Analysis of the Impact of the Use of Rehabilitation Services on Employment Outcomes for People With Disabilities: Based on Propensity Score Matching Authors: Mutian Ma, Xue Lan, Hongxia Zhou, Kaixi Zhong, Jing Wen Published: 2026 ABSTRACT ObjectiveTo evaluate the association between rehabilitation services and employment among persons with physical disabilities in China, and to examine other disability types in secondary analyses. Using propensity score matching on data from a northwestern province, we provide empirical evidence to guide policy development.DesignCross-sectional study using propensity score matching, logistic regression analysis, the chi-square test, and the t test.SettingPersons with disabilities in urban and rural areas of the Ningxia Hui Autonomous Region, China.ParticipantsA total of N=1870 persons with disabilities aged 16-60 years holding a disability certificate issued by the People’s Republic of China.InterventionsThis study employed propensity score matching to approximate a randomized controlled trial, designating persons with disabilities receiving rehabilitation services as the experimental group and those not receiving rehabilitation services as the control group.Main Outcome MeasuresThe primary outcomes were employment status and use of rehabilitation services among persons with disabilities, while demographic characteristics were secondary outcomes.ResultsRehabilitation services were positively associated with employment among persons with disabilities. Subgroup analyses by disability type revealed that this association was strongest among individuals with physical disabilities; those receiving rehabilitation services increased the employment rate by 8.8 percentage points (95% CI, 0.8-16.3; P=.025). In contrast, no statistically significant associations were found for other disability types, including visual, hearing, intellectual, and multiple disabilities (all P>.05). At the disability level, the largest increase was observed among individuals with level 3 (moderate) disabilities, whose employment rate rose by 22.1 percentage points (95% CI, 10.5-35.2; P<.001), whereas no statistically significant associations were found for disability levels 1, 2, and 4 (all P>.05). Significant associations were also observed among rural residents, men, unmarried individuals, those in nuclear families, and those reporting poor subjective health (all P<.05).ConclusionsRehabilitation services have a facilitating effect on the employment of persons with disabilities, which is regulated by the complexity of the nature of household registration, marital status, family structure, disability level, disability type and subjective health status. According to data from the second national sample survey on persons with disabilities, the total number of persons with disabilities in China has exceeded 85 million. Their quality of life, educational resources, medical care, and employment status are directly related to the happiness index and the quality of life for millions of families.1 In particular, the employment of people with disabilities has become a focus of interest in contemporary society.2 According to the 2019 Statistical Bulletin on Disability Affairs,3 China had 35.662 million certified persons with disabilities and 9.144 million employed persons with disabilities, with the employment rate for this population standing at only 23.98%. Promoting the employment of people with disabilities is not only an important means to improve their standard of living and quality of life, but also a key measure to narrow the gap between people with disabilities and those without disabilities at the economic, social, and psychological levels. In light of this, the Chinese Government has adopted a series of supportive measures, including rehabilitation services, to enable persons with disabilities to integrate into society and re-enter the workforce. Rehabilitation services are designed to help people with disabilities recover or compensate for functions; reduce their functional impairment; enhance their ability to take care of themselves; participate in society by comprehensively applying medical, educational, social, psychological, and assistive devices after the onset of their disabilities; explore their potential and promote their self-reliance; and integrate into society and the work force.4,5 Examples include providing physical and occupational therapy for individuals with physical disabilities, supplying hearing aids and speech training for those with hearing and speech impairments, and offering psychological counseling and cognitive-behavioral therapy for individuals with intellectual, mental, or psychosocial disabilities. Research indicates that rehabilitation services have a greater impact on the life satisfaction of chronic psychiatric patients than psychiatric treatment alone.6 Previous studies have