Journal of Physical Activity and Health, 2015, 12, 1031 -1038 http://dx.doi.org/10.1123/jpah.2014-0036 © 2015 Human Kinetics, Inc.
ORIGINAL RESEARCH
Correlates of Physical Activity Among People With Disabilities in South Korea: A Multilevel Modeling Approach Youngdeok Kim, Jaehoon Cho, Dana K Fuller, and Minsoo Kang Background: The purpose of this study was to examine the correlates of physical activity (PA) with personal and environmental factors among people with disabilities in South Korea. Methods: Data from the 2011 National Survey for Physical Activity and Exercise for the Disabled, conducted by Korea Sports Association for the Disabled, was used (n = 1478). The personal characteristics (age, gender, occupation, types of disabilities, family income) and the numbers of public PA-related facilities (welfare center, public indoor gym, and public outdoor facilities) and social sports/exercise clubs for people with disabilities across 16 local areas were also obtained. Hierarchical generalized linear model was used to examine subjectively measured PA in relation to personal and environmental factors. Results: The likelihood of engaging in PA was significantly lower for women with disabilities. People with hearing and intellectual disabilities were less likely to engage in PA compared with those with physical disabilities. The availability of sports/exercise clubs for people with disabilities was the only environmental factor that was significantly associated with PA. Conclusion: These findings suggest the need of systematic intervention strategies based upon personal characteristics of people with disabilities. Further public efforts to promote sports/exercise club activities should be encouraged in this population. Keywords: built environment, exercise, disability, public health
People with disabilities who have substantial limitations in activities of daily living due to physical or mental impairment1 are at a higher risk of having poor health outcomes compared with the general population.2 Over the past few decades, there has been a growing demand for health promotion and disease prevention in this population and reducing health disparities for people with disabilities has emerged as an important public health priority.1,3 Regular participation in physical activity (PA) is a modifiable lifestyle behavior that is the single most effective way for people with disabilities to improve their health and functioning status across their lifespan.4–6 However, a large body of literature cited lower levels of PA among people with disabilities,2,7 leading to poorer health outcomes and higher incidence of secondary and comorbid conditions that are directly or indirectly associated with their primary disability.8,9 A majority of evidence indicated that people with disabilities faced more substantial barriers to accessing PA opportunities compared with general population, discouraging them from engaging in regular PA.10–12 Based on the International Classification of Functioning, Disability, and Health (ICF) which provides the standard framework and language to describe health and healthrelated states,13 such barriers are classified into 2 contextual factors: personal and environmental factors. Personal factors, which have not yet been clearly defined due to the large variations in social and cultural differences among individuals,14,15 generally include factors that influence how disabilities are presented in the individuals such Kim ([email protected]) is with the Dept of Health, Exercise, and Sport Sciences, Texas Tech University, Lubbock, TX. Cho is with the Dept of Adapted Physical Activity, Korea Nazareth University, Cheonan, Chungnam, South Korea; and was at Middle Tennessee State University as a visiting professor supported by Korea Nazarene University, South Korea, during part of the study. Fuller is with the Dept of Psychology, Middle Tennessee State University, Murfreesboro, TN. Kang is with the Dept of Health and Human Performance, Middle Tennessee State University, Murfreesboro, TN.
