Neighborhood Disadvantage in a Nationally Representative Sample of Community-Living Older US Adults Authors: Thomas M. Gill, Linda Leo-Summers, Brent Vander Wyk, Robert D. Becher, Jingchen Liang Published: 2024 ABSTRACT This cohort study explores the prevalence of neighborhood disadvantage and its association with individual characteristics and mortality among Medicare beneficiaries aged 65 years or older. Introduction Neighborhood disadvantage is a well-established social determinant of health and functional well-being.1 A neighborhood may be disadvantaged based on indicators of education, employment, housing quality, and poverty. Prior research has shown that living in a socioeconomically disadvantaged neighborhood confers increased risk for several adverse outcomes, including mortality,2 hospital readmission,3 incidence and severity of delirium after major surgery,4 Alzheimer disease neuropathology,5 and reduced active life expectancy,6 among others. Since the development and release of the Neighborhood Atlas in 2018,1 hundreds of articles have been published on neighborhood disadvantage at the US Census-block level using the area deprivation index (ADI). Prior studies had largely evaluated area deprivation at less granular geographic units such as county7 or Census tract,8 which are too large and heterogeneous to represent discrete neighborhoods.9 Despite the burgeoning literature on the ADI at the Census-block level, little is known about the prevalence of neighborhood disadvantage in nationally representative samples of older adults in the US. In 2023, the National Health and Aging Trends Study (NHATS), a prospective, population-based longitudinal study of Medicare beneficiaries aged 65 years or older,10 made block group files available under limited circumstances to facilitate linkage to contextual data, such as the ADI. Using these linked data from NHATS, we set out to accomplish 3 objectives: (1) to estimate the prevalence of neighborhood disadvantage among a nationally representative sample of community-living older adults; (2) to identify how these prevalence estimates differ based on relevant demographic, socioeconomic, geographic, clinical, and geriatric characteristics; and (3) to evaluate the association between neighborhood disadvantage and all-cause mortality. The results of this study are intended to advance the understanding of neighborhood disadvantage, a granular contextual indicator of socioeconomic deprivation, at the national level and establish a strong foundation for future population-based research on health disparities. Methods Study Population On September 30, 2010, NHATS drew a random sample of persons aged 65 years or older living in the contiguous US (excluding Alaska, Hawaii, and Puerto Rico) from the Medicare enrollment file. Counties were sampled from regional strata, and non-Hispanic Black persons and those aged 90 years or older were oversampled within zip codes. Baseline (round 1) assessments, completed from May through November 2011, yielded a sample of 8245 persons with a 71% weighted response rate. Proxy respondents were interviewed when the participant could not respond (n = 583, or 5.8% weighted). This cohort has been linked to Medicare and geographic data.10 These restricted NHATS data are available through a Data Use Agreement, which requires an approved application and data protection plan. The Johns Hopkins University Institutional Review Board approved the NHATS protocol, and all NHATS participants provided written informed consent. The Yale University Institutional Review Board approved the current cohort study and its use of NHATS data. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.11 Data Collection During round 1 of NHATS, information was collected on demographic, socioeconomic, geographic, clinical, and geriatric characteristics.10,12 The demographic characteristics included age, sex, and self-reported race and ethnicity. Race and ethnicity (ie, Hispanic, non-Hispanic Black [hereafter Black], non-Hispanic White [hereafter White], and other [Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, other, do not know, or more than 1 race and ethnicity]) data were analyzed for descriptive purposes. The socioeconomic characteristics included educational level, annual income, and Medicaid eligibility, which was obtained from the Medicare records. Annual