Individualized and institutionalized residential place-based discrimination and self-rated health: a cross-sectional study of the working-age general population in Osaka city, Japan

Understanding the impact of place on health is a key element of epidemiologic investigation [1, 2]. Various aspects of residential place, such as green space, walkability, social capital, poverty, unemployment and deprivation, have been studied as the environmental determinants of health [3‐9]. The concept of an association between place and discrimination such as racial/ethnic residential segregation is also one of the most investigated research fields in the USA [10, 11]. Residential segregation based on race is associated with poor health outcomes among African Americans in the USA. Economic segregation and poverty concentrations are closely linked with such residential segregation [10]. However, studies based on this concept outside the USA are scarce, especially in Asian countries. On the other hand, the concept of discrimination due to geographical place of residence (place-based discrimination; PBD) has been explored as “territorial stigmatization”. Within the fields of sociology and geography, this exploration has focused on the bottom level of the hierarchy of place, in areas such as St Paul’s in Bristol, UK and the banlieues in France [12, 13].

In Japan, the Buraku people are a minority group that continues to face discrimination because their predecessors, the eta and hinin, were considered outcasts during the Japanese Edo period (1600–1868) when they were employed socially unacceptable jobs (such as slaughtering animals or treating leather). In present day Japan this discrimination occurs in many aspects of life, including employment, housing and education. Further, discrimination against Buraku district residents, which is based on the place itself regardless of residents’ ancestry, continues [14]. At the same time, a new type of discrimination against residents of the Nishinari ward in Osaka city, which includes the largest Buraku district, so-called Nishinari discrimination has emerged [14, 15]. This refers to the stigmatization of the ward for reasons of slums, poverty, crime, dilapidated dwellings and insanitary conditions. These two forms of discrimination, Buraku and Nishinari, could be categorized as PBD, which may affect the health of individuals who live in the specified area [16].

Many studies of racial/ethnic residential segregation have compared residents in a segregated area with those in other areas [10, 11]. However, some discrimination studies focused only on the bottom level of the hierarchy of place, such as our previous study [16]. We reported that interpersonal perceived PBD, such as Buraku or Nishinari discrimination [15, 17], was associated with poor mental health in a stigmatized area in Osaka, Japan [16]. However, this previous study treated only the individual-level perceived discrimination (individualized pathway), which is unlikely to capture the full complexity of PBD [18]. It is therefore important, when considering the mechanisms between PBD and health, that both individualized and institutionalized pathways are taken into account [19, 20]. PBD may not be merely the actions and prejudices of individuals against individuals; institutionalized PBD may be perpetuated by organizations and represent processes built into social entities, in a similar way to institutionalized racism [19]. Thus, even if residents do not recognize the existence of interpersonal PBD, they may suffer inherent institutionalized PBD and, in turn, their health may be compromised; i.e., the concept that living in a stigmatized neighborhood affects neighborhood health. Our objective was to examine the association of both individualized and institutionalized PBD with health, using the self-rated health (SRH) index, in Osaka city. This city has a wide ranging population from the very poor to the affluent of Japan and includes some stigmatized areas, identified from a priori knowledge of Buraku and Nishinari discrimination [9, 14]. We focused on area-level contextual impact on health, comparing the health impact of institutionalized PBD with that of other area-level indicators relating to socioeconomic status (SES), after adjusting for individual-level factors including perceived PBD and SES.

Basic characteristics of the study subjects and the proportions of perceived PBD and poor SRH are shown in Table 1. Women were more likely to answer “yes” to experiencing PBD than men. A higher proportion of perceived PBD was observed among subjects who were less educated, had no spouse or had friends. The proportion of poor SRH significantly differed for all factors. A higher proportion of poor SRH was observed among subjects who had low SES or low social relationships.

Mean, median, range and correlation among area-based indicators used in this study are shown in Table 2. The mean of ALPBD was 6.9 with a range from 0.0 to 50.0%. ALPBD had moderate correlation with area-level unemployment by census (ALUEC) (0.55) and ALDI (0.63). Two combinations of ALUEC and ALDI, and area-level not-home-owner by aggregated responses (ALNHA) and area-level not-home-owner by census (ALNHC) had an especially high correlation of 0.90 and 0.80, respectively. The distribution of ALPBD was consistent with our a priori knowledge of PBD in areas such as Buraku or Nishinari except for the low ALPBD areas where only one or two subjects answered “yes” to PBD (data not shown) [16].

Table 3 shows the proportions of perceived PBD and poor SRH according to the quartiles of area-level indicators with their ranges. Participants who lived in disadvantaged places (higher quartiles) were more likely to answer “yes” to experiencing PBD. The proportion of poor SRH significantly differed by all area-level indicators except for ALNHC.

