Introduction

The relationship between neighborhood context and socioeconomic status (SES) on health is widely recognized [1,2]. Previous studies found that persons living in lower socioeconomic status (SES) neighborhoods have increased risks of worse health outcomes, including chronic care management and physical functioning [3–5], mental health [6–8], and mortality [9–13].

It is thought that neighborhoods impact health through both the physical and social environment [2]. The physical environment influences health through exposure to pollution, poor housing, public transportation, and local infrastructures such as food markets, recreational facilities, medical centers, and education [9]. Social environment influences health through social cohesion, collective efficacy, and the availability of services that promote social organizations or social interaction [1]. For older adults, the role of neighborhoods on health may be even more important because they have less mobility and smaller social networks, making one more dependent on the local neighborhood [1,3,4,6,8,14,15]. Among older adults, previous studies have found that living in lower SES neighborhoods is associated with greater depressive symptoms [6,8,16] and with poorer physical health [5,15].

It is important to note that health is also strongly correlated with individual socioeconomic status. Low income has been consistently linked with a higher risk of mortality [17–19] and shorter life expectancy [20]. The associations between low-income and poorer health persist among older adults [15,21,22]. As well as income, homeownership is also widely appreciated its association with health conditions, including better self-rated health [23,24], lower likelihood of depression [25], and lower risks of mortality [24,26]. The dynamics of neighborhood and individual-level SES are not well understood. It is thought that higher SES neighborhoods could ameliorate the adverse effect of low individual SES on health by providing a richer physical and social environment [27,28]. On the other hand, low SES individuals residing in high SES neighborhoods could experience higher risks to their health due to exacerbated relative poverty and lack of access to expensive neighborhood resources [10,11,15].

Upon the mixed evidence in the literature, this study examined the cross-level interaction of neighborhood and individual SES on mortality, in the context of Medicare Advantage (MA) beneficiaries responding to the Medicare Health Outcomes Survey (HOS) in 2014 and 2015. This study used two measures of individual-level SES: annual income and wealth (homeownership). Income is an essential measure of SES that is known to have a large influence on health [15,27,29], but income may not solely represent one’s economic conditions, particularly among older adults [25,30]. Income tends to generally decrease later in the life course, which may result in smaller variations in income inequality. In addition, income and wealth may not substitute each other interchangeably [25,26]. Also, homeownership is considered to reflect the lifetime income or accumulated past savings [24,26,30]. Thus, this study used homeownership as a measure for the wealth aspect of individual SES, as well as annual income.

The objective of this study was (1) to examine the association of neighborhood SES, represented by the Area Deprivation Index (ADI), and individual SES, defined by self-reported household income and homeownership, with mortality among older adults (2) to determine how the relationship of neighborhood SES may differ by individual SES. The findings of this study would contribute to the understanding of the cross-level dynamics of individual SES and area-level measures.

Materials and methods

Results

Table 1 shows the descriptive statistics of the sample and the p-values from the chi-2 tests. In the study sample, 6.6% died within two years of the baseline survey. Low-income respondents had a higher 2-year mortality rate than higher-income respondents (8.9% versus 5.3%). One-third (35.8%) reported household incomes less than $20,000; 5.8% of low-income respondents and 8.9% of higher-income respondents resided in the least disadvantaged neighborhood (ADI deciles group 1); 14.1% of low-income respondents and 3.8% of higher-income respondents resided in the most disadvantaged neighborhood (ADI deciles group 10).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics.

https://doi.org/10.1371/journal.pone.0267542.t001

As for homeownership, slightly less than a third (31.7%) were nonhomeowners, and they had a higher 2-year mortality rate (9.1%) than homeowners (5.3%). Similar proportions of respondents were living in the least advantaged neighborhood (ADI deciles group 1) among homeowners (7.6%) and nonhomeowners (8.5%), but the proportions of those living in the most disadvantaged neighborhood (ADI deciles group 10) differed by almost three times (homeowners 4.9% vs. nonhomeowners 13.2%).

