The Brain Vulnerability Index: Development and Validation of a Machine Learning–Derived, Community-Informed Geospatial Risk Score for Cognitive Impairment

The Brain Vulnerability Index: Development and Validation of a Machine Learning–Derived, Community-Informed Geospatial Risk Score for Cognitive Impairment | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article The Brain Vulnerability Index: Development and Validation of a Machine Learning–Derived, Community-Informed Geospatial Risk Score for Cognitive Impairment Yusuf Tamer, Autumn Noon, Zhiyi Yang, Kexin Yu, Amil Shah, Albert Karam, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9381770/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 12 You are reading this latest preprint version Abstract Background Social determinants of health (SDOH) are increasingly recognized as important contributors to cognitive impairment, including Alzheimer’s disease and related dementias. Existing indices are heavily weighted toward financial indicators and are not validated against cognitive outcomes. We aimed to develop and validate a novel brain-specific SDOH index that identifies high-risk communities for cognitive impairment. Methods The Brain Vulnerability Index (BVI) integrates patient-level electronic health record data with neighborhood-level SDOH from the Community Vulnerability Compass, a population health analytics framework. In the model development phase, electronic health record data were obtained from patients seen at Parkland Health (Dallas, TX). Cases were defined as individuals with at least one ICD-10 diagnosis suggestive of cognitive impairment from 2015 to 2023 and controls were age-matched with no ICD-10 diagnosis. External validation against serial cognitive performance measured by the Montreal Cognitive Assessment (MoCA) and clinically adjudicated consensus diagnosis of cognitive impairment of any etiology was performed in community-based (Dallas Heart Study) and clinically based (Alzheimer’s Disease Research Center) cohorts. Results The model development sample included 39,570 cases and 192,060 controls. Derived BVI at the block group level achieved detection of electronic health documentation suggestive of cognitive impairment at a balanced accuracy of 54.6%. In external validation (n = 1,395), higher BVI values were associated with lower MoCA scores (β = −0.35; p = 0.009) and with faster cognitive decline (− 1.04 vs − 0.42 MoCA points/year in high- vs low-risk groups; p < 0.0001). BVI was also associated with adjudicated cognitive impairment in the community-based cohort only (p = 0.04). Across analyses, BVI correlated with but outperformed existing neighborhood indices. Conclusions BVI, a novel multidimensional brain-specific SDOH index, is clinically anchored and is externally validated. It may assist in earlier identification of at-risk communities with accelerated cognitive decline and aid in resource allocation for targeted prevention strategies. Health sciences/Diseases Health sciences/Health care Health sciences/Medical research Health sciences/Neurology Biological sciences/Neuroscience Figures Figure 1 Figure 2 Background Alzheimer’s disease (AD) and related dementias (ADRD) are neurodegenerative diseases with AD being the fifth-leading cause of death in the US. 1,2 AD impacted an estimated 6.9 million Americans 65 and older in 2024 and is expected to affect 13.8 million Americans by 2060. 3 The percentage of Americans with AD increases with age: 5.0% of people aged 65–74, 13.2% of people aged 75 to 84, and 33.4% of people 85 and older. 1,2 Despite great advances in our understanding of ADRD, nearly half of cognitive impairment risk remains unexplained. 4 More recently, risk models suggest a strong contribution of non-medical drivers to cognitive impairment risk. 5 Identifying elevated risk groups, especially beyond traditional risk factors, improves efficiencies of prevention programs as well as case ascertainments and identifying those that would benefit most from therapeutic programs. Increasingly, social determinants of health (SDOH), and more broadly non-medical drivers of health (NMDOH), have been identified as important, potentially modifiable risk factors for AD and ADRD. 6–8 Currently available measures of SDOH are not brain-specific, emphasize financial and structural vulnerability, and are rarely operationalized as patient-level tools due to the lack of interpretable thresholds for use in clinical practice. 9,10 For example, the Area Deprivation Index (ADI) is widely used to quantify neighborhood socioeconomic disadvantage in ADRD research, However, it provides generic, ranking of areas masking within‑neighborhood heterogeneity and lacks calibration as an individual‑level brain health risk tool. 11 , 12 To achieve progress towards clinical implementation, we need new tools encompassing a wide variety of measures that are highly relevant to ADRD and that identify patient-level risk. One emerging approach is to use machine learning (ML) and other artificial intelligence (AI) methods that jointly analyze high‑dimensional clinically and non‑clinically derived data, including a wide range of SDOH indicators, to identify an individual’s risk of cognitive impairment. 13 Fragmented diverse social data sources lacking clinical context have hindered the creation of more advanced SDOH indices. The Parkland Center for Clinical Innovation (PCCI), a North Texas nonprofit specializing in data-driven solutions, developed the Community Vulnerability Compass (CVC) to bridge the gap between socioeconomic data and medical records. Through its unique partnership with Parkland Health, PCCI integrates deep clinical datasets with various public and private sources. This allows for the sophisticated 'deep phenotyping' of SDOH within a clinical context, creating a more accurate picture of patient vulnerability. Aligned with the Healthy People 2030 SDOH framework, the CVC integrates expansive state and national datasets to characterize SDOH at the census-block group level. 14 By bridging this infrastructure with electronic health record (EHR) data from a safety-net hospital, PCCI overcomes previous limitations in SDOH measurement. This linkage creates a multidimensional framework capable of generating actionable insights anchored in disease-specific clinical outcomes to characterize individual- and neighborhood-level SDOH. 15 This approach has been applied to identify risk indices for other conditions. 16 , 17 Here we apply similar methodology to clinically identified diagnoses suggestive of cognitive impairment. Therefore, we aimed to leverage CVC's data infrastructure linked with Parkland Health EHR data and advanced ML methods to develop the Brain Vulnerability Index (BVI), a multidimensional measure reflecting individual SDOH-related risk for clinically identifiable cognitive impairment. We then sought to validate BVI's associations with cognitive performance and impairment across clinical- and population-based cohort studies. Methods This study employed a two-stage hybrid design. In Stage 1 (BVI Derivation), a retrospective case-control study was conducted using Parkland Health EHR data, integrated with census block group–level SDOH data from diverse national datasets, to develop and internally validate the BVI. In Stage 2 (External Validation), BVI performance was evaluated in two independent and geographically overlapping cohort studies in the Dallas–Fort Worth (DFW) metropolitan area. Both cohort studies were approved by the UT Southwestern Institutional Review Board, and written informed consent was obtained from all participants in the validation cohorts. Data used for BVI development : Clinical data were extracted from Parkland Health’s EHR, covering the period November 1, 2015, through November 30, 2023. Parkland Health is a 900-bed safety-net hospital with an integrated network of community-based primary care clinics serving the DFW metropolitan area. Cases were defined as patients with at least one of the following: a documented International Classification of Diseases, Ninth or Tenth Revision (ICD-9/ICD-10) code for mild cognitive impairment, dementia, or a related cognitive disorder; a corresponding Systematized Nomenclature of Medicine – Clinical Terms (SNOMED CT) designation; or a dispensed prescription for ADRD pharmacotherapy (donepezil, rivastigmine, galantamine, memantine, or anti-amyloid monoclonal antibodies), consistent with previously described EHR- approaches. 18 , 19 The complete ICD-10 and SNOMED CT code inventory is provided in Supplementary Table S1 . For each identified case, up to five controls were frequency-matched within the same 10-year age stratum and residing in the same metropolitan statistical area, yielding an approximate 1:5 case-to-control ratio. Extracted EHR variables included: sociodemographic characteristics (age, sex, race/ethnicity, smoking status, and marital status); comorbidity burden, quantified using the Charlson Comorbidity Index (CCI); 20 and health care utilization in the 12 months preceding the index date (number of primary care, emergency department, and neurology encounter, hospitalizations, and mode of arrival to the healthcare system). All case and control patients were geo-linked to their corresponding U.S. Census block group using their residential address on record during the study period. For each block group, multidimensional SDOH data were obtained from the CVC. 16 , 17 The CVC integrates data from eight authoritative national and local datasets: the American Community Survey; 21 the Center for Neighborhood Technology Housing and Transportation Index, 22 Environmental Protection Agency National Walkability Index, 23 United States Small-Area Life Expectancy Estimates Project, 24 Program for the International Assessment of Adult Competencies Literacy, 25 Air pollution from OpenWeather, 26 Greenspace from ParkServe, 27 and neighborhood safety data from Applied Geographic Solution’s Crimerisk. 28 For model development, 26 SDOH indicators derived from these sources were retained and organized into four theoretically informed thematic domains: (1) Household Essentials (food security, health insurance coverage, income predictability, and household income); (2) Good Health (neighborhood-level prevalence of alcohol use disorder, chronic disease burden, cancer burden, life expectancy, and mental health status); (3) Equitable Communities (affordable housing availability, air quality, greenspace access, unemployment rates, and neighborhood safety); and (4) Empowered People (broadband connectivity, walkability, vehicle access, and adult literacy). 15 Data used for external validation of BVI Two cohort studies enrolling participants residing within the BVI model development geographic area were selected for external validation. The Dallas Heart Study and Mind Study is an extension of the multiethnic Dallas Heart Study (DHS), an ongoing population-based probability sample recruited from 1999–2000 that has been described in detail. 29–43 DHS serves as the community-based (population-representative) validation cohort. In the 2021–2024 DHS follow-up wave, participants underwent standardized cognitive assessment using the Montreal Cognitive Assessment (MoCA) 44 as well as a supplementary cognitive battery. Participants were classified as either cognitively unimpaired or cognitively impaired based on aggregate performance across all administered assessments, independent of impairment severity or etiology. The Alzheimer’s Disease Center (ADC) at UT Southwestern served as the clinical validation cohort. ADC participants underwent standardized global cognitive assessment using the MoCA or Mini-Mental State Examination (MMSE), supplemented by a comprehensive neuropsychological battery. Cognitive status was adjudicated using a consensus diagnosis process and classified dichotomously as impaired or unimpaired, independent of the degree or underlying etiology of impairment. For participants assessed with the MMSE, scores were converted to MoCA-equivalent values using the validated Roalf conversion equation (MoCA = 1.13 × MMSE − 6.45), which has demonstrated strong concordance between the two measures across the cognitive spectrum. 45 . A local validation study in this population confirmed a concordance coefficient of 0.85 (p < 0.001). 