Volume 126, Issue 12 p. 2849-2858
Original Article
Free Access

Nativity, ethnic enclave residence, and breast cancer survival among Latinas: Variations between California and Texas

Salma Shariff-Marco PhD, MPH

Corresponding Author

Salma Shariff-Marco PhD, MPH

Greater Bay Area Cancer Registry, San Francisco, California

Department of Epidemiology and Biostatistics, University of California at San Francisco, San Francisco, California

Helen Diller Family Comprehensive Cancer Center, University of California at San Francisco, San Francisco, California

Corresponding Author: Salma Shariff-Marco, PhD, MPH, Department of Epidemiology and Biostatistics, University of California at San Francisco, 550 16th St, San Francisco, CA 94158 ([email protected]).Search for more papers by this author
Scarlett Lin Gomez PhD, MPH

Scarlett Lin Gomez PhD, MPH

Greater Bay Area Cancer Registry, San Francisco, California

Department of Epidemiology and Biostatistics, University of California at San Francisco, San Francisco, California

Helen Diller Family Comprehensive Cancer Center, University of California at San Francisco, San Francisco, California

Search for more papers by this author
Alison J. Canchola MS

Alison J. Canchola MS

Greater Bay Area Cancer Registry, San Francisco, California

Department of Epidemiology and Biostatistics, University of California at San Francisco, San Francisco, California

Search for more papers by this author
Hannah Fullington MPH

Hannah Fullington MPH

Department of Population and Data Sciences, University of Texas Southwestern Medical Center, Dallas, Texas

Search for more papers by this author
Amy E. Hughes PhD

Amy E. Hughes PhD

Department of Population and Data Sciences, University of Texas Southwestern Medical Center, Dallas, Texas

Search for more papers by this author
Hong Zhu PhD

Hong Zhu PhD

Department of Population and Data Sciences, University of Texas Southwestern Medical Center, Dallas, Texas

Harold C. Simmons Cancer Center, Dallas, Texas

Search for more papers by this author
Sandi L. Pruitt PhD, MPH

Sandi L. Pruitt PhD, MPH

Department of Population and Data Sciences, University of Texas Southwestern Medical Center, Dallas, Texas

Harold C. Simmons Cancer Center, Dallas, Texas

Search for more papers by this author
First published: 17 March 2020
Citations: 17
Cancer data were provided by the Texas Cancer Registry, Cancer Epidemiology and Surveillance Branch, Texas Department of State Health Services (https://www.dshs.texas.gov/tcr/).

Abstract

Background

Among Latinas with breast cancer, residence in an ethnic enclave may be associated with survival. However, findings from prior studies are inconsistent.

Methods

The authors conducted parallel analyses of California and Texas cancer registry data for adult (aged ≥18 years) Latinas who were diagnosed with invasive breast cancer from 1996 to 2005, with follow-up through 2014. Existing indices applied to tract-level 2000 US Census data were used to measure Latinx enclaves and neighborhood socioeconomic status (nSES). Multivariable Cox proportional hazard models were fit for all-cause and breast cancer–specific survival adjusted for year of diagnosis, patient age, nativity (with multiple imputation), tumor stage, histology, grade, size, and clustering by census tract.

Results

Among 38,858 Latinas, the majority (61.3% in California and 70.5% in Texas) lived in enclaves. In fully adjusted models for both states, foreign-born women were found to be more likely to die of breast cancer and all causes when compared with US-born women. Living in enclaves and in neighborhoods with higher SES were found to be independently associated with improved survival from both causes. When combined into a 4-level variable, those in low nSES nonenclaves had worse survival for both causes compared with those living in low nSES enclaves and, in the all-cause but not breast cancer–specific models, those in high nSES neighborhoods, regardless of enclave status, had improved survival from all causes.

Conclusions

Applying the same methods across 2 states eliminated previously published inconsistent associations between enclave residence and breast cancer survival. Future studies should identify specific protective effects of enclave residence to inform interventions.

Introduction

Among US Latinas, breast cancer is the most commonly diagnosed cancer and is the leading cause of cancer death.1, 2 The majority of studies published to date, but not all, have demonstrated that, once diagnosed with breast cancer, Latinas do not survive as long as non-Latina white women.3-6 Although there is growing recognition that neighborhoods play some role in outcomes across the cancer continuum, to our knowledge the extent to which the features of a residential neighborhood may influence survival among Latinas with breast cancer is unclear.