The results indicate that lower-income and not homeowning respondents were significantly more likely to be older, female, racial/ethnic minority, either obese or under-weight, to have more chronic conditions and difficulties in ADLs, and smokers compared to higher-income and homeowning respondents.

Table 2 shows the logistic regression results on mortality rates including interaction terms between neighborhood status and individual SES (income, homeownership). The full results with covariates is reported in S5 Table. The first two columns show the model using income as the measure of individual SES (Model 1), and the last two columns show the model using homeownership as the measure of individual SES (Model 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Logistic regression results on 2-year mortality.

https://doi.org/10.1371/journal.pone.0267542.t002

Those who had low income (odds ratio [OR]: 1.40, 95% confidence interval [CI]: 1.19–1.65) or were nonhomeowners (OR: 1.47, 95% CI: 1.32–1.63) had higher odds of 2-year mortality than those with higher income or homeownership. Living in a higher ADI decile neighborhood was associated with higher risks of death in both models; for example, living in a neighborhood of ADI decile group 10 had an odds ratio of 1.29 (95% CI: 1.16–1.44) of 2-year mortality in Model 1 and 1.38 (95% CI: 1.20–1.58) in Model 2. The interaction terms between individual SES and neighborhood status had decreasing odds of mortality as it goes to higher ADI decile groups. For example, in both models, the interaction terms of ADI groups 2–7 were not significantly different from the interaction term of ADI group 1, and the interaction terms of ADI groups 8–10 are significant with odds ratios smaller than 1. The logistic regression results without the interaction terms are provided in S5 Table, side by side with the results of the models with the interaction terms. The two sets of results suggest that the full picture on the cross-level relationships between individual-level and neighborhood-level SES is not captured without the interaction terms in the model.

For other covariates, older age was associated with higher mortality, and the magnitude increased according to age in both models. Living with more chronic conditions contributed to a higher risk of mortality. ADL deficits and smoking were also associated with higher mortality rates. Protective factors include being female, non-Hispanic Black (only significant among the low-income), Hispanic, or Asian compared to non-Hispanic white, and obese (S5 Table).

To further examine the cross-level interaction between individual SES and neighborhood status, the adjusted predicted mortalities are displayed in Table 3 and Figs 1 and 2. Table 3 Panel A and Fig 1 show the predicted mortality rates among higher-income and low-income by ADI decile groups, and Table 3 and Fig 2 illustrate among homeowners and nonhomeowners. In Table 3, the first four columns show the higher individual SES group’s the predicted mortality rates and comparisons between ADI group 1 and higher ADI groups within the same individual SES group (for example, comparing higher-income ADI group 1 vs. higher-income ADI group 7). The next four columns shoe the lower individual SES group’s information. The two right-most columns display the differences and the p-values comparing the predicted mortalities between higher SES and lower SES groups within the same ADI group (for example, comparing homeowners in ADI group 3 vs. nonhomeowners in ADI group 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Predicted mortality rates by ADI decile groups, higher-income vs. low-income among older adults ages 65+.

Source: Medicare Health Outcomes Survey, 2014–2015. Note: The predicted mortalities were obtained based on the regression models presented in Table 2. The models were controlled for age (65–69, 70–74, 80–84, 85 or older), sex (male, female), race/ethnicity (non-Hispanic white, non-Hispanic Black, non-Hispanic Asian, Hispanic, other), the number of chronic conditions (none, 1–2, 3–5, 6 or more), BMI (normal/overweight, obese, underweight), difficulties in ADL (none, 1 or more), smoking status (not smoking, smoking), and survey year (2014, 2015).

https://doi.org/10.1371/journal.pone.0267542.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Predicted mortality rates by ADI decile groups, homeowners vs. nonhomeowners among older adults ages 65+.

