Abstract
Background and Objectives
Patients of low individual socioeconomic status (SES) are at a greater risk of unfavorable health outcomes. However, the association between neighborhood socioeconomic deprivation and health outcomes for patients with neurologic disorders has not been studied at the population level. Our objective was to determine the association between neighborhood socioeconomic deprivation and 30-day mortality and readmission after hospitalization for various neurologic conditions.
Methods
This was a retrospective study of nationwide Medicare claims from 2017 to 2019. We included patients older than 65 years hospitalized for the following broad categories based on diagnosis-related groups (DRGs): multiple sclerosis and cerebellar ataxia (DRG 058–060); stroke (061–072); degenerative nervous system disorders (056–057); epilepsy (100–101); traumatic coma (082–087), and nontraumatic coma (080–081). The exposure of interest was neighborhood SES, measured by the area deprivation index (ADI), which uses socioeconomic indicators, such as educational attainment, unemployment, infrastructure access, and income, to estimate area-level socioeconomic deprivation at the level of census block groups. Patients were grouped into high, middle, and low neighborhood-level SES based on ADI percentiles. Adjustment covariates included age, comorbidity burden, race/ethnicity, individual SES, and sex.
Results
After exclusions, 905,784 patients were included in the mortality analysis and 915,993 were included in the readmission analysis. After adjustment for age, sex, race/ethnicity, comorbidity burden, and individual SES, patients from low SES neighborhoods had higher 30-day mortality rates compared with patients from high SES neighborhoods for all disease categories except for multiple sclerosis: magnitudes of the effect ranged from an adjusted odds ratio of 2.46 (95% CI 1.60–3.78) for the nontraumatic coma group to 1.23 (95% CI 1.19–1.28) for the stroke group. After adjustment, no significant differences in readmission rates were observed for any of the groups.
Discussion
Neighborhood SES is strongly associated with 30-day mortality for many common neurologic conditions even after accounting for baseline comorbidity burden and individual SES. Strategies to improve health equity should explicitly consider the effect of neighborhood environments on health outcomes.
Individual socioeconomic status (SES) and area-level or neighborhood-level socioeconomic deprivation have both been independently linked to a broad range of health outcomes.1 In neurology, area SES has been linked with outcomes for patients with conditions such as multiple sclerosis, epilepsy, and stroke.2-4 Furthermore, individual-level SES is associated with incidence and outcomes for a variety of neurologic conditions, including stroke and dementia.5,6 However, no studies have evaluated the contribution of neighborhood SES with outcomes for several common neurologic conditions, after adjusting for key risk factors such as clinical comorbidities, age, sex, and individual poverty using a large cohort of population-based data.
Understanding the contribution of neighborhood socioeconomic deprivation to outcomes for patients with neurologic diseases would have significant public policy and clinical practice implications. Currently, reimbursement systems implemented by the Centers for Medicare and Medicaid Services (CMS) evaluate hospital performance after adjusting for demographics and health status, but not individual-level or area-level SES. Furthermore, clinicians and health systems could use information about the association between area-level SES and 30-day mortality and readmission to devise and implement interventions to improve outcomes and health equity. Therefore, we studied the association between neighborhood SES and 30-day mortality and readmission, after adjusting for key confounding factors, in a nationwide cohort of Medicare beneficiaries. We hypothesized that neighborhood socioeconomic deprivation would be linked with worse outcomes for a broad range of common inpatient neurologic conditions.
Methods
Standard Protocol Approvals, Registrations, and Patient Consents
This study was approved by the Duke University Institutional Review Board.
Data Source
We used 2017–2019 data available from the CMS for a 100% sample of Medicare beneficiaries for this analysis. Inpatient claims data were used to select study-eligible admissions and to identify readmissions. Inpatient, outpatient, and professional (carrier) claims were used to identify comorbid conditions. Medicare Beneficiary Summary Files were used to determine beneficiary demographics, enrollment periods, program eligibility, and mortality. We obtained ZIP + 4 codes from CMS and linked these to census block groups for the purpose of determining a beneficiary's area-level SES.
