Ann Rehabil Med Search

CLOSE


Ann Rehabil Med > Volume 43(3); 2019 > Article
Lee, Hong, Lee, and Park: Identifying Risk of Depressive Symptoms in Adults With Physical Disabilities Receiving Rehabilitation Services: Propensity Score Approaches

Abstract

Objective

To compare the risk of depressive symptoms in adults with physical disabilities between rehabilitation-receivers and non-receivers.

Methods

A total of 3,568 adults with physical disabilities were retrieved from the 2014 Korean National Survey on People with Disabilities database. Four covariate adjustment methods (a multivariable regression model, inverse probability of treatment weighting [IPTW] adjusted for normalized weight, IPTW with stabilized weight, and greedy algorithm with 1:1 propensity score matching) were used to estimate the odds of having depressive symptoms. The dependent variable was depressive symptoms and the independent variable was the use of rehabilitation services. Baseline covariates were 19 demographic variables and 10 chronic condition variables.

Results

The four covariate adjustment methods revealed that adults with physical disabilities receiving rehabilitation services had a higher risk of depressive symptoms than those who did not receive these services (adjusted odds ratio, 1.191–1.294).

Conclusion

Our findings suggest that adults with physical disabilities receiving rehabilitation services have higher risk of developing depressive symptoms. Therefore, rehabilitation professionals need to pay attention to depressive symptoms and establish therapeutic strategies that can reduce such risk in rehabilitation settings.

INTRODUCTION

Physical disability defined as a condition in which an individual experiences deviation or loss in their body function or structure [1] can result in difficulty, restriction, or dependence in performing basic and instrumental activities of daily living (ADLs and IADLs) [2,3]. In South Korea, with increasing number of people with physical disabilities [4], the proportion of population with physical disabilities receiving rehabilitation services is also increasing. Therefore, developing a deeper understanding of people with physical disabilities is indispensable for rehabilitation professionals to provide more successful interventions for this population.
The International Classification of Functioning, Disability, and Health (ICF) based on a bio-psycho-social model recommends that rehabilitation professionals need to consider not only physical functions, but also psychological symptoms such as depressive symptoms or anxiety to better understand people with physical disabilities and deliver better interventions [1]. To be specific, depressive symptoms can exacerbate individuals’ physical disability symptoms [5] which may result in low health-related quality of life [6] and higher healthcare expenditures [7]. Considering adverse effects of depressive symptoms, rehabilitation professionals should monitor and track the risk of depressive symptoms and understand impacts of rehabilitation services on depressive symptoms to successfully manage health of people with physical disabilities.
Previous studies have acknowledged the importance of identifying the risk of depressive symptoms and reported a higher prevalence rate of depressive symptoms or depression in people with disabilities compared to the general population [5,8,9]. These previous studies have revealed that disabilities can increase the risk of depressive symptoms itself. To investigate whether receiving rehabilitation services can affect the risk of depressive symptoms among adults with physical disabilities, it is imperative to compare the risk of depressive symptoms between rehabilitation receivers and non-receivers within the same patient group. However, to the best of our knowledge, no research has compared the risk of depressive symptoms between these two groups.
To accurately compare the risk of depressive symptoms, risk factors for depressive symptoms should be controlled to minimize effects of confounding factors. For instance, there are well-known risk factors for depressive symptoms or depression, including low income [10], social support [11], osteoarthritis [12], low back pain [13], age [14], and sex [15]. Among patients receiving rehabilitation, depressive symptoms are also associated with higher comorbidity and impairment, lower social support, lower self-efficacy, and specific work-related problems [16]. Since previous studies have demonstrated various factors associated with depressive symptoms, it is important to effectively control for those factors related to depressive symptoms when investigating the risk of depressive symptoms.
The ICF model is an effective model for understanding and controlling for related risk factors of depressive symptoms. According to the ICF model, depressive symptoms are defined as impairments in an individual’s emotional function which belongs to the category of body function. Since impairments in body function can be affected by various factors such as health condition, personal and environmental factors, activities, and participation, related factors of depressive symptoms can be categorized according to the ICF model. Thus, information about relationships between depressive symptoms and its risk factors identified from the ICF model will be useful when estimating the risk of depressive symptoms by controlling for these risk factors.
Methodologically, propensity score (PS) matching methods are robust statistical methods effectively accounting for covariates. Although traditional regression adjustment methods are effective for controlling for confounders, these methods have limitations in addressing selection bias which is a typical limitation of observational studies [17-20]. In contrast, PS matching methods can mimic a randomized clinical trial by matching two comparison groups using conditional probability of receiving exposure given a set of covariates. For instance, various PS matching methods have been demonstrated to be robust after controlling for selection bias and confounders, such as inverse probability of treatment weighting adjustment (IPTW), IPTW with stabilized weight, and 1:1 matching [18-22].
Therefore, the objective of this study was to compare the risk of depressive symptoms between adults with physical disabilities who received rehabilitation services and those who did not receive these services by utilizing four different covariate adjustment methods.