income for the participant (and partner, if applicable) was assessed using a composite from Social Security benefits; Veterans Affairs benefits; pension plans; retirement plans; mutual funds, stocks, and bonds; and checking and savings accounts.13 Missing data on income were imputed using values provided by NHATS.12 An annual income less than $15 000 was classified as low.14,15 The geographic characteristics included Census division (eFigure in Supplement 1) and rural residence based on a 2013 Rural-Urban Continuum Code of 4 to 9.16 The clinical characteristics included 9 self-reported, physician-diagnosed chronic conditions: heart attack, high blood pressure, arthritis, osteoporosis, diabetes, lung disease, stroke, cancer, and hip fracture since age 50 years. The geriatric characteristics included frailty and dementia. Participants were categorized as nonfrail, prefrail, and frail (according to the Fried phenotype17) and as having no dementia, possible dementia, or probable dementia (based on a validated assessment strategy developed by NHATS18,19). Neighborhood disadvantage was assessed with the ADI, a Census-based socioeconomic index.1,3 As shown in the eTable in Supplement 1, the ADI includes 17 educational, employment, housing quality, and poverty indicators obtained from the American Community Survey (ACS). These indicators are weighted and summed to yield a score for each neighborhood at the Census-block level.7 For the current study, we used the Neighborhood Atlas to link block-group data from NHATS to the 2013 ADI file, representing ACS data from 2009 to 2013.3 The ADI national percentiles were categorized into national deciles, with higher decile scores indicating greater socioeconomic disadvantage. Based on its established threshold effects,3 these scores were dichotomized at the eighth decile to distinguish the disadvantaged group (9-10 deciles) from the nondisadvantaged groups (1-8 deciles). Because of privacy concerns, NHATS does not permit analysis or reporting of ADI scores as percentiles.20 Analytic Sample and Outcome Among the 7609 NHATS participants who were living in settings other than nursing homes (ie, community living) at the time of their round 1 assessment, 104 (1.4%) lacked ADI scores, leaving 7505 participants in the analytic sample. ADI scores were not available for 2 reasons: missing values in the source ACS data (n = 86) and high group quarters population or low population and/or housing (n = 18), as explained elsewhere.21 For the longitudinal analysis, we evaluated mortality rates over the 10-year follow-up and time to death within 10 years. Deaths were ascertained from the Medicare enrollment files.22 Statistical Analysis NHATS survey weights were used for all analyses,23 thereby permitting national estimates. Descriptive statistics were calculated for participants in the analytic sample but could not be compared with those of the 104 excluded participants because of small, unweighted cell sizes.24 Descriptive characteristics were also calculated for the ADI decile scores, including mean (SD), median (IQR), number (%) for each decile, and number (%) for the worst quintile, denoting neighborhood disadvantage. The prevalence of neighborhood disadvantage was calculated within subgroups defined on the basis of the demographic, socioeconomic, geographic, clinical, and geriatric characteristics. Adjusted risk ratios (with 95% CIs) were subsequently calculated. The values for age were adjusted for sex, the value for sex was adjusted for age, and all other values were adjusted for age and sex. For consistency and to facilitate interpretation, comparisons for each characteristic were generally made to the group with the lowest prevalence. The corresponding P values were adjusted for multiple comparisons using the Benjamini and Hochberg procedure.25 Ten-year mortality rates were calculated, and cumulative mortality curves were generated26,27 according to the presence of neighborhood disadvantage. The independent association between neighborhood disadvantage and mortality was evaluated using a series of hierarchical Cox proportional hazards regression models that sequentially adjusted for the demographic, socioeconomic, geographic, clinical, and geriatric characteristics. Robust SEs and their corresponding 95% CIs were calculated by fitting each Cox proportional hazards regression