Study Population
We selected hospitalizations for common neurologic disorders and grouped them into 6 broad categories based on diagnosis-related groups (DRGs): multiple sclerosis and cerebellar ataxia (DRG 058–060); stroke, which included acute ischemic stroke, intracranial hemorrhage, transient ischemia, and nonspecific cerebrovascular disorders (DRG 061–072); degenerative nervous system disorders (DRG 056–057); epilepsy (DRG 100–101); nontraumatic coma (DRG 080–081); and traumatic coma (DRG 082–087). We used DRGs—which are themselves based on ICD codes—to group hospitalizations instead of individual ICD code algorithms for 2 reasons: first, DRGs are specifically tied to the principal reason for admission and effectively aggregate many closely related ICD codes, allowing us to representatively capture patients admitted with similar conditions without fractionating the population into an excessive number of individual groups; second, DRGs are used to drive reimbursement for inpatient admissions and are used for influential ranking and quality metrics, including internal CMS quality metrics and US News and World Report rankings. Common ICD codes in the primary diagnosis position for each group are summarized in eTable 1 (links.lww.com/WNL/C638). To be included in the analysis, the beneficiaries associated with these hospitalizations must have been aged 65 years or older at admission and must have been continuously enrolled in fee-for-service Medicare for the 365 days before admission.
We created 2 analysis cohorts from these hospitalizations, one for the mortality analysis and the other for the readmission analysis. There were additional exclusions specific to each of these cohorts. For the mortality cohort, we required that beneficiaries were continuously enrolled in fee-for-service Medicare for the 30 days after admission or until death if they died in this 30-day window. We also excluded hospitalizations that started as transfers from another acute-care hospital (including the emergency department of another hospital). For the readmission cohort, we required that beneficiaries were continuously enrolled in fee-for-service Medicare for the 30 days after discharge or until death if they died in this 30-day window. We also excluded hospitalizations that resulted in a transfer to another acute-care hospital and hospitalizations where the patient died before discharge, left against medical advice, or were discharged on the same day as the admission, concordant with CMS methodology; we recognized that these exclusion criteria may introduce bias regarding study population (e.g., some of the sickest patients may be excluded because of being transferred to a higher level of care); however, we performed our analysis in line with CMS methodology to maximize relevance to payment systems and health policy. In addition, any hospitalization that itself was a readmission within 30 days of discharge from another hospitalization within the same category was excluded from the readmission cohort. In both cohorts, for patients hospitalized more than once in the study period, a single random hospitalization per person per admission category was selected for analysis, similar to CMS methodology for risk adjustment.7
Neighborhood SES
The primary exposure of interest is the area deprivation index (ADI), which uses area-level information about socioeconomic indicators, such as educational attainment, unemployment, infrastructure access, and income, to estimate area-level socioeconomic deprivation at the level of census block groups.8 For analysis, the ADI was classified into high SES (ADI 1–15), middle SES (ADI 16–85), and low SES (ADI 86–100), based on findings from prior research.1 Discharges missing information necessary to calculate the ADI were excluded from the analysis.
Outcomes
The 2 outcomes of interest were 30-day mortality, from admission, and 30-day unplanned readmission, from discharge. We used the CMS methodology for classifying admissions as planned or unplanned.9
Other Covariates
Other information used for risk adjustment included beneficiary demographics (age, sex, and race/ethnicity), Medicaid dual eligibility status (as a proxy for individual SES), end-stage renal disease status, discharge year, and the 29 Elixhauser comorbid conditions.10,11 Comorbidities were defined using claims data over the 365 days before admission. For summary metrics of comorbidity burden in the mortality and readmission cohorts, we additionally calculated the Elixhauser comorbidity index for mortality and readmission, respectively. Comorbidities that were either universally absent or nearly universally present in a given admission group were excluded from the logistic regression modeling: autoimmune deficiency syndrome, end-stage renal disease, and other neurologic disorders were excluded for the multiple sclerosis group; hypertension was excluded from the stroke group; other neurologic disorders were excluded from the epilepsy group. We likewise assessed other covariates for collinearity, particularly Medicaid dual eligibility and the ADI, and found that these variables were not collinear.