Tables 4 and 5 show the results of the multilevel logistic analyses. All analyzed individual-level factors showed significant associations with poor SRH in the age, sex, PBD and SES-adjusted model (model 4). For example, ORs (95%CI) were 1.89 (1.33, 2.81), 2.28 (1.50, 3.42) and 1.83 (1.45, 2.30) for perceived PBD, unemployment and not-home-owner, respectively (Table 4). Regarding associations between area-level indicators and poor SRH (Table 5), only area-level indicators were included in model 1. The ORs for the third and highest quartiles of ALPBD, ALNHA and ALNHC, third quartile of area-level unemployment by aggregated responses (ALUEA), and highest quartile of ALDI were statistically significant compared with the lowest quartile: for example, ORs for the highest quartiles of ALPBD, ALNHA, ALNHC and ALDI were 1.84 (1.35, 2.52), 1.57 (1.14, 2.20), 1.65 (1.18, 2.33) and 1.78 (1.30, 2.44), respectively. Results from an age and sex-adjusted model (model 2) did not vary compared with those from model 1. After adjustment for age, sex and a corresponding individual-level factor (model 3), the ORs for highest quartile of ALPBD, ALNHC and ALDI were attenuated, but remained statistically significant: i.e., 1.57 (1.13, 2.18), 1.53 (1.11, 2.12) and 1.66 (1.20, 2.28), respectively. In the age, sex, individual-level PBD and SES-adjusted model (model 4), the ORs for the third and highest quartiles of ALPBD were marginally significant: 1.31 (95%CI: 0.93, 1.84) and 1.32 (0.95, 1.86), respectively, while those of the highest quartiles of ALNHC and ALDI were attenuated to 1.26 and 1.21, respectively, and not significant. In the further social relationship-adjusted model (model 5), the ORs for the highest quartile of ALPBD, ALNHA, ALNHC and ALDI were attenuated to 1.28, 1.10, 1.25 and 1.13, respectively, and all were non-significant, although the ORs for the third quartile of ALUEC were 0.74 and marginally significant. ORs for the third quartile of ALUEA retained significance throughout all models: i.e., 0.65 (0.46, 0.93) for model 1, 0.64 (0.45, 0.90) for model 2, 0.60 (0.42, 0.85) for model 3, 0.65 (0.46, 0.92) for model 4 and 0.67 (0.47, 0.95) for model 5.

Our results showed that not only individualized PBD but also institutionalized PBD might be social determinants of health after individual-level factors such as SES had been taken into account, although the health impact of institutionalized PBD was not fully statistically significant and was lower than that of the individualized PBD. In our previous study, the association between individual-level PBD and poor mental health was observed after adjustment for SES such as employment, housing tenure and education, and stronger among the highly educated than among the less educated [16], although low education levels were positively associated with poor health in general [28]. In a study in New Zealand, health differences between ethnic minorities and majorities were considerably attenuated after adjustment for perceived discrimination in addition to SES [31]. These two results indicated that individual-level discrimination may affect health independently of SES, consistent with the results in the current study.

In the individual-level PBD and SES-adjusted model (model 4), the ORs of ALPBD for poor SRH were marginally significant, those of other area-level SES indicators were attenuated (i.e., ALNHC and ALDI) and did not show consistent tendency of quartile order (i.e., ALUEA, ALUEC and ALNHA). The lack of hierarchical association between area-level unemployment and poor health was unexpected, because unemployment is a key social factor in major deprivation indices used worldwide [5, 29, 32]. Because ALUEA was derived from the proportion of an average of 1.6 unemployed persons, ranging between zero and five, in tracts which consist of an average of 29.6 persons, this might be unstable and biased (Additional file 1: Table S1), although ALUEA and ALUEC indicated similar results in model 4 (key model). Because the proportion of unemployed persons within the area is generally low, area-level unemployment may not have wide range of distribution or strong relationship with residents’ health. Thus, the magnitude of area-level indicators should be compared carefully as the ranges of these measurements should be different. As this is conducted in Japan where severe poverty and areal segregation based on living standards are not highly likely, the range of areal deprivation may be narrower than those in other places with wider social disparities such as the USA. The patterning of gradients in health detected by the area-level indicators may reflect both the different meanings of the areas investigated and the different pathways by which diverse aspects of SES (PBD) influence health [5, 33].

There is no consensus on an optimal scale of perceived discrimination [34], and various patterns of questions have been used to assess self-reported racial/ethnic discrimination. A single item scale for self-reported discrimination has often been used, but may be unreliable and may understate the extent of discrimination [35], although these problems tend to bias associations between self-reported discrimination and health toward the null (i.e., type II error) [19]. In our results, the distribution of ALPBD was consistent with our a priori knowledge of PBD in areas such as Buraku or Nishinari. Although our single question about perceived discrimination has not been validated, the fact that our results matched with a priori knowledge may imply an acceptable validity of the scale used for perceived PBD in the study.

Institutionalized PBD may represent an aspect of place which has a greater impact on health than other area-level SES indicators such as ALDI, after adjustment for individual-level factors, in Osaka, Japan. The role of ALPBD in explaining spatial gradients in health may be smaller than that of individualized PBD. However, even if all interpersonal discrimination is eliminated, the institutionalized structures that shape people’s opportunities in life from early childhood may be so powerful that we should continue to investigate the association between ALPBD and health.