To What Extent Do Neighborhood Differences Mediate Racial Disparities in Participation After Spinal Cord Injury?

Accepted Manuscript To What Extent Do Neighborhood Differences Mediate Racial Disparities in Participation after Spinal Cord Injury? Amanda L. Botticello, PhD, MPH, Mike Boninger, MD, Susan Charlifue, PhD, Yuying Chen, MD, PhD, Denise Fyffe, PhD, Allen Heinemann, PhD, Jeanne M. Hoffman, PhD, Alan Jette, PhD, MPH, Claire Kalpakjian, PhD, MS, Tanya Rohrbach, MS PII:

S0003-9993(16)30135-6

DOI:

10.1016/j.apmr.2016.04.007

Reference:

YAPMR 56534

To appear in:

ARCHIVES OF PHYSICAL MEDICINE AND REHABILITATION

Received Date: 16 November 2015 Revised Date:

5 April 2016

Accepted Date: 10 April 2016

Please cite this article as: Botticello AL, Boninger M, Charlifue S, Chen Y, Fyffe D, Heinemann A, Hoffman JM, Jette A, Kalpakjian C, Rohrbach T, To What Extent Do Neighborhood Differences Mediate Racial Disparities in Participation after Spinal Cord Injury?, ARCHIVES OF PHYSICAL MEDICINE AND REHABILITATION (2016), doi: 10.1016/j.apmr.2016.04.007. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

1

Running Title: Neighborhood and Race Disparities in Participation after SCI

2 To What Extent Do Neighborhood Differences Mediate Racial Disparities in Participation after Spinal Cord Injury?

5 6 7 8 9

Amanda L. Botticello, PhD, MPH 1,2, Mike Boninger, MD 3, Susan Charlifue, PhD 4, Yuying Chen, MD, PhD 5 , Denise Fyffe, PhD 1,2, Allen Heinemann, PhD 6, Jeanne M. Hoffman, PhD 7, Alan Jette, PhD, MPH 8, Claire Kalpakjian, PhD, MS 9, Tanya Rohrbach, MS 9

RI PT

3 4

1

Kessler Foundation, West Orange, NJ

11

2

Department of Physical Medicine and Rehabilitation, Rutgers New Jersey Medical School, Newark

12 13

3

14

4

Craig Hospital, Denver, CO

15

5

University of Alabama, Birmingham

16 17

6

18

7

University of Washington, Seattle

19

8

Boston University

20

9

University of Michigan, Ann Arbor

21

10

SC

10

M AN U

Departments of Physical Medicine and Rehabilitation, Bioengineering, and Rehabilitation Science and Technology, University of Pittsburgh School of Medicine, Pittsburgh, PA

TE D

Rehabilitation Institute of Chicago and Department of Physical Medicine and Rehabilitation, Feinberg School of Medicine, Northwestern University

22

EP

Department of Science and Engineering, Raritan Valley Community College, Branchburg, NJ

Acknowledgement:

24

This research was supported by funding from the Eunice Kennedy Shriver National Institute of Child

25

Health and Development (grant number: 4R00HD065957-05) and the National Institute on Disability and

26

Rehabilitation Research (grant number: 90SI5011-01-00).

27

We would like to thank Ms. Nicolette Cobbold, BS and Ms. Rachel Byrne, MA for their assistance with

28

the preparation of the data and editorial comments, respectively.

AC C

23

29

Page 1 of 2

ACCEPTED MANUSCRIPT

30

Conflicts of interest: None

Corresponding Author:

33

Amanda L. Botticello, PhD, MPH

34

Senior Research Scientist

35

Outcomes and Assessment Research

36

Kessler Foundation

37

1199 Pleasant Valley Way

38

West Orange, NJ 07052

39

Phone: 973-243-6973

40

Email: [email protected]

41

EP

TE D

Word count: 3,452

AC C

42

M AN U

SC

32

RI PT

31

Page 2 of 2

ACCEPTED MANUSCRIPT

*This Manuscript is a submission for the Archives of Physical Medicine and Rehabilitation Special Issue that will feature Spinal Cord Injury Model Systems research.

RI PT

ABSTRACT Objective: To examine the role of residential neighborhood characteristics in accounting for race disparities in participation among a large sample of community-living adults with chronic

SC

spinal cord injury (SCI).

M AN U

Design: Secondary analysis of cross-sectional survey data from the national Spinal Cord Injury Model Systems (SCIMS) database linked with national survey and spatial data. Setting: SCIMS database participants enrolled at 10 collaborating centers active in follow up between 2000 and 2014.

TE D

Participants: The sample consisted of 6,892 persons with SCI in 5,441 Census tracts from 50 states and the District of Columbia.

EP

Intervention: Not applicable.

Main Outcome Measure: The Craig Handicap Assessment and Reporting Technique (CHART)

AC C

was used to measure full participation across four domains—physical independence (PI), mobility, occupation, and social integration (SI). Results: Racial minority groups had lower odds of reporting full participation relative to Whites across all domains, suggesting that Blacks and Hispanics are at risk for poorer community reintegration following SCI. Neighborhood characteristics—notably differences in socioeconomic advantage-- reduced race group differences in the odds of full occupational, and

Page 1 of 25

ACCEPTED MANUSCRIPT

social integration, suggesting that the race disparities in community reintegration after SCI are partially attributable to variation in the economic characteristics of the places where people

RI PT

live. Conclusion: This investigation suggests that addressing disadvantage at the neighborhood level may modify gaps in community participation following medical rehabilitation and provides

SC

further support for the role of the environment in the experience of disability.

Abbreviations: ACS – American Community Survey

M AN U

Keywords: Spinal Cord Injury; Participation; Race Disparities; Neighborhood; Mediation

ASIA – American Spinal Cord Injury Association

CHART – Craig Handicap Assessment and Reporting Technique

TE D

FIM – Functional Independence Measure

FIPS – Federal Information Processing Standard

EP

LU/LC – land use and land cover

NHGIS – National Historical Geographic Information System

AC C

PHQ-2 – Patient Health Questionnaire-2 PI – physical independence SCI – Spinal Cord Injury

SCIMS – Spinal Cord Injury Model Systems SES – socioeconomic status

Page 2 of 25

ACCEPTED MANUSCRIPT

SI – social integration

AC C

EP

TE D

M AN U

SC

RI PT

USGS – United States Geological Survey

Page 3 of 25

ACCEPTED MANUSCRIPT

Research in health disparities has brought needed attention to the added vulnerability of

2

specific populations, such as racial and ethnic minorities, to the adverse consequences of spinal

3

cord injury (SCI). Studies generally find that when compared to Whites, minority groups such as

4

Blacks and Hispanics are more likely to report medical complications including pain,1,2

5

depression,3 more days in poor health, and lengthier and more frequent hospitalizations4.

6

Adjustment following SCI as indicated by perceived health and well-being,5, 6 the adequacy of

7

assistive technologies,7 and employment rates8–11 is also found to be worse among minorities.