Statistical Analysis
We summarized the characteristics of study population using frequencies and percentages for categorical variables and mean values with SDs for continuous variables.
We used generalized estimating equation methods (to account for clustering of patients within hospitals) to estimate logistic regression models for both outcomes. Unadjusted models estimated the association between ADI groups and outcomes only, while adjusted models controlled for age, sex, race/ethnicity as defined by the CMS data source as Asian, Black, White, Hispanic, and Other or Unknown, Medicaid dual eligibility status, end-stage renal disease status, discharge year, and the Elixhauser conditions. We presented the results as odds ratios with 95% CIs.
Data Availability
Data may be obtained after execution of a data use agreement from the CMS.
Results
A total of 905,784 admissions were eligible for and included in the mortality analysis, and 915,993 admissions were included in the readmission analysis (eFigure 1, links.lww.com/WNL/C638). The stroke group was the largest group, and the multiple sclerosis group was the smallest for both the mortality and readmission analyses. Demographic and available clinical characteristics of the cohorts, stratified by disease group and ADI group, are summarized in Table 1 for the mortality cohort and Table 2 for the readmission cohort. In both the mortality and readmission cohorts and across all disease groups, the low ADI (high neighborhood SES) groups were older, had a higher percentage of male beneficiaries, had a lower percentage of patients dually eligible for Medicare and Medicaid, and had lower Elixhauser mortality index and readmission index scores, reflecting a lower comorbidity burden compared with the high ADI (low neighborhood SES) groups (p < 0.001 for all associations for the pooled cohort).
Table 1.
Characteristics of the Mortality Cohort, by Reason for Admission Group, by Neighborhood SES
Table 2.
Characteristics of the Readmission Cohort, by Reason for Admission Group, by Neighborhood SES
Unadjusted estimates of the association between neighborhood SES and 30-day mortality are summarized in Table 3. Unadjusted 30-day mortality rates were higher for patients from the lowest neighborhood SES group than the highest neighborhood SES group for patients admitted for traumatic coma (odds ratio [OR] 1.26, 95% CI [1.18–1.36]), nontraumatic coma (OR 1.79 [1.20–2.68]), and epilepsy (OR 1.13 [1.02, 1.26]), but there were no significant differences for the other admission groups. After adjustment for age, Medicare-Medicaid dual eligibility (as a proxy for individual SES), sex, race/ethnicity, and medical comorbidities, all disease groups except for multiple sclerosis/cerebellar ataxia showed greater odds of 30-day mortality in middle and low SES groups compared with those in high SES groups (Figure 1). Adjusted odds ratios ranged from 2.46 (1.60–3.78) for the nontraumatic coma group to 1.23 (1.19–1.28) for the stroke group.
Table 3.
Unadjusted Associations Between Neighborhood Socioeconomic Status and 30-Day Mortality and 30-Day Readmission, Stratified by Admission Group
Figure 1. Adjusted Associations Between Area Socioeconomic Status and 30-Day Mortality, Stratified by Admission Group.
ADI = area deprivation index; OR = odds ratio; SES = socioeconomic status.
Unadjusted estimates of the association between readmission and 30-day mortality are summarized in Table 3. Unadjusted 30-day readmission rates were higher for patients from the lowest neighborhood SES group than the highest neighborhood SES group for patients admitted for stroke (OR 1.17 [1.13, 1.21], epilepsy (OR 1.13 [1.04, 1.22], neurodegenerative diseases (1.12 [1.03, 1.22]), and traumatic coma (OR 1.13 [1.05, 1.21]). After multivariable adjustment, no difference in odds of 30-day readmission were seen for any group (Figure 2).
Figure 2. Adjusted Associations Between Area Socioeconomic Status and 30-Day Readmission, Stratified by Admission Group.
ADI = area deprivation index; OR = odds ratio; SES = socioeconomic status.
