Impact of rehabilitation services on employment outcomes for individuals with physical disabilities: a propensity score matching analysis | BMC Public Health

Data and participants

This cross-sectional study analyzed data from the 2020 National Survey of Disabled Persons conducted by the Ministry of Health and Welfare and the Korea Institute for Health and Social Affairs. This is a nationally representative survey of community-dwelling people with disabilities in South Korea conducted every three years. The National Survey of Disabled Persons data are available on the health and welfare data portal website (https://data.kihasa.re.kr/kihasa/kor/contents/ContentsList.html). In 2020, 7,025 registered individuals with disabilities participated in the survey without considering household sampling due to the 2019 pandemic [3].

The inclusion criterion for the current study was individuals with physical disabilities aged \(\ge\)20 years. All individuals with missing values in their survey were excluded because propensity score matching should not include missing values. A total of 1,757 eligible participants were extracted from a dataset of 7,025 samples (Appendix A).

Measures

Dependent variable

The question for employment was “Did you work for more than an hour for income last week?” with two possible responses: “yes” and “no.” If the answer was “yes,” it was defined as being employed.

Independent variable

Based on previous studies, the independent variables were rehabilitation services [8,9,10], age [8, 9, 11, 12], gender [8,9,10,11, 13], spouse [8,9,10, 12, 14], education [8, 10,11,12], monthly household income [11, 14], degree of disability [8, 9, 11, 13, 14], and subjective health status [11, 14]. The question for rehabilitation services was, “Is there any rehabilitation you are currently undergoing?” and was recorded as a binary value, with “yes” responses classified as “used rehabilitation services (= 1)” and “no” responses as “did not use rehabilitation services (= 0).” The question for age, “What is your age?” was recorded as a continuous variable. The question for gender was, “What is your gender?” This variable was also recorded as a binary value with two responses, “man (= 1)” and “woman (= 0).” The answer to the marital status question, “Are you married?” was recoded as a binary variable. A response of “yes” was classified as “has spouse (= 1),” and “widowed, divorced, separated, or never married” as “no spouse (= 0).” The question for education was, “What is your highest level of education?” and was recorded as follows: “pre-school, no school, elementary school (= 0)”; “elementary or under elementary graduation (= 1)”; “middle school” and “middle school graduation (= 2)”; “high school,” recorded as “high school graduation (= 3)”; “college, university, over graduation school,” recorded as “over college (= 4).” The question for monthly household income, “What was the average monthly household income during 2019 (2019.01.01 to 2019.12.31)?” was recorded as a continuous variable from 0 Korean won. The question for the degree of disability was, “What is the registered degree of disability?” This was a binary variable, and the answers were defined as “severe (= 1)” and “mild (= 0).” Subjective health status was assessed using a single question—“How do you feel about your health in general?”—on a 5-point Likert scale from “very bad (= 0)” to “very good (= 5).” A higher score indicated a better subjective health status.

Statistical analysis

Data analysis was conducted using SPSS Statistics 23.0 (IBM Corp., Armonk, NY, USA) and SAS 9.4. First, we conducted propensity score matching to reduce selection bias between the group with rehabilitation services and the group without rehabilitation services. In this study, the nearest-neighbor matching method and caliper matching method were combined and applied. This study applied 1:1 ratio matching, and the caliper range was 0.01. To verify the results of propensity score matching, a paired t-test was conducted and the standardized mean difference was determined [15]. In the propensity score matching analysis, the dependent variable was set as the use of rehabilitation services. Based on previous studies, the independent variables were gender [16, 17], age [16, 18], monthly household income [16, 18], degree of disability [16], and disability origin [17]. Appendix B provides further details about the description of the variables used in propensity score matching. Second, the chi-square test and independent t-test were conducted to check whether there was a difference in employment according to general characteristics in the propensity score-matched data. Finally, logistic regression analysis was performed to estimate the effect of rehabilitation services on the employment of individuals with physical disabilities. Furthermore, factors influencing the employment of individuals with disabilities, such as gender (male, female), degree of disability (mild, severe), and subjective health status (good, bad), were divided into subgroups. Subgroup analysis was then conducted to determine whether the characteristics of each subgroup affected employment. Sandwich estimators were applied to the logistic regression analysis.