as gender, age, socioeconomic status.13,14 Environmental factors consist of the physical (or built), social and attitudinal environments.13 The most influential environmental factors related to PA participation among people with disabilities may include the built and social environments in which they conduct their lives.12,15,16 Built environment refers to the physical form of communities that include large- and small-scale built and natural features, land-use patterns and the transportation system.16 These factors can facilitate or constrain PA participation among people with disabilities depending on the accessibility features of the respective built environment.11 Accessibility is defined as “approachable, functional and usable by persons with disabilities, independently, safely, and with dignity”17(p151) and has been identified as a key element of successful accommodation for people with disabilities in regular PA participation. Specifically, the presence of accessible PA-related facilities for people with disabilities at the community level provide greater opportunities to engage in regular PA.12,18 The social environment, on the other hand, refers to the social relationships with groups of people among whom one lives,19 which include but not limited to family, friends, health care professionals, and social support groups. In the context of this research, social environment specifically refers to the presence of social support group, a collection of people who share a common life stressor (eg, disability),20 that may provide mutual support to facilitate PA among people with disabilities. Public demand for a nationwide effort to facilitate the access of PA-related built and social environments for people with disabilities has increased across the world.2 In South Korea, since the Anti-Discrimination Against and Remedies for Persons with Disabilities Act (Public Law No. 8341) was first enacted in 2007 (Enforcement Decree was then enacted in April 2008—Presidential Decree No. 20766),21 public effort to provide legitimate accommodation for people with disabilities related to PA has been made under the Article 16 (Prohibition of Discrimination in Relation to Physical Activities) of the Enforcement Decree of the Act. Specifically, municipalities have been forced to make mandatory modifications/installations to public PA-related facilities to improve 1031
Downloaded by University of California on 09/16/16, Volume 12, Article Number 7
1032 Kim et al
the accessibilities of the PA-related built environment for people with disabilities. Furthermore, the Korea Sports Association for the Disabled (KOSAD), founded in 2005 for the aims of promoting PA among people with disabilities, has supported the establishment of social sports/exercise clubs at the community level for people with disabilities. As a result of such endeavors, there has been an increase in PA among people with disabilities; however, the prevalence of PA is still lower than the general population.22 Despite increased interest in PA promotion among people with disabilities, there has been a lack of epidemiological studies examining the correlates of built and social environments with regular participation in PA among people with disabilities in South Korea. This has become particularly important given the implementation of a national Act promoting PA participation among people with disabilities in South Korea. Therefore, the purpose of this study was to examine the correlates of regular participation in PA with personal and environmental factors among people with disabilities in South Korea.
Methods Survey and Sample Data for this study came from the 2011 National Survey for Physical Activity and Exercise for the Disabled,22 which was a nationwide cross-sectional survey implemented by KOSAD during a December 2011, and used to examine the prevalence of PA among people with disabilities. Data included a total of 1500 people with disabilities selected by a proportional, convenience quota sampling method stratified by age, gender, and types of disabilities at 16 local census areas in South Korea. The survey was primarily administrated through a telephone interview, for which household phone numbers were systematically chosen from the database of the Korea Differently Abled Federation which is a federation of 28 organizations for people with disabilities subdivided by types of disabilities and regions. In the case of persons with disabilities who were unable to complete a phone interview on their own, assistance by their caregiver or family was allowed. Informed verbal consent was obtained before each interview and ethics approval was granted by Statistics Korea (Approval No. 11320). We restricted the initial sample of 1500 to adults with disabilities aged greater than 20 years; hence, the final sample consisted of 1478 persons with disabilities.