Discussion
This study evaluated the association between neighborhood socioeconomic deprivation and 30-day mortality and readmission for 5 common groups of neurologic disorders (multiple sclerosis and cerebellar ataxia, stroke, degenerative nervous system disorders, epilepsy, and coma) in a 100% sample of Medicare claims from 2017 to 2019, after adjusting for key covariates such as age, sex, race/ethnicity, individual SES, and comorbid medical conditions. Our principal findings were as follows: (1) after adjustment, 30-day mortality was substantially higher in all disease groups except for multiple sclerosis, for patients both in low and middle neighborhood SES groups compared with the high area SES group; (2) after adjustment, 30-day readmission was not different for any group; (3) multivariable adjustment resulted in the relationship between neighborhood SES and 30-day readmission for patients with stroke, neurodegenerative diseases, epilepsy, and coma no longer being statistically significant, but increased the apparent magnitude of the relationship between neighborhood SES and 30-day mortality for these 5 disease groups compared with unadjusted estimates.
The fact that 30-day mortality was strongly associated with lower neighborhood SES for all disease groups has major implications for practice and policy. The magnitude of the effect of neighborhood SES on outcomes for these conditions are substantial: after adjustment, 30-day mortality rates were 23% higher for stroke, 38% higher for degenerative nervous system disorders, 34% higher for epilepsy, 44% higher for traumatic coma, and 146% higher for nontraumatic coma in the low neighborhood SES group than the high neighborhood SES group. These effect sizes are especially shocking in the context of existing transformative interventions in neurology: for example, the use of mechanical thrombectomy combined with best medical therapy is associated with a 17% relative decrease in 3-month mortality for patients with acute large vessel ischemic stroke.12 Notably, the increased rate of 30-day mortality in the stroke disease group is the smallest of the disease groups; by contrast, the 146% increased 30-day mortality for patients from low SES neighborhoods with nontraumatic coma is especially striking.
We suggest that 2 broad mechanisms may contribute to the neighborhood socioeconomic disparities in 30-day mortality seen in our study: (1) patients from low SES neighborhoods may be at a higher risk of mortality due to higher baseline risk or differing underlying pathology, and (2) patients from low SES neighborhoods may receive different clinical care or access to protective community and societal resources than patients from high SES neighborhoods, affecting mortality risk. In the first mechanism, it is possible that patients in areas with high amounts of socioeconomic deprivation also have lower personal incomes, inhibiting access to health care and healthy living resources; however, we adjusted for Medicare-Medicaid dual eligibility status, a coarse metric for individual poverty. Furthermore, decreased neighborhood-level access to healthcare resources and community supports and an increased exposure to unfavorable neighborhood conditions, such as heavy metals, pesticides, and noise pollution, could increase the underlying comorbidity burden and decrease the baseline health status of patients from high ADI areas.13-15
In the second mechanism, it is possible that patients from low SES neighborhoods have decreased access to high-quality health care, either preadmission or during their hospital admission, or that patients from low SES areas receive different clinical care due to altered decision-making or bias on the part of clinicians. Both mechanisms are consistent with the growing consensus in the literature that structural, crosscutting systems of inequity establish and propagate health disparities in several domains, notably race, ethnicity, and SES.