shown that social skills training helps patients with schizophrenia return to society, prevents the decline of social functioning, and promotes mental recovery. In a previous study,7 the re-employment rate in the social skills training group was significantly higher than that in the general physical therapy group. Other studies have shown that rehabilitation services enhance the sense of achievement and self-confidence of persons with disabilities, laying a foundation for their smooth integration into society and successful employment. Therefore, rehabilitation services are expected to have a positive, significant impact on employment for persons with disabilities. Despite the Chinese Government’s ongoing efforts to promote rehabilitation services for persons with disabilities, little is known about whether these services actually improve employment outcomes. This study aimed to address this knowledge gap by examining the effects of rehabilitation services on employment among persons with disabilities in China. Given the nonrandom assignment of rehabilitation services, we employed propensity score matching (PSM) to reduce selection bias and approximate a randomized controlled trial using cross-sectional survey data. Given that China has the world’s largest population of persons with disabilities, the findings of this study may offer valuable insights to other countries, particularly those in the Global South that are facing similar challenges in disability employment and the provision of rehabilitation services. This study aimed to explore the relationship between rehabilitation services currently implemented in China and employment among people with disabilities, using PSM based on survey data. Methods Data and participants This study employed a cross-sectional design and a multistage sampling approach, drawing on the disability population management database of a northwestern province in China. First, stratification was conducted across 5 cities (A, B, C, D, E). Within each city, individuals with disability certificates were sorted by their certificate numbers, and a systematic sampling method was used to select 1900 persons with disabilities in the employable age range. After excluding invalid questionnaires and those containing missing values, a final sample of 1870 eligible participants was included (detailed results are provided in supplemental fig S1, available online only at http://www.archives-pmr.org/). The sample included individuals with visual, hearing, physical, intellectual, mental, and multiple disabilities. Physical disabilities constituted the largest subgroup, accounting for 46.8% of the sample. To assess potential bias from excluding 30 questionnaires (1.6%), we compared the available demographic and disability-related characteristics between the excluded and included participants. Because of extensive missing data in the excluded questionnaires, the number of participants with nonmissing values varied across age (n=8), disability grade (n=10), and sex (n=12) (detailed results are provided in supplemental table S1, available online only at http://www.archives-pmr.org/). We compared age using the Wilcoxon rank-sum test, and disability grade and sex using the Fisher exact test (appropriate for small sample sizes). No statistically significant differences were found between the 2 groups for any of these variables (all P>.05). For other variables (eg, disability type, income, education), the excluded questionnaires had insufficient data to permit meaningful comparison. Given the very low exclusion rate and the absence of differences among the 3 variables examined, we considered the potential for substantial bias to be minimal. Measures Measures: dependent variables Employment status was measured by the question: “As of last week, were you earning income from work?” Responses were dichotomized as “Yes” (employed) or “No” (unemployed). Measures: independent variables Receipt of rehabilitation services was assessed by the question: “Have you received rehabilitation services?” Responses were coded as “Yes” (received) or “No” (did not receive) (detailed definitions and coding of all covariates are provided in supplemental table S2, available online only at http://www.archives-pmr.org/). Statistical analysis All statistical analyses in this study were conducted using R version 4.4.1.a PSM was performed to control for selection bias between the rehabilitation and nonrehabilitation service groups. We used a combination of 1:1 caliper matching (with a caliper width of 0.05) and 1:1 nearest-neighbor matching. The dependent variable in the PSM model was the use of rehabilitation services. Based on previous studies, the matching variables included sex, age, education level, disability type, disability grade, marital status, annual per capita household income, government assistance, receipt of employment