MATERIALS AND METHODS

Study sample

We retrieved records of 104,703 adults with physical disabilities from the 2014 Korea National Survey on People with Disabilities (NSPD) database [23]. The inclusion criteria for our study were: (1) the presence of physical impairments in the upper and/or lower extremities and/or spinal cord injury and (2) age older than 20 years. Physical impairments were defined as the state of being physically limited due to amputation, a medical disorder, or impairments of joints, extremities, or the spinal column. The study data set is de-identified and publicly available.

Definition of depressive symptoms and rehabilitation services

The presence of depressive symptoms was defined as a depressive mood or hopelessness that continued for 2 or more weeks and affected the participant’s daily activities during the past 12 months (September 2013 to August 2014). The survey participants were also asked if they had received rehabilitation services (treatment group vs. control group), including physical, occupational, speech, music, play, art, behavioral, and/or other therapies (e.g., exercise, horticultural therapy, therapeutic riding, and pottery psychotherapy). In other words, if participants received any of these rehabilitation services, we considered them to be receiving rehabilitation services.

Covariates based on the ICF model

Included covariates were categorized based on the ICF model.

Health condition

Health condition includes issues associated with chronic or acute illnesses [1]. In this study, history of falls, self-rated health (1 ‘very poor’ to 5 ‘very good’), and chronic conditions present for 3 months or more were included as factors for health condition.

Body function

Body function was classified as physiological functions of body systems according to the ICF model [1]. We selected physical function score as a factor for body function. It was estimated using an ADL instrument validated by an item response theory model [24].

Activity

The ICF model defines activity as the execution of a task or action by an individual [1]. Participation in physical activity (at least 3 days/week) was defined as a factor of activity.

Participation

Participation refers to involvement in life situations as noted in the ICF [1]. Disability grades (1 ‘severe’ to 6 ‘mild’) assigned by the Korean government include assessment of participation levels. Thus, disability grade was categorized as a factor for participation.

Personal factors

Personal factors are defined as characteristics of an individual, including race, age, and health conditions [1]. Thus, demographic variables such as age, years since primary impairment onset, sex, marital status (single or married/living with a partner), employment (currently working), low income (less than $1,388/month), education, smoking, and drinking were included as personal factors.

Environmental factors

The ICF model defines environmental factors as physical, social, and attitudinal environments in which people live and interact, including personal care providers and personal assistants, health professionals, and education and training services [1]. Thus, in this study, outpatient visits (past 2 weeks including local hospitals, health centers, or dental clinics), preventative health services (education programs related to exercise, nutrition, smoking, and drinking in the past 12 months), mental health services (in the past 12 months), currently receiving depressive symptoms treatment, and social support (living with family member(s) or partner) were defined as environmental factors.