Measures The survey covered various aspects of PA participation among people with disabilities. The participants were initially asked to answer whether they had engaged in PA, defined as any purposefully engaged structured PA (ie, exercise), during the past year. The participants who responded ‘yes’ were interviewed further to obtain detailed information regarding their PA habits (ie, frequency, duration, and purpose). Frequency of PA participation was measured by the question “How often are you engaged in PA?” using a 6-point Likert-type scale (less than once per month, 2 to 3 days per month, once per week, 2 to 3 days per week, 4 to 5 days per week, and almost every day per week). Duration of PA participation was measured by the question “How long do you engage in PA each time?” using a 5-point Likert-type scale (< 30 minutes, 30–60 minutes, 60–90 minutes, 90–120 minutes, and > 120 minutes). Using the recommended level of PA participation for people with disabilities suggested by KOSAD (ie, 2 to 3 days/week and at least a 30-minute duration), the participants that engaged in PA during the past years
were subcategorized into 2 levels, insufficient PA (I-PA) and sufficient PA (S-PA). In addition, based on the response on a question asking the primary purpose of PA engagement during the past year, the participants who were in S-PA were further categorized into 2 groups (ie, for rehabilitation or for leisure time activities). The participants were, therefore, assigned to 4 categories: 1) No-PA, those who did not engage in PA during the past year; 2) I-PA, those who had engaged in insufficient level of PA during the past year; 3) S-PA-R, those who had engaged in sufficient level of PA primarily for rehabilitation during the past year; and 4) S-PA-L, those who have engaged in sufficient level of PA primarily for leisure time activities during the past year. Demographic characteristics of the participants were obtained in the questionnaire, including gender, age (20–29, 30–39, 40–49, 50–59, and ≥ 60 years old), occupation (occupied and nonoccupied), type of disability [physical, cerebral palsy, visual, hearing, intellectual, and others (eg, renal, heart, hepatic disorders, etc)], and family income [< 1000k, 1000k–1999k, 2000k–2999k, and ≥ 3000k in Korean Won (KRW)], and were included in the statistical models to explain the participation in PA at the personal level. To examine the correlates of environmental factors with PA participation among people with disabilities, we obtained the relative numbers of public PA-related facilities that might be accessible to the people with disabilities within the local area. Specifically, the number of 1) welfare centers for people with disabilities (ie, a center that is specialized for people with disabilities to provide a comprehensive welfare service), 2) public indoor gyms, and 3) public outdoor recreational facilities across 16 local areas was obtained from 2 nationwide reports.23,24 In addition, the number of social sports/exercise clubs for people with disabilities that were officially supported by KOSAD was obtained from KOSAD database25 as social sports/exercise clubs are the PA-related social environment factor that may provide the support to facilitate PA participation among people with disabilities.
Data Analysis Given that the primary focus of this study was to examine the correlates of PA participation (ie, No-PA, I-PA, S-PA-R, and S-PA-L) with personal and environmental characteristics at local area, a multilevel multinomial logistic regression model with logit link function (ie, hierarchical generalized linear model) was chosen for the main analysis.26 Using No-PA as the reference category, 3 sets of independent logit models that were contrasted with the reference category were created with logit link function. Each logit model identified the log-odds of being in one of the PA categories (I-PA, S-PA-R, or S-PA-L) relative to No-PA. Three sequential steps were taken to develop the final models. First, a simple unconditional model in which no explanatory variables were included was established to estimate the unadjusted variance components for random effects on the log-odds of being I-PA, S-PA-R, or S-PA-L compared with No-PA. The unadjusted intraclass correlation coefficients (ICCs) were calculated across the set of logit models as a relative proportion of the estimated random variance over the total variance that was the sum of the estimated random variance and standard logistic distribution (ie, π2/3 = 3.29).27 Random intercept models were then established with demographic characteristics as the fixed explanatory variables at the personal level. Finally, the intercept as outcome models were examined using local level variables in addition to demographic characteristics at the personal level to predict the log-odds of being in respective categories compared with No-PA. The adjusted ICCs for random intercept and intercept as outcome models were calculated.
JPAH Vol. 12, No. 7, 2015
Physical Activity Among People With Disabilities 1033
HLM version 7.00 (Scientific Software International Inc., Lincolnwood, IL, USA) was used to fit the models based upon a penalized quasi-likelihood approximation with a second-order Taylor linearization procedure.28 The prevalence estimates of PA participation and its crude associations with demographic characteristics based on the chi-square tests of independence were also examined using SAS v 9.2 (SAS Institute Inc., Cary, NC, USA).