services, and household registration (details are in supplemental table S3, available online only at http://www.archives-pmr.org/). Matching quality was assessed using t tests, balance tests, and the common support hypothesis. After matching, chi-square tests and independent t tests were used to compare employment status between the 2 groups. Logistic regression analysis was then performed to estimate the association between rehabilitation services and employment. Subgroup analyses were conducted to examine whether the association between rehabilitation services and employment varied across participant characteristics. These characteristics included sex (men/women), household registration (urban/rural), marital status (married/unmarried/divorced), and subjective health status (good/poor). Additionally, we estimated the treatment effects separately by disability grade (levels 1-4) and disability type (visual, speech, physical, intellectual, mental, multiple) to assess heterogeneity across these clinically relevant subgroups. The average treatment effect (ATT) on the treated, estimated using PSM, is the percentage point difference in employment rates between the treatment group (which received rehabilitation services) and the matched control group (which did not receive rehabilitation services). The ATT is a specific type of effect size used to quantify the magnitude of the association. Ethics approval This study complies with the relevant national regulations on biomedical research ethics review and was approved by the Ethics Committee of Ningxia Medical University (approval no. 2023-G135). Results Analysis of variance between groups receiving and not receiving rehabilitation services The results of the analysis to determine significant differences in general aspects of people with disabilities between the groups receiving and not receiving rehabilitation services before PSM are presented in table 1. The results showed significant differences between people with disabilities in the group receiving rehabilitation services and those not receiving rehabilitation services in education level, disability level, disability type, family annual per capita income, and whether they received employment services, all significant at the α=0.05 level (P<.05).Table 1Prematch balance test between groups receiving and not receiving rehabilitation services.Table 1 dummy alt textVariableCategoryReceived Rehabilitation Services (n=1338),n (%)Did Not Receive Rehabilitation Services (n=532),n (%)Balance Testχ²/tP ValueSexMen790 (59.0)313 (59.0)2.51.28Women548 (41.0)219 (41.0)Age (y)49 (38.56)47 (38.54)0.29.77Education levelPrimary school698 (52.2)247 (46.4)9.88.04*Junior high school360 (26.9)162 (30.6)Senior high school142 (10.6)49 (9.2)Junior college91 (6.8)45 (8.4)Undergraduate47 (3.5)29 (5.4)Disability typeVision168 (12.6)48 (9.0)10.8.05Hearing/speech126 (9.4)39 (7.3)Physical626 (46.8)271 (55.9)Intellectual166 (12.4)66 (12.4)Mental161 (12.0)79 (14.8)Multiple91 (6.8)29 (5.5)Disability level1182 (13.6)67 (12.6)13.5.003†2428 (32.0)215 (40.4)3294 (21.0)112 (21.0)4434 (32.4)138 (25.9)Marital statusMarried899 (61.2)350 (67.4)4.38.22Single295 (22.0)131 (23.1)Divorced144 (10.8)51 (9.5)Government aidYes702 (52.5)284 (53.4)0.94.75No636 (47.5)248 (46.6)Employment serviceYes202 (15.1)112 (21.1)9.24.002†No1136 (84.9)420 (78.9)Household registrationRural926 (69.7)355 (66.7)0.97.32Urban412 (30.3)177 (33.3)Family structureNucleus family842 (63.0)336 (63.2)0.75.86Stem family322 (24.0)134 (25.2)Joint family76 (5.7)27 (5.1)Other98 (7.3)35 (6.5)DiscriminationYes684 (51.1)266 (50.0)0.39.53No654 (48.9)266 (50.0)EmploymentYes271 (20.3)155 (29.1)15.5<.001‡No1067 (79.7)377 (70.9)NOTE. Age is presented as median (interquartile range).*P<.05.†P<.01.‡P<.001. Table 1Prematch balance test between groups receiving and not receiving rehabilitation services.Table 1 dummy alt textVariableCategoryReceived Rehabilitation Services (n=1338),n (%)Did Not Receive Rehabilitation Services (n=532),n (%)Balance Testχ²/tP ValueSexMen790 (59.0)313 (59.0)2.51.28Women548 (41.0)219 (41.0)Age (y)49 (38.56)47 (38.54)0.29.77Education levelPrimary school698 (52.2)247 (46.4)9.88.04*Junior high school360 (26.9)162 (30.6)Senior high school142 (10.6)49 (9.2)Junior college91 (6.8)45 (8.4)Undergraduate47 (3.5)29 (5.4)Disability typeVision168 (12.6)48 (9.0)10.8.05Hearing/speech126 (9.4)39 (7.3)Physical626 (46.8)271 (55.9)Intellectual166 (12.4)66 (12.4)Mental161 (12.0)79 (14.8)Multiple91 (6.8)29 (5.5)Disability level1182 (13.6)67 (12.6)13.5.003†2428 (32.0)215 (40.4)3294 (21.0)112 (21.0)4434 (32.4)138 (25.9)Marital statusMarried899 (61.2)350 (67.4)4.38.22Single295 (22.0)131 (23.1)Divorced144 (10.8)51 (9.5)Government aidYes702 (52.5)284 (53.4)0.94.75No636 (47.5)248 (46.6)Employment serviceYes202 (15.1)112 (21.1)9.24.002†No1136 (84.9)420 (78.9)Household registrationRural926 (69.7)355 (66.7)0.97.32Urban412 (30.3)177 (33.3)Family structureNucleus family842 (63.0)336 (63.2)0.75.86Stem family322 (24.0)134 (25.2)Joint family76 (5.7)27 (5.1)Other98 (7.3)35 (6.5)DiscriminationYes684 (51.1)266 (50.0)0.39.53No654 (48.9)266 (50.0)EmploymentYes271 (20.3)155 (29.1)15.5<.001‡No1067 (79.7)377 (70.9)NOTE. Age is presented as median (interquartile range).