Statistical analysis

Descriptive analyses were conducted to describe demographic characteristics and covariates. To compare the two groups (rehabilitation receivers and non-receivers), chi-square test was used for categorical variables and Wilcoxon rank sum non-parametric test was used for continuous variables. Odds of depressive symptoms were estimated by using the four covariate adjustment methods (a traditional multivariable logistic regression and three PS matching methods) to examine the robustness of study results. The three PS matching methods were: (1) IPTW using propensity score with normalized weight, (2) IPTW with stabilized weight, and (3) greedy algorithms with 1:1 matching of the PS [18,21,25].
First, we used a complete pooling approach in PS analyses [26]. PS was estimated based on the likelihood of having depressive symptoms from a logistic regression model that accounted for participant’s characteristics (Table 1). Normalized weights for the IPTW matching method were applied to avoid extreme values of weight by dividing each individual propensity score by the mean of all propensity scores. This allows a comparison to samples representative of specific populations [18]. Therefore, normalized weight in the IPTW matching method does not result in a loss of observations.
While the robustness of IPTW with normalized weight has been demonstrated in observational studies, the null hypothesis is frequently rejected due to inflated sample size [22]. In other words, this method often produces large variance estimates that result in high type I error rates. Thus, we also used IPTW with stabilized weight method to account for high propensity scores in treatment and control groups and validated point estimation from normalized weight for the IPTW method.
The most common PS matching method is the 1:1 matching method [18,19]. In this study, we used a greedy 1:1 matching algorithm without replacement using a caliper of width equal to 0.10 of the standard deviation (SD) of the logit of complete pooling PS [25]. Consequently, this method allows matched sets of survey participants between treatment and control groups to share similar PS values estimated using covariates listed in Table 1.
After performing the three PS matching methods, we re-calculated descriptive statistics between the two comparison groups to examine if covariates listed in Table 1 were balanced. We also performed two PS diagnostic tests for the 1:1 matching method, including absolute standardized differences in means of all covariates [19,27] and percentage bias reduction of covariates [26]. We considered an absolute standardized difference of less than 0.2 in means of all covariates as indicative of a good match between the two groups [27]. If there were unbalanced covariates from the three PS matching methods, we additionally controlled for these covariates using logistic regression models. Lastly, we conducted sensitivity analysis to examine the magnitude of hidden bias in the risk of depression symptoms caused by unobserved covariates [18,28,29]. Sensitivity analysis result was reported with a value of gamma at two-sided α=0.05 level. All analyses were performed using SAS version 9.4 software (SAS Institute Inc., Cary, NC, USA). Statistical significance was determined at an alpha level of 0.05 in two-sided tests.

RESULTS

We analyzed records of a total of 3,568 adults with physical disabilities. The average age was 62.7 years (SD=13.6). Among retained study sample, 775 participants were identified as having depressive symptoms, including 312 (27.3%) in the treatment group and 463 (18.9%) in the control group. The majority (n=1,126, 99.8%) of patients in the treatment group received physical therapy. Baseline covariate differences between treatment and control groups before matching are presented in Table 1. Before matching, 23 out of 29 covariates were significantly different between the treatment group (n=1,126, 31.5%) and the control group (n=2,442, 68.5%).
PS ranged from 0.049 to 0.841 in the treatment group and from 0.031 to 0.834 in the control group (Appendix 1). Baseline covariate differences between the two groups after performing the three matching methods are shown in Table 2. After applying the IPTW with normalized weight and stabilized weight methods, all baseline covariates were balanced between the two comparison groups (all p>0.05).
After applying the 1:1 PS matching method, seven covariates were significantly different between treatment and control groups, including age, marital status, low income, current smoker, social support, osteoarthritis, and back pain (all p<0.05). However, these variables satisfied the criterion of minimum reduction in absolute standardized differences (Fig. 1).
Total percentage bias reduction of covariates by the 1:1 PS matching method was 75.8%. We controlled for these seven significantly different variables when estimating odds of depressive symptoms. Table 3 shows risks of depressive symptoms in adults with disabilities who have received rehabilitation services after applying covariate adjustment and the PS method plus further covariate adjustments.
Unadjusted risk of depressive symptoms was 63.9% (odds ratio=1.639; 95% confidence interval [CI], 1.389–1.933). However, this risk decreased to 19.1%–29.4% according to the four covariate adjustment methods, including a traditional multivariable logistic regression (adjusted odds ratio [AOR]=1.233; 95% CI, 1.013–1.501), IPTW normalized with logistic regression adjustment (AOR=1.191, 95% CI, 1.014–1.400), IPTW stabilized weight with logistic regression adjustment (AOR=1.191; 95% CI, 1.002–1.417), and 1:1 PS matching method with logistic regression adjustment (AOR=1.294; 95% CI, 1.065–1.574). Sensitivity analysis for the matched sample resulted in a gamma value of 1.11 at two-sided α=0.05 level, indicating that unobserved covariates could change our study results.