model using the balanced repeated replication approach, as previously recommended.28,29 This approach accounts for the complex survey design,23 and it overcomes potential bias in estimated variance when proportional hazards assumptions are violated.30 All analyses were performed using SAS, version 9.4 (SAS Institute Inc) and R, version 4.4.0 (R Core Team). Statistical significance was defined as a 2-tailed P < .05. Data analysis was conducted from February to July 2024. Results As shown in Table 1, the weighted mean (SD) age of the 7505 NHATS participants in the analytic sample was 75.3 (7.4) years. Among these participants, 56.8% were females and 43.2% were males, of whom 8.2% self-identified as Black, 6.6% as Hispanic, 81.7% as White, and 3.5% as other race and ethnicity. Altogether, 71.2% did not complete college, 21.0% had an annual income less than $15 000, and 13.4% were Medicaid eligible. The most common and least common Census divisions were South Atlantic (19.5%) and Mountain (2.7%), respectively. A total of 18.1% of participants were rural residents, 19.9% had more than 4 chronic conditions, 13.8% were categorized as frail, and 10.0% had probable dementia. Table 1. Baseline Characteristics of the Analytic SampleaCharacteristicParticipants, %UnweightedWeightedNo. of participants750534 742 661Demographic Age, mean (SD), y77.7 (7.9)75.3 (7.4) Age group, y 65-6918.527.9 70-7420.724.8 75-7919.919.1 80-8419.814.7 85-8912.69.2 ≥90 8.64.3 Sex Female58.556.8 Male41.543.2 Race and ethnicityb Hispanic5.96.6 Non-Hispanic Black22.18.2 Non-Hispanic White69.181.7 Otherc2.93.5Socioeconomic Educational level High school diploma20.021.6 ≥College degree 25.128.9 Annual income, $ Median (IQR)26 903 (14 376-50 000)31 992 (16 732-59 865) <15 00026.121.0 Medicaid eligible16.513.4Geographic US Census division New England4.85.9 Middle Atlantic13.613.4 South Atlantic20.819.5 East North Central13.813.7 East South Central7.16.4 West North Central9.69.4 West South Central10.810.9 Mountain2.52.7 Pacific16.918.1 Rural residence18.518.1Clinical Heart attack15.414.1 High blood pressure67.464.1 Arthritis56.053.8 Osteoporosis20.521.2 Stroke11.810.0 Diabetes25.323.9 Lung disease15.215.4 Cancer25.725.8 Hip fracture since age 50 y5.04.1No. of chronic conditions 08.89.9 118.920.6 226.926.9 323.922.7 ≥4 21.619.9Geriatric Frailty phenotype Nonfrail34.539.9 Prefrail48.346.3 Frail17.213.8 Dementia status No dementia73.379.1 Possible dementia13.110.9 Probable dementia13.610.0Abbreviation: GED, General Educational Development. a All values were based on National Health and Aging Trends Study (NHATS) survey weights except for the unweighted numbers. A total of 104 participants were excluded from the analytic sample because scores for the area deprivation index could not be obtained. b Self-reported by NHATS participants. c Includes those who self-identified as Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, other, do not know, or more than 1 race and ethnicity. Baseline Characteristics of the Analytic Samplea Abbreviation: GED, General Educational Development. All values were based on National Health and Aging Trends Study (NHATS) survey weights except for the unweighted numbers. A total of 104 participants were excluded from the analytic sample because scores for the area deprivation index could not be obtained. Self-reported by NHATS participants. Includes those who self-identified as Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, other, do not know, or more than 1 race and ethnicity. The mean (SD) and median (IQR) ADI decile scores in the analytic sample were 5.3 (2.8) and 5 (2-7), respectively. The distribution of ADI scores by decile is shown in Figure 1. Scores in the first 7 deciles were overrepresented, while those in the last 3 deciles were underrepresented. The prevalence of neighborhood disadvantage, defined as the worst 2 deciles (or quintile), was 15.8% (95% CI, 14.9%-16.7%). Distribution of Weighted Area Deprivation Index (ADI) Scores by Decile Among Community-Living Persons in the 2011 National Health and Aging Trends Study Cohort The weighted percentages are accompanied by SE bars. Table 2 shows the prevalence rates of neighborhood disadvantage according to relevant demographic, socioeconomic, geographic, clinical, and geriatric characteristics. Prevalence ranged from 4.7% for participants in the Pacific Census