8

The processes and conditions that may influence racial and ethnic disparities (hereafter

9

referred to as race disparities) in SCI are not well understood. Some SCI studies demonstrate

M AN U

SC

RI PT

1

that differences in individual socioeconomic status (SES) and access to resources such as

11

healthcare explain the gaps between Whites and Blacks in secondary complications,

12

rehospitalization, and indicators of well-being.4,12 Studies in the general population suggest that

13

neighborhoods may also influence race disparities in chronic conditions because place of

14

residence is strongly patterned by SES and race, leading to different environmental exposures

15

by social position.13–15 A series of studies of Blacks and Whites living in socially and

16

economically disadvantaged neighborhoods report similarly elevated rates of health-risk

17

behaviors (e.g., drinking, smoking, obesity, and physical inactivity) and adverse outcomes such

18

as diabetes, hypertension, and activity limitations.16–21 Other reports indicate that hypertension

19

disparities are less evident in non-segregated, low SES neighborhoods22 whereas explanatory

20

studies of disparate rates of diabetes23,24 and poor self-rated health25 between Blacks and

21

Whites demonstrate the importance of neighborhood SES. Although not extensive, these

22

findings support the relevance of place to race disparities in chronic conditions and have been

AC C

EP

TE D

10

Page 4 of 25

ACCEPTED MANUSCRIPT

similarly supported by research investigations of participation and activity limitations for

24

persons with disabilities.26,27 For instance, better quality of the built environment (i.e., mixed

25

land use, better street conditions) is associated with fewer reported activity limitations and

26

more participation among older adults 28,29 and more green space is associated with better

27

quality of life and greater participation for persons with SCI.30,31 Neighborhood social conditions

28

such as a perceived lack of safety, high crime rates, and residential instability are linked to

29

more activity limitations and lower quality of life in middle-aged and older adults 32–34 and in

30

adults with SCI.35,36 Similarly, socioeconomic deprivation in residential areas is detrimental to

31

health and well-being in both the general and SCI populations.37,38

M AN U

SC

RI PT

23

32

Conceptual models of disability emphasize the role of environmental factors in the disabling

34

process whereas social and ecological theories more directly elaborate the mechanisms that

35

connect the environment to health and disability. 39 Processes such as stress reduction, positive

36

coping, and healthy habits are posited as pathways connecting the environment to adaptive

37

health behaviors such as participation. 40 For instance, physical and socioeconomic

38

environments that provide few opportunities to access community locations in terms of

39

connectedness, destinations for activity, and transportation may lead to individuals constricting

40

the range of activity undertaken outside the home. The natural aspects of the environment are

41

theorized to positively impact participation by providing opportunity for activity, increasing the

42

social interaction that fosters social cohesion and the transfer of information relevant to

43

employment, healthcare activity, and social activity, and offering exposure to nature that

44

mitigates stress. In addition, many neighborhoods are segregated by characteristics of

AC C

EP

TE D

33

Page 5 of 25

ACCEPTED MANUSCRIPT

45

residents, such as SES and race, or isolated by location resulting in restricted resources and

46

opportunities that are contribute to the development and maintenance of race disparities.14

RI PT

47

Given the need to understand and address persistent inequalities in long-term outcomes from

49

SCI, this investigation sought to examine the contribution of neighborhood factors to observed

50

race disparities in community participation. Specifically, we hypothesized that accounting for

51

neighborhood SES and racial segregation would reduce observed gaps in overall community

52

participation. We hypothesized that aspects of the built environment indicative of insufficient

53

resources—that is, low food and transportation access—would reduce gaps between whites

54

and minorities in reported physical independence and mobility whereas the exposure to the

55

that natural environment would explain gaps in the social aspects of community participation

56

(i.e., occupational and social integration). To accomplish this, we linked survey data from the

57

Spinal Cord Injury Model Systems (SCIMS) database with residential characteristics derived

58

from national survey and spatial data to assess the contribution of social, economic, and

59

physical neighborhood characteristics to race disparities in community participation following

60

SCI.

M AN U

TE D

EP

AC C

61

SC

48

62

METHODS

63

Sample

64

This investigation uses cross-sectional data from participants enrolled at 10 SCIMS centers that

65

collaborated in a study that linked SCIMS data collected between 2000 and 2014 with Census

Page 6 of 25

ACCEPTED MANUSCRIPT

and geographic data. The protocol was approved by the Institutional Review Boards of all 10

67

sites. Approximately 15,800 individuals were active in SCIMS in this timeframe and 73% were

68

from the 10 collaborating sites. Case inclusion criteria were chronic impairment, age 18 years or

69

older when injured, not institutionalized, and valid residential address. Only data from the most

70

recently completed interview were used for analysis. Of the 8,351 cases meeting these criteria,

71

residential Census-tracts were successfully identified for 7,449 (89.2%). A small portion of the

72

cases were missing information on key measures (i.e., either the outcome measures, race, or

73

SCI diagnosis) and excluded (5.3%). The majority of the sample (95%) resided in tracts that were

74

sparsely clustered (i.e., one to two people). In order to minimize potential bias due to clustered

75

data, one case was sampled from the few tracts with five or more people resulting in the

76

exclusion of 52 cases. The final sample consisted of 6,892 persons in 5,441 Census tracts.

M AN U

SC

RI PT

66

TE D

77

Sources of Neighborhood Data

79

Census tracts defined by the US Census as small and relatively stable geographic areas with an

80

average of 4,000 residents41 were used to define neighborhoods. Data on the neighborhood

81

location, racial composition, SES, and concentrated poverty collected for the American

82

Community Survey (ACS) were extracted from the National Historical Geographic Information

83

System (NHGIS)42 and the United States Department of Agriculture Economic Research Service

84

Food Atlas data.43 Surveys obtained prior to 2008 were linked to the 2009 ACS and surveys

85

obtained in 2008 and after were linked to the 2013 ACS.44,45 Spatial data on land use and land

86

cover (LU/LC) attributes collected by the United States Geological Survey (USGS) was obtained

AC C

EP

78

Page 7 of 25

ACCEPTED MANUSCRIPT

87

from the 2006 and 2011 data releases and similarly linked to SCIMS data collected before and

88

after 2008.

RI PT

89

Measures

91

Participation was measured by four domains of the Craig Handicap Assessment and Reporting

92

Technique (CHART)46,47—physical independence (PI), mobility, occupation, and social integration

93

(SI). Domain scores range from 0 to 100 with maximum scores indicating participation

94

analogous to that of an able-bodied person. Due to ceiling effects, binary variables were

95

created using a cut point of 95 and above (1=full; 0=restricted), as described in other

96

approaches.30,48

M AN U

SC

90

97

Race was based on two self-reported variables assigning priority to Hispanic background,

99

followed by Black, and then other minority backgrounds. Due to small cell sizes, respondents

TE D

98

identifying as Asian, Native American, or an unidentified “other” racial background were

101

combined into a single group as a methodological control. The final variable included mutually

102

exclusive categories for White, Black, Hispanic, and Other.

AC C

103

EP

100

104

Participant demographic characteristics included current age, gender, education level, current

105

employment status, and marital status. SCI characteristics included American Spinal Injury

106

Association (ASIA) classification at discharge,49 duration of injury, and assistive device use.

107

Binary measures of self-rated health, depressive symptoms measured by the Patient Health

108

Questionnaire-2 (PHQ-2),50 and a continuous measure of motor Functional Independence

Page 8 of 25

ACCEPTED MANUSCRIPT

109

Measure (FIM) 51 scores that was divided by the number of motor items to retain the metric of

110

the original rating scale were used to control for differences in health and functioning.

RI PT

111 112

Neighborhood controls included Census designated regional and urban location (urban = ≥

113

2,500 residents, rural = < 2,500 residents).43

SC

114

Neighborhood mediating variables included racial composition assessed by the proportion of

116

White residents, categorized using tertile scores. Indicators of the built environment were

117

created from LU/LC data coded using the modified Anderson Classification System.52 Geospatial

118

Modelling Environment was used to calculate the proportion of developed open space and

119

green space within each tract and then binary variables based on a median split indicated a

120

large versus small proportion of developed open space and green space. The built environment

121

was also assessed by measures of access to food and transportation resources. Tracts were

122

categorized as low (versus high) food access if the majority of residents lived a far distance from

123

a supermarket (1 mile for urban and 10 miles for rural).43 Tracts were classified as low (versus

124

high) vehicle access if more than 100 households reported not having access to a vehicle and

125

are more than a half mile from a supermarket.43 Neighborhood SES was constructed from

126

standardized and summated ACS economic characteristics: median household income, median

127

home values, and the proportion of residents receiving investment income, graduating high

128

school, graduating four-year colleges, and employed in managerial or professional occupations.