Source: Medicare Health Outcomes Survey, 2014–2015. Note: The predicted mortalities were obtained based on the regression models presented in Table 2. The models were controlled for age (65–69, 70–74, 80–84, 85 or older), sex (male, female), race/ethnicity (non-Hispanic white, non-Hispanic Black, non-Hispanic Asian, Hispanic, other), the number of chronic conditions (none, 1–2, 3–5, 6 or more), BMI (normal/overweight, obese, underweight), difficulties in ADL (none, 1 or more), smoking status (not smoking, smoking), and survey year (2014, 2015).

https://doi.org/10.1371/journal.pone.0267542.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Predicted mortality rates by individual and neighborhood SES.

https://doi.org/10.1371/journal.pone.0267542.t003

Discussion

In this study, this paper analyzed the cross-level effects of neighborhood SES and individual SES, using two measures (income and homeownership), on mortality with a large-sized, diverse sample of Medicare Advantage beneficiaries.

In the findings, first, neighborhood-level SES and individual-level SES were significantly associated with mortality rate. Mortality rates were significantly higher for low-income individuals and nonhomeowners than higher-income or homeowning individuals across neighborhoods in all ADI deciles. Second, the association of neighborhood context on mortality was significantly different by individual-SES (income and homeownership). Living in a less disadvantaged neighborhood did not attenuate the adverse effect of having lower individual SES, while among higher-income or homeowning individuals, living in less disadvantaged neighborhoods had consistent associations with lower risks of death compared to living in more disadvantaged neighborhoods. Third, the patterns of cross-level interactions of neighborhood context with individual SES were different when using income or homeownership as the measure of individual SES, especially among lower-SES.

The present paper’s findings are consistent with previous studies examining cross-level interactions between individual SES and neighborhood SES, when it comes to using income as the measure of individual SES. In a study of adults aged 25–74 years in California, Winkleby et al. (2006) found that the mortality of low-income individuals did not decrease when living in higher SES areas compared to when living in lower SES areas, unlike high-income individuals. Roos et al. (2004) also found different trends in mortality between low-income and high-income individuals aged 18–75 in Canada in disadvantaged and advantaged neighborhoods classified using various indicators such as neighborhood household income and education, and their results also showed that the differences in mortality rates based on income were greater than that the differences based on neighborhood status. This paper adds to the previous literature by focusing on a national sample of older adults enrolled in the Medicare Advantage program, and leveraging the large sample size to examine the relationship between mortality and individual/neighborhood SES at more granular levels.

One explanation for these results is that income is a stronger predictor of mortality than neighborhood SES among low-income older adults. Even though less disadvantaged neighborhoods may have more resources such as access to healthy foods, medical centers, and transportation, these resources may not be meaningful for low-income individuals if they cannot afford them. In addition, low-income individuals living in advantaged neighborhoods may experience higher relative disadvantage connected to social stigma overlaid on the low material level and a greater sense of deprivation [36].

As for homeownership, little is known about the cross-level interactions on health of homeownership with neighborhood SES. Previous studies have examined the relationships between homeownership and perception on neighborhood [37,38] or the relationships between homeownership and health outcomes [25,26,39,40], but not how homeownership interacts with neighborhood-level SES when affects health. This study contributes to the literature by examining the cross-level interactions between neighborhood SES and individual SES by using the measure of homeownership.

For the different patterns of associations among neighborhood SES, individual SES, and mortality between using income and using homeownership, Ross & Mirowsky (2008) argued that homeownership distinguishes economic well-being at the middle to high end of the continuum rather than the low. The nature of income and homeownership might operate differently with health among older adults, and further study is needed.

As for the decreasing mortality rates in ADI groups 9 and 10 among nonhomeowners, residents in more disadvantaged neighborhoods might have different characteristics than others. For example, residents in ADI groups 9 and 10 tend to be younger than residents in less disadvantaged neighborhoods. Residents in ADI group 10 were significantly more likely not to own the house they lived in (56.4%), compared to the other ADI groups (the highest proportion of nonhomeowners was 37.2%). These characteristics may have affected the decreased risks of mortality in ADI groups 9 and 10 among nonhomeowners.

Also, previous studies found that sense of control, residential stability, and perceived relative standing in the neighborhood mediates homeownership’s relationships with mental health and perceived neighborhood disorder [37–39]. These factors may operate differently for nonhomeowners in neighborhood SES decile groups.

