Introduction

The failure to detect breast cancer until an advanced stage increases mortality risk and reduces the long-term survival of women with breast cancer [1–4] Several characteristics of patients are known to be risk factors for advanced stage at diagnosis, including low socioeconomic status, low educational attainment, membership in a minority race/ethnic group, and foreign-born or immigrant status [5–11]. Immigrant women, in particular, may be at greater risk of being diagnosed with breast cancer at a later stage since they are more likely to have limited English proficiency, to experience social exclusion, to lack health insurance, to have less knowledge about cancer prevention and to consequently be less likely to receive appropriate cancer screening services [12–23].

In recent years there has been a dramatic increase in the size of the US foreign-born population. For instance, between the 1990 and 2000 censuses, the foreign born proportion of the Cook County, Illinois population rose from 14.1 to 19.8%; of the foreign born, 40.4% entered between 1990 and 2000 (data computed from Tables DP-2, Social Characteristics, STF 3 for 1990 and SF 3 for 2000 [24]). Also, new patterns of immigration have been emerging. It has been observed that the educational attainment and economic power of new immigrants has improved [25, 26], and they have increasingly tended to settle in suburban neighborhoods [27–29]. Immigrants historically were attracted to enclaves of their own ethnic group in large urban centers, areas that often had high rates of poverty and neighborhood disadvantage [27].

Research also shows that many health status indicators, including general mortality and self-assessed health, consistently have appeared more favorable for immigrant populations compared to the US-born, generally supporting the “healthy immigrant” hypothesis [30, 31]. Over the last decade, foreign-born women have also improved on several important health indicators including 2-year mammography screening and cancer mortality rates [32]. These changes in immigrant populations lead us to explore the effects of neighborhood immigrant composition on the risk of late stage diagnosis.

Neighborhoods with high proportions of immigrants have been posited to provide health-conducive structural contexts such as the presence of ethnic food stores relevant for healthy diet, and to diminish the deleterious effects of poverty [33, 34]. It has been also suggested that the increased population density of immigrant neighborhoods can boost commercial activity and reduce social isolation of residents, immigrants as well as non-immigrants, especially elderly persons [35]. Neighborhood immigrant composition with these structural elements therefore potentially can affect the health of the residents independently of individual-level factors.

Studies have also linked neighborhood indicators of poverty and disadvantage to differences in stage at diagnosis, and individuals living in neighborhoods with low socioeconomic levels have had an increased probability of late stage diagnosis [36–38]. The potential effects of changes in immigrant population concentration on health outcomes in general, and breast cancer diagnosis in particular, however, have been largely unexplored. This study investigates the association between changes in immigrant population distributions in neighborhoods and the likelihood of late (regional or distant metastasis) stage at diagnosis of breast cancer among women in Cook County, Illinois, one of the immigrants’ leading ports of entry in the nation, which consists of the City of Chicago and surrounding suburban areas.

Between 1990 and 2000, census figures show that the foreign-born population of Cook County swelled from 717,300 to 1,064,700, or an increase of about 67%. The dominant elements of the foreign-born population have been those from Latin America, Europe and Asia. While European (27.1%) and Asian migration (21.8%) is significant in Cook County, almost half of its foreign born population (48.7%) originated in Latin America, and 82% of that Latin American-born population was Mexican in 2000 [24].

Methods

Study Population and Measurements

The analysis employed breast cancer stage at diagnosis data from the Illinois State Cancer Registry (ISCR) [39]. Our analysis is based on 42,714 breast cancer cases diagnosed between 1994 and 2003 among women residing in Cook County, Illinois. ISCR data are estimated to be 97 to 100% complete using the North American Association of Central Cancer Registry method for case ascertainment and classification [40]. Cancer registry records provided standardized information on age, race, Hispanic ethnicity, Surveillance Epidemiology and End Results (SEER) general summary stage and year of diagnosis [41]. Cases with unknown stage (5.3% of all the reported cases) were excluded from the analysis. Demographic characteristics of all included cases are presented in Table 1. As an outcome measure, distant metastasis and regional stage at diagnosis were coded 1, and in situ and localized stage at diagnosis were coded 0. Cancer cases were geocoded by ISCR to the census block level using residential address at the time of diagnosis. Complete and valid address information was available for more than 98% of the breast cancer cases. Using this information, we were able to code each case to a Cook County census tract.