46 The ADC sample contributed serial MoCA measurements (median: 3 assessments per participant), enabling longitudinal trajectory analysis. Statistical Methods Model Development and BVI Derivation. A supervised binary classification model was employed to generate clinically anchored, outcome-referenced risk scores from CVC-derived SDOH indicators, which were subsequently aggregated into a neighborhood-level brain vulnerability measure. All features were standardized to Z-scores prior to modeling to ensure harmonized scaling across heterogeneous data sources. To account for the potential multicollinearity among SDOH indicators, Least Absolute Shrinkage and Selection Operator (LASSO) regularization was applied for feature selection, with the regularization parameter λ optimized via cross-validation. Class imbalance inherent to the case-control design was addressed using the Synthetic Minority Oversampling Technique (SMOTE ) . 47 The final model was specified as regularized logistic regression. Data were partitioned into training (75%) and held-out internal validation (25%) sets using stratified random splitting to preserve class proportions. Model performance was characterized using macro-averaged F1-score, precision, recall, and balanced accuracy, defined as (sensitivity + specificity) / 2, metrics prioritizing stable contributions suitable for index derivation rather than discrimination performances. The latter were calculated on the held-out validation set including area under the receiver operating characteristic (ROC) curve (AUC), 48 positive predictive value (PPV),and negative predictive value (NPV). Given that the primary analytical objective was the derivation of a geospatial neighborhood-level index, a moderate individual-level AUC was anticipated and is consistent with established methodology for area-based SDOH indices. 49 , 50 Global feature contributions were characterized using log-odds coefficients (b) from the final regularized logistic regression model. The sign and magnitude of each coefficient (b) reflect the relative contribution of each feature to the model-derived risk score underlying BVI construction, rather than indicating absolute or independent risk of cognitive impairment. Feature-level contributions to individual predictions were further quantified using Shapley Additive Explanations (SHAP) analysis applied to a randomly selected subsample of 10,000 training observations. 51 To derive our geospatial index BVI, individual patient-level risk scores were aggregated to the census block group level by computing the median risk score across all patients attributed to each block group. This median value constitutes the BVI score, which was subsequently advanced to external validation. External Validation Analyses. In the index-building context, validation approaches included construct validity, external criterion validity (associations with cognitive outcomes), and incremental validity (superior performance vs. comparator indices). BVI scores were assigned to validation cohort participants by linking each participant's residential address at the time of cognitive assessment to the corresponding census block group using the U.S. Census Bureau's Geocoder ( https://geocoding.geo.census.gov/geocoder ). Each participant received the model-derived BVI score for their matched block group. For construct validity assessment, two established geospatial vulnerability indices were calculated for each participant: the ADI, a composite measure of neighborhood socioeconomic disadvantage derived from census-based variables including income, educational attainment, employment, and housing quality, with higher scores reflecting greater deprivation; 52 and the Social Vulnerability Index (SVI), a CDC-developed composite indicator characterizing a community’s susceptibility to adverse health outcomes from external stressors, constructed from census tract–level data on socioeconomic status, household composition, minority status, language barriers, housing type, and transportation access. 53 Construct validity was assessed by examining the Pearson and Spearman correlations between BVI and each comparator index. To assess whether BVI explained variance in MoCA scores beyond established ADI and SVI, a hierarchical general linear model fit within the combined cohort. Descriptive statistics are reported as mean (± standard deviation) for continuous variables and as frequency (percentage) for categorical variables. Between-group comparisons were performed using independent samples t-tests and chi-squared tests, as appropriate. Model development was implemented in R version 3.5.0 within Azure Databricks Runtime 15.4 (glmnet 54 , 55 , randomForest 56 e1071 57 , and pROC 48 ), and Python 3.11 in Azure Databricks Runtime 16.4 (scikit-learn 23 , imbalanced-learn 24 and pandas packages 25 ). External validation was conducted in the combined sample and stratified by cohort (ADC and DHS). Cross-sectional associations between BVI and cognitive function were examined using generalized linear models (GLM) for continuous MoCA scores and binary logistic regression for dichotomous cognitive impairment status. Models were specified as unadjusted (BVI only, or BVI + cohort indicator in combined analyses) and multivariable adjusted (additionally controlling for age, sex, and, in combined analyses, cohort), given that these variables were not incorporated into the BVI model development. Effect sizes were quantified using Cohen’s d. Longitudinal cognitive trajectories were examined using linear mixed-effects models with random intercepts for participants to account for within-subject correlation across repeated assessments. An optimal BVI risk stratification threshold for predicting accelerated cognitive decline was identified using Classification and Regression Tree (CART) analysis. 58 The CART algorithm recursively partitioned the sample based on BVI values to identify the split point that minimized within-group variance in longitudinal MoCA scores. This data-driven approach identified a threshold of BVI ≥ 0.563. Linear mixed-effects models with random intercepts were subsequently fit to test the hypothesis that participants residing in high-risk neighborhoods (BVI ≥ 0.563) experienced significantly greater rates of cognitive decline compared to those in lower-risk neighborhoods, with the Index × Visit interaction term as the primary parameter of interest. Construct validity of the BVI was assessed by computing Pearson product-moment correlations between block group–level BVI scores and corresponding SVI and ADI values. Comparative predictive performance across indices was evaluated by contrasting standardized regression coefficients (β), model coefficients of determination (R²), and AUC values in parallel cross-sectional and longitudinal models. All statistical tests were two-tailed, with a significance threshold of α = 0.05. Results Model Development Study Sample : The BVI derivation cohort comprised 231,630 patients with at least one encounter at Parkland Health during the study period, of whom 39,570 (17.1%) met criteria for EHR-documented cognitive impairment (cases) and 192,060 (82.9%) served as matched controls. Those with EHR based cognitive impairment diagnoses differed significantly from controls in all demographics, social habits, and across all sociodemographic and SDOH domains (all p < 0.001; Table 1 ). They had less primary care visits, emergency encounters but more mental health and neurology encounters (all p<0.001). Table S2 in supplementary materials summarizes the differences between impaired and unimpaired groups across all 4 thematic domains and 26 features included in this analysis. Model performance and BVI scores : Of the 26 CVC-derived SDOH indicators entered into LASSO-regularized logistic regression, 17 features were retained in the final model ( Table S3 ; Figure 1 ). Neighborhood-level chronic disease burden emerged as the dominant feature in the derived index (logistic regression coefficient b = 0.627) and the highest SHAP value, indicating that residence in areas with elevated chronic disease prevalence most strongly increased the predicted probability of cognitive impairment. Education and literacy was the second-ranked feature (β = 0.346), followed by life expectancy risk (β = 0.244) and alcohol abuse burden (β = 0.160); all with positive coefficients, reflecting their impact of an increased risk for cognitive impairment. Conversely, community vulnerability (β = −0.278), obesity prevalence (β = −0.224), overall good health index (β = −0.201), education level (β = −0.199), and high blood pressure burden (β = −0.138) carried negative coefficients, suggesting that these features, as operationalized within the CVC framework, exerted inverse or compensatory contributions to the model output. The directionality of individual features is further illustrated in the SHAP summary plot ( Figure 1 ), which depicts both the magnitude and direction of each feature's contribution to individual-level predictions, with chronic disease burden demonstrating the widest distribution of SHAP values and the clearest separation by feature value. On the 25% held-out validation set, the model achieved macro-averaged F1-score = 0.50, precision = 0.50, recall = 0.49, balanced accuracy = 54.6%, NPV = 86.1%, and AUC = 0.57. Balanced accuracy above chance and high NPV confirm stable, balanced feature contributions appropriate for ecological index derivation, with low-risk areas reliably identified. Patient-level risk scores were aggregated to the census block group and the resulting BVI distribution across Dallas County ( Figure S1 , map panel) shows high-risk concentrations in central Dallas, DeSoto, and Lancaster versus low-to-moderate risk in suburban Carrollton and Richardson, aligning with known geospatial signatures of socioeconomic disadvantage. External Validation of BVI (cross sectional) : Table 2 shows the key characteristics of our validation sample. After adjusting for age and sex, higher BVI was associated with lower MoCA scores in the combined sample (β=−0.35, 95% CI (−0.63, −0.07), R 2 =0.06) and in DHS (β=−1.06, 95% CI (−1.41, −0.71), R 2 =0.19) but not in the baseline ADC sample (β=0.14, 95% CI (−0.26, 0.53)). Using the top quintile cut-off (≥ 0.5282), those in the highest risk group had lower MoCA scores (Cohen’s d=-0.33, p<0.001) in the combined analysis and only in DHS. Participants residing in block groups with higher BVI were also more likely to have cognitive impairment in the DHS cohort only (OR: 1.55 95% CI (1.04–2.33), p=0.0335). These results are shown in Table 3 and Figure S2 and S3 in supplemental materials. Construct Validity and Incremental Validity : BVI demonstrated strong positive correlations with both SVI (r = 0.723, R² = 0.522, p < 0.001) and ADI (r = 0.700, R² = 0.491, p < 0.001), indicating substantial shared variance with established measures of neighborhood disadvantage. These results are shown in Figure S4 in supplemental materials. SVI showed an association with MoCA scores (β = −0.42, 95% CI(−0.59, −0.25)) but not ADI (β = −0.14, 95% CI (−0.34, 0.05). Neither ADI nor SVI were associated with cognitive impairment risk in the combined or individual cohort analyses using the top quintile cutoffs. ( Table 3 ) The base model including ADI, SVI, age, gender, and cohort accounted for 7.7% of variance in MoCA scores (R² = 0.077). Addition of BVI to this model produced a statistically significant increment in explained variance (ΔR² = 0.004, p = 0.019) suggesting BVI has unique incremental contribution to cognition beyond traditional measures. Longitudinal analysis in ADC cohort : CART analysis identified an optimal BVI threshold of ≥0.563 that maximally discriminated longitudinal cognitive trajectories. This threshold classified 152 participants (17.6%) as high-risk in the ADC longitudinal cohort. Linear mixed-effects models revealed that those with an address in the block group in the high-risk BVI demonstrated accelerated cognitive decline ( Figure 2 ). High-risk individuals (BVI ≥0.563) exhibited a significantly faster rate of MoCA decline compared to low-risk individuals (β = −0.68, 95% CI [−0.97, −0.39], p < 0.0001). This translates to an additional 0.68-point decline per visit among high-risk individuals, representing 2.5-fold faster cognitive decline compared to the low-risk group. SVI but not ADI was associated with accelerated decline in MoCA scores (SVI: β = −0.42, 95% CI (−0.59, −0.25), p < 0.0001; ADI: β = −0.14, 95% CI (−0.34, 0.05), p = 0.146; Figure 2 ). Discussion Although prior evidence supports the role of SDOH in dementia risks, 59–61 this study addresses a persistent gap in this field: the absence of a clinically anchored, brain-specific geospatial risk index capable of identifying high-risk communities for cognitive impairment using routinely available data at population scale. Leveraging a ML framework that integrates EHR-derived