Many Latinxs live in ethnic enclaves, defined as culturally distinct neighborhoods with high concentrations of individuals of the same ethnic origin, high linguistic isolation, a large share of recent immigrants, and ethnic-specific businesses and resources. Ethnic enclaves are hypothesized to contribute to outcomes across the cancer continuum through multiple pathways, some positively and some negatively.7, 8 Co-ethnic residents within an enclave often maintain cultural norms and behaviors (eg, diet and physical activity, social support) that may be health-promoting. Enclaves may facilitate communication and information sharing due to greater access to linguistic resources; they also may reduce exposure to discrimination and thus limit the use of unhealthy coping behaviors (eg, smoking and/or drinking) and reduce stress.9-13

In contrast, some features of enclaves could contribute to worse health. For example, neighborhoods with large Latinx and/or foreign-born populations face disproportionately higher poverty.14-16 Residence in neighborhoods with low socioeconomic status (SES) is associated with worse survival among patients with cancer.17, 18 Low neighborhood SES (nSES) may influence unhealthy behaviors and worse health through pathways associated with adverse social, built, and physical environments. For example, low nSES neighborhoods may have high crime and poor safety; greater social isolation; low walkability resulting from high traffic density and poor street conditions; poor food environments with high concentrations of fast food restaurants, tobacco outlets, or liquor stores; and greater proximity to environmental pollutants. Thus, to elucidate the association between ethnic enclave residence and cancer survival, nSES must be considered.19, 20

Prior studies regarding the association between enclave residence and cancer survival have demonstrated mixed results, as underscored in a recent literature review concerning ethnic density and cancer outcomes.21 Five of the reviewed studies examined associations between Latinx ethnic density with the survival of Latinas with breast cancer. Associations varied, with 2 null studies and others documenting both increased (1 study) and decreased (2 studies) survival among Latinas residing within ethnic enclaves compared with Latinas residing in nonenclave areas.19, 22-25 These studies applied various measures of neighborhood ethnic density or ethnic enclave residence and different analytic strategies, including adjustments for patient nativity and nSES. Because all 5 studies were limited to single states or metropolitan areas, it is unclear whether the observed inconsistencies were a result of different analytic methods or true regional differences in enclave effects.

Because nativity often is missing in cancer registry data, it can be imputed using various approaches.23, 25 Perhaps as a result, findings are varied and demonstrate that foreign-born Latinas with breast cancer, compared with US-born Latinas, have worse survival, no difference in survival, or improved survival.19, 22

Given published inconsistencies, an advanced understanding of the role of enclaves can help to inform neighborhood-level interventions designed to improve survival among Latinas. In the current study, we aimed to investigate the independent associations of ethnic enclaves with survival after a diagnosis of breast cancer among Latinas, accounting for patient nativity and nSES. We applied consistent measures and analytic methods to parallel analyses of cancer registry data from California and Texas to compare and contrast the effects of ethnic enclaves.

Materials and Methods

Data

The data for these analyses were obtained from 2 population-based cancer registries: the California Cancer Registry and the Texas Cancer Registry. These 2 states have 2 of the largest Latinx populations in the United States. Both registries collect demographic and clinical data regarding incident cancers diagnosed in the state in accordance with North American Association of Central Cancer Registries standards.

A total of 50,696 Latina adults (those aged ≥18 years) with a first primary breast cancer diagnosed between January 1, 1996, and December 31, 2005, were identified by the California (29,217 patients) and Texas (21,479 patients) cancer registries. The total number of Latinas eligible for analyses from California and Texas were 23,281 and 15,577, respectively (see Fig. 1 for inclusion and/or exclusion criteria). We chose these years to anchor cancer diagnoses on 2000 US Census data for our neighborhood variables and to have sufficient follow-up to accrue enough events and/or deaths, given that the 5-year survival for breast cancer is approximately 90%.

Details are in the caption following the image
Inclusion and exclusion criteria and the final sample from the California (CA) and Texas (TX) cancer registries.

Registry geocodes were used to append 2000 US Census tract-level data to ascertain nSES and Latinx enclave.8 The nSES index used in the current study is a validated and well-established composite measure of 7 SES indicators, including education, occupation, employment, household income, poverty, rent, and house values.5, 6, 26-31 Latinx enclaves were defined using an established multidimensional index of 7 measures (percentage of residents who are Latinx, foreign-born, recent immigrants, and linguistically isolated [general and of those who speak Spanish], and with limited English proficiency [general and of those who speak Spanish]).32 We classified each index into state-specific quintiles. For nSES, quintile 5 (Q5) represents the highest SES neighborhoods, whereas Q1 represent the lowest SES neighborhoods. For ethnic enclaves, Q5 represents the most ethnically distinct neighborhoods whereas Q1 represents the least distinct.

