (aORs). The variance inflation
variables (income and place of residence), and the presence factor score ranged from 1.002 to 2.339, and there were
of chronic diseases were selected as predisposing, enabling, no large directional switches between the crude and
and need-for-care factors, respectively, and used as control aORs, suggesting that multicollinearity was not an issue
variables in the analysis. (Menard, 2002).
Age and income were measured using tercile scales In this study, statistical significance was set at a
that divided the sample into three equal proportions. p-value of 0.05. For the BLR models, the odds ratio and
A higher tercile indicates higher income and older age. 95% confidence interval (CI) were used to determine
To measure the income tercile, we used the equivalence the directional relationship and statistical significance,
scale income, which is income adjusted for a one-person respectively. All analyses were conducted using the SAS
household. Equivalent income was computed using version 9.2 software.
Equation I (Organization for Economic Co-operation and
Development, 2008): 3. Results
3.1. Descriptive analysis
Total household income ÷ household members (I)
Table 1 reveals the descriptive analysis results. According
The income thresholds for Terciles 1 and 2 (T1 and T2) to the results derived from the matched sample (i.e., the
were 3,536 and 8,660 baht (equivalent to approximately 106 sample obtained by the PSM method), all p-values of the
and 260 USD, respectively, based on a baht-USD exchange Chi-squared tests were higher than 0.05, indicating that all
rate of 1 bath = 0.03 USD). The age thresholds for T1 and sociodemographic characteristics were statistically equal
T2 were 70 and 77 years old, respectively. between older people living alone and those who did not.
The remaining control variables were measured as The results derived from the matched sample revealed
binary variables: sex (male versus female), education that the prevalence of UHN was significantly higher in

Volume 11 Issue 2 (2025) 67 https://doi.org/10.36922/ijps.1218