clinical outcomes with multidimensional neighborhood-level SDOH from the CVC, we developed and externally validated the BVI. 62 , 63 Across two independent geographically overlapping cohorts with complementary epidemiological designs, BVI demonstrated consistent associations with concurrent cognitive performance and with the trajectory of cognitive decline over time. BVI was equivalent to or outperformed established SDOH neighborhood indices in detecting at-risk communities but offered the advantage of being an absolute risk discriminator, which allows for improved clinical implementation at the patient level. Prior risk indices provide relative ranking of geographic areas and not actual patient-level predictive probabilities. BVI fills this gap in applying SDOH in clinical workflows. BVI followed an outcome-anchored derivation that ensured that the features retained in the model (and their corresponding assigned weights) were empirically linked to the health endpoint of interest rather than applied post hoc to a domain for which they were not designed. The strong convergent correlations between BVI and both ADI and SVI confirm that BVI captures the same broad landscape of neighborhood disadvantage, while its incremental predictive performance in longitudinal analyses confirms meaningful discriminant validity beyond these existing measures. Neighborhood-level chronic disease burden and educational attainment emerged as the dominant predictive features in the BVI model, consistent with the growing literature identifying cardiometabolic risk factors and cognitive reserve as two of the most modifiable upstream determinants of ADRD. 4 , 62 The inclusion of environmental exposures (air quality, greenspace), behavioral risk factors (alcohol use disorder prevalence), and social infrastructure indicators (connectivity, walkability, literacy) reflects the multidimensional nature of neighborhood influence on brain health and represents a meaningful expansion beyond the financial and structural variables that dominate existing indices. 64 This breadth of domain coverage is a core feature of the CVC framework upon which BVI is built, and it may explain BVI's ability to capture cognitive-specific neighborhood risk that generic indices may miss. The longitudinal findings carry translational significance. The ability of a passively derived, location-based index to discriminate meaningfully different rates of cognitive decline without requiring clinical measurement underscores the scalability of BVI for proactive population brain health surveillance. 60 A data-driven risk stratification threshold, identified through CART analysis, provides a clinically interpretable decision boundary that can be operationalized in health system workflows using routinely available residential address data, a key prerequisite for real-world deployment at scale. This study has multiple strengths. The two-stage hybrid design separating BVI derivation in a large retrospective EHR case-control study from external validation in two independent cohorts rigorously protects against overfitting and ensures that performance metrics reflect true out-of-sample generalizability rather than in-sample model fit, a standard that prior risk scores may have not met. The dual-cohort approach, rarely employed in prior geospatial ADRD research, provides a more complete and honest characterization of BVI's performance envelope than single-cohort validation alone. BVI incorporates a neurologically informed, multidimensional set of SDOH domains spanning environmental exposures, cognitive enrichment, health behaviors, and material deprivation. This breadth aligns with the emerging scientific consensus that ADRD risk is shaped by a constellation of upstream social, environmental, and behavioral factors rather than any single deprivation dimension. 65 Balancing this high dimensionality, we implemented the combination of LASSO regularization and SHAP interpretability analyses providing methodological rigor and clinical interpretability. 51 , 66 The former controls for multicollinearity and overfitting in a high-dimensional feature space; the latter enables feature-level explanation of neighborhood risk contributions, supporting the kind of mechanistic understanding needed to translate index findings into targeted community interventions. This study also has multiple limitations. The key limitation is its geographic narrowness. CVC-derived SDOH indicators were developed and calibrated within the North Texas region, and the performance of the BVI in other geographic contexts where the prevalence, intercorrelation, and health impact of neighborhood risk factors may differ substantially remains untested. However, given that most of the factors included in BVI are derived from national data, it is feasible and highly possible to generate a national-level BVI that is generalizable to all states. Case ascertainment in the derivation cohort relied on ICD-based EHR documentation and ADRD-specific prescriptions, both of which are subject to systematic underdiagnosis. 67 , 68 Differential diagnostic intensity across racial and ethnic groups is well documented in the ADRD literature and may have introduced ascertainment bias into the development sample, potentially influencing which neighborhood features carried the greatest predictive weight. 69 However, our ability to validate BVI in a rigorous research context unreliant on EHR measurement bias suggests that BVI is reliable despite this constraint. As a geolocation-level index, the BVI is inherently subject to the ecological fallacy: block-group-level risk scores reflect average neighborhood exposure and cannot capture within-neighborhood heterogeneity or individual-level social circumstances that may deviate substantially from the group median. 70 Accordingly, BVI should complement and not replace individual-level social risk screening in clinical encounters, consistent with best practices. Nevertheless, BVI is more precise than generic SDOH indices given its included patient-level risk derived from medical encounters. Finally, generalization to individuals outside of the demographic and clinical distribution of our development or validation cohorts should be done with caution given the inherent variability of cognitive impairment risk by age, sex, and race. Conclusion This study demonstrates that a clinically anchored, machine learning-derived geospatial index can identify communities and individuals at elevated risk for cognitive impairment and accelerated cognitive decline using routinely available health system and publicly accessible SDOH data. BVI provides a scalable, brain-specific, equity-focused neighborhood risk tool that outperforms existing generic deprivation indices in predicting longitudinal cognitive outcomes across both community-based and clinical settings. By enabling passive, address-based risk stratification at population scale, BVI offers a practical instrument for guiding targeted prevention, early identification, and equitable allocation of resources toward communities bearing the greatest burden of neighborhood-attributable risk for cognitive impairment. Abbreviations ADC Alzheimer's Disease Center DHS Dallas Heart Study SD standard deviation MoCA Montreal Cognitive Assessment BVI Brain Vulnerability Index SVI Social Vulnerability Index ADI Area Deprivation Index. Declarations Author Contributions: Concept and design: YT, ACN, AK, IH Acquisition, analysis, or interpretation or data: YT, ACN, AK for Model development; IH, ZY, AS for model validation data and analysis Drafting of the manuscript: All authors Critical revision of the manuscript for the important intellectual content: All Authors Statistical Analysis: YT, ACN, JY, IH Obtained funding: IH Supervision: IH, AK Funding/Support: National Institute of Health funding to P30AG092752, R01AG070861, K24AG062786 to IH Role of the Funder/Sponsor: None Data Availability Data is Available upon request References Alzheimer's disease facts and figures. Alzheimer's & Dementia 2023;19(4):1598–1695. Alzheimer's Disease Facts and Figures. 2024. Alzheimer's disease facts and figures. Alzheimer's & Dementia 2025;21(4):e70235. Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. The Lancet 2024;404(10452):572–628. Ganguli M, Jacobsen E, Song R, et al. Social determinants of health and mild cognitive impairment in a diverse community sample. Journal of the American Geriatrics Society 2025;73(2):367–378. Krasnova A, Tom SE, Valeri L, Crane PK, Bennett DA. Direct Effect of Life-Course Socioeconomic Status on Late-Life Cognition and Cognitive Decline in the Rush Memory and Aging Project. Am J Epidemiol 2023;192(6):882–894. (In eng). Llibre-Guerra JJ, Jiang M, Acosta I, et al. Social determinants of health but not global genetic ancestry predict dementia prevalence in Latin America. Alzheimers Dement 2024;20(7):4828–4840. (In eng). Walsh S, Klee M, Hui EK, et al. Social determinants of dementia: A scoping review. Alzheimers Dement 2025;21(7):e70524. (In eng). Vassilaki M, Petersen RC, Vemuri P. Area Deprivation Index as a Surrogate of Resilience in Aging and Dementia. Front Psychol 2022;13:930415. (In eng). Dow D. Mapping Health Disparities:: Leveraging Area-Based Deprivation Indices for Targeted Chronic Disease Intervention. Dela J Public Health 2024;10(1):106–110. (In eng). D’Abreu A, Chernyavskiy P. Does the Area Deprivation Index (ADI) influence dementia evaluation and diagnosis in Virginia? (S29.004). Neurology 2023;100(17_supplement_2):2410. Hooten M, Ortega M, Oyeyemi A, Yu F, Ofori E. Investigating the relationships between motor skills, cognitive status, and area deprivation index in Arizona: a pilot study. Frontiers in Public Health 2024;Volume 12 - 2024 (Brief Research Report) (In English). Tan WY, Hargreaves C, Chen C, Hilal S. A Machine Learning Approach for Early Diagnosis of Cognitive Impairment Using Population-Based Data. J Alzheimers Dis 2023;91(1):449–461. (In eng). Rine CM. Healthy People 2030: Tools for Action. Health & Social Work 2023;48(4):225–226. Pengetnze Y, Sundaram V, Tamer Y, et al. The Community Vulnerability Compass: a novel, scalable approach for measuring and visualizing social determinants of health insights. JAMIA Open 2025;8(4):ooaf059. Tamer YT, Karam A, Roderick T, Miff S. Know Thy Patient: A Novel Approach and Method for Patient Segmentation and Clustering Using Machine Learning to Develop Holistic, Patient-Centered Programs and Treatment Plans. Catalyst non-issue content 2022;3(4). Naeem J, Salazar-Contreras E, Sundaram V, Wainwright L, Kosel K, Miff S. The Dallas Accountable Health Community: Its Impact on Health-Related Social Needs, Care, and Costs. NEJM Catalyst 2022;3(9):CAT.22.0149. Vassilaki M, Fu S, Christenson LR, et al. Characterizing Performance Gaps of a Code-Based Dementia Algorithm in a Population-Based Cohort of Cognitive Aging. J Alzheimers Dis 2023;95(3):931–940. (In eng). Bhattacharyya J, Barnes LL, Chen Y, et al. Evaluating linked ICD-10 Medicare claims data as a method of dementia case ascertainment in research settings. Alzheimers Dement 2025;21(5):e70200. (In eng). Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373–83. U.S. Census Bureau. Understanding and using American Community Survey data. The Center for Neighborhood Technology. The Center for Neighborhood Technology’s Housing and Transportation (H+T®) Affordability Index United States. Environmental Protection A. National walkability index user guide and methodology. Washington, D.C.: United States Environmental Protection Agency, 2021. National Center for Health S. U.S. small-area life expectancy estimates project : methodology and results summary. Hyattsville, Maryland: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics, 2018. U.S. Department of Education. Institute of Education Sciences NCfES. Highlights of the 2017 U.S. Program for the International Assessment of Adult Competencies (PIAAC) Results (NCES 2020-777). OpenWeather. OpenWeather OWHL™ model. Trust for Public Land. ParkServe. Solution AG. Crimerisk. 