Downloaded by University of California on 09/16/16, Volume 12, Article Number 7
Results Demographic characteristics of the participants across the subcategories of PA participation are presented in Table 1. Overall, 29.5% of participants answered No-PA during the past year. Of the participants who had engaged in PA during the past year, 28.6% had not sufficiently participated in PA. A majority of the participants who had engaged in sufficient levels of PA reported rehabilitation as a primary reason for their involvement (S-PA-R = 31.7%) while 10.2% of participants reported leisure time activity (S-PA-L). The results of unconditional and random intercept models are presented in Table 2. The unadjusted ICCs for the unconditional model which represented the percentage of the variations accounted
for by the local level discrepancies in the log-odds of being with respective categories compared with the reference category (NoPA) were 4.44%, 2.55%, and 1.50% for I-PA, S-PA-R, and S-PA-L, respectively. Random variations for S-PA-L was not statistically significant at alpha level of .05 (χ2 = 20.9, P = .140). Considering that there is no exact agreement with respect to the significance level of random variations in multilevel logistic regression models29 and the argument suggested by Snijders and Bosker,27 that a P-value of .20 or .25 should be used for testing the significance of random variation in a multilevel logistic regression model, we concluded that the odds of being in one of the respective PA categories compared with No-PA were dependent upon the local areas. After adjusting for personal characteristics of the participants, the ICCs were slightly reduced to 4.42% and 2.40% for I-PA and S-PA-R, respectively, whereas the ICC for S-PA-L was increased to 1.76%. Table 3 represents the result of the set of random intercept as outcome models using No-PA as a reference category. At the personal level, after adjusting for other personal and local level predictors, there were systematic gender effects across the logit models for S-PA-R and S-PA-L. Women with disabilities were less likely to engage in sufficient PA for both rehabilitation and leisure time activity (OR = 0.53, 95% CI = 0.39–0.71, for S-PA-R; OR =
Table 1 Prevalence of Physical Activity (PA) Participation By Demographic Characteristics Total
No-PA
I-PA
S-PA-R
S-PA-L
Men
989 (66.9%)
258 (26.1%)
278 (18.8%)
345 (23.3%)
108 (7.3%)
Women
489 (33.1%)
178 (36.4%)
145 (29.7%)
124 (25.4%)
42 (8.6%)
20–29 yrs
108 (7.3%)
27 (25.0%)
47 (43.5%)
24 (22.2%)
10 (9.3%)
30–39 yrs
168 (11.4%)
62 (36.9%)
46 (27.4%)
45 (26.8%)
15 (8.9%)
40–49 yrs
356 (24.1%)
104 (29.2%)
97 (27.3%)
113 (31.7%)
42 (11.8%)
50–59 yrs
425 (28.8%)
103 (24.2%)
119 (28.0%)
151 (35.5%)
52 (12.2%)
> 60 yrs
421 (28.5%)
140 (33.3%)
114 (27.1%)
136 (32.3%)
31 (7.4%)
Occupied
1007 (68.1%)
316 (31.4%)
269 (26.7%)
323 (32.1%)
99 (9.8%)
Nonoccupied
471 (31.9%)
120 (25.5%)
154 (32.7%)
146 (31.0%)
51 (10.8%)
Gender
ORIGINAL RESEARCH
Correlates of Physical Activity Among People With Disabilities in South Korea: A Multilevel Modeling Approach Youngdeok Kim, Jaehoon Cho, Dana K Fuller, and Minsoo Kang Background: The purpose of this study was to examine the correlates of physical activity (PA) with personal and environmental factors among people with disabilities in South Korea. Methods: Data from the 2011 National Survey for Physical Activity and Exercise for the Disabled, conducted by Korea Sports Association for the Disabled, was used (n = 1478). The personal characteristics (age, gender, occupation, types of disabilities, family income) and the numbers of public PA-related facilities (welfare center, public indoor gym, and public outdoor facilities) and social sports/exercise clubs for people with disabilities across 16 local areas were also obtained. Hierarchical generalized linear model was used to examine subjectively measured PA in relation to personal and environmental factors. Results: The likelihood of engaging in PA was significantly lower for women with disabilities. People with hearing and intellectual disabilities were less likely to engage in PA compared with those with physical disabilities. The availability of sports/exercise clubs for people with disabilities was the only environmental factor that was significantly associated with PA. Conclusion: These findings suggest the need of systematic intervention strategies based upon personal characteristics of people with disabilities. Further public efforts to promote sports/exercise club activities should be encouraged in this population. Keywords: built environment, exercise, disability, public health
People with disabilities who have substantial limitations in activities of daily living due to physical or mental impairment1 are at a higher risk of having poor health outcomes compared with the general population.2 Over the past few decades, there has been a growing demand for health promotion and disease prevention in this population and reducing health disparities for people with disabilities has emerged as an important public health priority.1,3 Regular participation in physical activity (PA) is a modifiable lifestyle behavior that is the single most effective way for people with disabilities to improve their health and functioning status across their lifespan.4–6 However, a large body of literature cited lower levels of PA among people with disabilities,2,7 leading to poorer health outcomes and higher incidence of secondary and comorbid conditions that are directly or indirectly associated with their primary disability.8,9 A majority of evidence indicated that people with disabilities faced more substantial barriers to accessing PA opportunities compared with general population, discouraging them from engaging in regular PA.10–12 Based on the International Classification of Functioning, Disability, and Health (ICF) which provides the standard framework and language to describe health and healthrelated states,13 such barriers are classified into 2 contextual factors: personal and environmental factors. Personal factors, which have not yet been clearly defined due to the large variations in social and cultural differences among individuals,14,15 generally include factors that influence how disabilities are presented in the individuals such Kim ([email protected]) is with the Dept of Health, Exercise, and Sport Sciences, Texas Tech University, Lubbock, TX. Cho is with the Dept of Adapted Physical Activity, Korea Nazareth University, Cheonan, Chungnam, South Korea; and was at Middle Tennessee State University as a visiting professor supported by Korea Nazarene University, South Korea, during part of the study. Fuller is with the Dept of Psychology, Middle Tennessee State University, Murfreesboro, TN. Kang is with the Dept of Health and Human Performance, Middle Tennessee State University, Murfreesboro, TN.