division to 41.0% for Black participants. Within subgroups, large absolute differences in prevalence (ie, exceeding 10%) were observed for race and ethnicity (from 7.4% for other to 41.0% for Black), educational level (from 8.2% or college degree and higher to 27.3% for less than high school), annual income (from 13.0% for $15 000 and higher to 26.4% for less than $15 000), Medicaid (from 13.6% for not eligible to 29.9% for eligible), Census division (from 4.7% for Pacific to 34.5% for West South Central), and residence (from 12.7% for nonrural to 30.1% for rural). In relative terms, the largest subgroup differences, adjusted for age and sex, were observed for Black compared with White participants (rate ratio [RR], 3.11; 95% CI, 2.56-3.79), less than a high school diploma vs college degree and higher educational level (RR, 3.47; 95% CI, 2.75-4.39), and West South Central vs Pacific Census divisions (RR, 7.31; 95% CI, 2.98-17.90). Differences by age group were modest, with the only significant difference observed for age groups 75 to 79 years vs 90 years or older (sex-adjusted RR, 1.34; 95% CI, 1.05-1.72). Table 2. Prevalence of Neighborhood Disadvantage by Relevant Demographic, Socioeconomic, Geographic, Clinical, and Geriatric CharacteristicsaCharacteristicPrevalence, %Adjusted RR (95% CI)bP valuecDemographic Age group, y 65-6916.8 (14.8-18.8)1.33 (1.00-1.75).06 70-7414.7 (12.9-16.5)1.16 (0.90-1.50).28 75-7917.1 (15.2-19.0)1.34 (1.05-1.72).03 80-8415.2 (13.4-17.0)1.19 (0.93-1.52).17 85-8915.5 (13.1-17.8)1.21 (0.96-1.52).13 ≥90 12.9 (10.2-15.6)1 [Reference]NA Sex Female16.4 (15.3-17.6)1.10 (0.99-1.23).11 Male15.0 (13.7-16.4)1 [Reference]NA Race and ethnicityd Hispanic21.5 (17.3-25.6)1.62 (1.07-2.47).04 Non-Hispanic Black41.0 (38.4-43.6)3.11 (2.56-3.79)<.001 Non-Hispanic White13.2 (12.2-14.1)1 [Reference]NA Othere7.4 (3.5-11.2)0.55 (0.26-1.19).15Socioeconomic Educational level High school diploma14.4 (12.4-16.3)1.77 (1.37-2.28)<.001 ≥College degree8.2 (6.9-9.5)1 [Reference]NA Annual income, $ <15 00026.4 (24.1-28.6)2.07 (1.82-2.36)<.001 ≥15 000 13.0 (12.1-14.0)1 [Reference]NA Medicaid eligibility Yes29.9 (26.9-32.9)2.21 (1.92-2.54)<.001 No13.6 (12.7-14.5)1 [Reference]NAGeographic US Census division Middle Atlantic and New Englandf7.3 (5.9-8.7)1.55 (0.51-4.76).47 South Atlantic19.3 (17.1-21.5)4.09 (1.73-9.70).004 East North Central16.7 (14.2-19.1)3.57 (1.52-8.38).008 East South Central26.8 (22.6-31.0)5.70 (2.06-15.80).003 West North Central20.4 (17.3-23.4)4.34 (1.29-14.60).03 West South Central34.5 (30.9-38.1)7.31 (2.98-17.90)<.001 Mountain5.1 (1.9-8.4)1.08 (0.39-3.03).90 Pacific4.7 (3.3-6.0)1 [Reference]NA Rural residence Yes30.1 (27.5-32.7)2.39 (1.35-4.21).006 No12.7 (11.8-13.6)1 [Reference]NAClinical30.1 (27.5-32.7) Specific chronic conditions Heart attack Yes20.2 (17.6-22.7)1.38 (1.17-1.62)<.001 No15.1 (14.2-16.1)1 [Reference]NA High blood pressure Yes17.4 (16.3-18.6)1.35 (1.17-1.55)<.001 No13.0 (11.6-14.4)1 [Reference]NA Arthritis Yes17.4 (16.2-18.6)1.24 (1.10-1.40).002 No14.0 (12.7-15.2)1 [Reference]NA Osteoporosis Yes16.4 (14.4-18.4)1.01 (0.86-1.18).95 No15.7 (14.7-16.7)1 [Reference]NA Stroke Yes20.6 (17.6-23.6)1.38 (1.22-1.57)<.001 No15.3 (14.4-16.2)1 [Reference]NA Diabetes Yes20.2 (18.2-22.2)1.41 (1.25-1.59)<.001 No14.4 (13.5-15.4)1 [Reference]NA Lung disease Yes19.8 (17.3-22.4)1.30 (1.12-1.52).003 No15.1 (14.2-16.0)1 [Reference]NA Cancer Yes14.1 (12.5-15.8)0.87 (0.77-1.00).06 No16.4 (15.3-17.4)1 [Reference]NA Hip fracture since age 50 y Yes18.1 (13.7-22.5)1.18 (0.96-1.46).14 No15.7 (14.8-16.6)1 [Reference]NA No. of chronic conditions 012.4 (9.8-15.0)1 [Reference]NA 112.0 (10.2-13.7)0.97 (0.75-1.26).87 216.5 (14.8-18.3)1.35 (1.04-1.74).04 315.8 (14.0-17.6)1.30 (1.04-1.63).04 ≥420.6 (18.4-22.8)1.70 (1.33-2.18)<.001Geriatric Frailty phenotype Nonfrail12.6 (11.3-14.0)1 [Reference]NA Prefrail16.9 (15.6-18.3)1.38 (1.19-1.59)<.001 Frail21.4 (18.9-23.9)1.81 (1.50-2.17)<.001 Dementia status No dementia14.8 (13.8-15.8)1 [Reference]NA Possible dementia19.4 (16.8-22.1)1.38 (1.13-1.69).004 Probable dementia19.7 (17.0-22.3)1.44 (1.20-1.72)<.001Abbreviations: GED, General Educational Development; NA, not applicable; RR, rate ratio. a All values were based on National Health and Aging Trends Study (NHATS) survey weights. b The values for age were adjusted for sex, the value for sex was adjusted for age, and all