*P<.05.†P<.01.‡P<.001. Prematch balance test between groups receiving and not receiving rehabilitation services. NOTE. Age is presented as median (interquartile range). *P<.05. †P<.01. ‡P<.001. PSM The matching model included the variables that differed in the aforementioned balance test as control variables. The results demonstrated successful matching of 512 individuals per group (Table 2). While significant prematching differences (P<.05) existed in disability level, disability type, educational attainment, annual per capita household income, and receipt of employment services, these were eliminated postmatching (P>.05). Balance testing confirmed that confounding variables had standardized mean differences below 10% (fig 1). The variables used for matching included disability type, disability level, educational attainment, annual per capita household income, and whether rehabilitation employment services were received. Figure 2 shows the distributions of the sum density functions before and after matching, primarily to test the common support assumption. Red indicates the group that received rehabilitation services, and pink indicates the group that did not. After matching, the kernel density functions of the 2 groups overlapped, and there were no differences in the distributions of the variables between the 2 groups. Figure 3 shows the cumulative distribution of propensity scores for the group that received and did not receive rehabilitation services. The red line represents the group that did not receive rehabilitation services, and the green line represents the group that did. After matching, the cumulative distributions of the 2 groups were almost identical. Fig 1, Fig 2, Fig 3 indicate that PSM was effective, and there were no statistically significant differences in the distributions of the variables between the rehabilitation and nonrehabilitation service groups after matching (P>.05).Table 2Balance test after matching the rehabilitation service group with the nonrehabilitation service group.Table 2 dummy alt textStratified by RehabilitationVariableCategoryYes (n=512)No (n=512)P Valuet-SMDEducation levelPrimary school249 (48.6)246 (48.0).8670.07Junior high school156 (30.5)156 (30.5)Senior high school47 (9.2)44 (8.6)Junior college42 (8.2)41 (8.0)Undergraduate18 (3.5)25 (4.9)Disability typeVision45 (8.8)48 (9.4).9990.029Hearing / Speech36 (7.0)37 (7.2)Physical264 (51.6)263 (51.4)Intellectual66 (12.9)63 (12.3)Mental77 (15.0)76 (14.8)Multiple24 (4.7)25 (4.9)Disability level162 (12.1)66 (12.9).9520.0372207 (40.4)205 (40.0)3101 (19.7)105 (20.5)4142 (27.7)136 (26.6)Family per capita income10,391.70 (14,209.81)10,178.51 (9887.24).5490.037Employment serviceYes102 (19.9)101 (19.7)>.990.005EmploymentYes108 (21.1)143 (27.9).0140.159NOTE. Data are presented as n (%).Abbreviations: tSMD, transformed Standardized Mean Difference.Fig 1Standardized mean differences plot. Blue circles represent standardized mean differences of variables before matching; red circles represent standardized mean differences of variables after matching. Variables used for matching include Education level, Disability categories, Disability level, and Family per capita income.Fig 1 dummy alt textFig 2Kernel density function plots before and after matching. Plots were used to compare differences in variable characteristics between the rehabilitation and nonrehabilitation service groups. Red represents rehabilitation services, and pink denotes no rehabilitation services.Fig 2 dummy alt textFig 3Cumulative distribution of logit propensity scores. The cumulative distribution of logit propensity scores compares 2 groups: those who received rehabilitation services and those who did not. The red line represents the group that did not receive rehabilitation services, and the green line represents the group that did. The lower half of the coordinate axis denotes the logit propensity score, and the left side shows the cumulative percentage.Fig 3 dummy alt text Table 2Balance test after matching the rehabilitation service group with the nonrehabilitation service group.Table 2 dummy alt textStratified by RehabilitationVariableCategoryYes (n=512)No (n=512)P Valuet-SMDEducation levelPrimary school249 (48.6)246 (48.0).8670.07Junior high school156 (30.5)156 (30.5)Senior high school47 (9.2)44 (8.6)Junior college42 (8.2)41 (8.0)Undergraduate18 (3.5)25 (4.9)Disability typeVision45 (8.8)48 (9.4).9990.029Hearing / Speech36 (7.0)37 (7.2)Physical264 (51.6)263 (51.4)Intellectual66 (12.9)63 (12.3)Mental77 (15.0)76 (14.8)Multiple24 (4.7)25 (4.9)Disability level162 (12.1)66 (12.9).9520.0372207 (40.4)205 (40.0)3101 (19.7)105 (20.5)4142 (27.7)136 (26.6)Family per capita income10,391.70 (14,209.81)10,178.51 (9887.24).5490.037Employment serviceYes102 (19.9)101 (19.7)>.990.005EmploymentYes108 (21.1)143 (27.9).0140.159NOTE. Data are presented as n (%).Abbreviations: tSMD, transformed Standardized Mean Difference. Balance test after matching the rehabilitation service group with the nonrehabilitation service group. NOTE. Data are presented as n (%). Abbreviations: tSMD, transformed Standardized Mean Difference. Standardized mean differences plot. Blue circles represent standardized mean differences of variables before matching; red circles represent standardized mean differences of variables after matching. Variables used for matching include Education level, Disability categories, Disability level, and Family per capita income. Kernel density function plots before and after matching. Plots were used to compare differences in variable characteristics between the rehabilitation and nonrehabilitation service groups. Red represents rehabilitation services, and pink denotes no rehabilitation services. Cumulative distribution of logit propensity scores. The cumulative distribution of logit propensity scores compares 2 groups: those who received rehabilitation services and those who did not. The red line represents the group that did not receive rehabilitation services, and the green line represents the group that did. The lower half of the coordinate axis denotes the logit propensity score, and the left side shows the cumulative percentage. Differences in characteristics between employed and unemployed individuals with disabilities After PSM, chi-square tests or t tests to examine the characteristics of employed and unemployed individuals with disabilities found significant differences between the 2 groups for sex (χ²=19.8, P<.001), age (t=3.92, P<.001), disability level (χ²=96.9, P<.001), disability type (χ²=17.2, P<.05), education (χ²=82.2, P<.001), family per capita income (t=−6.05, P<.001), household registration status (χ²=10.5, P<.05), and subjective health status (χ²=8.88, P<.05) (table 3).Table 3Differences in individual characteristics between employed and unemployed persons with disabilities.Table 3 dummy alt textVariableCategoryEmployment (n=251)Nonemployment (n=773)X2/tP ValueSexMen175 (69.7)416 (53.8)19.8<.001‡Women76 (30.3)357 (46.2)Age (y)43±1047±123.92<.001‡Disability level112 (4.8)116 (15.0)96.9<.001‡252 (22.7)355 (45.9)367 (24.7)144 (18.6)4120 (47.8)158 (20.5)Disability typeSight23 (9.2)70 (9.1)17.2.004†Hearing/speech21 (8.4)52 (6.7)Physical151 (60.2)376 (48.7)Intellectual26 (10.4)103 (13.3)Mental22 (8.8)131 (16.9)Multiple8 (3.2)41 (5.3)Education levelPrimary school68 (27.1)427 (55.2)82.2<.001‡Junior high school92 (36.7)220 (28.5)Senior high school29 (11.5)62 (8.0)Junior college37 (14.7)46 (6.0)Undergraduate25 (10.0)18 (2.3)Family per capita income14,891±15,1858636±8782−6.05<.001‡Household registrationRural149 (59.6)546 (70.6)10.5.001†Urban102 (40.4)227 (29.4)Marital statusMarried180 (71.7)487 (63.0)8.55.03*Single51 (20.3)201 (26.0)Divorced20 (8.0)85 (11.0)Subjective health statusVery poor17 (6.8)69 (8.9)8.88.03*Poor72 (28.7)160 (20.7)Good84 (33.5)313 (40.5)Very good78 (31.0)231 (29.9)NOTE. Data are presented as mean ± SD or n (%).*P<.05.†P<.01.‡P<.001. Table 3Differences in individual characteristics between employed and unemployed persons with disabilities.Table 3 dummy alt textVariableCategoryEmployment (n=251)Nonemployment (n=773)X2/tP ValueSexMen175 (69.7)416 (53.8)19.8<.001‡Women76 (30.3)357 (46.2)Age (y)43±1047±123.92<.001‡Disability level112 (4.8)116 (15.0)96.9<.001‡252 (22.7)355 (45.9)367 (24.7)144 (18.6)4120 (47.8)158 (20.5)Disability typeSight23 (9.2)70 (9.1)17.2.004†Hearing/speech21 (8.4)52 (6.7)Physical151 (60.2)376 (48.7)Intellectual26 (10.4)103 (13.3)Mental22 (8.8)131 (16.9)Multiple8 (3.2)41 (5.3)Education levelPrimary school68 (27.1)427 (55.2)82.2<.001‡Junior high school92 (36.7)220 (28.5)Senior high school29 (11.5)62 (8.0)Junior college37 (14.7)46 (6.0)Undergraduate25 (10.0)18 (2.3)Family per capita income14,891±15,1858636±8782−6.05<.001‡Household registrationRural149 (59.6)546 (70.6)10.5.001†Urban102 (40.4)227 (29.4)Marital statusMarried180 (71.7)487 (63.0)8.55.03*Single51 (20.3)201 (26.0)Divorced20 (8.0)85 (11.0)Subjective health statusVery poor17 (6.8)69 (8.9)8.88.03*Poor72 (28.7)160 (20.7)Good84 (33.5)313 (40.5)Very good78 (31.0)231 (29.9)NOTE. Data are presented as mean ± SD or n (%).*P<.05.†P<.01.