1996–2026. Bernardo K, Rossetti H, Weiner M, Cullum C, De Lemos J, Lacritz L. Pro-Inflammatory Substances and Cognition in the Dallas Heart Study. Open Journal of Psychiatry 2016;6:95–101. Victor RG, Haley RW, Willett DL, et al. The Dallas Heart Study: a population-based probability sample for the multidisciplinary study of ethnic differences in cardiovascular health. Am J Cardiol 2004;93(12):1473–80. Das SR, Drazner MH, Dries DL, et al. Impact of body mass and body composition on circulating levels of natriuretic peptides: results from the Dallas Heart Study. Circulation 2005;112(14):2163–8. Murthy VL, Wu KC. Advances in cardiac MRI: The Dallas Heart Study. Curr Cardiol Rep 2007;9(1):43. Albert MA, Ravenell J, Glynn RJ, Khera A, Halevy N, de Lemos JA. Cardiovascular risk indicators and perceived race/ethnic discrimination in the Dallas Heart Study. Am Heart J 2008;156(6):1103–9. Lucarelli RT, Peshock RM, McColl R, et al. MR imaging of hippocampal asymmetry at 3T in a multiethnic, population-based sample: results from the Dallas Heart Study. AJNR Am J Neuroradiol 2013;34(4):752–7. Chandra A, Neeland IJ, Berry JD, et al. The relationship of body mass and fat distribution with incident hypertension: observations from the Dallas Heart Study. J Am Coll Cardiol 2014;64(10):997–1002. Maroules CD, Khera A, Ayers C, et al. Cardiovascular outcome associations among cardiovascular magnetic resonance measures of arterial stiffness: the Dallas heart study. J Cardiovasc Magn Reson 2014;16(1):33. McClelland RL, Jorgensen NW, Budoff M, et al. 10-Year Coronary Heart Disease Risk Prediction Using Coronary Artery Calcium and Traditional Risk Factors: Derivation in the MESA (Multi-Ethnic Study of Atherosclerosis) With Validation in the HNR (Heinz Nixdorf Recall) Study and the DHS (Dallas Heart Study). J Am Coll Cardiol 2015;66(15):1643–53. Yano Y, Vongpatanasin W, Ayers C, et al. Regional Fat Distribution and Blood Pressure Level and Variability: The Dallas Heart Study. Hypertension 2016;68(3):576–83. Carlson SM, Kim J, Khan DA, et al. Hippocampal volume in patients with asthma: Results from the Dallas Heart Study. J Asthma 2017;54(1):9–16. Ceasar JN, Ayers C, Andrews MR, et al. Unfavorable perceived neighborhood environment associates with less routine healthcare utilization: Data from the Dallas Heart Study. PLoS One 2020;15(3):e0230041. Tejani S, McCoy C, Ayers CR, et al. Cardiometabolic Health Outcomes Associated With Discordant Visceral and Liver Fat Phenotypes: Insights From the Dallas Heart Study and UK Biobank. Mayo Clin Proc 2022;97(2):225–237. Yuan C, Palka JM, Rohatgi A, et al. The Relationship Between Coronary Artery Calcification and Carotid Intima Media Thickness and Hippocampal Volume: An Analysis From the Dallas Heart Study. J Acad Consult Liaison Psychiatry 2023;64(3):218–225. Rossetti HC, Weiner M, Hynan LS, Cullum CM, Khera A, Lacritz LH. Subclinical atherosclerosis and subsequent cognitive function. Atherosclerosis 2015;241(1):36–41. (In eng). Nasreddine ZS, Phillips NA, Bédirian V, et al. The Montreal Cognitive Assessment, MoCA: A Brief Screening Tool For Mild Cognitive Impairment. Journal of the American Geriatrics Society 2005;53(4):695–699. Roalf DR, Moberg PJ, Xie SX, Wolk DA, Moelter ST, Arnold SE. Comparative accuracies of two common screening instruments for classification of Alzheimer's disease, mild cognitive impairment, and healthy aging. Alzheimers Dement 2013;9(5):529–37. (In eng). Falkowski JA, Hynan LS, Krishnan K, et al. Conversion of MoCA to MMSE scores. Alzheimers Dement (Amst) 2015;1(1):125. (In eng). Blagus R, Lusa L. SMOTE for high-dimensional class-imbalanced data. BMC Bioinformatics 2013;14:106. Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011;12:77. Bensken WP, McGrath BM, Gold R, Cottrell EK. Area-level social determinants of health and individual-level social risks: Assessing predictive ability and biases in social risk screening. Journal of Clinical and Translational Science 2023;7(1):e257. Lokar K, Zagar T, Zadnik V. Estimation of the Ecological Fallacy in the Geographical Analysis of the Association of Socio-Economic Deprivation and Cancer Incidence. Int J Environ Res Public Health 2019;16(3) (In eng). Lundberg SM, Lee S-I. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems; 2017. Kind AJH, Buckingham WR. Making Neighborhood-Disadvantage Metrics Accessible — The Neighborhood Atlas. New England Journal of Medicine 2018;378(26):2456–2458. Flanagan BE, Gregory EW, Hallisey EJ, Heitgerd JL, Lewis B. A Social Vulnerability Index for Disaster Management. Journal of Homeland Security and Emergency Management;8(1). Tay JK, Narasimhan B, Hastie T. Elastic Net Regularization Paths for All Generalized Linear Models. Journal of Statistical Software 2023;106(1):1 – 31. Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 2010;33(1):1–22. Breiman L. Random Forests. Machine Learning 2001;45(1):5–32. Meyer D. e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. 2024. Krzywinski M, Altman N. Classification and regression trees. Nature Methods 2017;14(8):757–758. Zahodne LB, Kraal AZ, Zaheed AB, others. Neighborhood disadvantage reduces cognitive reserve independent of ethnicity, education, and occupation. Neurology 2024;102(5):e209047. Meyer OL, Tsai J, Marcum ZA, others. Neighborhood disadvantage and the incidence of dementia in US older adults. Alzheimer's & Dementia 2025. Kind AJH, others. Area Deprivation Index and Progression to Dementia. Neurology 2022;98(18 Supplement):1244. Alzheimer's A. Social Determinants of Health and Dementia Risk - Compiled Report. 2023. Livingston G, Huntley J, Liu XY, others. Dementia prevention, intervention, and care: 2024 report of the Lancet Commission. The Lancet 2024;404(10452):572–628. Centers for Disease C, Prevention. Non-Medical Factors that Affect Alzheimer's Disease and Related Dementias. 2025. Smith ER, Broughton M, Baker R, others. Social determinants of dementia: A scoping review. Alzheimer's & Dementia 2025. Wang Y, Wang D, Ye X, others. Practical guide to SHAP analysis: Explaining supervised machine learning models in healthcare. Communications Medicine 2024;4(1):206. Savage SA. Characterizing Performance Gaps of A Code-Based Dementia Algorithm. The University of Texas Health Science Center at Houston; 2022. Zhu Y, Wang L, Yin P, others. Automated phenotyping of mild cognitive impairment and dementias using electronic health records. International Journal of Medical Informatics 2025;193:105649. Mayeda ER, Gilsanz P, Haneuse S, others. Regional variation in diagnostic intensity of dementia among older Medicare beneficiaries: 2016-2019. Alzheimer's & Dementia 2024. Richardson S, Montfort C. Estimation of the Ecological Fallacy in the Geographical Analysis of the Association of Socio-economic Deprivation and Cancer Incidence. Spatial and Spatio-temporal Epidemiology 2019;28:1–9. Tables Table 1. Characteristics of Study Population Used for the Development of the Brain Vulnerability Index Source: Authors' own analyses. Overall EHR-Evidence for Cognitive Impairment= No EHR-Evidence for Cognitive Impairment=Yes N=231,630 N=192,060 N=39,570 p-value* Age 61.87 (14.91) 66.70 (18.61) 60.87 (13.82) <0.001 Sex <0.001 Female 126,894 (54.78%) 105,605 (54.99%) 21,289 (53.80%) Male 104,736 (45.22%) 86,455 (45.01%) 18,281 (46.20%) First Race <0.001 American Indian 643 (0.28%) 528 (0.27%) 115 (0.29%) Asian 7,740 (3.34%) 6,347 (3.30%) 1,393 (3.52%) Black or African American 72,028 (31.10%) 57,640 (30.01%) 14,388 (36.36%) Other Pacific Islander 727 (0.31%) 605 (0.32%) 122 (0.31%) Unkown 21,158 (9.13%) 20,291 (10.56%) 867 (2.19%) White 129,334 (55.84%) 106,649 (55.53%) 22,685 (57.33%) Ethnicity <0.001 Hispanic 90,241 (38.96%) 77,060 (40.12%) 13,181 (33.31%) Non-Hispanic 118,723 (51.26%) 93,232 (48.54%) 25,491 (64,42%) Unknown 22,666 (9.79%) 21,768 (11.33%) 898 (2.27%) Smoking Status <0.001 Ever 67,571 (29.17%) 50,613 (26.35%) 16,958 (42.86%) Never 164,059 (70.83%) 141,447 (73.65%) 22,612 (57.14%) <0.001 Marital Status <0.001 Divorced or Legally Separated 27,371 (11.82%) 21,389 (11.14%) 5,982 (15.12%) Married, Common Law or Significant Other 78,834 (34.03%) 68,250 (35.54%) 10,584 (26.75%) Single 81,640 (35.17%) 65,084 (33.89%) 16,376 (31.38%) Unknown 28,161 (12.16%) 27,415 (14.27%) 746 (1.89%) Widowed 15,804 (6.82%) 9,922 (5.17%) 5,882 (14.86%) Charlson Comorbidity Index (Age-Adjusted) 0.61 (1.58) 2.00 (2.56) 0.32 (1.09) <0.001 SDOH (Average Index/Indicator Value) ** <0.001 Community Vulnerability Index) 0.55 (0.27) 0.54 (0.27) 0.59 (0.26) <0.001 Household Essentials Index 0.54 (0.27) 0.53 (0.27) 0.58 (0.26) <0.001 Empowered People Index 0.51 (0.26) 0.5 (0.26) 0.55 (0.24) <0.001 Equitable Communities Index 0.61 (0.24) 0.61 (0.24) 0.64 (0.23) <0.001 Good Health Index 0.48 (0.25) 0.47 (0.25) 0.52 (0.25) <0.001 Healthcare Utilization, 1 year prior to diagnosis Primary Care Visits 136085 (0.59±3.29 per patient) 86758 (0.45±2.64 per patient) 49327 (1.25±5.37 per patient) <0.001 ED Visits 1505 (0.01±0.08 per patient) 868 (0±0.07 per patient) 637 (0.02±0.13 per patient) <0.001 Hospital Admissions 276 (0±0.03 per patient) 50 (0±0.02 per patient) 226 (0.01±0.08 per patient) <0.001 Mental Health Encounters 1250 (0.01±0.24 per patient) 370 (0±0.13 per patient) 880 (0.02±0.5 per patient) <0.001 Neurology Encounters 5732 (0.02±0.51 per patient) 1951 (0.01±0.29 per patient) 3781 (0.1±1.04 per patient) <0.001 Rheumatology Encounter 1464 (0.01±0.21 per patient) 911 (0±0.18 per patient) 553 (0.01±0.3 per patient) <0.001 Gastroenterology Encounter 3928 (0.02±0.36 per patient) 2136 (0.01±0.27 per patient) 1792 (0.05±0.64 per patient) <0.001 Pulmonology Encounter 1921 (0.01±0.23 per patient) 959 (0±0.18 per patient) 962 (0.02±0.42 per patient) <0.001 * Student t-test performed for mean difference between Alzheimer's Dementia Status ** Higher the value, higher the vulnerability Table 2. Key Characteristics of External Validation Study Participants Source: Authors' own analyses. Characteristic ADC (N=863) DHS (N=532) Combined (N=1,395) Demographics Age, years, mean (SD) 70.1 (8.3) 60.3 (10.6) 66.4 (10.4) Female, n (%) 440 (51.0) 322 (60.5) 762 (54.6) Race/Ethnicity, n (%) White 713 (82.6) 161 (30.3) 874 (62.7) African American 139 (16.1) 310 (58.3) 449 (32.2) Other 11 (1.3) 61 (11.5) 72 (5.2) Cognitive Measures MoCA score, mean (SD) 22.3 (5.7) 22.1 (4.4) 22.2 (5.2) Cognitive impairment, n (%) 551 (63.9) 156 (29.3) 707 (50.7) Impairment Category, n (%) Normal cognition 312 (36.2) 376 (70.7) 688 (49.3) Mild cognitive impairment 218 (25.3) 153 (28.8) 371 (26.6) Dementia 333 (38.6) 3 (0.6) 336 (24.1) Neighborhood Indices BVI (block group), mean (SD) 0.45 (0.07) 0.48 (0.08) 0.46 (0.08) SVI, mean (SD)ᵃ 0.37 (0.28) 0.54 (0.30) 0.44 (0.30) ADI national percentile, mean (SD)ᵃ 38.7 (25.8) 56.5 (26.1) 45.7 (27.3) Abbreviations: ADC = Alzheimer's Disease Center; DHS = Dallas Heart Study; SD = standard deviation; MoCA = Montreal Cognitive Assessment; BVI = Brain Vulnerability Index; SVI = Social Vulnerability Index; ADI = Area Deprivation Index. ᵃ SVI available for N=772 (ADC) and N=496 (DHS); ADI available for N=771 (ADC) and N=495 (DHS). Table 3. Cross-sectional Association of Neighborhood Vulnerability Indices with Cognitive Outcomes Source: Authors' own analyses. Characteristic ADC (N=863) DHS (N=532) Combined (N=1,395) MoCA scores Unadjusted, β (95% CI)ᵃ 0.14 (−0.25, 0.54) −1.36 (−1.72, −1.01)*** −0.47 (−0.76, −0.19)** Adjusted, β (95% CI)ᵇ 0.14 (−0.26, 0.53) −1.06 (−1.41, −0.71)*** −0.35 (−0.63, −0.07)* R² 0.028 0.191 0.061 Risk Stratification (BVI ≥0.5282) High risk, n (%) 119 (13.8) 159 (29.9) 278 (19.9) MoCA difference, unadjusted −0.63 (−1.73, 0.47) −2.73 (−3.52, −1.93)* −1.72 (−2.42, −1.02)* MoCA difference, adjustedᵇ −0.82 (−1.91, 0.28) −2.21 (−2.97, −1.45)*** −1.55 (−2.23, −0.87)*** Cohen's d −0.11 −0.64 −0.33 Cognitive Impairment risk for top quintile BVI (≥0.5282) High risk, n (%) 122 (14.1) 159 (29.9) 281 (20.1) Unadjusted OR (95% CI) 0.92 (0.62–1.37) 1.61 (1.08–2.40)* 1.22 (0.92–1.63) Adjusted OR (95% CI)ᵇ 1.07 (0.71–1.60) 1.55 (1.04–2.33)* 1.27 (0.95–1.70) SVI (≥0.60) High risk, n (%) 186 (24.1) 231 (46.6) 417 (32.9) Unadjusted OR (95% CI) 1.03 (0.73–1.44) 1.39 (0.94–2.06) 1.17 (0.91–1.52) Adjusted OR (95% CI)ᵇ 1.10 (0.77–1.56) 1.39 (0.93–2.06) 1.20 (0.93–1.57) ADI (≥60) High risk, n (%) 174 (22.6) 236 (47.7) 410 (32.4) Unadjusted OR (95% CI) 0.98 (0.69–1.39) 1.45 (0.98–2.14) 1.17 (0.90–1.52) Adjusted OR (95% CI)ᵇ 1.12 (0.78–1.60) 1.48 (0.99–2.20) 1.25 (0.96–1.63) Abbreviations: BVI = Brain Vulnerability Index; SVI = Social Vulnerability Index; ADI = Area Deprivation Index; OR = odds ratio; CI = confidence interval. ᵃ Values are regression coefficients from the general linear models (GLM) with MoCA score modeled as a continuous dependent variable; BVI entered as a continuous standardized BVI along with study indicator in combined analysis. ᵇ Adjusted for age and sex (study indicator added in combined analysis). *p<0.05, **p<0.01, ***p<0.001 Additional Declarations No competing interests reported. Supplementary Files supplementmaterials.pdf Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 02 May, 2026 Reviews received at journal 01 May, 2026 Reviews received at journal 24 Apr, 2026 