129

Higher index scores indicate higher SES based on prior approaches53.

AC C

EP

TE D

M AN U

115

Page 9 of 25

ACCEPTED MANUSCRIPT

130

Statistical Analysis

132

All statistical analyses were conducted using Stata/SE 14.0. Bivariate tests of the relationships

133

between race, neighborhood characteristics, and participation were used to assess the initial

134

associations.54,55 Low to moderate values for initial pairwise correlations between the

135

neighborhood characteristics ranged from -0.04 to 0.56 and VIF values ranged between 1.02

136

and 2.35, indicating that multicollinearity among these predictors was not of concern. The

137

multivariate analyses first regressed participation on race to estimate the disparities among the

138

groups in each domain. Next, the models were adjusted for individual and residential location

139

controls. Several criteria were used to evaluate the most parsimonious model for each outcome

140

including the size and significance of the effect (i.e., odds ratios and 95% confidence intervals),

141

percent change in odds ratios between models, and overall tests of model fit (i.e., the Hosmer-

142

Lemeshow goodness-of-fit statistic). Covariates meeting each criterion were retained for the

143

mediation analyses of race differences in participation neighborhood characteristics. Mediation

144

was evaluated by observed reductions in the effect of race (calculations of percent change),

145

comparisons of model fit statistics, and testing of indirect effects using bootstrapping

146

procedures with 1,000 samples.56 Data on complete cases were used, resulting in differences in

147

the analytic n for each participation domain. Specifically, case completeness varied across

148

several covariates in including poor health, depressive symptoms, functional independence,

149

and wheelchair status. To assess potential complete-case bias for these covariates, missing

150

imputation analyses using the Stata “mi” commands were conducted post-hoc for each

AC C

EP

TE D

M AN U

SC

RI PT

131

Page 10 of 25

ACCEPTED MANUSCRIPT

151

outcome. No differences in the patterns of effect sizes or model fit were observed, therefore all

152

final reported parameter estimates are based on models estimated from complete data.

RI PT

153

RESULTS

155

Whites were significantly more likely than Blacks or Hispanics to report full participation across

156

all domains (Table 1). On average, Black and Hispanic participants were younger, more likely to

157

be male, had less education, were less likely to be employed post-injury, reported a shorter

158

duration of injury, and were more likely to report poor health than Whites. A higher proportion

159

of Blacks reported being unmarried, had complete paraplegia, and were wheelchair users. The

160

racial minority groups represented in this sample relative to Whites were more likely to reside

161

in places categorized as urban which include a larger proportion of minority residents, more

162

developed open space, less green space, low vehicle access, and low average neighborhood SES

163

(Table 1). Whites were more likely to live in communities characterized by low food access.

M AN U

TE D

164

SC

154

The unadjusted analyses demonstrated differences in participation by neighborhood

166

characteristics (Table 2). Regional variation was observed across all four domains. Residing in an

167

urban area was marginally associated with a lower likelihood of PI compared to living in a rural

168

area. Participation across all domains was significantly less likely to be reported by persons

169

residing in neighborhoods with a larger proportion of minority residents. The likelihood of

170

reporting full mobility or SI was greater among persons living in places with a large (versus

171

small) proportion of developed open space whereas a large (versus small) proportion of green

172

space was positively associated with full PI and SI. Significant positive relationships between

AC C

EP

165

Page 11 of 25

ACCEPTED MANUSCRIPT

living in neighborhoods with low food access and full PI, mobility, and SI were observed

174

whereas living in neighborhoods with a low vehicle access significantly decreased the likelihood

175

of reporting full mobility, occupation, and SI. More SES advantage in the neighborhood

176

significantly increased the likelihood of reporting full participation across all four domains.

RI PT

173

177

Tables 3a-3d show the results of a series of logistic regression analyses analyzing the role of key

179

neighborhood variables in observed race disparities in physical, mobility, occupational, and

180

social independence, respectively. As shown on Table 3a, Blacks and Hispanics had significantly

181

lower odds than Whites of full PI after SCI and these gaps were even more apparent after

182

controlling for systematic differences by race in background, SCI, health, and region of

183

residence (Model 2). One community characteristics was retained for the PI analysis. The

184

results of Model 3 indicated, counterintuitively, that persons living in neighborhoods with low

185

food access were significantly more likely to report full PI. However, accounting for

186

neighborhood differences in infrastructure resulted in a small to negligible reduction on the

187

race disparities in PI and bootstrapping results indicated that the indirect relationship between

188

race and food access only attained moderate statistical significance for the difference between

189

Whites and blacks, accounting for 2.68 % (95% CI: 0.002—0.052) of the total effect.

190

Furthermore, neither adjusting for participant nor neighborhood differences improved model

191

fit, suggesting that race disparities in PI may be influenced by other, unexamined processes.

AC C

EP

TE D

M AN U

SC

178

192 193

Blacks and Hispanics were also significantly less likely to report full Mobility post-injury relative

194

to Whites (59% and 52% respectively; Table 3b, Model 1). Accounting for participant

Page 12 of 25

ACCEPTED MANUSCRIPT

differences in background and area of residence reduced the size of mobility disparities for

196

Blacks by 49% and Hispanics by 27% (Model 2). In Model 3, living in advantaged neighborhoods

197

increased the likelihood of reporting full mobility whereas persons with SCI living in

198

neighborhoods with a high proportion of green space were 17% less likely to report full

199

mobility. The negative relationship between mobility and neighborhood green space observed

200

in the multivariate analysis suggests that this difference was suppressed when relevant

201

individual predictors—namely race—were excluded from the model. However, accounting for

202

neighborhood differences had little effect on the observed associations between race and

203

mobility, reducing the odds ratios by 2-3% and explaining 4.57 % (95% CI: -0.076 — -0.015) and

204

4.83% (95% CI: -0.084 — -0.073) of the total effect, respectively, using bootstrapping

205

techniques. More robust support for the mediating role of neighborhood SES was found, with

206

SES explaining 20.1 % (95% CI: 0.151 — -0.262) and 17.7% (95% CI: 0.108 — 0.073) of the total

207

difference between Blacks and Hispanics, respectively, relative to Whites.

SC

M AN U

TE D

208

RI PT

195

Full occupational participation post-injury was significantly less likely for Blacks and Hispanics

210

compared to Whites (Model 1 in Table 3c). A large portion of these gaps were reduced after

211

adjusting for differences in participants’ backgrounds (Model 2, 48% and 13% for Blacks and

212

Hispanics, respectively). Accounting for neighborhood differences in SES (Model 3) further

213

reduced the occupation disparity between Blacks and Whites by 7% and Hispanics and Whites

214

by 4%. Bootstrapping results indicated that the indirect relationship between race and

215

neighborhood SES was statistically significant accounting for 15.9 % (95% CI: 0.094—0.225) of

216

the total difference between Blacks and Whites and 17.9% (95% CI: 0.046—0.311) of the total

AC C

EP

209

Page 13 of 25

ACCEPTED MANUSCRIPT

difference between Hispanics and Whites suggesting that neighborhood differences in SES

218

partially mediate the disparity between Blacks, Hispanics, and Whites in employment and other

219

productive activities following SCI.