DISCUSSION

In this study, we compared the risk of depressive symptoms in adults with physical disabilities who received rehabilitation services to those who did not receive such services by utilizing four different covariate adjustment methods (a traditional regression model and three different propensity score matching methods). Our study findings revealed that adults with physical disabilities receiving rehabilitation services had higher risk of developing depressive symptoms (AOR, 1.191–1.294). To the best of our knowledge, this is the first study to compare the risk of depressive symptoms between rehabilitation receivers and non-receivers among the same adult population with physical disabilities. For instance, a recent study in Germany has investigated the prevalence of depression in rehabilitation settings (i.e., neurological, orthopedic, cancer rehabilitation, and others) and reported a higher prevalence of depression in those who have received rehabilitation than healthy individuals [16]. However, this previous study only compared people with disabilities to healthy individuals.
In a rehabilitation setting, patients encounter unavoidable physical and emotional challenges such as pain, setbacks, and frustration [30]. In particular, if people experience permanent physical disabilities, they could be exposed to these types of personal challenges while receiving rehabilitation. For example, patients with physical disabilities are likely to experience challenges such as frustration and depression when their recovery does not progress as rapidly as they expect [30]. These challenges might increase the risk and likelihood of depressive symptoms in rehabilitation receivers. Our findings suggest that rehabilitation receivers require special care for depressive symptoms.
The ICF model proposes that rehabilitation professionals should follow a multidisciplinary approach to determining the best options for patients [1]. However, in South Korea, rehabilitation services are primarily designed to provide services targeting physical functions or functional performance, not psychological symptoms such as depressive symptoms. This current rehabilitation practice is partially due to the current reimbursement policy and the lack of attention to or care for depressive symptoms in rehabilitation settings. Nevertheless, rehabilitation professionals can indirectly provide therapeutic interventions such as recreational activities to ameliorate clients’ depressive symptoms. Therefore, the current reimbursement policy needs to be revised for practitioners to administer systematic professional therapeutic strategies for people with psychological symptoms.
It is important to note that the identified risk of depressive symptoms in adults with physical disabilities who received rehabilitation services might stem from particular depressive symptoms covariates. After conducting the 1:1 PS matching method, the seven covariates (age, sex, low income, current smoker, social support, osteoarthritis, and back pain) showed significant differences between rehabilitation receivers and non-receivers within adults having physical disabilities (all p<0.05). All of these seven uncontrolled covariates are associated with depressive symptoms in adults with physical disabilities. Numerous studies examining risk factors for depressive symptoms support these findings [10-15]. While some of these risk factors are non-modifiable personal factors (age, sex, low income), other relevant variables are potentially modifiable factors, such as environmental factors (social support) and health condition (osteoarthritis and back pain). Therefore, rehabilitation professionals might need to address modifiable risk factors to reduce the risk of depressive symptoms in rehabilitation settings.
Of note, gamma level of 1.11 was relatively small. In other words, there are high probability that hidden bias can affect our results. Thus, a prospective study that controls for various depressive symptom covariates that are currently unavailable in this study is needed in the future to validate our current study findings.
In conclusion, Korean adults with physical disabilities who receive rehabilitation services show risk of developing depressive symptoms. Because depressive symptoms can cause a myriad of challenges for both the society and individuals, depressive symptoms in rehabilitation receivers should be carefully monitored to provide appropriate interventions.

CONFLICTS OF INTEREST

No potential conflict of interest relevant to this article was reported.

ACKNOWLEDGMENTS

This research was supported in part by grant (No. K12 HD055929) from the National Institutes of Health (NIH). Its contents are solely the responsibility of the authors. They do not necessarily represent official views of the NIH.

AUTHOR CONTRIBUTION

Conceptualization: all authors. Methodology: all authors. Formal analysis: Hong I. Funding acquisition: Hong I, Park HY. Project administration: Park HY. Visualization: Hong I, Lee MJ. Writing – original draft: Lee Y, Hong I. Writing – review and editing: all authors. Approval of final manuscript: all authors.