Reviews received at journal 23 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers invited by journal 17 Apr, 2026 Editor assigned by journal 16 Apr, 2026 Submission checks completed at journal 15 Apr, 2026 First submitted to journal 10 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9381770","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":629093260,"identity":"dade6d06-be33-4e92-a9eb-caaf8c426e08","order_by":0,"name":"Yusuf Tamer","email":"","orcid":"","institution":"Parkland Center for Clinical Innovation","correspondingAuthor":false,"prefix":"","firstName":"Yusuf","middleName":"","lastName":"Tamer","suffix":""},{"id":629093262,"identity":"4057f50e-c97f-44a2-a69a-fbd9316ec45f","order_by":1,"name":"Autumn Noon","email":"","orcid":"","institution":"Parkland Center for Clinical Innovation","correspondingAuthor":false,"prefix":"","firstName":"Autumn","middleName":"","lastName":"Noon","suffix":""},{"id":629093266,"identity":"cb611426-4dfc-4d71-a5fc-53402805925b","order_by":2,"name":"Zhiyi Yang","email":"","orcid":"","institution":"University of Texas Southwestern Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Zhiyi","middleName":"","lastName":"Yang","suffix":""},{"id":629093269,"identity":"a8cb1fbb-f0bd-4338-89ae-43d75a945c1f","order_by":3,"name":"Kexin Yu","email":"","orcid":"","institution":"University of Texas Southwestern Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Kexin","middleName":"","lastName":"Yu","suffix":""},{"id":629093270,"identity":"47ddb018-ca1c-4cd5-9cd6-33a000dc57db","order_by":4,"name":"Amil Shah","email":"","orcid":"","institution":"University of Texas Southwestern Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Amil","middleName":"","lastName":"Shah","suffix":""},{"id":629093273,"identity":"4a592739-94ed-46fe-a8b9-1724466b5a27","order_by":5,"name":"Albert Karam","email":"","orcid":"","institution":"Parkland Center for Clinical Innovation","correspondingAuthor":false,"prefix":"","firstName":"Albert","middleName":"","lastName":"Karam","suffix":""},{"id":629093274,"identity":"ed6b2753-ba36-4ee8-9a57-0b70353952c3","order_by":6,"name":"Ihab Hajjar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABMklEQVRIie3Qv0rDQBzA8TsOrkuS+UJL8goXAskifZYLgWTpJuhQSAOFc7F0jVjfwUkcK4G4VF+ggpkyVUgXqZjBiykoJBZHwftClt/lw/0BQCb7g2ECMIAxQOKDeTNTvq0rhwmigNWjz9/ojwTsyZ7/hmj9aZHD22EPnT0UY3UXGfR5lpGyqgz3Mob5C28fbJC5FK58BM9DZ62w1KaZFugXnNqDpyWyrjoIYQ6BHIm7BFiQpXedKc5ajamXEIb7ahcJXwWZIDgv8LHCooZUFZ0I0nvvJKN6lxTBJMBIYaghAFNGxC6om5xSj98LUiB9EaS2no3ctxm3rYR4U33x2CJmEt7kWz72rXkAy81RZGjpyqa7yjAJ8e/KzUn7lesYAL4Vt+ewY/bV0Dy0KpPJZP+7DzWdYzkkw0BzAAAAAElFTkSuQmCC","orcid":"","institution":"University of Texas Southwestern Medical Center","correspondingAuthor":true,"prefix":"","firstName":"Ihab","middleName":"","lastName":"Hajjar","suffix":""}],"badges":[],"createdAt":"2026-04-10 16:38:53","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9381770/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9381770/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107916732,"identity":"c12a0735-8f10-42f7-a196-39e3f0ca5933","added_by":"auto","created_at":"2026-04-27 14:18:55","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2722645,"visible":true,"origin":"","legend":"\u003cp\u003eFeature Contributions to Brain Vulnerability Index Predictions.\u003c/p\u003e\n\u003cp\u003eSource: Authors' own analyses.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLegend:\u003c/strong\u003e SHAP (SHapley Additive exPlanations) values for individual features from LASSO-regularized logistic regression model predicting EHR-documented cognitive impairment (n=10,000 subsample). Horizontal position indicates direction and magnitude of feature impact on model output—right (positive SHAP) increases predicted risk, left (negative SHAP) decreases risk. Bar length shows mean absolute SHAP value (feature importance ranking). Color intensity reflects feature value (blue=low, red=high). Chronic disease burden exhibits highest impact (widest SHAP distribution), followed by education/literacy. Negative contributions (community vulnerability, obesity) reflect model-derived protective effects within CVC framework.\u003c/p\u003e","description":"","filename":"Fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-9381770/v1/50310bdc26634cce32434a89.png"},{"id":107916734,"identity":"07b35990-8b03-4b4e-a1d0-2408cbec9995","added_by":"auto","created_at":"2026-04-27 14:18:55","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":74470,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eForest Plot of Index × Visit Interactions Predicting Longitudinal Cognitive Decline.\u003c/strong\u003e Mixed-effects models examined whether neighborhood vulnerability indices predicted differential rates of MoCA decline over time. The interaction coefficient (β) represents the additional change in MoCA score per visit for high-risk compared to low-risk individuals. Negative values indicate accelerated cognitive decline in the high-risk group. BVI demonstrated the largest effect (β = −0.68), followed by SVI (β = −0.42), while ADI did not reach statistical significance (β = −0.14, p = 0.146). Filled circles indicate significant interactions (p \u0026lt; 0.05); open circles indicate non-significant interactions. Error bars represent 95% confidence intervals. Models adjusted for age and sex with random intercepts for participants. Abbreviations: BVI, Brain Vulnerability Index; SVI, Social Vulnerability Index; ADI, Area Deprivation Index; CI, confidence interval.\u003c/p\u003e","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-9381770/v1/75b011547a1eb63244c38ff3.png"},{"id":108008712,"identity":"5a9fd779-2182-47e8-88ef-05d74e48cdfc","added_by":"auto","created_at":"2026-04-28 13:08:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2852754,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9381770/v1/e9654e59-cd24-4ad6-8395-2ee758cabc8f.pdf"},{"id":108006916,"identity":"931e010c-9c99-412e-a66f-7fc080a5b56e","added_by":"auto","created_at":"2026-04-28 12:57:54","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1198911,"visible":true,"origin":"","legend":"","description":"","filename":"supplementmaterials.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9381770/v1/66e7e2388523f16b42620e88.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"The Brain Vulnerability Index: Development and Validation of a Machine Learning–Derived, Community-Informed Geospatial Risk Score for Cognitive Impairment","fulltext":[{"header":"Background","content":"\u003cp\u003eAlzheimer\u0026rsquo;s disease (AD) and related dementias (ADRD) are neurodegenerative diseases with AD being the fifth-leading cause of death in the US.\u003csup\u003e1,2\u003c/sup\u003e AD impacted an estimated 6.9\u0026nbsp;million Americans 65 and older in 2024 and is expected to affect 13.8\u0026nbsp;million Americans by 2060.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e The percentage of Americans with AD increases with age: 5.0% of people aged 65\u0026ndash;74, 13.2% of people aged 75 to 84, and 33.4% of people 85 and older. \u003csup\u003e1,2\u003c/sup\u003e Despite great advances in our understanding of ADRD, nearly half of cognitive impairment risk remains unexplained.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e More recently, risk models suggest a strong contribution of non-medical drivers to cognitive impairment risk.\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eIdentifying elevated risk groups, especially beyond traditional risk factors, improves efficiencies of prevention programs as well as case ascertainments and identifying those that would benefit most from therapeutic programs. Increasingly, social determinants of health (SDOH), and more broadly non-medical drivers of health (NMDOH), have been identified as important, potentially modifiable risk factors for AD and ADRD. \u003csup\u003e6\u0026ndash;8\u003c/sup\u003e Currently available measures of SDOH are not brain-specific, emphasize financial and structural vulnerability, and are rarely operationalized as patient-level tools due to the lack of interpretable thresholds for use in clinical practice. \u003csup\u003e9,10\u003c/sup\u003e For example, the Area Deprivation Index (ADI) is widely used to quantify neighborhood socioeconomic disadvantage in ADRD research, However, it provides generic, ranking of areas masking within‑neighborhood heterogeneity and lacks calibration as an individual‑level brain health risk tool.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e To achieve progress towards clinical implementation, we need new tools encompassing a wide variety of measures that are highly relevant to ADRD and that identify patient-level risk. One emerging approach is to use machine learning (ML) and other artificial intelligence (AI) methods that jointly analyze high‑dimensional clinically and non‑clinically derived data, including a wide range of SDOH indicators, to identify an individual\u0026rsquo;s risk of cognitive impairment.\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e Fragmented diverse social data sources lacking clinical context have hindered the creation of more advanced SDOH indices.\u003c/p\u003e \u003cp\u003eThe Parkland Center for Clinical Innovation (PCCI), a North Texas nonprofit specializing in data-driven solutions, developed the Community Vulnerability Compass (CVC) to bridge the gap between socioeconomic data and medical records. Through its unique partnership with Parkland Health, PCCI integrates deep clinical datasets with various public and private sources. This allows for the sophisticated 'deep phenotyping' of SDOH within a clinical context, creating a more accurate picture of patient vulnerability. Aligned with the Healthy People 2030 SDOH framework, the CVC integrates expansive state and national datasets to characterize SDOH at the census-block group level.\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e By bridging this infrastructure with electronic health record (EHR) data from a safety-net hospital, PCCI overcomes previous limitations in SDOH measurement. This linkage creates a multidimensional framework capable of generating actionable insights anchored in disease-specific clinical outcomes to characterize individual- and neighborhood-level SDOH.\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e This approach has been applied to identify risk indices for other conditions.\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e Here we apply similar methodology to clinically identified diagnoses suggestive of cognitive impairment.\u003c/p\u003e \u003cp\u003eTherefore, we aimed to leverage CVC's data infrastructure linked with Parkland Health EHR data and advanced ML methods to develop the Brain Vulnerability Index (BVI), a multidimensional measure reflecting individual SDOH-related risk for clinically identifiable cognitive impairment. We then sought to validate BVI's associations with cognitive performance and impairment across clinical- and population-based cohort studies.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eThis study employed a two-stage hybrid design. In Stage 1 (BVI Derivation), a retrospective case-control study was conducted using Parkland Health EHR data, integrated with census block group\u0026ndash;level SDOH data from diverse national datasets, to develop and internally validate the BVI. In Stage 2 (External Validation), BVI performance was evaluated in two independent and geographically overlapping cohort studies in the Dallas\u0026ndash;Fort Worth (DFW) metropolitan area. Both cohort studies were approved by the UT Southwestern Institutional Review Board, and written informed consent was obtained from all participants in the validation cohorts.\u003c/p\u003e \u003cp\u003e \u003cem\u003eData used for BVI development\u003c/em\u003e: Clinical data were extracted from Parkland Health\u0026rsquo;s EHR, covering the period November 1, 2015, through November 30, 2023. Parkland Health is a 900-bed safety-net hospital with an integrated network of community-based primary care clinics serving the DFW metropolitan area. Cases were defined as patients with at least one of the following: a documented International Classification of Diseases, Ninth or Tenth Revision (ICD-9/ICD-10) code for mild cognitive impairment, dementia, or a related cognitive disorder; a corresponding Systematized Nomenclature of Medicine \u0026ndash; Clinical Terms (SNOMED CT) designation; or a dispensed prescription for ADRD pharmacotherapy (donepezil, rivastigmine, galantamine, memantine, or anti-amyloid monoclonal antibodies), consistent with previously described EHR- approaches.