In contrast to the disparities seen in mortality, low neighborhood SES was not associated with 30-day readmission rates for any condition. It is possible that once patients survive an incident admission for the conditions in the study, their neighborhood discharge environment has comparatively little impact on subsequent readmission. For example, prior research evaluating predictors for readmission after traumatic brain injury suggests that prehospital factors such as age, mechanism of injury (falls vs motor vehicle collision), severity of injury, increased comorbidity burden, and skilled nursing facility or long-term acute care discharge disposition were most strongly associated with 30-day readmission.16,17
Our results are overall consistent with results seen in other jurisdictions, including the United Kingdom, Australia, China, and the Netherlands, where cohort studies have generally shown that neighborhood disadvantage is linked to the incidence of neurologic disorders and unfavorable outcomes for a variety of neurologic conditions, including dementia and stroke; however, 1 study based in South Korea did not observe any relationship between neighborhood deprivation and mortality for patients with Parkinson disease.18-22
Strengths of our study include its large geographically representative sample size of more than 1 million beneficiaries and its estimation of ADI at the census block group level, allowing for more spatial precision than other methods of geographic linkage. Additional strengths include the ability to control for confounding demographic and available clinical variables and the fact that Medicare data capture hospitalizations at any facility, preventing the lack of ascertainment due to out-of-network readmissions that can be seen in electronic health record or commercial claim–based studies. Limitations of our study include its use of Medicare data, meaning that results may not be generalizable to other patient populations or patients younger than 65 years. Our adjustment covariate for individual SES, Medicare-Medicaid dual eligibility, is a relatively crude measure (in contrast to more precise metrics such as household income and educational attainment), and unfortunately, more precise means of adjusting for individual SES are unavailable in Medicare data. Furthermore, our study, similar to all studies of Medicare claims data, lacks more granular information about patient disease severity at presentation and data about lifestyle characteristics. For example, we were unable to incorporate clinical information (e.g., the Expanded Disability Status Scale for disease severity in MS) and lifestyle factors.
Future studies using registry or electronic health record data would enrich the literature on this topic. In addition, our study was designed to investigate the effect of neighborhood SES on outcomes independent of the effect of several baseline covariates that merit further exploration, particularly race/ethnicity (as a proxy for exposure to racism), sex identity (as a proxy for exposure to sex-based discrimination), and individual SES; Medicare data have limitations in assessing these relationships, partly because of poor data definition (e.g., Medicare data report legal sex and not sex identity, preventing more thoughtful analysis of sex disparities). Finally, our study was a retrospective observational study, preventing determination of causality.
In conclusion, our study provides strong evidence that, even after adjustment for age, sex, race/ethnicity, individual-level SES, and baseline comorbid condition burden, lower neighborhood SES is associated with a substantially higher 30-day mortality for patients admitted for epilepsy, stroke, neurodegenerative diseases, and coma. In addition, lower neighborhood SES is associated with substantially higher 30-day readmission rates for patients with stroke and neurodegenerative diseases. These findings have broad implications for clinical practice, research, and future quality initiatives: researchers, policymakers, and practitioners should consider the effect of neighborhood socioeconomic deprivation when designing risk adjustment models, considering the effect of individual SES or performing place-based analyses of risk factors and outcomes.
Glossary
- ADI
area deprivation index
- CMS
Centers for Medicare and Medicaid Services
- DRGs
diagnosis-related groups
- ICD
International Classification of Diseases
- OR
odds ratio
- SES
socioeconomic status
Appendix. Authors
Study Funding
This work was internally funded by the Duke University Health System. The sponsor had no role in the interpretation of the study results.
Disclosure
The authors report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.
References