Fig. 1.
Standardized difference of means of propensity score covariates (y-axis) before matching (all data, left on the x-axis) and after matching (matched data, right on the x-axis).
arm-2019-43-3-250f1.jpg
Table 1.
Demographic characteristics of adults with physical disabilities who received rehabilitation services and those who did not receive rehabilitation services
Variable ICF category Rehabilitation (n=1,126) No rehabilitation (n=2,442) p-value
Agea) (yr) PF 67.5±11.2 60.4±14.0 <0.001*
Years since primary impairment onseta) PF 15.9±14.0 21.4±16.8 <0.001*
Maleb) PF 409 (36.3) 1,529 (62.6) <0.001*
Marriedb) PF 648 (57.5) 1,619 (66.3) <0.001*
Current employmentb) PF 385 (34.1) 1,216 (49.8) <0.001*
Low incomeb) PF 692 (61.4) 1,055 (43.2) <0.001*
Educationb) PF <0.001*
 Less than elementary school 685 (60.8) 987 (40.0)
 Middle school 185 (16.4) 387 (15.8)
 High school 172 (15.2) 674 (27.6)
 College beyond 84 (7.4) 394 (16.1)
Current smokerb) PF 144 (12.8) 614 (25.1) <0.001*
Current drinkerb) PF 771 (68.4) 1,252 (51.2) <0.001*
Physical function scoresa) BF 49.6±7.3 50.1±7.2 0.007*
Physical activityb) (3 days/wk) A 539 (47.8) 1,096 (44.8) 0.096
Disability gradesb) P <0.001*
 1 22 (1.9) 64 (2.6)
 2 48 (4.2) 145 (5.9)
 3 120 (10.6) 321 (13.1)
 4 240 (21.3) 481 (19.7)
 5 375 (33.3) 660 (27.0)
 6 278 (24.6) 715 (29.2)
 NA 43 (3.8) 56 (2.2)
Social supportb) EF 816 (72.4) 2,025 (82.9) <0.001*
Outpatient visit (past 2 wk)b) EF 858 (76.2) 1,131 (46.3) <0.001*
Preventative health services (past 12 mo)b) EF 47 (4.1) 66 (2.8) 0.034*
Mental health services (past 12 mo)b) EF 30 (2.6) 41 (1.7) 0.074
Current depression treatmentb) EF 61 (5.4) 79 (3.2) 0.002*
History of fallsb) HC 48 (4.2) 86 (3.5) 0.279
Self-rated healthb) HC <0.001*
 Very poor 216 (19.1) 275 (11.2)
 Poor 607 (53.9) 953 (39.0)
 Fair 246 (21.8) 824 (33.7)
 Good 53 (4.7) 362 (14.8)
 Very good 4 (0.3) 28 (1.1)
Chronic conditionsb) HC
 Hypertension 581 (61.6) 931 (38.1) <0.001*
 Dyslipidemia 136 (12.0) 210 (8.6) 0.001*
 Stroke 47 (4.1) 89 (3.6) 0.443
 Heart problem 113 (10.0) 147 (6.0) <0.001*
 Osteoarthritis 578 (51.3) 557 (22.8) <0.001*
 Rheumatoid arthritis 62 (5.5) 77 (3.1) <0.001*
 Back pain 357 (31.7) 334 (13.6) <0.001*
 Lung problem 51 (4.5) 100 (4.1) 0.549
 Diabetes 238 (21.1) 414 (16.9) 0.003*
 Cancer 42 (3.7) 81 (3.3) 0.530

Values are presented as mean±standard deviation or number (%).

ICF, International Classification of Functioning, Disability, and Health; PF, personal factors; BF, body functions; A, activities; P, participation; EF, environmental factors; HC, health condition; NA, not applicable.

a) Wilcoxon rank sum test,

b) chi-square test.

* p<0.05.