as gender, age, socioeconomic status.13,14 Environmental factors consist of the physical (or built), social and attitudinal environments.13 The most influential environmental factors related to PA participation among people with disabilities may include the built and social environments in which they conduct their lives.12,15,16 Built environment refers to the physical form of communities that include large- and small-scale built and natural features, land-use patterns and the transportation system.16 These factors can facilitate or constrain PA participation among people with disabilities depending on the accessibility features of the respective built environment.11 Accessibility is defined as “approachable, functional and usable by persons with disabilities, independently, safely, and with dignity”17(p151) and has been identified as a key element of successful accommodation for people with disabilities in regular PA participation. Specifically, the presence of accessible PA-related facilities for people with disabilities at the community level provide greater opportunities to engage in regular PA.12,18 The social environment, on the other hand, refers to the social relationships with groups of people among whom one lives,19 which include but not limited to family, friends, health care professionals, and social support groups. In the context of this research, social environment specifically refers to the presence of social support group, a collection of people who share a common life stressor (eg, disability),20 that may provide mutual support to facilitate PA among people with disabilities. Public demand for a nationwide effort to facilitate the access of PA-related built and social environments for people with disabilities has increased across the world.2 In South Korea, since the Anti-Discrimination Against and Remedies for Persons with Disabilities Act (Public Law No. 8341) was first enacted in 2007 (Enforcement Decree was then enacted in April 2008—Presidential Decree No. 20766),21 public effort to provide legitimate accommodation for people with disabilities related to PA has been made under the Article 16 (Prohibition of Discrimination in Relation to Physical Activities) of the Enforcement Decree of the Act. Specifically, municipalities have been forced to make mandatory modifications/installations to public PA-related facilities to improve 1031
Downloaded by University of California on 09/16/16, Volume 12, Article Number 7
1032 Kim et al
the accessibilities of the PA-related built environment for people with disabilities. Furthermore, the Korea Sports Association for the Disabled (KOSAD), founded in 2005 for the aims of promoting PA among people with disabilities, has supported the establishment of social sports/exercise clubs at the community level for people with disabilities. As a result of such endeavors, there has been an increase in PA among people with disabilities; however, the prevalence of PA is still lower than the general population.22 Despite increased interest in PA promotion among people with disabilities, there has been a lack of epidemiological studies examining the correlates of built and social environments with regular participation in PA among people with disabilities in South Korea. This has become particularly important given the implementation of a national Act promoting PA participation among people with disabilities in South Korea. Therefore, the purpose of this study was to examine the correlates of regular participation in PA with personal and environmental factors among people with disabilities in South Korea.