RI PT

217

220

Of the race disparities observed in participation, gaps in reported social integration were the

222

most pronounced with Blacks 65% less likely and Hispanics 41% less likely to report full SI

223

compared to Whites (Model 1, Table 3d). In Model 2, adjusting for the participant and area

224

controls mitigated a large portion of these gaps in SI (69% and 22%, for Blacks and Hispanics,

225

respectively). SI differences were further attenuated in Model 3 (10% and 8% reduction in the

226

odds ratios for Blacks and Hispanics, respectively, suggesting that disparities between

227

minorities and Whites in SI are partially mediated by differences in neighborhood SES. This was

228

supported by the bootstrapping results which showed that SES accounted for 20.7% of the total

229

effect between Blacks and Whites (95% CI: 0.046—0.311) and 27.6% of the total effect between

230

Hispanics and Whites (95% CI: 0.140—0.412)

231

EP

TE D

M AN U

SC

221

DISCUSSION

233

This study sought to understand race disparities in participation limitations after SCI by

234

investigating the role of neighborhood characteristics. The gaps in participation by race

235

observed in this large, comprehensive sample of persons with chronic SCI were substantial,

236

with Blacks and Hispanics reporting lower odds of full participation across all domains

237

compared to Whites. The patterns observed in this study corroborated prior reports of race

238

disparities in participation in SCI5 as well as prior reports noting the partial attenuation of race

AC C

232

Page 14 of 25

ACCEPTED MANUSCRIPT

disparities by differences in individual background.12 Accounting for differences in

240

neighborhood SES, in particular, significantly reduced the disparities in the social domains of

241

participation (i.e., occupation and integration), suggesting that the minorities with SCI may be

242

vulnerable to poorer adjustment in the long-term due to potentially constrained opportunities

243

for community activity in economically disadvantaged neighborhoods. Contrary to

244

expectations, the neighborhood characteristics examined here did little to account for the race

245

disparities observed in the physical aspects of participation. Furthermore, the hypothesized

246

intervening effects of neighborhood racial segregation as well as indicators of the quality of the

247

neighborhood infrastructure and greenspace were not supported, suggesting that observed

248

inequalities in participation after SCI may be attributed to other, unexamined pathways.

M AN U

SC

RI PT

239

249

The contribution of neighborhood characteristics was most relevant in explaining gaps in

251

community participation observed between Blacks and Whites. A possible explanation for this

252

pattern is that Blacks may be more influenced by neighborhood characteristics due to

253

accumulated disadvantage from likely lifelong exposure to adverse environmental conditions.

254

Although research suggests that Blacks and Hispanics are more likely to live in similar

255

neighborhoods, 57 other influences such as heterogeneity in nativity and generational status

256

may influence Hispanics rendering individual background and interpersonal circumstances

257

more relevant than neighborhoods. Furthermore, the proportion of Hispanics and the groups

258

composing the “Other” category were relatively small which posed an analytic constraint.

AC C

EP

TE D

250

259

Page 15 of 25

ACCEPTED MANUSCRIPT

The lack of influence of the built environment characteristics was a departure from previous

261

studies in SCI which reported improved outcomes among residents of areas characterized by

262

higher proportions of green space.30,31 These analyses differed from the current investigation in

263

the assessment of spatial scale, sources of spatial data, and investigation of a small area (i.e.,

264

metro New Jersey) suggesting that relationships between the physical environment and

265

disability may differ by area of the country. The influence of residential racial segregation on

266

participation was largely subsumed by the individual differences in background characteristics

267

whereas the influence of neighborhood SES was robust in the final models, highlighting the

268

importance of economic resources, both at the individual and neighborhood level, in

269

supporting adjustment to chronic disability from SCI. Finally, the positive relationship between

270

PI and living in a low food access tract potentially reflects the fact that proximity to food is less

271

consequential than individual resources, such as individual SES and vehicle access that give

272

individuals the resources to acquire food. These factors were unavailable for this analysis but

273

will be important to include in future analyses of the relationship between this indicator and

274

disability-related outcomes.

SC

M AN U

TE D

EP

275

RI PT

260

The key limitation of this study is the cross-sectional design which limits our ability to account

277

for the duration of neighborhood exposure, individual migration, and changes to

278

neighborhoods over time. The pooling of data obtained over a decade ensured an adequate

279

sample size and geographic representation but may have obscured changes in the

280

characteristics of residential context over time. Data sparseness at the spatial scale used to

281

operationalize neighborhoods combined with the unique enrollment and follow-up structure of

AC C

276

Page 16 of 25

ACCEPTED MANUSCRIPT

the SCIMS database precluded the use of multilevel analytic techniques to obtain estimates of

283

tract-level variation in participation by race as well as the testing of cross-level interactions to

284

attempt to disentangle the interrelationships between individual and neighborhood

285

characteristics. The inclusion of an aggregate “Other” race category and a combined Hispanic

286

group for analysis was less than ideal. However, prior work in health disparities in SCI is largely

287

limited to investigating White and non-White differences, thus obscuring the diversity in the

288

SCIMS database and in the SCI population as a whole. Bias due to unmeasured and therefore

289

omitted variables such as individual income, health insurance status, and household vehicle

290

ownership may also result in biased parameter estimates for neighborhood characteristics.

291

However, this study examined a comprehensive group of neighborhood characteristics

292

documented to differ by race. Analytically accounting for neighborhood differences is likely to

293

reduce bias in estimating race disparities by eliminating an important source of confounding.

294

TE D

M AN U

SC

RI PT

282

Conclusions

296

Neighborhood characteristics may be critical in understanding race disparities in community

297

outcomes following SCI. It is important to identify barriers to community reintegration

298

following SCI that may result in inequalities in health, disability, and quality of life. This is

299

especially important among historically disadvantaged and marginalized groups of people

300

residing in areas with adverse conditions have less personal and economic resources to

301

overcome environmental barriers14 and are in greater need of interventions. Neighborhood

302

differences are modifiable. Research focused on understanding the role of residential context in

AC C

EP

295

Page 17 of 25

ACCEPTED MANUSCRIPT

303

the experience of disability will allow us to address persistent inequalities in health by

304

improving the environment with informed public policy.

AC C

EP

TE D

M AN U

SC

RI PT

305

Page 18 of 25

ACCEPTED MANUSCRIPT

306

Suppliers:

307

StataCorp. Stata SE 14.0 for Windows (64-bit x86-64). (software).2014

308

Geospatial Modelling Environment (Version 0.7.2.0). (software). 2013

AC C

EP

TE D

M AN U

SC

RI PT

309

Page 19 of 25

ACCEPTED MANUSCRIPT

References

311 312 313

1.

Goossens D, Dousse M, Ventura M, Fattal C. Chronic neuropathic pain in spinal cord injury patients: What is the impact of social and environmental factors on care management? Ann. Phys. Rehabil. Med. 2009;52:173–9.

314 315 316

2.

Cardenas DD, Bryce TN, Shem K, Richards JS, Elhefni H. Gender and minority differences in the pain experience of people with spinal cord injury. Arch. Phys. Med. Rehabil. 2004;85:1774–81.

317 318

3.

Krause JS, Kemp B, Coker J. Depression after spinal cord injury: Relation to gender, ethnicity, aging, and socioeconomic. Arch. Phys. Med. Rehabil. 2000;81:1099–109.

319 320 321

4.

Krause JS, Broderick LE, Saladin LK, Broyles J. Racial disparities in health outcomes after spinal cord injury: mediating effects of education and income. J. Spinal Cord Med. 2006;29:17–25.

322 323

5.

Krause JS, Broderick L. Outcomes after spinal cord injury: Comparisons as a function of gender and race and ethnicity. Arch. Phys. Med. Rehabil. 2004;85:355–62.

324 325 326

6.

Krause JS, Saladin LK, Adkins RH. Disparities in subjective well-being, participation, and health after spinal cord injury: A 6-year longitudinal study. NeuroRehabilitation. 2009;24:47–56.

327 328 329 330

7.