We calculated follow-up as the number of days from diagnosis to either death or December 31, 2014, whichever occurred first. For breast cancer–specific survival, we censored follow-up at the date of death for those individuals dying of another cause, and we excluded those with an unknown cause of death.

Imputation

We imputed missing birthplace data (22% for California and 44% for Texas) to US-born or foreign-born using multiple imputation separately by state. We used maximum likelihood logistic regression to impute nativity using variables available from both states, including age at diagnosis, year of diagnosis, tumor stage, tumor grade, tumor histology, tumor size (continuous variable with an indicator for missing data), reporting source, diagnosis and/or treatment at the reporting facility (vs elsewhere), microscopic tumor confirmation, Hispanic origin, quintile categories of nSES and Latinx enclave and all component continuous census-level measures, time from diagnosis to death or December 31, 2014, and status at the end of the study (alive, died of breast cancer, died of another cause, or unknown cause of death).33 We fit imputation models 20 times, creating 20 data sets of imputed nativity. We excluded those individuals who were missing data for any covariate (except for tumor size or tumor grade) from imputation models (1257 individuals in California and 4272 individuals in Texas). For descriptive analyses, we defined patients who were missing birthplace as foreign-born if nativity was imputed to foreign-born in >10 imputation runs, and US-born otherwise.

Statistical Analysis

We fit multivariable Cox regression models on each of the 20 imputation data sets separately by state and combined regression results across all 20 imputed data sets to estimate hazard ratios and 95% confidence intervals for associations with mortality risk, using the rules developed by Rubin.34 The proportional hazards assumption did not hold for stage and tumor grade. Therefore, stage of disease was included as a stratifying variable in all Cox regressions, allowing baseline hazards to vary by stage. In addition, because stratifying by grade did not meaningfully change hazard ratios for nativity, nSES, or enclave, grade was included as a covariate. Minimally adjusted models included age (continuous) and year of diagnosis (continuous). Fully adjusted models also included histology (ductal, lobular, other, or unknown), grade (1, 2, 3/4, or unknown), tumor size (in cm; continuous, with an indicator variable for other/missing), and census tract clustering (ie, using a sandwich estimator of the covariance structure that accounts for intracluster dependence). In California and Texas, there was a median of 3 cases per census tract with interquartile ranges of 2 to 5 cases and 1 to 6 cases, respectively. We performed Wald tests for trend across quintile categories.

We initially allowed the variables of interest, nSES and Latinx enclave, categorized by quintiles, to be entered into models separately. Given their high correlation as continuous measures (correlation coefficient, −0.76 in California and correlation coefficient, −0.72 in Texas) and an observed statistically significant interaction (California overall survival: P for interaction = .004; and Texas breast cancer–specific survival, P for interaction = .013), we created a 4-level combined variable of nSES (low vs high) and enclave residence (no vs yes). Based on sample distributions, we defined high nSES as the top 3 state-specific quintiles and Latinx enclave as the top 2 state-specific quintiles. We did not observe statistically significant interactions between nativity and nSES nor enclave in either state.

Finally, to facilitate comparison of survival across the multiple independent and joint associations of interest, we calculated 5-year survival probabilities and associated 95% confidence intervals from the fully adjusted Cox models with covariates set to their reference level or mean value and stage entered into the model as a covariate. Survival probability estimates first were normalized using the complementary log-log transformation before combining the results across the 20 multiple imputation runs, and the combined results then were back-transformed.35

Results

Table 1 shows how the percentage of US- and foreign-born Latinas changed after imputation. Table 2 shows patient characteristics. Texas had a higher percentage of cases living in ethnic enclaves than California. In both states, more foreign-born compared with US-born Latinas lived in ethnic enclaves and in low SES neighborhoods.