other values were adjusted for age and sex. For consistency and to facilitate interpretation, the reference group was generally the subgroup with the lowest prevalence. c All values were corrected for multiple comparisons, with P < .05 denoting statistical significance. d Self-reported by NHATS participants. e Includes those who self-identified as Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, other, do not know, or more than 1 race and ethnicity. f New England had to be combined with Middle Atlantic because of its small unweighted cell size.24 Prevalence of Neighborhood Disadvantage by Relevant Demographic, Socioeconomic, Geographic, Clinical, and Geriatric Characteristicsa Abbreviations: GED, General Educational Development; NA, not applicable; RR, rate ratio. All values were based on National Health and Aging Trends Study (NHATS) survey weights. The values for age were adjusted for sex, the value for sex was adjusted for age, and all other values were adjusted for age and sex. For consistency and to facilitate interpretation, the reference group was generally the subgroup with the lowest prevalence. All values were corrected for multiple comparisons, with P < .05 denoting statistical significance. Self-reported by NHATS participants. Includes those who self-identified as Asian, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, other, do not know, or more than 1 race and ethnicity. New England had to be combined with Middle Atlantic because of its small unweighted cell size.24 Many of the chronic conditions were associated with neighborhood disadvantage, including heart attack, high blood pressure, arthritis, stroke, diabetes, and lung disease, with age- and sex-adjusted RRs ranging from 1.24 (95% CI, 1.10-1.40) for arthritis to 1.41 (95% CI, 1.25-1.59) for diabetes. Significant differences were also observed for 2 (RR, 1.35; 95% CI, 1.04-1.74), 3 (RR, 1.30; 95% CI, 1.04-1.63), and 4 or more (RR, 1.70; 95% CI, 1.33-2.18) chronic conditions vs 0. For the geriatric characteristics, prefrail and frail vs nonfrail as well as possible dementia and probable dementia vs no dementia were all associated with neighborhood disadvantage, with age- and sex-adjusted RRs of 1.81 (95% CI, 1.50-2.17) for the frail phenotype and 1.44 (95% CI, 1.20-1.72) for probable dementia. Over the 10-year follow-up period, there were 3980 deaths, representing 15 389 064 survey-weighted deaths and 44.3% (95% CI, 43.0%-45.6%) mortality. The mortality rates were 48.5% (95% CI, 44.6%-52.1%) and 43.5% (95% CI, 42.2%-44.7%) among participants in a disadvantaged and nondisadvantaged neighborhood, respectively. Cumulative mortality for these 2 groups is shown in Figure 2. In a series of hierarchical Cox proportional hazards regression models, shown in Table 3, neighborhood disadvantage was associated with mortality over 10 years in the unadjusted analysis (hazard ratio [HR], 1.18; 95% CI, 1.05-1.33) and after adjustment for the demographic characteristics (HR, 1.25; 95% CI, 1.11-1.40). This finding was no longer statistically significant after adjustment for the socioeconomic characteristics (HR, 1.11; 95% CI, 0.98-1.25) but did not change much after additional adjustment for the geographic, clinical, and geriatric characteristics, respectively. Cumulative Mortality Over 10 Years by Neighborhood Disadvantage National Health and Aging Trends Study–weighted values are shown, accompanied by SE bands. Fewer values are provided for the disadvantaged group because of small unweighted cell sizes.24 For all participants, the median (IQR) time to death was 5 (3-8) years. Table 3. Associations Between Neighborhood Disadvantage and Mortality Over 10 YearsModel No.CovariatesaHazard ratio (95% CI)bP valuec1Unadjusted1.18 (1.05-1.33).0082Demographic characteristics1.25 (1.11-1.40)<.0013Demographic and socioeconomic characteristics1.11 (0.98-1.25).094Demographic, socioeconomic, and geographic characteristics1.07 (0.95-1.21).265Demographic, socioeconomic, geographic, and clinical characteristics1.04 (0.92-1.16).566Demographic, socioeconomic, geographic, clinical, and geriatric characteristics1.06 (0.94-1.19).32 a Demographic characteristics included age, sex, and self-reported race and ethnicity. Socioeconomic characteristics included educational level, annual income, and Medicaid eligibility. Geographic characteristics included US Census division