Hunt PC, Boninger ML, Cooper R a., Zafonte RD, Fitzgerald SG, Schmeler MR. Demographic and socioeconomic factors associated with disparity in wheelchair customizability among people with traumatic spinal cord injury. Arch. Phys. Med. Rehabil. 2004;85:1859–64.

331 332

8.

Meade M a., Lewis A, Jackson MN, Hess DW. Race, employment, and spinal cord injury. Arch. Phys. Med. Rehabil. 2004;85:1782–92.

333 334 335

9.

Arango-Lasprilla JC, Ketchum JM, Francis K, Lewis A, Premuda P, Wehman P, et al. Race, ethnicity, and employment outcomes 1, 5, 10 years after spinal cord injury: a longitudinal analysis. PM&R. 2010;2:901–10.

336 337

10.

Pflaum C, McCollister G, Strauss DJ, Shavelle RM, DeVivo MJ. Worklife after traumatic spinal cord injury. J. Spinal Cord Med. 2006;29:377–86.

338 339

11.

340 341 342

12.

343 344

13.

Williams DR, Collins C. Racial residential segregation: a fundamental cause of racial disparities in health. Public Health Rep. 2001;116:404–16.

345 346

14.

Diez Roux A V, Mair C. Neighborhoods and health. Ann. N. Y. Acad. Sci. 2010;1186:125– 45.

AC C

EP

TE D

M AN U

SC

RI PT

310

Lidal IB, Huynh TK, Biering-Sørensen F. Return to work following spinal cord injury: a review. Disabil. Rehabil. 2007;29:1341–75. Saunders LL, Krause JS, Acuna J. Association of race, socioeconomic status, and health care access with pressure ulcers after spinal cord injury. Arch. Phys. Med. Rehabil. 2012;93:972–7.

Page 20 of 25

ACCEPTED MANUSCRIPT

15.

Massey DS, Tannen J. A Research Note on Trends in Black Hypersegregation. Demography. 2015;52:1025–34.

349 350 351

16.

LaVeist T, Thorpe R, Bowen-Reid T, Jackson J, Gary T, Gaskin D, et al. Exploring health disparities in integrated communities: Overview of the EHDIC study. J. Urban Heal. 2008;85:11–21.

352 353 354

17.

LaVeist T, Pollack K, Thorpe R, Fesahazion R, Gaskin D. Place, not race: Disparities dissipate in Southwest Baltimore when blacks and whites live under similar conditions. Health Aff. 2011;30:1880–7.

355 356 357 358

18.

Thorpe RJ, Bowie JV, Smolen JR, Bell CN, Jenkins ML, Jackson J, et al. Racial disparities in hypertension awareness and management: Are there differences among African Americans and Whites living in similar social and healthcare resource environments? Ethn. Dis. 2014;24:269–75.

359 360 361

19.

LaVeist T a., Thorpe RJ, Galarraga JE, Bower KM, Gary-Webb TL. Environmental and socioeconomic factors as contributors to racial disparities in diabetes prevalence. J. Gen. Intern. Med. 2009;24:1144–8.

362 363 364

20.

LaVeist T a., Thorpe RJ, Mance G a., Jackson J, William C, Richardson NF. Overcoming confounding of race with socio-economic status and segregation to explore race disparities in smoking. Addiction. 2007;102:65–70.

365 366

21.

Bleich SN, Thorpe RJ, Sharif-Harris H, Fesahazion R, LaViest T. Social context explains race disparities in obesity among women. J. Epidemiol. Community Heal. 2010;64:465–9.

367 368 369

22.

Kershaw KN, Diez Roux A V., Burgard S a., Lisabeth LD, Mujahid MS, Schulz AJ. Metropolitan-level racial residential segregation and black-white disparities in hypertension. Am. J. Epidemiol. 2011;174:537–45.

370 371 372

23.

Gaskin DJ, Thorpe RJ, McGinty EE, Bower K, Rohde C, Young JH, et al. Disparities in Diabetes: The Nexus of Race, Poverty, and Place. Am. J. Public Health. 2014;104:2147– 55.

373 374 375

24.

Piccolo RS, Duncan DT, Pearce N, McKinlay JB. The role of neighborhood characteristics in racial/ethnic disparities in type 2 diabetes: Results from the Boston Area Community Health (BACH) Survey. Soc. Sci. Med. 2015;130:79–90.

376 377

25.

378 379

26.

380 381 382

27.

Whiteneck G, Meade MA, Dijkers M, Tate DG, Bushnik T, Forchheimer MB. Environmental factors and their role in participation and life satisfaction after spinal cord injury. Arch. Phys. Med. Rehabil. 2004;85:1793–803.

383 384

28.

Clarke P, Ailshire JA, Bader M, Morenoff JD, House JS. Mobility Disability and the Urban Built Environment. Am. J. Epidemiol. 2008;168:506–13.

AC C

EP

TE D

M AN U

SC

RI PT

347 348

Robert S a, Ruel E. Racial Segregation and Health Disparities Between Black and White Older Adults. J. Gerontol. 2006;61B4:S203–11. Rosso AL, Auchincloss AH, Michael YL. The Urban Built Environment and Mobility in Older Adults: A Comprehensive Review. J. Aging Res. 2011;2011:1–10.

Page 21 of 25

ACCEPTED MANUSCRIPT

29.

Clarke PJ, Ailshire JA, Nieuwenhuijsen ER, de Kleijn – de Vrankrijker MW. Participation among adults with disability: the role of the urban environment. Soc. Sci. Med. 2011;72:1674–84.

388 389 390

30.

Botticello AL, Rohrbach T, Cobbold N. Disability and the built environment: an investigation of community and neighborhood land uses and participation for physically impaired adults. Ann. Epidemiol. 2014;24:545–50.

391 392 393

31.

Botticello AL, Rohrbach T, Cobbold N. Differences in the Community Built Environment Influence Poor Perceived Health among Persons with Spinal Cord Injury. Arch. Phys. Med. Rehabil. 2015;96:1583–90.

394 395

32.

Balfour JL, Kaplan GA. Neighborhood environment and loss of physical function in older adults: Evidence from the Alameda County study. Am. J. Epidemiol. 2002;155:9.

396 397 398

33.

Beard JR, Blaney S, Cerda M, Frye V, Lovasi GS, Ompad D, et al. Neighborhood Characteristics and Disability in Older Adults. Journals Gerontol. Ser. B Psychol. Sci. Soc. Sci. 2009;64B:252–7.

399 400

34.

Freedman VA, Grafova IB, Schoeni RF, Rogowski J. Neighborhoods and disability in later life. Soc. Sci. Med. 2008;66:2253–67.

401 402 403

35.

Liang H, Tomey K, Chen D, Savar NL, Rimmer JH, Braunschweig CL. Objective measures of neighborhood environment and self-reported physical activity in spinal cord injured men. Arch. Phys. Med. Rehabil. 2008;89:1468–73.

404 405 406

36.

Roach MJ. Community Social Structure as an Indicator of Social Integration and Its Effect on Quality of Life for Persons with a Spinal Cord Injury. Top. Spinal Cord Inj. Rehabil. 2002;7:101–11.

407 408 409

37.

Botticello AL, Chen Y, Cao Y, Tulsky DS. Do communities matter after rehabilitation? The effect of socioeconomic and urban stratification on well-being after spinal cord injury. Arch. Phys. Med. Rehabil. 2011;92:464–71.

410 411 412 413

38.

Botticello AL, Chen Y, Tulsky DS. Geographic variation in participation for physically disabled adults : The contribution of area economic factors to employment after spinal cord injury Extra-Individual Factors Functional Limitations Intra-Individual Factors. Soc. Sci. Med. 2012;75.

414 415 416 417

39.

418 419

40.