‡P<.001. Differences in individual characteristics between employed and unemployed persons with disabilities. NOTE. Data are presented as mean ± SD or n (%). *P<.05. †P<.01. ‡P<.001. Effects of rehabilitation services on employment among persons with disabilities This study employed PSM to investigate the effects of rehabilitation services on employment outcomes among persons with disabilities. The results indicated that receiving rehabilitation services was positively associated with employment (odds ratio [OR]=1.45; 95% CI, 1.09, 1.93; P=.01). Subgroup analyses revealed that this positive association was particularly pronounced among men (OR=1.51; 95% CI, 1.06, 2.17; P=.02), rural residents (OR=1.61; 95% CI, 1.11, 2.32; P=.01), unmarried individuals (OR=3.26; 95% CI, 1.66, 6.40; P<.001), and those reporting poor subjective health status (OR=2.48; 95% CI, 1.48, 4.18; P<.001) (table 4 and fig 4). However, the wide CI for the unmarried subgroup indicates considerable uncertainty in this estimate, suggesting cautious interpretation (Table 4, Table 5 and fig 4).Table 4Subgroup analysis of rehabilitation services on employment (propensity score-matched samples).Table 4 dummy alt textSubgroupCategoryReceived Rehabilitation ServicesβSEOR95% CIP ValueLowerUpperSexMenYes0.410.081.511.062.17.02WomenYes0.290.251.340.812.20.25Household registrationRuralYes0.480.191.611.112.32.01*UrbanYes0.200.241.220.771.95.40Marital statusMarriedYes0.100.171.110.791.56.56SingleYes1.180.343.261.666.40<.001†DivorcedYes0.640.551.910.655.57.23Family structureNucleus familyYes0.690.182.001.392.88<.001†Stem familyYes−0.060.290.940.531.68.84Joint familyYes−0.180.610.830.252.74.76OtherYes−0.940.670.390.101.44.16Subjective health statusGoodYes0.100.181.100.771.56.59PoorYes0.910.272.481.484.18<.001†NOTE. All estimates are based on propensity score-matched samples.*P<.05.†P<.001.Fig 4Forest plot for the effects of rehabilitation services on employment (subgroup 1).Fig 4 dummy alt textTable 5Logistic regression analysis of the association between rehabilitation services and employment.Table 5 dummy alt textVariableCategoryβSEOR95% CIP ValueLowerUpperIntercept−1.310.110.270.220.34<.001†Received rehabilitationYes (reference: no)0.370.151.451.091.93.01*NOTE. Analysis is based on propensity score-matched samples (N=1024).*P<.05.†P<.001. Table 4Subgroup analysis of rehabilitation services on employment (propensity score-matched samples).Table 4 dummy alt textSubgroupCategoryReceived Rehabilitation ServicesβSEOR95% CIP ValueLowerUpperSexMenYes0.410.081.511.062.17.02WomenYes0.290.251.340.812.20.25Household registrationRuralYes0.480.191.611.112.32.01*UrbanYes0.200.241.220.771.95.40Marital statusMarriedYes0.100.171.110.791.56.56SingleYes1.180.343.261.666.40<.001†DivorcedYes0.640.551.910.655.57.23Family structureNucleus familyYes0.690.182.001.392.88<.001†Stem familyYes−0.060.290.940.531.68.84Joint familyYes−0.180.610.830.252.74.76OtherYes−0.940.670.390.101.44.16Subjective health statusGoodYes0.100.181.100.771.56.59PoorYes0.910.272.481.484.18<.001†NOTE. All estimates are based on propensity score-matched samples.*P<.05.†P<.001. Subgroup analysis of rehabilitation services on employment (propensity score-matched samples). NOTE. All estimates are based on propensity score-matched samples. *P<.05. †P<.001. Forest plot for the effects of rehabilitation services on employment (subgroup 1). Table 5Logistic regression analysis of the association between rehabilitation services and employment.Table 5 dummy alt textVariableCategoryβSEOR95% CIP ValueLowerUpperIntercept−1.310.110.270.220.34<.001†Received rehabilitationYes (reference: no)0.370.151.451.091.93.01*NOTE. Analysis is based on propensity score-matched samples (N=1024).*P<.05.†P<.001. Logistic regression analysis of the association between rehabilitation services and employment. NOTE. Analysis is based on propensity score-matched samples (N=1024). *P<.05. †P<.001. Effects of rehabilitation services on employment by disability type Subgroup analyses by disability type revealed that receiving rehabilitation services was significantly associated with employment among individuals with physical disabilities (ATT=8.8%; 95% CI, 0.8%, 16.3%; P=.025) and those with mental health conditions (ATT=15.8%; 95% CI, 5.6%, 27.1%; P=.008). In contrast, no statistically significant association was observed for individuals with visual impairments, hearing impairments, intellectual disabilities, or multiple disabilities (all P>.05) (table 6 and fig 5). The rehabilitation services available to persons with different types of disabilities can be found in the supplemental table S5.Table 6Effects of rehabilitation services on employment by disability type (propensity score-matched analysis).Table 6 dummy alt textDisability typeTreatment Group (%)Control Group (%)ATT95% CIP Value(100%)LowerUpperVisual22.926.7−3.8−21.815.6.68Hearing/speech18.938.9−20.0−39.00.2.06Physical33.024.28.80.816.3.025*Intellectual26.913.613.3−0.225.1.06Mental health conditions22.36.515.85.627.1.008†Multiple16.016.7−0.7−23.520.0.95NOTE. All estimates are based on propensity score-matched samples.*P<.05.