- 1.Kind AJH, Jencks S, Brock J. Neighborhood socioeconomic disadvantage and 30 day rehospitalizations: an analysis of Medicare DataNeighborhood socioeconomic disadvantage and 30-day rehospitalization: a retrospective cohort study. Ann Intern Med. 2014;161(11):765-774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Calocer F, Dejardin O, Kwiatkowski A, et al. Socioeconomic deprivation increases the risk of disability in multiple sclerosis patients. Mult Scler Relat Disord. 2020;40:101930. [DOI] [PubMed] [Google Scholar]
- 3.Chen R, McKevitt C, Rudd AG, Wolfe CDA. Socioeconomic deprivation and survival after stroke. Stroke. 2014;45(1):217-223. [DOI] [PubMed] [Google Scholar]
- 4.Cihan E, Hesdorffer DC, Brandsoy M, et al. Socioeconomic disparities in SUDEP in the US. Neurology. 2020;94(24):e2555-e2566. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Russ TC, Stamatakis E, Hamer M, Starr JM, Kivimäki M, Batty GD. Socioeconomic status as a risk factor for dementia death: individual participant meta-analysis of 86 508 men and women from the UK. Br J Psychiatry. 2013;203(1):10-17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Marshall IJ, Wang Y, Crichton S, McKevitt C, Rudd AG, Wolfe CDA. The effects of socioeconomic status on stroke risk and outcomes. Lancet Neurol. 2015;14(12):1206-1218. [DOI] [PubMed] [Google Scholar]
- 7.Center for Medicare and Medicaid Services. Mortality measures methodology [online]. 2021. Accessed February 17, 2022. qualitynet.cms.gov/inpatient/measures/mortality/methodology
- 8.Singh GK. Area deprivation and widening inequalities in US mortality, 1969-1998. Am J Public Health. 2003;93(7):1137-1143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Centers for Medicare and Medicaid Services. Readmission measures methodology [online]. 2022. Accessed May 4, 2022. qualitynet.cms.gov/inpatient/measures/readmission/methodology
- 10.Lewis VA, Joynt Maddox K, Austin AM, Gottlieb DJ, Bynum JPW. Developing and validating a measure to predict poverty in Medicare administrative data. Med Care. 2019;57(8):601-607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130-1139. [DOI] [PubMed] [Google Scholar]
- 12.Katsanos AH, Malhotra K, Goyal N, et al. Mortality risk in acute ischemic stroke patients with large vessel occlusion treated with mechanical thrombectomy. J Am Heart Assoc. 2019;8(21):e014425. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Weuve J, D'Souza J, Beck T, et al. Long-term community noise exposure in relation to dementia, cognition, and cognitive decline in older adults. Alzheimer's Demen J Alzheimer's Assoc. 2021;17(3):525-533. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Fluegge K, Fluegge K. Air pollution and risk of hospitalization for epilepsy: the role of farm use of nitrogen fertilizers and emissions of the agricultural air pollutant, nitrous oxide. Arq Neuropsiquiatr. 2017;75(9):614-619. [DOI] [PubMed] [Google Scholar]
- 15.Kulick ER, Elkind MSV, Boehme AK, et al. Long-term exposure to ambient air pollution, APOE-ε4 status, and cognitive decline in a cohort of older adults in northern Manhattan. Environ Int. 2020;136:105440. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Pollifrone M, Callender L, Bennett M, et al. Predictors for 30-day readmissions after traumatic brain injury. J Head Trauma Rehabil. 2021;36(3):E178-E185. [DOI] [PubMed] [Google Scholar]
- 17.Hoffman H, Furst T, Jalal MS, Chin LS. Costs and predictors of 30-day readmissions after craniotomy for traumatic brain injury: a nationwide analysis. J Neurosurg. 2019;133(3):875-883. [DOI] [PubMed] [Google Scholar]
- 18.de Koning JS, Klazinga N, Koudstaal PJ, et al. Deprivation and systematic stroke prevention in general practice: an audit among general practitioners in the Rotterdam region, The Netherlands. Eur J Public Health. 2003;13(4):340-346. doi: 10.1093/eurpub/13.4.340. [DOI] [PubMed] [Google Scholar]
- 19.Hamilton CA, Matthews FE, Erskine D, Attems J, Thomas AJ. Neurodegenerative brain changes are associated with area deprivation in the United Kingdom: findings from the Brains for Dementia Research study. Acta Neuropathol Commun. 2021;9(1):198. doi: 10.1186/s40478-021-01301-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ouyang F, Wang Y, Huang W, et al. Association between socioeconomic status and post-stroke functional outcome in deprived rural southern China: a population-based study. BMC Neurol. 2018;18(1):12. doi: 10.1186/s12883-018-1017-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Yoon SY, Shin J, Chang JS, Lee SC, Kim YW. Effects of socioeconomic status on mortality after Parkinson's disease: a nationwide population-based matched cohort study in Korean populations. Parkinsonism Relat Disord. 2020;80:206-211. [DOI] [PubMed] [Google Scholar]
- 22.Pase MP, Rowsthorn E, Cavuoto MG, et al. Association of neighborhood-level socioeconomic measures with cognition and dementia risk in Australian adults. JAMA Netw Open. 2022;5(3):e224071. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data may be obtained after execution of a data use agreement from the CMS.