Methods Survey and Sample Data for this study came from the 2011 National Survey for Physical Activity and Exercise for the Disabled,22 which was a nationwide cross-sectional survey implemented by KOSAD during a December 2011, and used to examine the prevalence of PA among people with disabilities. Data included a total of 1500 people with disabilities selected by a proportional, convenience quota sampling method stratified by age, gender, and types of disabilities at 16 local census areas in South Korea. The survey was primarily administrated through a telephone interview, for which household phone numbers were systematically chosen from the database of the Korea Differently Abled Federation which is a federation of 28 organizations for people with disabilities subdivided by types of disabilities and regions. In the case of persons with disabilities who were unable to complete a phone interview on their own, assistance by their caregiver or family was allowed. Informed verbal consent was obtained before each interview and ethics approval was granted by Statistics Korea (Approval No. 11320). We restricted the initial sample of 1500 to adults with disabilities aged greater than 20 years; hence, the final sample consisted of 1478 persons with disabilities.
Measures The survey covered various aspects of PA participation among people with disabilities. The participants were initially asked to answer whether they had engaged in PA, defined as any purposefully engaged structured PA (ie, exercise), during the past year. The participants who responded ‘yes’ were interviewed further to obtain detailed information regarding their PA habits (ie, frequency, duration, and purpose). Frequency of PA participation was measured by the question “How often are you engaged in PA?” using a 6-point Likert-type scale (less than once per month, 2 to 3 days per month, once per week, 2 to 3 days per week, 4 to 5 days per week, and almost every day per week). Duration of PA participation was measured by the question “How long do you engage in PA each time?” using a 5-point Likert-type scale (< 30 minutes, 30–60 minutes, 60–90 minutes, 90–120 minutes, and > 120 minutes). Using the recommended level of PA participation for people with disabilities suggested by KOSAD (ie, 2 to 3 days/week and at least a 30-minute duration), the participants that engaged in PA during the past years
were subcategorized into 2 levels, insufficient PA (I-PA) and sufficient PA (S-PA). In addition, based on the response on a question asking the primary purpose of PA engagement during the past year, the participants who were in S-PA were further categorized into 2 groups (ie, for rehabilitation or for leisure time activities). The participants were, therefore, assigned to 4 categories: 1) No-PA, those who did not engage in PA during the past year; 2) I-PA, those who had engaged in insufficient level of PA during the past year; 3) S-PA-R, those who had engaged in sufficient level of PA primarily for rehabilitation during the past year; and 4) S-PA-L, those who have engaged in sufficient level of PA primarily for leisure time activities during the past year. Demographic characteristics of the participants were obtained in the questionnaire, including gender, age (20–29, 30–39, 40–49, 50–59, and ≥ 60 years old), occupation (occupied and nonoccupied), type of disability [physical, cerebral palsy, visual, hearing, intellectual, and others (eg, renal, heart, hepatic disorders, etc)], and family income [< 1000k, 1000k–1999k, 2000k–2999k, and ≥ 3000k in Korean Won (KRW)], and were included in the statistical models to explain the participation in PA at the personal level. To examine the correlates of environmental factors with PA participation among people with disabilities, we obtained the relative numbers of public PA-related facilities that might be accessible to the people with disabilities within the local area. Specifically, the number of 1) welfare centers for people with disabilities (ie, a center that is specialized for people with disabilities to provide a comprehensive welfare service), 2) public indoor gyms, and 3) public outdoor recreational facilities across 16 local areas was obtained from 2 nationwide reports.23,24 In addition, the number of social sports/exercise clubs for people with disabilities that were officially supported by KOSAD was obtained from KOSAD database25 as social sports/exercise clubs are the PA-related social environment factor that may provide the support to facilitate PA participation among people with disabilities.
Data Analysis Given that the primary focus of this study was to examine the correlates of PA participation (ie, No-PA, I-PA, S-PA-R, and S-PA-L) with personal and environmental characteristics at local area, a multilevel multinomial logistic regression model with logit link function (ie, hierarchical generalized linear model) was chosen for the main analysis.26 Using No-PA as the reference category, 3 sets of independent logit models that were contrasted with the reference category were created with logit link function. Each logit model identified the log-odds of being in one of the PA categories (I-PA, S-PA-R, or S-PA-L) relative to No-PA. Three sequential steps were taken to develop the final models. First, a simple unconditional model in which no explanatory variables were included was established to estimate the unadjusted variance components for random effects on the log-odds of being I-PA, S-PA-R, or S-PA-L compared with No-PA. The unadjusted intraclass correlation coefficients (ICCs) were calculated across the set of logit models as a relative proportion of the estimated random variance over the total variance that was the sum of the estimated random variance and standard logistic distribution (ie, π2/3 = 3.29).27 Random intercept models were then established with demographic characteristics as the fixed explanatory variables at the personal level. Finally, the intercept as outcome models were examined using local level variables in addition to demographic characteristics at the personal level to predict the log-odds of being in respective categories compared with No-PA. The adjusted ICCs for random intercept and intercept as outcome models were calculated.