Table 1. Distribution of Patient Nativity With and Without Multiple Imputation Among Latinas With Breast Cancer Diagnosed in California or Texas, 1996 Through 2005, by State
  US Born Foreign Born Missing Total
No. Row % No. Row % No. Row % No. Row %
CA: No imputation 7876 33.83 10,217 43.89 5188 22.28 23,281 100
CA: With imputation 12,621 54.21 10,660 45.79 23,281 100
TX: No imputation 5965 38.29 2735 17.56 6877 44.15 15,577 100
TX: With imputation 11,897 76.38 3680 23.62 15,577 100
  • Abbreviations: CA, California; TX, Texas.
Table 2. Characteristics of Latinas With Breast Cancer Diagnosed in California or Texas, 1996 Through 2005, by State and Patient Nativity
  Texas California
US Born N = 11,897 No. (Column %) Foreign Born N = 3680 No. (Column %) All N = 15,577 No. (Column %) US Born N = 12,621 No. (Column %) Foreign Born N = 10,660 No. (Column %) All N = 23,281 No. (Column %)
Age at diagnosis, y                        
<40 1302 (10.9) 456 (12.4) 1758 (11.3) 1386 (11.0) 1381 (13.0) 2767 (11.9)
40-49 2983 (25.1) 908 (24.7) 3891 (25.0) 3217 (25.5) 2971 (27.9) 6188 (26.6)
50-59 3058 (25.7) 861 (23.4) 3919 (25.2) 3094 (24.5) 2644 (24.8) 5738 (24.6)
60-69 2201 (18.5) 702 (19.1) 2903 (18.6) 2463 (19.5) 1892 (17.7) 4355 (18.7)
≥70 2353 (19.8) 753 (20.5) 3106 (19.9) 2461 (19.5) 1772 (16.6) 4233 (18.2)
Diagnosis y                        
1996-2000 4863 (40.9) 1627 (44.2) 6490 (41.7) 5593 (44.3) 4713 (44.2) 10,306 (44.3)
2001-2005 7034 (59.1) 2053 (55.8) 9087 (58.3) 7028 (55.7) 5947 (55.8) 12,975 (55.7)
Summary stage                        
Local 6150 (51.7) 1565 (42.5) 7715 (49.5) 7419 (58.8) 5331 (50.0) 12,750 (54.8)
Regional 4522 (38.0) 1493 (40.6) 6015 (38.6) 4578 (36.3) 4523 (42.4) 9101 (39.1)
Distant 716 (6.0) 299 (8.1) 1015 (6.5) 509 (4.0) 562 (5.3) 1071 (4.6)
Unknown 509 (4.3) 323 (8.8) 832 (5.3) 115 (0.9) 244 (2.3) 359 (1.5)
Histology                        
Ductal 9440 (79.3) 2918 (79.3) 12,358 (79.3) 10,271 (81.4) 8456 (79.3) 18,727 (80.4)
Lobular 768 (6.5) 196 (5.3) 964 (6.2) 885 (7.0) 630 (5.9) 1515 (6.5)
Other 1531 (12.9) 482 (13.1) 2013 (12.9) 1439 (11.4) 1539 (14.4) 2978 (12.8)
Unknown 158 (1.3) 84 (2.3) 242 (1.6) 26 (0.2) 35 (0.3) 61 (0.3)
Grade                        
1 1253 (10.5) 296 (8.0) 1549 (9.9) 1991 (15.8) 1231 (11.5) 3222 (13.8)
2 3506 (29.5) 1109 (30.1) 4615 (29.6) 4333 (34.3) 3549 (33.3) 7882 (33.9)
3/4 5221 (43.9) 1566 (42.6) 6787 (43.6) 4976 (39.4) 4679 (43.9) 9655 (41.5)
Unknown 1917 (16.1) 709 (19.3) 2626 (16.9) 1321 (10.5) 1201 (11.3) 2522 (10.8)
Tumor size, cm                        
Mean 2.3   2.3   2.3   2.2   2.5   2.3  
SD 3.2   4.1   3.5   2.1   2.5   2.3  
Missing, no. (%) 2422 (20.4) 1005 (27.3) 3427 (22.0) 1243 (9.9) 1265 (11.9) 2508 (10.8)
nSES quintile                        
1: Low SES 3793 (31.9) 1544 (42.0) 5337 (34.3) 2973 (23.6) 3950 (37.1) 6923 (29.7)
2 2730 (22.9) 799 (21.7) 3529 (22.7) 3117 (24.7) 2706 (25.4) 5823 (25.0)
3 1772 (14.9) 440 (12.0) 2212 (14.2) 2765 (21.9) 1850 (17.4) 4615 (19.8)
4 1895 (15.9) 473 (12.9) 2368 (15.2) 2201 (17.4) 1229 (11.5) 3430 (14.7)
5: High SES 1707 (14.3) 424 (11.5) 2131 (13.7) 1565 (12.4) 925 (8.7) 2490 (10.7)
Ethnic enclave quintile                        
1: Least ethnically distinct 662 (5.6) 102 (2.8) 764 (4.9) 1405 (11.1) 525 (4.9) 1930 (8.3)
2 1365 (11.5) 255 (6.9) 1620 (10.4) 2015 (16.0) 943 (8.8) 2958 (12.7)
3 1935 (16.3) 338 (9.2) 2273 (14.6) 2602 (20.6) 1518 (14.2) 4120 (17.7)
4 2998 (25.2) 786 (21.4) 3784 (24.3) 3352 (26.6) 2778 (26.1) 6130 (26.3)
5: Most ethnically distinct 4937 (41.5) 2199 (59.8) 7136 (45.8) 3247 (25.7) 4896 (45.9) 8143 (35.0)
Joint nSES/enclave measure                        
Low nSES, no enclave 529 (4.4) 85 (2.3) 614 (3.9) 1108 (8.8) 465 (4.4) 1573 (6.8)
Low nSES, enclave 6900 (58.0) 2511 (68.2) 9411 (60.4) 4982 (39.5) 6191 (58.1) 11,173 (48.0)
High nSES, no enclave 2306 (19.4) 428 (11.6) 2734 (17.6) 4914 (38.9) 2521 (23.6) 7435 (31.9)
High nSES, enclave 2162 (18.2) 656 (17.8) 2818 (18.1) 1617 (12.8) 1483 (13.9) 3100 (13.3)
Vital status                        
Alive 7033 (59.1) 1764 (47.9) 8797 (56.5) 8399 (66.5) 6256 (58.7) 14,655 (62.9)
Dead 4864 (40.9) 1916 (52.1) 6780 (43.5) 4222 (33.5) 4404 (41.3) 8626 (37.1)
  • Abbreviation: nSES, neighborhood socioeconomic status.
  • For CA, ICD-O-3; for TX, cases diagnosed 1996-2000 used ICD-O-2 and 2001-2005 used ICD-O-3.