The significant gaps between higher-SES and lower-SES individuals in predicted mortality rates indicate the importance of considering the cumulative effect of poverty. As incomes in older ages heavily rely on Social Security, especially among low-income older adults, and the Social Security income in one’s older ages reflect incomes in their younger ages [41,42], those who are low-income in older years are more likely to have been low-income during their younger years. Homeownership is also considered as the accumulation of lifetime economic living standards [26,30]. The accumulated financial hardship may have a much stronger impact on mortality in the sample than neighborhood disadvantage [43–45]. This paper’s results suggest the cumulative effects of poverty in the sample. Low-income older adults were twice as likely to be living with 6 or more chronic conditions (18.6% vs. 9%), more likely to be obese (33.7% vs. 29.4%), and more likely to be a current smoker (13.9% vs. 8%) than higher-income older adults.

The findings of this paper stress the significance of individual SES on health, which suggests the importance of reducing poverty and lifetime economic hardship in promoting the health of older adults. Policies that affect poverty often require efforts outside the health care sector, including but not limited to housing, transportation, education, environment, criminal justice [46]. Such cross-sectional collaborative efforts are needed at the federal, regional, state, and local levels. One successful example of cross-sectional effort at the local level is the Maryland Health Enterprise Zone (HEZ) initiative [47]. The HEZ initiative promoted collaborations between health departments, local government agencies, and local health care providers. They recruited primary care physicians to most underserved areas in the state and provided state-level tax incentives. The initiative also offered increased access to licensed clinical social workers to address the underserved population’s non-medical health needs [47,48]. In addition, programs to reduce poverty among older adults, such as programs to reduce copayments and out-of-pocket costs, may be effective to provide health benefits for lower-income individuals.

This study has several limitations. First, this study examined the relationship between neighborhood and individual SES in a sample of older adults Medicare Advantage beneficiaries. These results may not be generalized to older adults in the Medicare Fee-For-Service (FFS). Previous studies on differences between MA and FFS Medicare enrollees found that there were some but not substantial differences in sociodemographic characteristics (e.g., more racial minority, more low-income, more living in metropolitan area) [49,50], lower mortality rates among MA beneficiaries than FFS but decreasing over time enrolled [51], and fewer health care utilizations [49,50]. Despite some differences in population characteristics from FFS Medicare, Medicare Advantage plans account for one-third of all Medicare beneficiaries enrolling more than 20 million people, so it is important to examine these relationships with the MA beneficiaries’ program.

Second, this study is not able to provide which aspect of neighborhood disadvantage affects mortality and interacts with income level since this study uses a composite measure of neighborhood context. However, the composite index also helps us to capture the complex contributions of neighborhood resources and structure on mortality. Third, this study utilized self-reported survey measures for income, homeownership, and health status, which are subject to recall bias and respondent bias. Forth, the homeownership measure for wealth only captures whether the respondent was living in a house that they owned but not the value of the house if they owned it. There is also a possibility that someone did own a house but was not living in that place where they do not own, which is captured as non-homeowner as per how the survey asked homeownership. However, a study found that housing tenure (ownership or rent) had similar results to when using net wealth measure [52].

Though not without limitations, this paper has several strengths and contributions. This study adds to the literature by examining the cross-level interactions between neighborhood and individual SES, which has been less studied, using two measures of individual SES—income and homeownership. This study was able to apply the research question to a large-sized, diverse sample of older adults.

Conclusion

Neighborhood-level SES and individual-level SES (income and homeownership) are significant predictors of mortality among older adults enrolled in MA plans. Income and homeownership were consistently a significant predictor of higher mortality across all neighborhood status. Among higher-income or homeowning individuals, mortality was higher in low SES neighborhoods compared to higher SES neighborhoods. On the other hand, among low-income or non-homeowning individuals, mortality did not increase with more disadvantaged neighborhood status. These findings suggest that the impact of neighborhood resources varies by individual income and homeownership, and approaches to improving health should recognize the differential impacts.

Supporting information

Acknowledgments

The author would like to thank Dr. Eva DuGoff for comments on an earlier draft of this paper.