Table 2.
Demographic characteristics after matching
Variable ICF category IPTW normalized weight
IPTW stabilized weight
1:1 Greedy PS matching
Rehabilitation (n=1,122) No rehabilitation (n=2,285) p-value Rehabilitation (n=1,122) No rehabilitation (n=2,285) p-value Rehabilitation (n=1,122) No rehabilitation (n=1,122) p-value
Weighted (n=1,684) Weighted (n=1,722) Weighted (n=1,056) Weighted (n=2,342)
Agea) (yr) PF 64.1±15.7 63.4±11.8 0.108 64.1±12.4 63.4±13.7 0.141 67.5±11.2 66.5±12.0 0.016*
Years since primary impairment onseta) PF 18.9±20.9 19.1±13.6 0.696 18.9±16.5 19.1±15.8 0.713 15.9±14.0 16.3±13.4 0.124
Maleb) PF 872 (51.7) 903 (52.4) 0.693 547 (51.7) 1,229 (52.4) 0.715 408 (36.3) 500 (44.5) <0.001*
Marriedb) PF 1,074 (63.7) 1,095 (63.6) 0.928 673 (63.7) 1,490 (63.6) 0.933 645 (57.4) 589 (61.4) 0.059
Current employmentb) PF 688 (40.8) 736 (42.7) 0.259 431 (40.8) 1,001 (42.7) 0.297 384 (34.2) 404 (36.0) 0.376
Low incomeb) PF 819 (48.6) 855 (49.6) 0.553 513 (48.6) 1,163 (49.6) 0.584 689 (61.4) 629 (56.0) 0.010*
Educationb) PF 0.944 0.956 0.086
 Less than elementary school 814 (48.3) 829 (48.1) 510 (48.3) 1128 (48.1) 684 (60.9) 624 (55.6)
 Middle school 273 (16.2) 282 (16.4) 171 (16.2) 384 (16.4) 183 (16.3) 207 (18.4)
 High school 392 (23.3) 391 (22.7) 532 (22.7) 246 (23.3) 172 (15.3) 196 (17.4)
 College beyond 204 (12.1) 218 (12.7) 297 (12.7) 128 (12.1) 83 (7.4) 95 (8.4)
Current smokerb) PF 337 (20.0) 345 (20.0) 0.981 211 (20.0) 470 (20.0) 0.983 144 (12.8) 177 (15.7) 0.047*
Current drinkerb) PF 1,005 (59.6) 1,007 (58.4) 0.477 426 (40.3) 972 (41.5) 0.511 768 (68.4) 736 (65.6) 0.151
Physical function scoresa) BF 49.4±9.4 49.8±6.5 0.173 49.4±7.4 49.8±7.6 0.205 49.6±7.3 49.3±7.8 0.532
Physical activityb) (3 days/wk) A 817 (48.5) 813 (47.2) 0.454 512 (48.5) 1,106 (47.2) 0.489 539 (48.0) 540 (48.1) 0.966
Disability gradesb) P 0.965 0.975 0.934
 1 49 (2.9) 42 (2.4) 30 (2.9) 57 (2.4) 22 (1.9) 30 (2.6)
 2 88 (5.2) 92 (5.3) 55 (5.2) 126 (5.3) 47 (4.1) 49 (4.3)
 3 196 (11.6) 208 (12.1) 123 (11.6) 283 (12.1) 120 (10.7) 129 (11.5)
 4 324 (19.2) 331 (19.2) 203 (19.2) 451 (19.2) 240 (21.3) 218 (19.4)
 5 525 (31.2) 520 (30.2) 329 (31.2) 707 (30.2) 374 (33.3) 380 (33.8)
 6 448 (26.5) 476 (27.6) 281 (26.6) 648 (27.6) 276 (24.6) 280 (24.9)
 NA 50 (2.9) 49 (2.8) 31 (2.9) 67 (2.8) 43 (3.8) 36 (3.2)
Social supportb) EF 1,356 (80.5) 1,370 (79.5) 0.484 850 (80.5) 1,863 (79.5) 0.519 813 (72.4) 866 (77.1) 0.010*
Outpatient visitb) (past 2 wk) EF 999 (59.3) 1,008 (58.5) 0.655 626 (59.3) 1,372 (58.5) 0.680 855 (76.2) 818 (72.9) 0.073
Preventative health servicesb) (past 12 mo) EF 55 (3.3) 56 (3.3) 0.996 34 (3.3) 77 (3.3) 0.996 47 (4.1) 36 (3.2) 0.219
Mental health servicesb) (past 12 mo) EF 33 (2.0) 36 (2.1) 0.813 21 (2.0) 49 (2.1) 0.827 30 (2.6) 25 (2.2) 0.495
Current depression treatmentb) EF 75 (4.5) 72 (4.2) 0.700 47 (4.5) 99 (4.2) 0.721 59 (5.2) 58 (5.1) 0.924
History of fallsb) HC 64 (3.8) 66 (3.8) 0.961 40 (3.8) 89 (3.8) 0.964 47 (4.1) 48 (4.2) 0.917
Self-rated healthb) HC 0.766 0.714 0.186
 Very poor 244 (14.5) 242 (14.1) 153 (14.5) 330 (14.1) 216 (19.2) 192 (17.1)
 Poor 794 (47.1) 780 (45.3) 498 (47.1) 1061 (45.3) 604 (53.8) 578 (51.2)
 Fair 472 (28.6) 509 (29.6) 296 (28.0) 693 (29.6) 245 (21.8) 289 (25.7)
 Good 163 (9.6) 174 (10.1) 102 (9.6) 237 (10.1) 53 (4.7) 60 (5.3)
 Very good 9 (0.5) 14 (0.8) 6 (0.5) 20 (0.8) 4 (0.3) 3 (0.2)
Chronic conditionsb) HC
 Hypertension 757 (44.9) 753 (43.7) 0.457 475 (44.9) 1,024 (43.7) 0.491 579 (51.6) 564 (50.2) 0.527
 Dyslipidemia 170 (10.1) 175 (10.2) 0.932 106 (10.1) 239 (10.2) 0.937 135 (12.0) 133 (11.8) 0.896
 Stroke 62 (3.7) 64 (3.7) 0.977 39 (3.7) 87 (3.7) 0.979 47 (4.1) 47 (4.1) 1.000
 Heart problem 131 (7.7) 130 (7.6) 0.862 82 (7.7) 178 (7.6) 0.872 111 (9.8) 94 (8.3) 0.213
 Osteoarthritis 575 (34.1) 574 (33.3) 0.638 360 (34.1) 781 (33.3) 0.663 575 (51.2) 469 (41.8) <0.001*
 Rheumatoid arthritis 78 (4.6) 72 (4.2) 0.552 49 (4.6) 99 (4.2) 0.579 61 (5.4) 60 (5.3) 0.926
 Back pain 367 (21.8) 354 (20.6) 0.387 230 (21.8) 482 (20.6) 0.422 354 (31.5) 275 (24.5) <0.001*
 Lung problem 80 (4.8) 76 (4.4) 0.609 50 (4.8) 103 (4.4) 0.634 51 (4.5) 52 (4.6) 0.920
 Diabetes 341 (20.2) 330 (19.2) 0.442 213 (20.2) 449 (19.2) 0.476 238 (21.2) 249 (22.1) 0.573
 Cancer 62 (3.7) 60 (3.5) 0.757 39 (3.7) 82 (3.5) 0.774 42 (3.7) 44 (3.9) 0.826