In minimally adjusted and fully adjusted models, nativity, Latinx enclave residence, and nSES were found to be independently associated with both outcomes in both states. Given the similarity in findings across models, we have presented associations for fully adjusted models in Table 3 (see Supporting Table 1 for minimally adjusted results). For all causes, foreign-born Latinas were found to have worse survival compared with US-born Latinas. Compared with those residing in the most ethnically distinct neighborhoods, those living in the least distinct neighborhoods had worse survival. Compared with those residing in the highest SES neighborhoods, those in the lowest SES neighborhoods had worse survival. Results were similar with regard to for breast cancer–specific survival.

Table 3. Independent Associations (HRs, 95% CIs; 5-Year Survival Probabilities) of Nativity, Ethnic Enclave, and nSES and Survival After Breast Cancer: California and Texas, 1996 Through 2005
  All-Cause Survival Breast Cancer–Specific Survival
Fully Adjusted HR (95% CI) 5-Year Survival Probability (95% CI) Fully Adjusted HR (95% CI) 5-Year Survival Probability (95% CI)
California                
Nativity                
US born (reference) 1.00   0.953 (0.946-0.959) 1.00   0.986 (0.983-0.989)
Foreign born 1.13 (1.08-1.19) 0.947 (0.940-0.953) 1.19 (1.11-1.26) 0.984 (0.980-0.987)
Ethnic enclave quintile                
1: Least ethnically distinct 1.18 (1.06-1.31) 0.944 (0.937-0.950) 1.16 (1.01-1.34) 0.984 (0.981-0.987)
2 1.21 (1.10-1.34) 0.942 (0.935-0.948) 1.21 (1.07-1.38) 0.983 (0.980-0.986)
3 1.11 (1.02-1.21) 0.947 (0.941-0.954) 1.03 (0.92-1.16) 0.986 (0.983-0.988)
4 1.06 (1.00-1.14) 0.950 (0.943-0.956) 1.05 (0.96-1.15) 0.986 (0.982-0.988)
5: Most ethnically distinct (reference) 1.00   0.953 (0.946-0.959) 1.00   0.986 (0.983-0.989)
P for trend   <.001       .008    
nSES quintile                
1: Low nSES 1.58 (1.41-1.76) 0.926 (0.919-0.932) 1.47 (1.27-1.70) 0.980 (0.976-0.983)
2 1.48 (1.34-1.64) 0.929 (0.921-0.935) 1.42 (1.24-1.62) 0.980 (0.977-0.983)
3 1.32 (1.20-1.46) 0.937 (0.930-0.944) 1.29 (1.14-1.47) 0.982 (0.979-0.985)
4 1.16 (1.06-1.28) 0.944 (0.937-0.950) 1.24 (1.10-1.41) 0.983 (0.979-0.986)
5: High nSES (reference) 1.00   0.953 (0.946-0.959) 1.00   0.986 (0.983-0.989)
P for trend   <.001       <.001    
Texas
Nativity                
US born (reference) 1.00   0.937 (0.927-0.945) 1.00   0.982 (0.977-0.985)
Foreign born 1.18 (1.10-1.26) 0.917 (0.905-0.928) 1.27 (1.16-1.38) 0.975 (0.969-0.980)
Ethnic enclave quintile                
1: Least ethnically distinct 1.28 (1.12-1.47) 0.918 (0.905-0.929) 1.19 (1.00-1.42) 0.978 (0.972-0.982)
2 1.11 (1.00-1.24) 0.929 (0.920-0.938) 0.97 (0.84-1.12) 0.982 (0.977-0.986)
3 1.15 (1.06-1.25) 0.926 (0.916-0.935) 1.16 (1.04-1.29) 0.978 (0.973-0.983)
4 1.12 (1.05-1.19) 0.929 (0.919-0.938) 1.07 (0.98-1.17) 0.980 (0.975-0.984)
5: Most ethnically distinct (reference) 1.00   0.937 (0.927-0.945) 1.00   0.982 (0.977-0.985)
P for trend   <.001       .115    
nSES quintile                
1: Low nSES 1.44 (1.29-1.60) 0.908 (0.899-0.917) 1.24 (1.08-1.43) 0.977 (0.972-0.981)
2 1.31 (1.18-1.46) 0.917 (0.908-0.925) 1.15 (1.00-1.31) 0.979 (0.974-0.983)
3 1.30 (1.17-1.45) 0.918 (0.908-0.927) 1.17 (1.02-1.35) 0.978 (0.973-0.983)
4 1.14 (1.03-1.27) 0.927 (0.918-0.935) 1.06 (0.93-1.21) 0.980 (0.976-0.984)
5: High nSES (reference) 1.00   0.937 (0.927-0.945) 1.00   0.982 (0.977-0.985)
P for trend   <.001       .001    
  • Abbreviations: 95% CI, 95% confidence interval; HR, hazard ratio; nSES, neighborhood socioeconomic status; SES, socioeconomic status.
  • Fully adjusted models included age at diagnosis, tumor grade, tumor size, year of diagnosis, histology, underlying stratification by stage of disease, and clustering by census tract.