\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e The complete ICD-10 and SNOMED CT code inventory is provided in \u003cb\u003eSupplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e. For each identified case, up to five controls were frequency-matched within the same 10-year age stratum and residing in the same metropolitan statistical area, yielding an approximate 1:5 case-to-control ratio. Extracted EHR variables included: sociodemographic characteristics (age, sex, race/ethnicity, smoking status, and marital status); comorbidity burden, quantified using the Charlson Comorbidity Index (CCI);\u003csup\u003e20\u003c/sup\u003e and health care utilization in the 12 months preceding the index date (number of primary care, emergency department, and neurology encounter, hospitalizations, and mode of arrival to the healthcare system).\u003c/p\u003e \u003cp\u003eAll case and control patients were geo-linked to their corresponding U.S. Census block group using their residential address on record during the study period. For each block group, multidimensional SDOH data were obtained from the CVC.\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e The CVC integrates data from eight authoritative national and local datasets: the American Community Survey;\u003csup\u003e21\u003c/sup\u003e the Center for Neighborhood Technology Housing and Transportation Index,\u003csup\u003e22\u003c/sup\u003e Environmental Protection Agency National Walkability Index,\u003csup\u003e23\u003c/sup\u003e United States Small-Area Life Expectancy Estimates Project,\u003csup\u003e24\u003c/sup\u003e Program for the International Assessment of Adult Competencies Literacy,\u003csup\u003e25\u003c/sup\u003e Air pollution from OpenWeather,\u003csup\u003e26\u003c/sup\u003e Greenspace from ParkServe,\u003csup\u003e27\u003c/sup\u003e and neighborhood safety data from Applied Geographic Solution\u0026rsquo;s Crimerisk.\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e For model development, 26 SDOH indicators derived from these sources were retained and organized into four theoretically informed thematic domains: (1) \u003cem\u003eHousehold Essentials\u003c/em\u003e (food security, health insurance coverage, income predictability, and household income); (2) \u003cem\u003eGood Health\u003c/em\u003e (neighborhood-level prevalence of alcohol use disorder, chronic disease burden, cancer burden, life expectancy, and mental health status); (3) \u003cem\u003eEquitable Communities\u003c/em\u003e (affordable housing availability, air quality, greenspace access, unemployment rates, and neighborhood safety); and (4) \u003cem\u003eEmpowered People\u003c/em\u003e (broadband connectivity, walkability, vehicle access, and adult literacy).\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eData used for external validation of BVI\u003c/strong\u003e \u003cp\u003eTwo cohort studies enrolling participants residing within the BVI model development geographic area were selected for external validation. The \u003cem\u003eDallas Heart Study and Mind Study\u003c/em\u003e is an extension of the multiethnic Dallas Heart Study (DHS), an ongoing population-based probability sample recruited from 1999\u0026ndash;2000 that has been described in detail. \u003csup\u003e29\u0026ndash;43\u003c/sup\u003e DHS serves as the community-based (population-representative) validation cohort. In the 2021\u0026ndash;2024 DHS follow-up wave, participants underwent standardized cognitive assessment using the Montreal Cognitive Assessment (MoCA)\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e as well as a supplementary cognitive battery. Participants were classified as either cognitively unimpaired or cognitively impaired based on aggregate performance across all administered assessments, independent of impairment severity or etiology.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eThe \u003cem\u003eAlzheimer\u0026rsquo;s Disease Center (ADC) at UT Southwestern\u003c/em\u003e served as the clinical validation cohort. ADC participants underwent standardized global cognitive assessment using the MoCA or Mini-Mental State Examination (MMSE), supplemented by a comprehensive neuropsychological battery. Cognitive status was adjudicated using a consensus diagnosis process and classified dichotomously as impaired or unimpaired, independent of the degree or underlying etiology of impairment. For participants assessed with the MMSE, scores were converted to MoCA-equivalent values using the validated Roalf conversion equation (MoCA\u0026thinsp;=\u0026thinsp;1.13 \u0026times; MMSE\u0026thinsp;\u0026minus;\u0026thinsp;6.45), which has demonstrated strong concordance between the two measures across the cognitive spectrum.\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. A local validation study in this population confirmed a concordance coefficient of 0.85 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003csup\u003e46\u003c/sup\u003e The ADC sample contributed serial MoCA measurements (median: 3 assessments per participant), enabling longitudinal trajectory analysis.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Methods\u003c/h2\u003e \u003cp\u003e \u003cem\u003eModel Development and BVI Derivation.\u003c/em\u003e A supervised binary classification model was employed to generate clinically anchored, outcome-referenced risk scores from CVC-derived SDOH indicators, which were subsequently aggregated into a neighborhood-level brain vulnerability measure. All features were standardized to Z-scores prior to modeling to ensure harmonized scaling across heterogeneous data sources. To account for the potential multicollinearity among SDOH indicators, Least Absolute Shrinkage and Selection Operator (LASSO) regularization was applied for feature selection, with the regularization parameter λ optimized via cross-validation. Class imbalance inherent to the case-control design was addressed using the Synthetic Minority Oversampling Technique (SMOTE\u003cem\u003e)\u003c/em\u003e.\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e The final model was specified as regularized logistic regression. Data were partitioned into training (75%) and held-out internal validation (25%) sets using stratified random splitting to preserve class proportions.\u003c/p\u003e \u003cp\u003eModel performance was characterized using macro-averaged F1-score, precision, recall, and balanced accuracy, defined as (sensitivity\u0026thinsp;+\u0026thinsp;specificity) / 2, metrics prioritizing stable contributions suitable for index derivation rather than discrimination performances. The latter were calculated on the held-out validation set including area under the receiver operating characteristic (ROC) curve (AUC),\u003csup\u003e48\u003c/sup\u003e positive predictive value (PPV),and negative predictive value (NPV). Given that the primary analytical objective was the derivation of a geospatial neighborhood-level index, a moderate individual-level AUC was anticipated and is consistent with established methodology for area-based SDOH indices.\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eGlobal feature contributions were characterized using log-odds coefficients (b) from the final regularized logistic regression model. The sign and magnitude of each coefficient (b) reflect the relative contribution of each feature to the model-derived risk score underlying BVI construction, rather than indicating absolute or independent risk of cognitive impairment. Feature-level contributions to individual predictions were further quantified using Shapley Additive Explanations (SHAP) analysis applied to a randomly selected subsample of 10,000 training observations.\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e To derive our geospatial index BVI, individual patient-level risk scores were aggregated to the census block group level by computing the median risk score across all patients attributed to each block group. This median value constitutes the BVI score, which was subsequently advanced to external validation.\u003c/p\u003e \u003cp\u003e \u003cem\u003eExternal Validation Analyses.\u003c/em\u003e In the index-building context, validation approaches included construct validity, external criterion validity (associations with cognitive outcomes), and incremental validity (superior performance vs. comparator indices). BVI scores were assigned to validation cohort participants by linking each participant's residential address at the time of cognitive assessment to the corresponding census block group using the U.S. Census Bureau's Geocoder (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://geocoding.geo.census.gov/geocoder\u003c/span\u003e\u003cspan address=\"https://geocoding.geo.census.gov/geocoder\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Each participant received the model-derived BVI score for their matched block group. For construct validity assessment, two established geospatial vulnerability indices were calculated for each participant: the ADI, a composite measure of neighborhood socioeconomic disadvantage derived from census-based variables including income, educational attainment, employment, and housing quality, with higher scores reflecting greater deprivation;\u003csup\u003e52\u003c/sup\u003e and the Social Vulnerability Index (SVI), a CDC-developed composite indicator characterizing a community\u0026rsquo;s susceptibility to adverse health outcomes from external stressors, constructed from census tract\u0026ndash;level data on socioeconomic status, household composition, minority status, language barriers, housing type, and transportation access.\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e Construct validity was assessed by examining the Pearson and Spearman correlations between BVI and each comparator index. To assess whether BVI explained variance in MoCA scores beyond established ADI and SVI, a hierarchical general linear model fit within the combined cohort.\u003c/p\u003e \u003cp\u003eDescriptive statistics are reported as mean (\u0026plusmn;\u0026thinsp;standard deviation) for continuous variables and as frequency (percentage) for categorical variables. Between-group comparisons were performed using independent samples t-tests and chi-squared tests, as appropriate. Model development was implemented in R version 3.5.0 within Azure Databricks Runtime 15.4 (glmnet \u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e,\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e, randomForest\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e e1071 \u003csup\u003e57\u003c/sup\u003e, and pROC\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e), and Python 3.11 in Azure Databricks Runtime 16.4 (scikit-learn\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e, imbalanced-learn\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e and pandas packages\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e). External validation was conducted in the combined sample and stratified by cohort (ADC and DHS). Cross-sectional associations between BVI and cognitive function were examined using generalized linear models (GLM) for continuous MoCA scores and binary logistic regression for dichotomous cognitive impairment status. Models were specified as unadjusted (BVI only, or BVI\u0026thinsp;+\u0026thinsp;cohort indicator in combined analyses) and multivariable adjusted (additionally controlling for age, sex, and, in combined analyses, cohort), given that these variables were not incorporated into the BVI model development. Effect sizes were quantified using Cohen\u0026rsquo;s d. Longitudinal cognitive trajectories were examined using linear mixed-effects models with random intercepts for participants to account for within-subject correlation across repeated assessments.\u003c/p\u003e \u003cp\u003eAn optimal BVI risk stratification threshold for predicting accelerated cognitive decline was identified using Classification and Regression Tree (CART) analysis.\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e The CART algorithm recursively partitioned the sample based on BVI values to identify the split point that minimized within-group variance in longitudinal MoCA scores. This data-driven approach identified a threshold of BVI\u0026thinsp;\u0026ge;\u0026thinsp;0.563. Linear mixed-effects models with random intercepts were subsequently fit to test the hypothesis that participants residing in high-risk neighborhoods (BVI\u0026thinsp;\u0026ge;\u0026thinsp;0.563) experienced significantly greater rates of cognitive decline compared to those in lower-risk neighborhoods, with the Index \u0026times; Visit interaction term as the primary parameter of interest.