Table 1 Frequency distribution of variables included in the analysis
Full size table

Neighborhood characteristics were measured using 1990 and 2000 US Census tract level data for Cook County, Illinois [42]. We constructed two indices to measure neighborhood characteristics using 1990 census data: concentrated immigration and concentrated disadvantage [43]. The concentrated immigration index was composed of two measures: percent foreign-born, and percent linguistically-isolated households (i.e., percent of households lacking at least one member who speaks English well or very well). Concentrated disadvantage was defined as the percentage of families in the census tract with incomes below the poverty line, and percentage of persons in the tract who are unemployed. Principal components analysis confirmed that the variables used for each index were highly interrelated and loaded on a single component. Following Sampson and colleagues [43], we first calculated z-scores for each variable (in all census tracts in Cook County, N = 1,343) to be used in constructing each of the two indices and averaged them with equal weighting to create each index. Each of these indices (Table 1, bottom panel) has a mean value close to zero and a standard deviation close to one; they deviate slightly from these norms because census tracts with no cases of female breast cancer between 1994 and 2003 were excluded from the multilevel analysis. A total of 1,308 census tracts were included in the final analysis.

Consistent with previous studies concerned with community change [44–48], standardized residual change scores were employed to measure neighborhood change. First, two additional status indices, concentrated immigration and concentrated disadvantage in 2000, were measured in a manner similar to that described above for the neighborhood status indices in 1990. Each of these indices from 2000 were then regressed on the 1990 status indices (i.e., concentrated immigration and concentrated disadvantage in 1990), and the standardized residuals of the 2000 variable were obtained. As Bohrnstedt [49] demonstrates, residual change scores are statistically independent of the initial status, and they represent the change in the variable that is beyond the expected level on the basis of the initial status. In addition, since all census tracts in Cook County are used to compute the regression equation, these scores represent changes that account for general trends between 1990 and 2000. Thus, a residual change score of “1” in immigration concentration, for example, indicates that there was a one standard deviation greater than expected level of immigrant concentration in that neighborhood in 2000, considering its preexisting status and changes experienced in Cook County as a whole. An alternate approach to conducting this analysis that we considered was to instead employ simple difference scores (i.e., difference between 1990 and 2000 index scores). In comparing these measures, the bivariate correlations between the residual change scores and simple difference scores of concentrated immigration and concentrated disadvantage were .936 and .957, respectively, and models with simple difference scores produce essentially the same results. Consequently, all analyses presented here use residual change scores.

Data Analysis

Multilevel logistic regression, accounting for the nesting of women within census tracts, was employed to predict the log odds of patients presenting with late stage diagnosis of breast cancer, [50]. We estimated two models: (1) a random intercept model with only patient level variables, i.e., age, race/ethnicity, and year of diagnosis; and (2) a random intercept model with both patient level variables and several neighborhood level compositional variables, including neighborhood concentrated immigration status in 1990, concentrated disadvantage status in 1990, and neighborhood changes in concentrated immigration and concentrated disadvantage status between 1990 and 2000. Noticing that the late stage diagnosis is overrepresented among younger minority women with different tumor characteristics [51, 52], we included interaction terms between patient’s age at diagnosis and race/ethnicity in the models. These age-race interaction effects may be confounded with neighborhood effects, and hence, needed to be controlled. Using this model, the effects of neighborhood changes in immigration concentration between 1990 and 2000 on the likelihood of late stage diagnosis of breast cancer were assessed. The study was approved by the institutional review board at the University of Illinois at Chicago (Protocol # 2004-0271).

Results

When patient level variables only were included in the model, each was significantly associated in the expected direction with the probability of late stage diagnosis. As shown in Equation 1 in Table 2, age had a curvilinear relationship with late stage diagnosis of breast cancer: the youngest and oldest patients were at increased risk. African Americans and Hispanics, compared with whites, were each more likely to have a late stage diagnosis. As indicated by the significant effects of interaction between age of patients and African American and Hispanic patients, the late stage diagnosis was overrepresented among younger patients of these two race/ethnic groups, relative to whites. The odds of late stage diagnosis also varied in a curvilinear manner with year of diagnosis, such that they decreased from 1994–2000 and then were essentially unchanged between 2001 and 2003.