Values are presented as mean±standard deviation or number (%).

ICF, International Classification of Functioning, Disability, and Health; IPTW, inverse probability of treatment weighting matching; PS, propensity score; PF, personal factors; BF, body functions; A, activities; P, participation; EF, environmental factors; HC, health condition; NA, not applicable.

a) Wilcoxon rank sum test,

b) chi-square test.

* p<0.05.

Table 3.
Odds of depressive symptoms in adults with physical disabilities receiving rehabilitation services
Estimation method Depressive symptoms
AORa) 95% CI
Unadjusted logistic regression 1.639* 1.389–1.933
Adjusted multivariable logistic regression 1.233* 1.013–1.501
Inverse probability of treatment weighting and normalized weight with logistic regression adjustment 1.191* 1.014–1.400
Inverse probability of treatment weighting and stabilized weight with logistic regression adjustment 1.191* 1.002–1.417
1:1 propensity score matching with logistic regression adjustment 1.294* 1.065–1.574

AOR, adjusted odds ratio; CI, confidence interval.

a) p-values for estimated odds ratios were calculated using multivariable logistic regression.

* p<0.05.

REFERENCES

1. World Health Organization. International classification of functioning, disability and health: ICF. Geneva: World Health Organization; 2001.

2. Kang LJ, Hwang AW, Chen CL,Participation and environmental factors of children with physical disabilities in Taiwan. Tan U, editors. Physical disabilities: therapeutic Implications. Rijeka, Croatia: InTech; 2017. p.43–56.