JPAH Vol. 12, No. 7, 2015
Physical Activity Among People With Disabilities 1033
HLM version 7.00 (Scientific Software International Inc., Lincolnwood, IL, USA) was used to fit the models based upon a penalized quasi-likelihood approximation with a second-order Taylor linearization procedure.28 The prevalence estimates of PA participation and its crude associations with demographic characteristics based on the chi-square tests of independence were also examined using SAS v 9.2 (SAS Institute Inc., Cary, NC, USA).
Downloaded by University of California on 09/16/16, Volume 12, Article Number 7
Results Demographic characteristics of the participants across the subcategories of PA participation are presented in Table 1. Overall, 29.5% of participants answered No-PA during the past year. Of the participants who had engaged in PA during the past year, 28.6% had not sufficiently participated in PA. A majority of the participants who had engaged in sufficient levels of PA reported rehabilitation as a primary reason for their involvement (S-PA-R = 31.7%) while 10.2% of participants reported leisure time activity (S-PA-L). The results of unconditional and random intercept models are presented in Table 2. The unadjusted ICCs for the unconditional model which represented the percentage of the variations accounted
for by the local level discrepancies in the log-odds of being with respective categories compared with the reference category (NoPA) were 4.44%, 2.55%, and 1.50% for I-PA, S-PA-R, and S-PA-L, respectively. Random variations for S-PA-L was not statistically significant at alpha level of .05 (χ2 = 20.9, P = .140). Considering that there is no exact agreement with respect to the significance level of random variations in multilevel logistic regression models29 and the argument suggested by Snijders and Bosker,27 that a P-value of .20 or .25 should be used for testing the significance of random variation in a multilevel logistic regression model, we concluded that the odds of being in one of the respective PA categories compared with No-PA were dependent upon the local areas. After adjusting for personal characteristics of the participants, the ICCs were slightly reduced to 4.42% and 2.40% for I-PA and S-PA-R, respectively, whereas the ICC for S-PA-L was increased to 1.76%. Table 3 represents the result of the set of random intercept as outcome models using No-PA as a reference category. At the personal level, after adjusting for other personal and local level predictors, there were systematic gender effects across the logit models for S-PA-R and S-PA-L. Women with disabilities were less likely to engage in sufficient PA for both rehabilitation and leisure time activity (OR = 0.53, 95% CI = 0.39–0.71, for S-PA-R; OR =
Table 1 Prevalence of Physical Activity (PA) Participation By Demographic Characteristics Total
No-PA
I-PA
S-PA-R
S-PA-L
Men
989 (66.9%)
258 (26.1%)
278 (18.8%)
345 (23.3%)
108 (7.3%)
Women
489 (33.1%)
178 (36.4%)
145 (29.7%)
124 (25.4%)
42 (8.6%)
20–29 yrs
108 (7.3%)
27 (25.0%)
47 (43.5%)
24 (22.2%)
10 (9.3%)
30–39 yrs
168 (11.4%)
62 (36.9%)
46 (27.4%)
45 (26.8%)
15 (8.9%)
40–49 yrs
356 (24.1%)
104 (29.2%)
97 (27.3%)
113 (31.7%)
42 (11.8%)
50–59 yrs
425 (28.8%)
103 (24.2%)
119 (28.0%)
151 (35.5%)
52 (12.2%)
> 60 yrs
421 (28.5%)
140 (33.3%)
114 (27.1%)
136 (32.3%)
31 (7.4%)
Occupied
1007 (68.1%)
316 (31.4%)
269 (26.7%)
323 (32.1%)
99 (9.8%)
Nonoccupied
471 (31.9%)
120 (25.5%)
154 (32.7%)
146 (31.0%)
51 (10.8%)
Gender