and rural residence. Clinical characteristics included number of physician-diagnosed chronic conditions. Geriatric characteristics included frailty and dementia. b Values were calculated in a series of hierarchical Cox proportional hazards regression models that sequentially adjusted for the demographic, socioeconomic, geographic, clinical, and geriatric characteristics. c P < .05 indicated statistical significance. Associations Between Neighborhood Disadvantage and Mortality Over 10 Years Demographic characteristics included age, sex, and self-reported race and ethnicity. Socioeconomic characteristics included educational level, annual income, and Medicaid eligibility. Geographic characteristics included US Census division and rural residence. Clinical characteristics included number of physician-diagnosed chronic conditions. Geriatric characteristics included frailty and dementia. Values were calculated in a series of hierarchical Cox proportional hazards regression models that sequentially adjusted for the demographic, socioeconomic, geographic, clinical, and geriatric characteristics. P < .05 indicated statistical significance. Discussion In this nationally representative sample of community-living Medicare beneficiaries aged 65 years or older, we found that the prevalence of neighborhood disadvantage at the Census-block level was 15.8%. The prevalence differed significantly, however, according to many demographic, socioeconomic, geographic, clinical, and geriatric characteristics, with the largest absolute and relative differences observed for subgroups defined on the basis of race and ethnicity, educational level, and Census division, although substantive differences were also observed according to annual income, rural residence, several individual and multiple chronic conditions, frailty, and dementia. Living in a disadvantaged neighborhood was associated with an 18% increase in mortality over 10 years. This elevation in mortality risk was modestly increased after adjustment for the demographic characteristics but was considerably decreased and became nonsignificant after adjustment for socioeconomic characteristics, such as educational level, annual income, and Medicaid eligibility. The results from this population-based study demonstrate substantial heterogeneity in the prevalence of neighborhood disadvantage across multiple important subgroups of older adults in the US and suggest that the association between neighborhood disadvantage and long-term mortality may be attributable in large part to individual-level socioeconomic characteristics. Population-based national estimates of neighborhood disadvantage are relatively sparse, with prior studies focusing on specific regions of the country31 or patient groups, such as veterans.32 To our knowledge, none has focused on community-living older persons or evaluated how neighborhood disadvantage differs according to relevant subgroups. Inherently, the prevalence of neighborhood disadvantage depends on the threshold used to dichotomize scores on the ADI. Although we chose the worst quintile, which has been commonly used in other studies,4,5,6 the prevalence of neighborhood disadvantage in the current study was only 15.8%. The lower-than-expected prevalence in this nationally representative sample could be due, at least in part, to the stratified multilevel sampling strategy of NHATS, which was based on select demographic and geographic factors.33 Although neighborhood disadvantage is a distinct social determinant of health, it mapped closely to participants’ race and ethnicity and to each of the individual-level socioeconomic characteristics, including educational level, income, and Medicaid eligibility. For example, 27.3% of participants with less than a high school education lived in a disadvantaged neighborhood vs only 8.2% of those with a college degree, leading to an adjusted RR of 3.47. These disparities for living in a disadvantaged neighborhood were also large for participants who had low annual income vs those who did not (26.4% vs 13.0%) and participants who were Medicaid eligible vs those who were not (29.9% vs 13.7%), leading to adjusted RRs that exceeded 2.0 for each characteristic. These results indicate that many older persons are disadvantaged socioeconomically at both the contextual and individual levels. The prevalence of neighborhood