Table 2 A multilevel logistic regression model of change in concentrated immigration and distant breast cancer stage diagnosis 1994–2003
Full size table

Equation 2 shows that after controlling for individual-level factors, concentrated immigration in 1990 was positively associated with greater odds of late stage diagnosis among breast cancer patients residing in those neighborhoods. For each unit increase (about one standard deviation) in a census tract’s level of concentration of immigrant population in 1990, the odds of late stage diagnosis among breast cancer patients in that neighborhood increased by 4% (odds ratio [OR] = 1.04; 95% confidence interval [CI] = 1, 1.07; P = .035). Concentrated disadvantage in 1990 was also independently associated with greater odds of late stage diagnosis among residents with breast cancer (OR = 1.19; 95% CI = 1.14, 1.24; P < .001).

Changes in concentrated disadvantage and concentrated immigration status within neighborhoods between 1990 and 2000 were also each independently associated with the odds of late stage diagnosis. Each one unit increase in neighborhood changes in concentrated immigration (i.e., one standard deviation from the expected level of concentrated immigration in 2000), was associated with an increase of 4% in the odds of late stage diagnosis among female breast cancer patients within that neighborhood (OR = 1.04; 95% CI = 1.01,1.08; P < .026). Changes in concentrated disadvantage were also associated with the probability of late stage diagnosis. Each unit increase in changes in neighborhood concentrated disadvantage brought about a 5% increase in the odds of late stage diagnosis among breast cancer patients in that neighborhood (OR = 1.05; 95% CI = 1.01–1.10; P < .013).

Discussion

Consistent with previous studies, this research confirms that level of neighborhood disadvantage is strongly associated with late stage diagnosis of breast cancer [52, 53]. Yet the dynamic effects of community change also appear to influence late stage diagnosis. Specifically, breast cancer patients residing in neighborhoods that became relatively more disadvantaged over the 1990–2000 decade experienced an additional risk of late stage diagnosis. These findings also demonstrated that increasing concentration of immigrant populations within neighborhoods over the course of that decade additionally contributed to the risk of late stage diagnosis of breast cancer among women in those communities. They further suggest that, as an important policy implication, cancer screening and follow-up resources should be directed to areas experiencing rapid increases in immigrant populations.

A potential mechanism by which changing neighborhood composition can influence the likelihood of cancer detection at a late stage is that increases in disadvantaged and immigrant populations may contribute to an added burden on the structural capacity of communities to successfully supply quality health care resources to residents. Our findings also suggest that the any potential protective effect of concentrated immigrants may not be applicable to some chronic illnesses indicators such as stage at breast cancer diagnosis, and that community-level structural barriers such as shortages of health care services may be counterbalancing and therefore diminishing any such positive effects.

To our knowledge, this is the first study to examine the effects of changes in neighborhood immigrant population composition on late stage breast cancer diagnosis. This is particularly important in light of a recent study that suggests that single-point-in-time estimates of neighborhood characteristics in multilevel models can mis-specify these effects because they may be changing over time [54]. As in our study, measuring neighborhood change shows whether the direction of change, as well as its initial level, is affecting the health outcomes of interest among the residents of the target area.

There are a number of limitations to our findings. First, we do not have some key patient level measures such as immigration status, as well as information regarding how long women lived in their neighborhood prior to their breast cancer diagnosis and, hence, how long they had been exposed to the neighborhood environment. Caution, therefore, is necessary when interpreting the results as evidence of a causal relationship between neighborhood compositional changes and late stage diagnosis. Additionally, the census tract, which is here used as a proxy for neighborhood, might not be the optimal geographic unit for capturing the effects of the distribution of and changes in immigrant populations. Finally, the available data do not allow us to directly examine issues such as acculturative stress, social support, and changes in access to care across neighborhoods that experience changes in immigrant population. Examining measures such as these would allow us to more clearly elucidate the mechanisms by which change may impact cancer diagnosis among immigrants as well as non-immigrants. Future research should focus on these dimensions in order to improve our understanding of how individual and neighborhood changes influence where, when and how all women obtain the health care necessary to benefit from earlier detection of breast cancer.