†P<.01.Fig 5Forest plot for the effects of rehabilitation services on employment by disability type.Fig 5 dummy alt text Table 6Effects of rehabilitation services on employment by disability type (propensity score-matched analysis).Table 6 dummy alt textDisability typeTreatment Group (%)Control Group (%)ATT95% CIP Value(100%)LowerUpperVisual22.926.7−3.8−21.815.6.68Hearing/speech18.938.9−20.0−39.00.2.06Physical33.024.28.80.816.3.025*Intellectual26.913.613.3−0.225.1.06Mental health conditions22.36.515.85.627.1.008†Multiple16.016.7−0.7−23.520.0.95NOTE. All estimates are based on propensity score-matched samples.*P<.05.†P<.01. Effects of rehabilitation services on employment by disability type (propensity score-matched analysis). NOTE. All estimates are based on propensity score-matched samples. *P<.05. †P<.01. Forest plot for the effects of rehabilitation services on employment by disability type. Effects of rehabilitation services on employment by disability level Subgroup analyses by disability level revealed that receiving rehabilitation services was significantly associated with employment among individuals with level 3 (moderate) disabilities (ATT=22.1%; 95% CI, 10.5%, 35.2%; P<.001). In contrast, no statistically significant association was observed among individuals with levels 1 (very severe), 2 (severe), or 4 (mild) disabilities (all P>.05) (table 7 and fig 6).Table 7Effects of rehabilitation services on employment by disability level (propensity score-matched analysis).Table 7 dummy alt textDisability levelTreatment Group (%)Control Group (%)ATT(100%)95% CIP ValueLowerUpper17.611.3−3.7−13.17.2.47217.010.66.4−0.113.1.06340.318.222.110.535.2<.001*444.142.21.9−9.913.1.75*P<.05Fig 6Forest plot for the effects of rehabilitation services on employment by disability level.Fig 6 dummy alt text Table 7Effects of rehabilitation services on employment by disability level (propensity score-matched analysis).Table 7 dummy alt textDisability levelTreatment Group (%)Control Group (%)ATT(100%)95% CIP ValueLowerUpper17.611.3−3.7−13.17.2.47217.010.66.4−0.113.1.06340.318.222.110.535.2<.001*444.142.21.9−9.913.1.75*P<.05 Effects of rehabilitation services on employment by disability level (propensity score-matched analysis). *P<.05 Forest plot for the effects of rehabilitation services on employment by disability level. Discussion The Chinese Government has consistently sought to encourage and assist persons with disabilities in rehabilitation and employment through various means; however, there has been limited research on the impact of disability rehabilitation services on employment outcomes.8 This study, which examined the effects of rehabilitation services on employment among people with disabilities aged 16-60 years via a survey and controlled for a series of influencing factors (eg, disability level, disability type, educational attainment, and household economic status) using PSM, found that receiving rehabilitation services had a positive effect on employment.9,10 This aligns with international research findings that vocational rehabilitation services promote employment through functional restoration and the accumulation of social capital.11,12 Research from Virginia Commonwealth University in the United States showed that vocational rehabilitation services can significantly enhance the occupational adaptability of people with disabilities and reduce employment barriers.12 Additionally, China’s “Disability Blue Book” emphasizes that rehabilitation services improve physical function, provide psychological support, and offer social integration training, helping people with disabilities overcome the “ability-opportunity” dual barriers and thereby increasing their employment prospects.13 Subgroup analyses by participant characteristics suggest that the significant benefits for men and rural populations may reflect structural biases in the existing service system. Rural areas have long suffered from insufficient rehabilitation services, but with recent improvements in rehabilitation policies, marginal benefits have increased significantly, and intervention effects have become more pronounced.13 Additionally, most individuals with disabilities in rural areas are employed in agriculture and animal husbandry, industries with relatively low educational and skill requirements. Research shows that individuals with disabilities are often excluded from digital life,13,14 making it difficult for them to find employment through the internet. Compared with individuals with disabilities who are married in nuclear families, those who are unmarried may rely more on public services to meet their employment needs. Additionally, Because of the smaller size of nuclear families, rehabilitation services have a more direct effect on alleviating the economic pressure on families.14 This aligns with the phenomenon of “heightened sensitivity to services among vulnerable groups” observed by Brazilian researchers, where