3. Lenze EJ, Rogers JC, Martire LM, Mulsant BH, Rollman BL, Dew MA, et al. The association of late-life depression and anxiety with physical disability: a review of the literature and prospectus for future research. Am J Geriatr Psychiatry 2001;9:113–35.
crossref pmid
4. Kim WH, Park YG, Shin HI, Im SH. The World Report on Disability and recent developments in South Korea. Am J Phys Med Rehabil 2014;93(1 Suppl 1): S58–62.
crossref pmid
5. Shen SC, Huang KH, Kung PT, Chiu LT, Tsai WC. Incidence, risk, and associated factors of depression in adults with physical and sensory disabilities: a nationwide population-based study. PLoS One 2017;12:e0175141.
crossref pmid pmc
6. Gallegos-Carrillo K, Garcia-Pena C, Mudgal J, Romero X, Duran-Arenas L, Salmeron J. Role of depressive symptoms and comorbid chronic disease on health-related quality of life among community-dwelling older adults. J Psychosom Res 2009;66:127–35.
crossref pmid
7. Bock JO, Hajek A, Weyerer S, Werle J, Wagner M, Maier W, et al. The Impact of Depressive Symptoms on Healthcare Costs in Late Life: Longitudinal Findings From the AgeMooDe Study. Am J Geriatr Psychiatry 2017;25:131–41.
crossref pmid
8. McDermott S, Moran R, Platt T, Issac T, Wood H, Dasari S. Depression in adults with disabilities, in primary care. Disabil Rehabil 2005;27:117–23.
crossref pmid
9. Robinson-Whelen S, Taylor HB, Hughes RB, Wenzel L, Nosek MA. Depression and depression treatment in women with spinal cord injury. Top Spinal Cord Inj Rehabil 2014;20:23–31.
crossref pmid pmc
10. Brunger M, Spyra K. Prevalence of comorbid depressive symptoms in rehabilitation: a cross-indication, nationwide observational study. J Rehabil Med 2016;48:903–8.
crossref pmid
11. Park JH, Kim KW, Kim MH, Kim MD, Kim BJ, Kim SK, et al. A nationwide survey on the prevalence and risk factors of late life depression in South Korea. J Affect Disord 2012;138:34–40.
crossref pmid
12. Muller R, Peter C, Cieza A, Geyh S. The role of social support and social skills in people with spinal cord injury: a systematic review of the literature. Spinal Cord 2012;50:94–106.
crossref pmid pdf
13. Hawker GA, Gignac MA, Badley E, Davis AM, French MR, Li Y, et al. A longitudinal study to explain the pain-depression link in older adults with osteoarthritis. Arthritis Care Res (Hoboken) 2011;63:1382–90.
crossref pmid
14. Hirsch O, Strauch K, Held H, Redaelli M, Chenot JF, Leonhardt C, et al. Low back pain patient subgroups in primary care: pain characteristics, psychosocial determinants, and health care utilization. Clin J Pain 2014;30:1023–32.
crossref pmid
15. Wild B, Herzog W, Schellberg D, Lechner S, Niehoff D, Brenner H, et al. Association between the prevalence of depression and age in a large representative German sample of people aged 53 to 80 years. Int J Geriatr Psychiatry 2012;27:375–81.
crossref pmid
16. Noh JW, Kwon YD, Park J, Oh IH, Kim J. Relationship between physical disability and depression by gender: a panel regression model. PLoS One 2016;11:e0166238.
crossref pmid pmc
17. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 2011;46:399–424.
crossref pmid pmc
18. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:41–55.
crossref
19. Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc 1984;79:516–24.
crossref
20. Rosenbaum PR, Rubin DB. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985;39:33–8.
crossref
21. Austin PC, Mamdani MM. A comparison of propensity score methods: a case-study estimating the effectiveness of post-AMI statin use. Stat Med 2006;25:2084–106.
crossref pmid
22. Xu S, Ross C, Raebel MA, Shetterly S, Blanchette C, Smith D. Use of stabilized inverse propensity scores as weights to directly estimate relative risk and its confidence intervals. Value Health 2010;13:273–7.
crossref pmid
23. Kim S, Lee Y, Hwang J, Oh M, Lee MK, Lee NH, et al. 2014 National Survey on people with disabilities. Sejong, Korea: Korean Institute for Health and Social Affairs; 2014.

24. Hong I, Lee MJ, Kim MY, Park HY. Item-level psychometrics of the ADL instrument of the Korean National Survey on persons with physical disabilities. Qual Life Res 2017;26:2867–75.
crossref pmid pdf
25. Coca-Perraillon M. Local and global optimal propensity score matching. In: Proceedings of the SAS Global Forum 2007; 2007;Apr 16-19 Orlando, FL. 1–9.

26. d’Agostino RB. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998;17:2265–81.
crossref pmid
27. Stuart EA. Matching methods for causal inference: a review and a look forward. Stat Sci 2010;25:1–21.
crossref pmid pmc
28. Rosenbaum PR. Observational studies. 2nd ed. New York: Springer; 2002.

29. Rosenbaum PR, Rubin DB. Assessing sensitivity to an unobserved binary covariate in an observational study with binary outcome. J R Stat Soc Series B Stat Methodol 1983;45:212–8.
crossref
30. Taylor J, Taylor S. Psychological approaches to sports injury rehabilitation. Gaithersburg, MD: Aspen Publishers; 1997.

Appendices

Appendix 1.

Histogram of propensity score for the two groups.
arm-2019-43-3-250-app.pdf
TOOLS
Share :
Facebook Twitter Linked In Google+ Line it
METRICS Graph View
  • 0 Crossref
  •   Scopus
  • 193 View
  • 29 Download
Related articles in ARM


ABOUT
ARTICLE CATEGORY

Browse all articles >

BROWSE ARTICLES
AUTHOR INFORMATION
Terms of Use   |   Privacy Polity
Editorial Office
Department of Rehabilitation Medicine, Seoul National University Boramae Medical Center,
20 Boramae-ro 5-gil, Dongjak-gu, Seoul, 07061, Korea
Tel: +82-2-870-2679    Fax: +82-2-870-2679    E-mail: edit@e-arm.org
Business Registration: 110-82-07460                

Copyright © 2019 by Korean Academy of Rehabilitation Medicine. All rights reserved.

Developed in M2community

Close layer
prev next