disadvantage was also substantially higher in rural than in nonrural communities and differed greatly by Census division in the contiguous US, with the highest rates observed in the West South Central and East South Central divisions and lowest rates observed in the Pacific and Mountain divisions. An earlier report showed similar results graphically by ADI deciles.1 Clinically, participants were more likely to live in a disadvantaged neighborhood if they had multimorbidity, defined as 2 or more chronic conditions, or if they had specific chronic conditions, such as heart attack, high blood pressure, stroke, diabetes, and lung disease, that disproportionately affect older persons with low socioeconomic status. Similar results were observed among participants who had frailty or dementia, 2 common geriatric conditions. Although temporal precedence cannot be established given the cross-sectional nature of these associations, the vulnerability to adverse outcomes among older persons living in disadvantaged neighborhoods may be further heightened when these clinical and geriatric risk factors are present. Living in a disadvantaged neighborhood was associated with an increased risk of death over the course of 10 years, even after adjustment for age, sex, and race and ethnicity but not after further adjustment for the socioeconomic, geographic, clinical, and geriatric characteristics. The results of the hierarchical models suggest that adjustment for the socioeconomic characteristics had the largest role in the reduction of mortality risk. While prior nationally representative studies have demonstrated associations between area deprivation at the county level and 1-year mortality,7 to our knowledge, no study has evaluated the association between neighborhood disadvantage at the Census-block level and long-term mortality or has attempted to ascertain whether such associations persist after serial adjustment for demographic, socioeconomic, geographic, clinical, and geriatric characteristics. As the smallest available geographic unit of analysis, typically ranging from 600 to 3000 persons, Census-block groups likely capture individual-level exposure to neighborhood disadvantage more precisely than larger Census units.9 These smaller geographic units may be more amenable to targeted interventions designed to address deficiencies in resource-poor environments.34 Two additional strengths enhance the generalizability, validity, and applicability of these findings. First, by linking ADI data from the Neighborhood Atlas to participants of a well-phenotyped, population-based cohort from the NHATS, we were able to generate nationally representative estimates of neighborhood disadvantage in numerous subgroups of community-living older adults across multiple domains. This wealth of data is not available in nationally representative administrative datasets, such as those from the Centers for Medicare & Medicaid Services. Second, the ascertainment of neighborhood disadvantage was nearly 100% complete, and the ascertainment of death over 10 years was 100% complete. Limitations This study has limitations. First, the results may not be applicable to the 4% of US adults aged 65 years or older who did not have Medicare in 2011 or to Medicare recipients outside of the contiguous US.10 Second, because NHATS does not permit analysis or reporting of ADI scores as percentiles, we were unable to evaluate whether or how the results might differ based on thresholds other than worst ADI quintile for identifying disadvantaged neighborhoods. The 2 most common alternative thresholds—worst 15th percentile and worst quartile—would lead to lower and higher prevalence estimates, respectively, but may not have substantively affected the subgroup differences or mortality results. Third, it was not possible to report the prevalence of neighborhood disadvantage (and associated RR) separately for the New England Census division because of its small unweighted cell size.24 Conclusions In this cohort study, the population-based estimates of neighborhood disadvantage differed greatly across multiple subgroups of community-living older adults in the US. This contextual indicator of socioeconomic deprivation was associated with long-term mortality, but the finding was no longer significant after accounting for individual-level socioeconomic characteristics.