Taylor SE, Repetti RL, Seeman TE. Health psychology:What is an unhealthy environment and how does it get under the skin? Annu. Rev. Psychol. 1997;48:37.

420 421

41.

U.S. Census Bureau. 2010 Geographic Terms and Concepts [Internet]. 2015;Available from: https://www.census.gov/geo/reference/gtc/gtc_ct.html

422

42.

Minnesota Population Center. National Historical Geographic Information System:

AC C

EP

TE D

M AN U

SC

RI PT

385 386 387

Magasi S, Wong A, Gray DB, Hammel J, Baum C, Wang CC, et al. Theoretical foundations for the measurement of environmental factors and their impact on participation among people with disabilities. Arch. Phys. Med. Rehabil. [Internet]. 2015;96:569–77. Available from: http://dx.doi.org/10.1016/j.apmr.2014.12.002

Page 22 of 25

ACCEPTED MANUSCRIPT

Version 2.0 [Internet]. 2015 [cited 2015 Jan 1];Available from: http://www.nhgis.org

423

43.

Economic Research Service (ERS), (USDA) USD of A. Food Access Research Atlas. 2011;Available from: http://www.ers.usda.gov/data-products/food-access-researchatlas.aspx.

427 428

44.

U.S. Census Bureau. American Community Survey [Internet]. 2009;Available from: http://www.nhgis.org

429 430

45.

U.S. Census Bureau. American Community Survey [Internet]. 2013;Available from: http://www.nhgis.org

431 432 433

46.

Whiteneck GG, Charlifule SW, Gerhart KA, Overholser JD, Richardson GN. Quantifying Handicap: A New Measure of Long-Term Rehabilitation Outcomes. Arch. Phys. Med. Rehabil. 1992;73:519–26.

434 435 436 437 438 439 440 441

47.

Hall KM, Dijkers M, Whiteneck G, Brooks C a, Krause JS, K.M. H, et al. The Craig Handicap Assessment and Reporting Technique (CHART): metric properties and scoring. Top. Spinal Cord Inj. Rehabil. [Internet]. 1998;4:16–30. Available from: http://ovidsp.ovid.com/ovidweb.cgi?T=JS&PAGE=reference&D=emed4&NEWS=N&AN=1 999026798\nhttp://search.ebscohost.com/login.aspx?direct=true&db=cin20&AN=19990 27499&site=ehostlive\nhttp://search.ebscohost.com/login.aspx?direct=true&db=cin20&AN=1999070141& site=

442 443 444

48.

Hall KM, Bushnik T, Lakisic-Kazazic B, Wright J, Cantagallo A. Assessing traumatic brain injury outcome measures for long-term follow-up of community-based individuals. Arch. Phys. Med. Rehabil. 2001;82:367–74.

445 446

49.

American Spinal Cord Injury Association. International Standards for Neurological Classification of Spinal Cord Injury. 2015;

447 448

50.

Kroenke K, Spitzer RL, Williams JB. The Patient Health Questionnaire-2: Validity of a TwoItem Depression Screener. Med. Care. 2003;41:184–1292.

449 450 451

51.

Hamilton B, Granger C, Sherwin F, Zielezny M, Tashman J. A uniform national data system for medical rehabilitation. In: Fuhrer M, editor. Rehabilitation outcomes: analysis and measurement. Baltimore, MD: Paul H. Brooks Publishing Company; 1987. p. 137–47.

452 453

52.

454 455 456

53.

457 458 459

54.

Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J. Pers. Soc. Psychol. 1986;51:1173–82.

460

55.

Aneshensel CS. Theory-Based Data Analysis for the Social Sciences. Second. Los Angeles,

AC C

EP

TE D

M AN U

SC

RI PT

424 425 426

Anderson BJR, Hardy EE, Roach JT, Witmer RE. A Land Use And Land Cover Classification System For Use With Remote Sensor Data. 2001;2001. Diez Roux A V, Merkin SS, Arnett D, Chambless L, Massing M, Nieto FJ, et al. Neighborhood of residence and incidence of coronary heart disease. N. Engl. J. Med. 2001;345:8.

Page 23 of 25

ACCEPTED MANUSCRIPT

CA: Sage; 2013.

461 462

56.

Buis M. NIH Public Access. Stata J. 2010;10:11–29.

463 464

57.

Glaeser L, Vigdor JL. Racial Segregation in the 2000 Census: Promising News in the 2000 Census: Promising News. Brookings insitute; Surv. Ser. 2001;1–16.

AC C

EP

TE D

M AN U

SC

RI PT

465

Page 24 of 25

ACCEPTED MANUSCRIPT

Table 1. Sample characteristics (2000-2014) by race/ethnicity (Mean (SD) or %) White

Black

(N = 6,892 )

(n = 4,948 )

RI PT

Total

51.4

% Full Mobility (vs. restricted) *

38.0

% Full Occupation (vs. restricted) * % Full Social Integration (vs. restricted) *

% Male (vs. female) * Education * % Less than high school % High school diploma

(n= 133)

44.7

53.4

43.2

23.5

26.2

33.6

36.6

40.8

23.9

29.3

35.9

63.3

69.3

44.5

56.0

59.7

40.9 (13.9)

46.3 (15.4)

45.7 (14.6)

EP

Age (years) *

AC C

Demographic characteristics

(n = 413 )

41.3

TE D

Participants

Other

54.8

M AN U

% Full Physical Independence (vs. restricted)*

SC

Participation Domains (full versus restricted)

(n = 1,398)

Hispanic

46.9 (14.4)

42.4 (14.4)

78.9

77.8

82.9

80.9

74.4

12.4

7.7

23.8

30.0

13.5

52.9

51.3

59.7

52.5

41.4

Page 1 of 17

% Some college or more

41.0

16.5

17.4

45.1

% Currently employed (vs. unemployed) *

25.2

30.3

9.5

15.0

31.6

% Currently married (vs. unmarried) *

36.1

40.6

20.3

32.9

46.6

% Complete tetraplegia

19.1

18.4

20.6

15.8

% Incomplete tetraplegia

32.2

33.5

29.8

26.2

33.1

% Complete paraplegia

29.2

27.4

34.7

31.7

29.3

% Incomplete paraplegia

19.6

20.1

17.2

21.6

21.8

14.1 (9.2)

12.9 (8.1)

13.5 (9.4)

SCI-related characteristics

19.2

TE D

M AN U

Injury Severity *

RI PT

34.7

SC

ACCEPTED MANUSCRIPT

Length of time injured (years) *

16.2 (10.4) 74.1

73.3

76.3

78.6

68.2

% Perceived poor health (vs. good health) *

23.4

21.0

29.5

28.6

30.5

% Depressive (vs. asymptomatic)

17.4

16.7

19.2

19.0

20.0

EP

% Wheelchair user (vs. other/no device)†

17.1 (10.7)

FIM Score

AC C

Health-related characteristics

4.9 (2.1)

5.0 (2.1)

4.8 (1.9)

4.9 (1.9)

5.0 (1.8)

Page 2 of 17

ACCEPTED MANUSCRIPT

Residential area controls

25.2

22.4

35.6

24.0

23.3

% Midwest

15.9

16.7

13.5

14.5

14.3

% South

27.6

23.8

44.3

21.3

12.8

% West

31.4

37.1

6.7

40.2

49.6

72.0

89.7

85.8

93.6

33.3

18.4

77.2

57.9

48.1

33.4

38.2

17.7

28.1

39.1

33.3

43.4

5.1

14.0

12.8

Developed open space (% large vs. small) *

53.1

53.9

53.1

42.1

56.4

Green space (% large vs .small) *

50.3

58.2

29.3

32.5

33.8

M AN U

% Urban (vs. rural) *

77.0

Social Environment

TE D

Racial Composition (tertiles) * % < 60 White

EP

% 60-85 White

AC C

% > 85 White Built Environment

SC

% Northeast

RI PT

Regional location *

Page 3 of 17

ACCEPTED MANUSCRIPT

43.0

45.6

35.8

36.4

46.4

Access to vehicle resources (% low vs .high) *

27.5

21.7

46.7

31.0

22.4

Economic Environment 0.0 (3.5)