vocational counseling and psychological support provided by rehabilitation services have become key enablers of employment.13,14 However, Jung et al15 indicated that social discrimination and cultural biases remain implicit barriers to employment for people with disabilities. Subgroup analyses by disability type suggest that rehabilitation services were associated with an 8.8 percentage point increase in employment among individuals with physical disabilities, aligning with previous research on vocational rehabilitation in this population. For instance, Dutta et al16 found that diagnostic and treatment services and rehabilitation technology services were significantly associated with employment outcomes for individuals with physical impairments in the United States. However, a study from South Korea on the effects of rehabilitation services on employment among individuals with physical disabilities found that although rehabilitation services had no direct effect on employment in this population, they had a positive effect on employment among individuals who reported a poor subjective health status.17 Furthermore, this study found that receiving rehabilitation services was positively associated with employment among individuals with mental health conditions,18 whereas no statistically significant association was observed for other disability types (all P>.05). Subgroup analyses by disability level suggest that individuals with level 3 disabilities may retain basic occupational abilities while also having functional deficits that can be addressed through rehabilitation services. This may also be related to improvements in subjective health status among persons with disabilities through rehabilitation services. As shown in subgroup analyses by participant characteristics, when persons with disabilities had a poorer subjective health status, rehabilitation services had a positive effect on employment. Subjective health is a subjective indicator that influences whether persons with disabilities are employed.17 According to surveys, a common reason for unemployment among persons with disabilities is “I lack the ability to work.”19 However, individuals with first- or second-level disabilities may have physical functional impairments that exceed the intervention threshold of conventional rehabilitation services, making it difficult for them to meet the basic capability standards required for employment even after receiving services.20 Individuals with fourth-level disabilities have relatively mild functional impairments and may achieve employment through natural adaptation or simple adjustments without relying on systematic rehabilitation services. Study limitations This study has several limitations. First, as a cross-sectional study, it cannot establish causal relationships between rehabilitation services and employment outcomes, nor can it track the long-term effects of these services. Second, we did not collect detailed information on the specific types, content, duration, or providers of rehabilitation services, all of which may influence their effectiveness. Future studies should incorporate more granular measures of rehabilitation services to better understand which components are most beneficial for employment. Third, although PSM was used to reduce selection bias, the matching process reduced the sample size, and subsequent subgroup analyses further decreased the number of participants within each stratum. This reduction in sample size led to larger SEs and, consequently, wider CIs for some subgroup estimates (eg, unmarried individuals: OR=3.26; 95% CI, 1.66, 6.40), indicating less precise effect estimates. Future research with larger initial samples is needed to obtain more stable and precise estimates for subgroup analyses. Finally, while we conducted sensitivity analyses to assess potential bias from excluded participants, the small number of excluded questionnaires with available data limited our ability to compare excluded and included participants across all key variables. Future studies should aim for more complete data collection and consider using multiple imputation to handle missing data. Conclusions Rehabilitation services positively influence employment among people with disabilities, though this effect is moderated by multiple factors, including household registration, marital status, family structure, disability level, and subjective health. This complex interplay highlights the dynamic nature of the bio-psychosocial model while revealing limitations in the adaptability of current policy systems. Supplier a.R, version 4.4.1; R Foundation for Statistical Computing. a.R, version 4.4.1; R Foundation for Statistical Computing. R, version 4.4.1; R Foundation for Statistical Computing. Disclosure The investigators have no financial or nonfinancial disclosures to make in relation to this project.