Page 48 - IJPS-11-5

International Journal of
Population Studies Need for social services among disabled older Chinese

included agriculture, forestry, animal husbandry, fishery, the sample reported that needs were not met. Thus, unmet
industrial worker, etc. Enabling factors included both need was a dummy variable with fully met coded as 0 and
economic and caregiving resources. Economic resources unmet coded as 1. Following previous studies (e.g., Zhu,
included economic independence (having a retirement 2015), severe ADL disability was measured by whether the
wage/pension and/or own earnings vs. no), whether or respondent was unable to perform at least four of the six
not the respondent received adequate medication (yes ADL tasks, with yes coded as 1 and no coded as 0.
vs. no), and whether medical expenses were paid by self
or family members (yes vs. no). Caregiving resource 2.2.3. Control variables
variables included whether the primary caregiver was a We included health behavior, region, and survey year as
family member (yes vs. no), the number of children alive, control variables. Health behavior refers to the frequency
whether the respondent coresidence with a child (yes vs. of engagement in leisure activities. Research has found that
no), and self-reported availability of social services in the social engagement activities are associated with the risk of
neighborhood. Self-reported availability of social services experiencing disabilities (Mendes de Leon et al., 2003),
in the neighborhood included eight items that are identical which may indirectly influence the need for social services.
to those used to ask about social service needs. Similar Furthermore, participating in leisure activities may improve
to the need for social services, we classified these eight social interactions with others (Silverstein & Parker, 2002),
available services into two summary indices, availability which may affect a respondent’s attitude and behavior
of basic care services and availability of social connection toward social services (Fisher et al., 2015). The measure of
services, with values ranging from 0 to 4. The reliability leisure activities included six activities: (a) playing cards/
coefficients are 0.68 for the availability of basic care services mahjong, (b) watching television, listening to the radio, or
and 0.80 for the availability of social connection services. accessing the Internet, (c) reading books or newspapers,
We restricted our sample to ADL-disabled respondents (d) raising pets or domestic poultry, (e) gardening or
who need assistance with daily care. A respondent was doing housework, and (f) engaging in outdoor activities
considered disabled if he/she needed help performing such as exercise or jogging. Each item was scored from 0
any of the following six ADLs: Bathing, dressing, indoor (no participation) to 4 (daily participation). These scores
transferring, toileting, eating, and continence. This were summed across all six items (total score: 0 – 24) to
definition is a common practice in the field (e.g., Feng et al., calculate the overall frequency of involvement in leisure
2016; Zhu, 2015). After limiting the study sample to those activities based on previous studies (Zhu, 2015). To better
who reported ADL disability, we included three variables capture the skewed distribution of this variable, the scores
to represent need as a predictor in the Andersen model: were grouped into three categories: 0 – 3 (low frequency),
Self-rated health, whether a respondent’s care needs were 4 – 9 (medium frequency), and 10+ (high frequency).
unmet, and severe ADL disability. Self-rated health was We also controlled for the region, which classified 23
measured by the question “how do you rate your overall provinces into five regions: North (Beijing, Tianjin, Hebei,
health?” with six response categories: very good, good, fair, and Shanxi), Northeast (Liaoning, Jilin, and Heilongjiang),
poor, very poor, and unable to answer. Following previous East (Shanghai, Jiangsu, Zhejiang, Anhui, Fujian, Jiangxi,
research (e.g., Feng et al., 2016), we combined both very poor and Shandong), Central and South (Henan, Hubei, Hunan,
and poor into one category, “poor,” and both good and very Guangdong, Guangxi, and Hainan), and West (Chongqing,
good into one category, “good,” because the proportions of Sichuan, and Shaanxi). The wave variable has five values:
respondents in the “very good” and “very poor” categories 2005, 2008, 2011, 2014, and 2018. All analyses controlled
are relatively small. We kept the response category “unable for this variable.
to answer” because this category accounted for about 20% 2.3. Analytical strategy
of the sample (unweighted). Thus, we created a measure of
self-rated health with four categories: Good, fair, poor, and We first performed descriptive analyses for the study
unable to answer. An alternative approach using multiple variables. We reported appropriate statistics such as
imputations to impute missing values associated with percentages and means for the total, rural, and urban
“unable to answer” produced very similar results, so we samples. We also performed Chi-square tests for rural–
kept “unable to answer” as a category to keep the original urban differences in study variables (t-tests for age and a
data intact as much as possible. Unmet need was measured number of available social services). Multilevel ordered
by the question, “Does the assistance provided by caregivers logistic regression models were used to examine whether
meet your needs?” with three response categories: Fully the predisposing, enabling, and need variables were
met, partially met, and not met. Partially met and not met associated with the dependent variables of need for basic
were combined into one category because only about 3% of care and social connection services. We first ran multilevel

Volume 11 Issue 5 (2025) 42 https://doi.org/10.36922/ijps.448