0.7 (3.4)

-2.1 (2.8)

-1.4 (3.2)

0.9 (4.0)

* p < 0.001; † p < 0.01;

AC C

EP

TE D

n and column percent presented for categorical variables

M AN U

SC

SES Advantage Index*

RI PT

Access to food resources (% low vs. high) *

Page 4 of 17

ACCEPTED MANUSCRIPT

Table 2. Unadjusted Odds Ratios of Full Participation by Neighborhood Characteristics

OR

95% CI

OR

95% CI

Region § Midwest

1.08

0.92 - 1.26

1.28†

1.09 -

South

1.50*

1.30 - 1.73

1.11

0.96 -

Social Integration

OR

95% CI

0.95 - 1.32

1.60* 1.36 - 1.89

0.92

0.79 - 1.06

1.11

2.01*

1.75 - 2.30

1.99* 1.73 - 2.30

0.95

0.83 - 1.07

0.92

1.43*

1.26 - 1.63

1.70* 1.50 - 1.92

1.12

M AN U

1.51

Occupation

RI PT

Model

Mobility

SC

Physical Independence

OR

95% CI

0.97 - 1.27

1.28

1.83*

1.59 - 2.10

1.69*

1.47 -

TE D

West

1.94

Racial Composition § % 60-85 White

0.78 - 0.99

1.12

EP

0.88‡

0.99 -

0.81 - 1.05

1.27

AC C

Urban (vs. Rural)

1.32*

1.16 - 1.49

1.54*

1.36 -

Page 5 of 17

ACCEPTED MANUSCRIPT

1.75 1.45*

1.28 - 1.64

1.60*

1.41 -

Large proportion

1.06

0.95 - 1.17

1.13‡

developed open

1.02 -

1.05

1.25

SC

1.82

1.23†

1.11 - 1.36

1.05

green space (vs.

1.11 - 1.37

1.10

1.20†

1.08 -

1.05

1.02†

EP

0.80 - 1.01

0.70*

AC C

0.90

1.00 - 1.03

1.09*

0.62 -

2.04* 1.80 - 2.32

0.95 - 1.16

1.14‡ 1.03 - 1.26

0.99 - 1.22

1.14 - 1.40 1.27*

0.94 - 1.17

1.34

(vs. high) SES advantage

TE D

1.23†

(vs. high) Low vehicle access

0.95 -

1.34 - 1.73

1.16

small) Low food access

M AN U

space (vs. small) Large proportion

1.52*

RI PT

% > 85 White

1.10 – 1.37 1.23*

0.73*

0.65 - 0.83

0.59*

0.53 - 0.67

1.05 - 1.09

1.12*

1.10 - 1.14

0.79 1.08 -

1.07*

1.11

Page 6 of 17

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

*p < 0.001; †p < 0.01; ‡p < 0.05; § Reference category: Northeast; § Reference category: % < 60 White.

Page 7 of 17

ACCEPTED MANUSCRIPT

Table 3a. Mediation of race disparities in physical independence (PI) participation by neighborhood characteristics

OR (SE)

Model 2

95% CI

OR

95% CI

SC

Race/ethnicity §

Model 3

RI PT

Model 1

OR

95% CI

0.57 (0.04)*

0.50-0.65

0.60 (0.06)*

0.49-0.72

0.61 (0.06)*

0.50-0.73

Hispanic

0.65 (0.07)*

0.52-0.81

0.58 (0.10)†

0.42-0.79

0.59 (0.10)†

0.43-0.81

Other

1.00 (0.20)

0.68-1.47

0.79 (0.23)

0.45-1.41

0.79 (0.23)

0.44-1.41

0.99 (0.00)*

0.98-0.99

0.99 (0.00)*

0.98-0.99

2.29 (0.24)*

1.87-2.81

2.28 (0.24)*

1.86-2.79

1.27 (0.11)†

1.06-1.51

1.25 (0.11)‡

1.05-1.50

1.05 (0.17)

0.77-1.44

1.05 (0.17)

0.76-1.43

1.53 (0.18)*

1.22-1.93

1.53 (0.18)*

1.22-1.92

Married Injury Severity ‖ Complete tetraplegia Incomplete paraplegia

EP

Paid Work

AC C

Age

TE D

Covariates

M AN U

Black

Page 8 of 17

ACCEPTED MANUSCRIPT

1.62-2.50

2.02 (0.22)*

1.62-2.51

Injury Duration

1.05 (0.00)*

1.04-1.06

1.05 (0.00)*

1.04-1.06

Wheelchair

0.38 (0.04)*

0.31-0.48

0.38 (0.04)*

0.31-0.48

FIM Total

3.71 (0.16)*

3.40-4.05

3.71 (0.16)*

3.40-4.05

M AN U

Region ¶ Midwest South West

1.23 (0.15)

0.97-1.55

1.23 (0.15)

0.97-1.55

1.47 (0.15)*

1.20-1.80

1.42 (0.15)†

1.16-1.74

2.60 (0.31)*

2.06-3.28

2.57 (0.31)*

2.00-3.25

1.24 (0.10)†

1.06-1.46

Χ2 =78.96

p <0.001

Low food access (vs. high) Fit statistics

n/a

Χ2 =78.69

p <0.001

AC C

n/a

EP

TE D

Neighborhood characteristics

Hosmer-Lemeshow Chi-Square

RI PT

2.01 (0.22)*

SC

Complete paraplegia

*p < 0.001; †p < 0.01; ‡p < 0.05; § Reference category: NHW; ‖ Reference category: Incomplete tetraplegia; ¶ Reference category: Northeast.

Page 9 of 17

ACCEPTED MANUSCRIPT

Table 3b. Mediation of race disparities in mobility participation by neighborhood characteristics

Mobility

OR (SE)

Model 2

95% CI

OR (SE)

95% CI

SC

Race/ethnicity §

Model 3

RI PT

Model 1

OR (SE)

95% CI

0.41 (0.03)*

0.35-0.47

0.64 (0.06)*

0.54-0.77

0.66 (0.06)*

0.55-0.80

Hispanic

0.48 (0.06)*

0.38-0.60

0.60 (0.09)*

0.45-0.79

0.61 (0.09)†

0.46-0.82

Other

0.76 (0.15)

0.52-1.12

0.59 (0.15)‡

0.36-0.97

0.56 (0.14) ‡

0.34-0.93

0.97 (0.00)*

0.96-0.97

0.97 (0.00)*

0.96-0.97

0.43 (0.53)*

0.33-0.55

0.46 (0.06)*

0.36-0.59

0.63 (0.05)*

0.54-0.73

0.67 (0.05)*

0.58-0.78

1.32 (0.11)†

1.12-1.55

1.33 (0.11)†

1.13-1.57

5.63 (0.45)*

4.81-6.58

5.46 (0.44)*

4.66-6.39

1.59 (0.12)*

1.38-1.84

1.59 (0.12)*

1.38-1.84

Covariates

High school graduate Male Paid Work Married

EP

Less than high school

AC C

Education ‖

TE D

Age

M AN U

Black

Page 10 of 17

ACCEPTED MANUSCRIPT

Injury Severity ¶ 1.06 (0.12)

Incomplete paraplegia

0.78 (0.07)†

Complete paraplegia

0.81 (0.08)‡

0.84-1.33

1.07 (0.13)