In fully adjusted models with enclave and nSES defined as a 4-category combination variable (Table 4) (see Supporting Table 2 for minimally adjusted results), foreign-born Latinas had worse all-cause and breast cancer–specific survival compared with US-born Latinas. In comparison with Latinas residing in enclaves with low nSES, those living in enclaves with high nSES had improved all-cause survival in both states regardless of enclave status. Latinas residing in low nSES nonenclave neighborhoods had worse all-cause survival (compared with those residing in low nSES enclaves). The results were similar with regard to breast cancer–specific survival, but statistical significance was observed only in California for those residing in high SES, nonenclave neighborhoods.

Table 4. Independent Association of Nativity and Joint Associations (HRs, 95% CIs; 5-Year Survival Probabilities) of nSES/Ethnic Enclave and Survival After Breast Cancer: California and Texas, 1996 Through 2005
  All-Cause Survival Breast Cancer–Specific Survival
Fully Adjusted HR (95% CI) 5-Year Survival Probability (95% CI) Fully Adjusted HR (95% CI) 5-Year Survival Probability (95% CI)
California                
Nativity                
US born (reference) 1.00   0.926 (0.919-0.932) 1.00   0.980 (0.976-0.983)
Foreign born 1.12 (1.07-1.17) 0.918 (0.911-0.924) 1.17 (1.10-1.25) 0.976 (0.973-0.980)
Joint nSES/enclave measure                
High nSES, enclave 0.89 (0.84-0.95) 0.934 (0.927-0.940) 0.95 (0.87-1.03) 0.981 (0.977-0.984)
High nSES, no enclave 0.85 (0.81-0.89) 0.936 (0.931-0.941) 0.89 (0.83-0.95) 0.982 (0.979-0.984)
Low nSES, enclave (reference) 1.00   0.926 (0.919-0.932) 1.00   0.980 (0.976-0.983)
Low nSES, no enclave 1.15 (1.06-1.26) 0.914 (0.904-0.922) 1.13 (1.00-1.28) 0.977 (0.972-0.981)
Texas
Nativity                
US born (reference) 1.00   0.909 (0.900-0.917) 1.00   0.977 (0.972-0.981)
Foreign born 1.17 (1.09-1.25) 0.895 (0.884-0.905) 1.26 (1.16-1.37) 0.971 (0.965-0.977)
Joint nSES/enclave measure                
High nSES, enclave 0.91 (0.85-0.97) 0.917 (0.909-0.925) 0.94 (0.86-1.02) 0.979 (0.974-0.982)
High nSES, no enclave 0.90 (0.85-0.96) 0.917 (0.909-0.925) 0.94 (0.87-1.02) 0.978 (0.974-0.982)
Low nSES, enclave (reference) 1.00   0.909 (0.900-0.917) 1.00   0.977 (0.972-0.981)
Low nSES, no enclave 1.09 (0.97-1.23) 0.900 (0.885-0.913) 1.06 (0.92-1.22) 0.975 (0.969-0.980)
  • Abbreviations: 95% CI, 95% confidence interval; HR, hazard ratio; nSES, neighborhood socioeconomic status.
  • Fully adjusted models included age at diagnosis, tumor grade, tumor size, year of diagnosis, histology, underlying stratification by stage of disease, and clustering by census tract.