\u003c/p\u003e \u003cp\u003eConstruct validity of the BVI was assessed by computing Pearson product-moment correlations between block group\u0026ndash;level BVI scores and corresponding SVI and ADI values. Comparative predictive performance across indices was evaluated by contrasting standardized regression coefficients (β), model coefficients of determination (R\u0026sup2;), and AUC values in parallel cross-sectional and longitudinal models. All statistical tests were two-tailed, with a significance threshold of α\u0026thinsp;=\u0026thinsp;0.05.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cem\u003eModel Development Study Sample\u003c/em\u003e: The BVI derivation cohort comprised 231,630 patients with at least one encounter at Parkland Health during the study period, of whom 39,570 (17.1%) met criteria for EHR-documented cognitive impairment (cases) and 192,060 (82.9%) served as matched controls. Those with EHR based cognitive impairment diagnoses differed significantly from controls in all demographics, social habits, and across all sociodemographic and SDOH domains (all p \u0026lt; 0.001; \u003cstrong\u003eTable 1\u003c/strong\u003e). They had less primary care visits, emergency encounters but more mental health and neurology encounters (all p\u0026lt;0.001). \u003cstrong\u003eTable S2\u003c/strong\u003e in supplementary materials summarizes the differences between impaired and unimpaired groups across all 4 thematic domains and 26 features included in this analysis. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eModel performance and BVI scores\u003c/em\u003e: \u0026nbsp; Of the 26 CVC-derived SDOH indicators entered into LASSO-regularized logistic regression, 17 features were retained in the final model (\u003cstrong\u003eTable S3\u003c/strong\u003e; \u003cstrong\u003eFigure 1\u003c/strong\u003e). Neighborhood-level chronic disease burden emerged as the dominant feature in the derived index (logistic regression coefficient b = 0.627) and the highest SHAP value, indicating that residence in areas with elevated chronic disease prevalence most strongly increased the predicted probability of cognitive impairment. Education and literacy was the second-ranked feature (\u0026beta; = 0.346), followed by life expectancy risk (\u0026beta; = 0.244) and alcohol abuse burden (\u0026beta; = 0.160); all with positive coefficients, reflecting their impact of an increased risk for cognitive impairment. Conversely, community vulnerability (\u0026beta; = \u0026minus;0.278), obesity prevalence (\u0026beta; = \u0026minus;0.224), overall good health index (\u0026beta; = \u0026minus;0.201), education level (\u0026beta; = \u0026minus;0.199), and high blood pressure burden (\u0026beta; = \u0026minus;0.138) carried negative coefficients, suggesting that these features, as operationalized within the CVC framework, exerted inverse or compensatory contributions to the model output. The directionality of individual features is further illustrated in the SHAP summary plot (\u003cstrong\u003eFigure 1\u003c/strong\u003e), which depicts both the magnitude and direction of each feature\u0026apos;s contribution to individual-level predictions, with chronic disease burden demonstrating the widest distribution of SHAP values and the clearest separation by feature value.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOn the 25% held-out validation set, the model achieved macro-averaged F1-score = 0.50, precision = 0.50, recall = 0.49, balanced accuracy = 54.6%, NPV = 86.1%, and AUC = 0.57. Balanced accuracy above chance and high NPV confirm stable, balanced feature contributions appropriate for ecological index derivation, with low-risk areas reliably identified.\u003c/p\u003e\n\u003cp\u003ePatient-level risk scores were aggregated to the census block group \u0026nbsp;and the resulting BVI distribution across Dallas County (\u003cstrong\u003eFigure S1\u003c/strong\u003e, map panel) shows high-risk concentrations in central Dallas, DeSoto, and Lancaster versus low-to-moderate risk in suburban Carrollton and Richardson, aligning with known geospatial signatures of socioeconomic disadvantage.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eExternal Validation of BVI (cross sectional)\u003c/em\u003e: \u003cstrong\u003eTable 2\u003c/strong\u003e shows the key characteristics of our validation sample. After adjusting for age and sex, higher BVI was associated with lower MoCA scores in the combined sample (\u0026beta;=\u0026minus;0.35, 95% CI (\u0026minus;0.63, \u0026minus;0.07), R\u003csup\u003e2\u003c/sup\u003e=0.06) and in DHS (\u0026beta;=\u0026minus;1.06, 95% CI (\u0026minus;1.41, \u0026minus;0.71), R\u003csup\u003e2\u003c/sup\u003e=0.19) but not in the baseline ADC sample (\u0026beta;=0.14, 95% CI (\u0026minus;0.26, 0.53)). Using the top quintile cut-off (\u0026ge; 0.5282), those in the highest risk group had lower MoCA scores (Cohen\u0026rsquo;s d=-0.33, p\u0026lt;0.001) \u0026nbsp;in the combined analysis and only in DHS. Participants residing in block groups with higher BVI were also more likely to have cognitive impairment in the DHS cohort only (OR: 1.55 95% CI (1.04\u0026ndash;2.33), p=0.0335). These results are shown in \u003cstrong\u003eTable 3 and\u003c/strong\u003e \u003cstrong\u003eFigure S2 and S3\u0026nbsp;\u003c/strong\u003ein supplemental materials.\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eConstruct Validity and Incremental Validity\u003c/em\u003e: BVI demonstrated strong positive correlations with both SVI (r = 0.723, R\u0026sup2; = 0.522, p \u0026lt; 0.001) and ADI (r = 0.700, R\u0026sup2; = 0.491, p \u0026lt; 0.001), indicating substantial shared variance with established measures of neighborhood disadvantage. These results are shown in \u003cstrong\u003eFigure S4\u0026nbsp;\u003c/strong\u003ein supplemental materials. SVI showed an association with MoCA scores (\u0026beta; = \u0026minus;0.42, 95% CI(\u0026minus;0.59, \u0026minus;0.25)) but not ADI (\u0026beta; = \u0026minus;0.14, 95% CI (\u0026minus;0.34, 0.05). Neither ADI nor SVI were associated with cognitive impairment risk in the combined or individual cohort analyses using the top quintile cutoffs. (\u003cstrong\u003eTable 3\u003c/strong\u003e) The base model including ADI, SVI, age, gender, and cohort accounted for 7.7% of variance in MoCA scores (R\u0026sup2; = 0.077). Addition of BVI to this model produced a statistically significant increment in explained variance (\u0026Delta;R\u0026sup2; = 0.004, p = 0.019) suggesting BVI has unique incremental contribution to cognition beyond traditional measures. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eLongitudinal analysis in ADC cohort\u003c/em\u003e: CART analysis identified an optimal BVI threshold of \u0026ge;0.563 that maximally discriminated longitudinal cognitive trajectories. This threshold classified 152 participants (17.6%) as high-risk in the ADC longitudinal cohort. Linear mixed-effects models revealed that those with an address in the block group in the high-risk BVI demonstrated \u0026nbsp;accelerated cognitive decline (\u003cstrong\u003eFigure 2\u003c/strong\u003e). High-risk individuals (BVI \u0026ge;0.563) exhibited a significantly faster rate of MoCA decline compared to low-risk individuals (\u0026beta; = \u0026minus;0.68, 95% CI [\u0026minus;0.97, \u0026minus;0.39], p \u0026lt; 0.0001). This translates to an additional 0.68-point decline per visit among high-risk individuals, representing 2.5-fold faster cognitive decline compared to the low-risk group. SVI but not ADI was associated with accelerated decline in MoCA scores (SVI: \u0026beta; = \u0026minus;0.42, 95% CI (\u0026minus;0.59, \u0026minus;0.25), p \u0026lt; 0.0001; ADI: \u0026beta; = \u0026minus;0.14, 95% CI (\u0026minus;0.34, 0.05), p = 0.146; \u003cstrong\u003eFigure 2\u003c/strong\u003e).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eAlthough prior evidence supports the role of SDOH in dementia risks,\u003csup\u003e59\u0026ndash;61\u003c/sup\u003e this study addresses a persistent gap in this field: the absence of a clinically anchored, brain-specific geospatial risk index capable of identifying high-risk communities for cognitive impairment using routinely available data at population scale. Leveraging a ML framework that integrates EHR-derived clinical outcomes with multidimensional neighborhood-level SDOH from the CVC, we developed and externally validated the BVI.\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e,\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e Across two independent geographically overlapping cohorts with complementary epidemiological designs, BVI demonstrated consistent associations with concurrent cognitive performance and with the trajectory of cognitive decline over time. BVI was equivalent to or outperformed established SDOH neighborhood indices in detecting at-risk communities but offered the advantage of being an absolute risk discriminator, which allows for improved clinical implementation at the patient level. Prior risk indices provide relative ranking of geographic areas and not actual patient-level predictive probabilities. BVI fills this gap in applying SDOH in clinical workflows.\u003c/p\u003e \u003cp\u003eBVI followed an outcome-anchored derivation that ensured that the features retained in the model (and their corresponding assigned weights) were empirically linked to the health endpoint of interest rather than applied post hoc to a domain for which they were not designed. The strong convergent correlations between BVI and both ADI and SVI confirm that BVI captures the same broad landscape of neighborhood disadvantage, while its incremental predictive performance in longitudinal analyses confirms meaningful discriminant validity beyond these existing measures.\u003c/p\u003e \u003cp\u003eNeighborhood-level chronic disease burden and educational attainment emerged as the dominant predictive features in the BVI model, consistent with the growing literature identifying cardiometabolic risk factors and cognitive reserve as two of the most modifiable upstream determinants of ADRD.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e The inclusion of environmental exposures (air quality, greenspace), behavioral risk factors (alcohol use disorder prevalence), and social infrastructure indicators (connectivity, walkability, literacy) reflects the multidimensional nature of neighborhood influence on brain health and represents a meaningful expansion beyond the financial and structural variables that dominate existing indices.\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e This breadth of domain coverage is a core feature of the CVC framework upon which BVI is built, and it may explain BVI's ability to capture cognitive-specific neighborhood risk that generic indices may miss.\u003c/p\u003e \u003cp\u003eThe longitudinal findings carry translational significance. The ability of a passively derived, location-based index to discriminate meaningfully different rates of cognitive decline without requiring clinical measurement underscores the scalability of BVI for proactive population brain health surveillance.\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e A data-driven risk stratification threshold, identified through CART analysis, provides a clinically interpretable decision boundary that can be operationalized in health system workflows using routinely available residential address data, a key prerequisite for real-world deployment at scale.\u003c/p\u003e \u003cp\u003eThis study has multiple strengths. The two-stage hybrid design separating BVI derivation in a large retrospective EHR case-control study from external validation in two independent cohorts rigorously protects against overfitting and ensures that performance metrics reflect true out-of-sample generalizability rather than in-sample model fit, a standard that prior risk scores may have not met. The dual-cohort approach, rarely employed in prior geospatial ADRD research, provides a more complete and honest characterization of BVI's performance envelope than single-cohort validation alone. BVI incorporates a neurologically informed, multidimensional set of SDOH domains spanning environmental exposures, cognitive enrichment, health behaviors, and material deprivation. This breadth aligns with the emerging scientific consensus that ADRD risk is shaped by a constellation of upstream social, environmental, and behavioral factors rather than any single deprivation dimension.