0.85-1.35

0.64-0.94

0.79 (0.08)‡

0.65-0.96

0.67-0.97

0.81 (0.08)‡

0.67-0.98

RI PT

Complete tetraplegia

1.02 (0.00)*

1.01-1.03

1.02 (0.00)*

1.01-1.03

Wheelchair

0.69 (0.07)*

0.58-0.84

0.70 (0.07)*

0.58- 0.85

0.45 (0.04)*

0.38-0.53

0.45 (0.04)*

0.38- 0.54

0.56 (0.05)*

0.46-0.67

0.56 (0.05)*

0.47- 0.68

1.64 (0.05)*

1.55-1.73

1.65 (0.05)*

1.56-1.74

1.17 (0.12)

0.95-1.44

1.23 (0.13)

1.00-1.52

1.05 (0.10)

0.88-1.25

1.23 (0.12) ‡

1.02-1.49

1.25 (0.12)‡

1.04-1.50

1.30 (0.12)†

1.08-1.56

0.83 (0.06)‡

0.73-0.96

M AN U

SC

Injury Duration

Poor Health Depress Symptomatic

South West Neighborhood characteristics Large proportion green space

EP

Midwest

AC C

Region #

TE D

FIM Total

Page 11 of 17

ACCEPTED MANUSCRIPT

(vs. low)

Fit statistics n/a

Χ2 = 9.82

n/a

p =0.28

1.05 (0.01)*

1.03-1.07

Χ2 = 7.25

p =0.51

SC

Hosmer-Lemeshow Chi-Square

RI PT

Socioeconomic advantage

AC C

EP

TE D

Incomplete tetraplegia; # Reference category: Northeast.

M AN U

*p < 0.001; †p < 0.01; ‡p < 0.05; § Reference category: NHW; ‖ Reference category: some college or more; ¶ Reference category:

Page 12 of 17

ACCEPTED MANUSCRIPT

Table 3c. Mediation of race disparities in occupation participation by neighborhood characteristics

OR (SE)

Model 2

95% CI

OR (SE)

95% CI

SC

Race/ethnicity §

Model 3

RI PT

Model 1

OR (SE)

95% CI

0.46 (0.03)*

0.40-0.53

0.70 (0.06)*

0.60-0.83

0.75 (0.06)†

0.63-0.88

Hispanic

0.62 (0.07)*

0.49-0.77

0.70 (0.09)†

0.54-0.90

0.73 (0.10)‡

0.56-0.94

Other

0.83 (0.16)

0.56-1.21

0.75 (0.17)

0.49-1.16

0.75 (0.17)

0.49-1.16

0.95 (0.00)*

0.95-0.96

0.95 (0.00)*

0.95-0.96

0.36 (0.04)*

0.29-0.45

0.38 (0.04)*

0.30-0.47

0.50 (0.03)*

0.44-0.56

0.51 (0.03)*

0.45-0.58

0.67 (0.05)*

0.58-0.77

0.67 (0.05)*

0.58-0.77

2.08 (0.14)*

1.82-2.37

2.05 (0.14)*

1.80-2.34

Covariates

High school graduate Male Married

EP

Less than high school

AC C

Education ‖

TE D

Age

M AN U

Black

Injury Severity ¶

Page 13 of 17

ACCEPTED MANUSCRIPT

1.28 (0.13)‡

1.05-1.57

1.29 (0.13)‡

1.05-1.58

Incomplete paraplegia

1.06 (0.09)

0.89-1.26

1.07 (0.10)

0.90-1.27

Complete paraplegia

1.17 (0.10)

0.98-1.39

1.17 (0.10)

0.99-1.39

1.04-1.05

1.04 (0.00)*

1.04-1.05

RI PT

Complete tetraplegia

1.04 (0.00)*

Wheelchair (versus non)

0.76 (0.07)†

0.64-0.90

0.76 (0.07)†

0.64-0.90

Poor Health (versus good

0.72 (0.06)*

0.62-0.84

0.72 (0.06)*

0.62-0.85

0.65 (0.06)*

0.55-0.77

0.65 (0.06)*

0.55-0.77

1.39 (0.03)*

1.32-1.46

1.39 (0.03)*

1.33-1.46

0.92 (0.09)

0.76-1.11

0.95 (0.09)

0.78-1.15

0.68 (0.06)*

0.58-0.81

0.72 (0.06)*

0.61-0.86

1.48 (0.13)*

1.26-1.75

1.51 (0.13)*

1.28-1.79

M AN U

SC

Injury Duration

health) Depression (versus

Midwest South West Neighborhood characteristics

EP

Region #

AC C

FIM Total

TE D

asymptomatic)

Page 14 of 17

ACCEPTED MANUSCRIPT

Socioeconomic advantage

1.03 (0.01)†

1.01-1.05

Χ2 =11.02

p =0.20

Hosmer-Lemeshow test

n/a

Χ2 =7.87

n/a

RI PT

Fit statistics p =0.45

M AN U

AC C

EP

TE D

Incomplete tetraplegia; # Reference category: Northeast.

SC

*p < 0.001; †p < 0.01; ‡p < 0.05; § Reference category: NHW; ‖ Reference category: some college or more; ¶ Reference category:

Page 15 of 17

ACCEPTED MANUSCRIPT

Table 3d. Mediation of race disparities in social integration (SI) participation by neighborhood characteristics

OR (SE)

Model 2

95% CI

OR (SE)

95% CI

SC

Race/ethnicity §

Model 3

RI PT

Model 1

OR (SE)

95% CI

0.35 (0.02)*

0.31-0.40

0.59 (0.04)*

0.51-0.68

0.65 (0.05)*

0.55-0.76

Hispanic

0.58 (0.06)*

0.47-0.72

0.71 (0.09)†

0.56-0.91

0.77 (0.10) ‡

0.60-0.98

Other

0.72 (0.14)

0.50-1.07

0.58 (0.13)‡

0.37-0.89

0.57 (0.13) ‡

0.37-0.89

0.97 (0.00)*

0.97-0.97

0.97 (0.00)*

0.97-0.97

0.36 (0.04)*

0.29-0.44

0.39 (0.04)*

0.31-0.47

0.58 (0.04)*

0.51-0.67

0.61 (0.04)*

0.53-0.70

2.38 (0.21)*

2.01-2.83

2.32 (0.20)*

1.96-2.76

4.69 (0.35)*

4.06-5.43

4.60 (0.34)*

3.97-5.32

0.62 (0.04)*

0.54-0.72

0.63 (0.04)*

0.54-0.72

Covariates

High school graduate Paid Work Married Poor Health

EP

Less than high school

AC C

Education ‖

TE D

Age

M AN U

Black

Page 16 of 17

ACCEPTED MANUSCRIPT

0.56 (0.04)*

0.48-0.66

0.57 (0.04)*

0.49-0.66

FIM Total

1.04 (0.02)‡

1.01-1.07

1.05 (0.02)†

1.01-1.08

1.23-1.78

1.56 (0.15)*

1.29-1.89

Region ¶

RI PT

Depression Symptomatic

1.48 (0.14)*

South

1.12 (0.09)

0.95-1.30

1.23 (0.10)‡

1.05-1.45

West

1.67 (0.14)*

1.42-1.97

1.73 (0.15)*

1.47-2.05

1.05 (0.01)*

1.03-1.07

Χ2 =5.31

p =0.72

M AN U

SC

Midwest

Neighborhood characteristics Socioeconomic advantage

n/a

n/a

Χ2 =9.57

p =0.30

EP

Hosmer-Lemeshow test

TE D

Fit statistics

Northeast.

AC C

*p < 0.001; †p < 0.01; ‡p < 0.05; § Reference category: NHW; ‖ Reference category: some college or more; ¶ Reference category:

Page 17 of 17