Adjusted survival probabilities demonstrated the differences between enclave and nSES quintiles, the 4-level variable, and between nativity groups. Differences in survival probabilities allowed for a more qualitative comparison among various categories, demonstrating little difference with regard to probability of survival for those residing in high SES neighborhoods, regardless of enclave status, but for those residing in low SES neighborhoods, we observed a lower survival probability for those living in nonenclave neighborhoods compared with those residing in enclaves. It is interesting to note that differences appeared larger for overall survival compared with breast cancer–specific survival.

Discussion

To address inconsistent associations between ethnic enclave and breast cancer survival noted in the literature, we used the same multilevel measures and analytic methods and found similar associations across 2 states, California and Texas. We observed consistent associations between survival, nSES, ethnic enclave residence, and nativity among Latinas with breast cancer across both states. We also demonstrated that foreign-born Latinas were more likely to live in areas with low nSES and in more ethnically distinct neighborhoods compared with US-born Latinas. Taken together, these results have provided a compelling rationale for continued attention to multilevel and place-based factors contributing to the survival of Latinas with breast cancer.

Enclave

When examined using quintiles, we observed associations between enclave residence, after accounting for nativity, nSES, and other covariates. Residence in more distinct ethnic neighborhoods was associated with improved survival in both states. We observed statistically significant trends for quintiles of all-cause survival in both states; however, for breast cancer–specific survival, although the direction of the point estimates was nearly always consistent, the trend across quintiles was found to be significant in California but not Texas.

Neighborhood SES

Latinas living in neighborhoods with lower SES faced worse survival from both causes in both states. This finding is consistent with a large body of literature demonstrating worse survival among patients with cancer who are living in neighborhoods with low SES, regardless of how nSES is measured.17-19, 22-24

Enclave and nSES

When compared with those in low nSES enclaves, Latinas in either type of high SES neighborhood (enclave or nonenclave) had improved survival and Latinas residing in low nSES nonenclaves had the worst survival, although this was not consistently statistically significant by cause of death or state. This finding demonstrated that living in enclaves may entail some benefits for residents that result in improved survival. Co-ethnic residents within enclaves often maintain lifestyles, cultural norms, and behaviors (eg, diet and physical activity, social networks, and social cohesion) that are health promoting. Enclaves may facilitate communication and information sharing due to greater access to linguistic resources, and also may reduce exposure to discrimination and thus limit the use of unhealthy coping behaviors (eg, smoking and/or drinking) and reduce levels of individual stress.9-13 Associations between enclave residence and survival for Texas, although trending in the same direction as the results for California, were somewhat more attenuated, which may reflect differences in sample distribution; historical patterns related to immigration and settlement; or numerous other social, political, and physical environment differences between the states.

Comparison With Prior Research

Prior breast cancer studies from California using the same nSES and enclave measures as in the current study also demonstrated worse survival for those in neighborhoods with low SES, regardless of enclave status.19, 22 Prior studies demonstrated differing results, both positive and negative, with regard to enclave residence. In 2 prior Texas studies, 2 different measures of neighborhood Latinx composition were found to be associated with worse survival.23 These prior studies may not be directly comparable to the current study given the different methods used (eg, differing lengths of follow-up; adjustment for different covariates; the inclusion of other racial/ethnic groups; and, in the Texas studies, the use of a single indicator of enclave [ie, neighborhood ethnic composition and/or segregation]).

We demonstrated the importance of using the same methods across states to ensure the findings are comparable and not artifacts of methodological differences. A multicomponent ethnic enclave index better captures the multiple dimensions of place that may be relevant for survival, going beyond single measures such as ethnic density. Our measure allows for the identification of enclaves that are both culturally and ethnically concentrated and distinct from the remainder of the state with regard to race and/or ethnicity, language, nativity, and recency of immigration.32

Nativity

When compared with US-born Latinas, foreign-born Latinas in both states had worse survival from both causes. Prior studies demonstrated inconsistent findings. Foreign-born Latinas with breast cancer had worse survival in a Texas study,24 slightly improved all-cause (but not cause-specific) survival in one California study,19 and survival that was equivalent to that of US-born Latinas in another California study.22 Differences in methodology, including nativity imputation methods, may explain these discrepancies. Consistent with prior research, we also demonstrated that foreign-born Latinas were more likely to live in ethnic enclaves and neighborhoods with low SES.19, 24

Implications

Future interventions and research should prioritize historically underserved populations and neighborhoods. More research regarding the pathways through which enclaves and nSES impact survival is needed to inform and tailor community interventions. Future research is needed in other states with Latinx populations who differ by race, country of origin, nativity, length of time in the United States, and residential settlement patterns.

Limitations

The current study has several limitations. Birthplace often is missing from registry data.36-39 Imputation is necessary because dropping patients for whom information regarding birthplace is missing reduces generalizability and introduces bias due to the unique reasons why the data are missing.33, 40, 41 Registry data often are missing at random (ie, missing conditional to other observed variables). For example, missing nativity and ethnicity are conditional on vital status, among other variables, because these data are obtained from death certificates.33, 36-39 Multiple imputation can be used to handle data that are missing at random or missing completely at random and has been applied and validated to impute missing cancer registry data (eg, stage of disease).42-45 Although a multiple imputation model validated in a study of cervical cancer was used in the current study,33 there may have been some misclassification. We calculated the sensitivity and specificity of our imputation using a sample of Latinas with known birthplace and found we could determine birthplace within the United States (93.5% in California and 86.2% in Texas) and foreign birthplace (90.7% in California and 77.0% in Texas) with good accuracy. It is interesting to note that when we did not impute nativity and kept unknown as a category, the direction of enclave and nSES variables remained unchanged. To the best of our knowledge, the imputation approach used in the current study is the best currently available method with which to examine nativity disparities in survival because any interpretation of nativity is flawed without imputation (given the reasons for missingness outlined above), and because we lacked gold-standard (ie, self-report) birthplace data. Future studies should collect self-report birthplace to allow for these validations. In addition, cancer registries should work with their reporting facilities to ensure that this information is collected in a systematic way given its importance in understanding patterns of cancer burden in rapidly growing populations in the United States.

Data were unavailable for several prognostic factors and length of neighborhood residence. When we repeated analyses with California treatment data (missing in Texas data), the findings were similar. These results may not be generalizable outside of Texas and California, in which Latinx predominantly are white, many are born in the United States, many are multigenerational, and most foreign-born individuals are from Mexico.46

Finally, we acknowledge that any potential beneficial or detrimental impacts of ethnic enclaves are likely to be highly contextualized and dependent on specific place-based historic and cultural patterns of immigration and assimilation. Despite these limitations, the results of the current study fill key gaps in the literature by using the same multilevel measures and imputation methods across 2 states with the largest US Latinx populations.

Conclusions

The results of the current study demonstrated consistently harmful effects of low nSES residence, but some evidence was found indicating that enclaves may have small protective effects. We also found synergistic effects between enclaves and nSES, and worse survival for foreign-born Latinas with breast cancer compared with US-born Latinas. Future place-based, mixed methods research and interventions within ethnic enclaves is warranted given the high concentration of underserved populations and overall lower nSES noted in enclaves compared with nonenclave areas. By engaging community members in future research, cancer prevention and control efforts can better leverage local assets, such as ethnic-specific businesses and social networks, to improve cancer outcomes within enclaves.

Funding Support

Supported by the California Department of Public Health pursuant to California Health and Safety Code Section 103885; the Centers for Disease Control and Prevention's National Program of Cancer Registries under cooperative agreement 5NU58DP006344; and the National Cancer Institute's Surveillance, Epidemiology, and End Results Program under contract HHSN261201800032I awarded to the University of California at San Francisco, contract HHSN261201800015I awarded to the University of Southern California, and contract HHSN261201800009I awarded to the Public Health Institute, Cancer Registry of Greater California. The ideas and opinions expressed herein are those of the author(s) and do not necessarily reflect the opinions of the State of California, Department of Public Health, the National Cancer Institute, and the Centers for Disease Control and Prevention or their contractors and subcontractors.

Conflict of Interest Disclosures

The authors made no disclosures.

Author Contributions

Salma Shariff-Marco and Sandi L. Pruitt conceived of the study idea, led the article writing, and supervised the analysis. Scarlett Lin Gomez helped to shape the original study idea and Hong Zhu provided guidance regarding imputation methods. Alison J. Canchola and Hannah Fullington conducted the analyses. Amy E. Hughes created the Texas neighborhood measures. All authors contributed to critically interpreting the findings and writing and editing the article, and have read and approved the final submitted article.