\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e Balancing this high dimensionality, we implemented the combination of LASSO regularization and SHAP interpretability analyses providing methodological rigor and clinical interpretability.\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e,\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e The former controls for multicollinearity and overfitting in a high-dimensional feature space; the latter enables feature-level explanation of neighborhood risk contributions, supporting the kind of mechanistic understanding needed to translate index findings into targeted community interventions.\u003c/p\u003e \u003cp\u003eThis study also has multiple limitations. The key limitation is its geographic narrowness. CVC-derived SDOH indicators were developed and calibrated within the North Texas region, and the performance of the BVI in other geographic contexts where the prevalence, intercorrelation, and health impact of neighborhood risk factors may differ substantially remains untested. However, given that most of the factors included in BVI are derived from national data, it is feasible and highly possible to generate a national-level BVI that is generalizable to all states.\u003c/p\u003e \u003cp\u003eCase ascertainment in the derivation cohort relied on ICD-based EHR documentation and ADRD-specific prescriptions, both of which are subject to systematic underdiagnosis.\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e,\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e Differential diagnostic intensity across racial and ethnic groups is well documented in the ADRD literature and may have introduced ascertainment bias into the development sample, potentially influencing which neighborhood features carried the greatest predictive weight.\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e However, our ability to validate BVI in a rigorous research context unreliant on EHR measurement bias suggests that BVI is reliable despite this constraint. As a geolocation-level index, the BVI is inherently subject to the ecological fallacy: block-group-level risk scores reflect average neighborhood exposure and cannot capture within-neighborhood heterogeneity or individual-level social circumstances that may deviate substantially from the group median.\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e Accordingly, BVI should complement and not replace individual-level social risk screening in clinical encounters, consistent with best practices. Nevertheless, BVI is more precise than generic SDOH indices given its included patient-level risk derived from medical encounters. Finally, generalization to individuals outside of the demographic and clinical distribution of our development or validation cohorts should be done with caution given the inherent variability of cognitive impairment risk by age, sex, and race.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study demonstrates that a clinically anchored, machine learning-derived geospatial index can identify communities and individuals at elevated risk for cognitive impairment and accelerated cognitive decline using routinely available health system and publicly accessible SDOH data. BVI provides a scalable, brain-specific, equity-focused neighborhood risk tool that outperforms existing generic deprivation indices in predicting longitudinal cognitive outcomes across both community-based and clinical settings. By enabling passive, address-based risk stratification at population scale, BVI offers a practical instrument for guiding targeted prevention, early identification, and equitable allocation of resources toward communities bearing the greatest burden of neighborhood-attributable risk for cognitive impairment.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eADC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAlzheimer's Disease Center\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDHS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDallas Heart Study\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003estandard deviation\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMoCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMontreal Cognitive Assessment\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBVI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eBrain Vulnerability Index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSVI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSocial Vulnerability Index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eADI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArea Deprivation Index.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConcept and design: YT, ACN, AK, IH\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAcquisition, analysis, or interpretation or data: YT, ACN, AK for Model development; IH, ZY, AS for model validation data and analysis\u003c/p\u003e\n\u003cp\u003eDrafting of the manuscript: All authors\u003c/p\u003e\n\u003cp\u003eCritical revision of the manuscript for the important intellectual content: All Authors\u003c/p\u003e\n\u003cp\u003eStatistical Analysis: YT, ACN, JY, IH\u003c/p\u003e\n\u003cp\u003eObtained funding: IH\u003c/p\u003e\n\u003cp\u003eSupervision: IH, AK\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding/Support:\u0026nbsp;\u003c/strong\u003eNational Institute of Health funding to P30AG092752, R01AG070861, K24AG062786 to IH\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRole of the Funder/Sponsor:\u0026nbsp;\u003c/strong\u003eNone\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData is Available upon request\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAlzheimer\u0026apos;s disease facts and figures. Alzheimer\u0026apos;s \u0026amp; Dementia 2023;19(4):1598\u0026ndash;1695.\u003c/li\u003e\n\u003cli\u003eAlzheimer\u0026apos;s Disease Facts and Figures. 2024.\u003c/li\u003e\n\u003cli\u003eAlzheimer\u0026apos;s disease facts and figures. Alzheimer\u0026apos;s \u0026amp; Dementia 2025;21(4):e70235.\u003c/li\u003e\n\u003cli\u003eLivingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the \u0026lt;em\u0026gt;Lancet\u0026lt;/em\u0026gt; standing Commission. The Lancet 2024;404(10452):572\u0026ndash;628.\u003c/li\u003e\n\u003cli\u003eGanguli M, Jacobsen E, Song R, et al. Social determinants of health and mild cognitive impairment in a diverse community sample. Journal of the American Geriatrics Society 2025;73(2):367\u0026ndash;378.\u003c/li\u003e\n\u003cli\u003eKrasnova A, Tom SE, Valeri L, Crane PK, Bennett DA. Direct Effect of Life-Course Socioeconomic Status on Late-Life Cognition and Cognitive Decline in the Rush Memory and Aging Project. Am J Epidemiol 2023;192(6):882\u0026ndash;894. (In eng).\u003c/li\u003e\n\u003cli\u003eLlibre-Guerra JJ, Jiang M, Acosta I, et al. Social determinants of health but not global genetic ancestry predict dementia prevalence in Latin America. Alzheimers Dement 2024;20(7):4828\u0026ndash;4840. (In eng).\u003c/li\u003e\n\u003cli\u003eWalsh S, Klee M, Hui EK, et al. Social determinants of dementia: A scoping review. Alzheimers Dement 2025;21(7):e70524. (In eng).\u003c/li\u003e\n\u003cli\u003eVassilaki M, Petersen RC, Vemuri P. Area Deprivation Index as a Surrogate of Resilience in Aging and Dementia. Front Psychol 2022;13:930415. (In eng).\u003c/li\u003e\n\u003cli\u003eDow D. Mapping Health Disparities:: Leveraging Area-Based Deprivation Indices for Targeted Chronic Disease Intervention. Dela J Public Health 2024;10(1):106\u0026ndash;110. (In eng).\u003c/li\u003e\n\u003cli\u003eD\u0026rsquo;Abreu A, Chernyavskiy P. Does the Area Deprivation Index (ADI) influence dementia evaluation and diagnosis in Virginia? (S29.004). Neurology 2023;100(17_supplement_2):2410.\u003c/li\u003e\n\u003cli\u003eHooten M, Ortega M, Oyeyemi A, Yu F, Ofori E. Investigating the relationships between motor skills, cognitive status, and area deprivation index in Arizona: a pilot study. Frontiers in Public Health 2024;Volume 12 - 2024 (Brief Research Report) (In English).\u003c/li\u003e\n\u003cli\u003eTan WY, Hargreaves C, Chen C, Hilal S. A Machine Learning Approach for Early Diagnosis of Cognitive Impairment Using Population-Based Data. J Alzheimers Dis 2023;91(1):449\u0026ndash;461. (In eng).\u003c/li\u003e\n\u003cli\u003eRine CM. Healthy People 2030: Tools for Action. Health \u0026amp; Social Work 2023;48(4):225\u0026ndash;226.\u003c/li\u003e\n\u003cli\u003ePengetnze Y, Sundaram V, Tamer Y, et al. The Community Vulnerability Compass: a novel, scalable approach for measuring and visualizing social determinants of health insights. JAMIA Open 2025;8(4):ooaf059.\u003c/li\u003e\n\u003cli\u003eTamer YT, Karam A, Roderick T, Miff S. Know Thy Patient: A Novel Approach and Method for Patient Segmentation and Clustering Using Machine Learning to Develop Holistic, Patient-Centered Programs and Treatment Plans. Catalyst non-issue content 2022;3(4).\u003c/li\u003e\n\u003cli\u003eNaeem J, Salazar-Contreras E, Sundaram V, Wainwright L, Kosel K, Miff S. The Dallas Accountable Health Community: Its Impact on Health-Related Social Needs, Care, and Costs. NEJM Catalyst 2022;3(9):CAT.22.0149.\u003c/li\u003e\n\u003cli\u003eVassilaki M, Fu S, Christenson LR, et al. Characterizing Performance Gaps of a Code-Based Dementia Algorithm in a Population-Based Cohort of Cognitive Aging. J Alzheimers Dis 2023;95(3):931\u0026ndash;940. (In eng).\u003c/li\u003e\n\u003cli\u003eBhattacharyya J, Barnes LL, Chen Y, et al. Evaluating linked ICD-10 Medicare claims data as a method of dementia case ascertainment in research settings. Alzheimers Dement 2025;21(5):e70200. (In eng).\u003c/li\u003e\n\u003cli\u003eCharlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373\u0026ndash;83.\u003c/li\u003e\n\u003cli\u003eU.S. Census Bureau. Understanding and using American Community Survey data.\u003c/li\u003e\n\u003cli\u003eThe Center for Neighborhood Technology. The Center for Neighborhood Technology\u0026rsquo;s Housing and Transportation (H+T\u0026reg;) Affordability Index \u003c/li\u003e\n\u003cli\u003eUnited States. Environmental Protection A. National walkability index user guide and methodology. Washington, D.C.: United States Environmental Protection Agency, 2021.\u003c/li\u003e\n\u003cli\u003eNational Center for Health S. U.S. small-area life expectancy estimates project : methodology and results summary. Hyattsville, Maryland: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics, 2018.\u003c/li\u003e\n\u003cli\u003eU.S. Department of Education. Institute of Education Sciences NCfES. Highlights of the 2017 U.S. Program for the International Assessment of Adult Competencies (PIAAC) Results (NCES 2020-777).\u003c/li\u003e\n\u003cli\u003eOpenWeather. OpenWeather OWHL\u0026trade; model.\u003c/li\u003e\n\u003cli\u003eTrust for Public Land. ParkServe.\u003c/li\u003e\n\u003cli\u003eSolution AG. Crimerisk. 1996\u0026ndash;2026.\u003c/li\u003e\n\u003cli\u003eBernardo K, Rossetti H, Weiner M, Cullum C, De Lemos J, Lacritz L. Pro-Inflammatory Substances and Cognition in the Dallas Heart Study. Open Journal of Psychiatry 2016;6:95\u0026ndash;101.\u003c/li\u003e\n\u003cli\u003eVictor RG, Haley RW, Willett DL, et al. The Dallas Heart Study: a population-based probability sample for the multidisciplinary study of ethnic differences in cardiovascular health. Am J Cardiol 2004;93(12):1473\u0026ndash;80.\u003c/li\u003e\n\u003cli\u003eDas SR, Drazner MH, Dries DL, et al. Impact of body mass and body composition on circulating levels of natriuretic peptides: results from the Dallas Heart Study. Circulation 2005;112(14):2163\u0026ndash;8.\u003c/li\u003e\n\u003cli\u003eMurthy VL, Wu KC. Advances in cardiac MRI: The Dallas Heart Study. Curr Cardiol Rep 2007;9(1):43.\u003c/li\u003e\n\u003cli\u003eAlbert MA, Ravenell J, Glynn RJ, Khera A, Halevy N, de Lemos JA. Cardiovascular risk indicators and perceived race/ethnic discrimination in the Dallas Heart Study. Am Heart J 2008;156(6):1103\u0026ndash;9.\u003c/li\u003e\n\u003cli\u003eLucarelli RT, Peshock RM, McColl R, et al. MR imaging of hippocampal asymmetry at 3T in a multiethnic, population-based sample: results from the Dallas Heart Study. AJNR Am J Neuroradiol 2013;34(4):752\u0026ndash;7.\u003c/li\u003e\n\u003cli\u003eChandra A, Neeland IJ, Berry JD, et al. The relationship of body mass and fat distribution with incident hypertension: observations from the Dallas Heart Study. J Am Coll Cardiol 2014;64(10):997\u0026ndash;1002.\u003c/li\u003e\n\u003cli\u003eMaroules CD, Khera A, Ayers C, et al. Cardiovascular outcome associations among cardiovascular magnetic resonance measures of arterial stiffness: the Dallas heart study. J Cardiovasc Magn Reson 2014;16(1):33.\u003c/li\u003e\n\u003cli\u003eMcClelland RL, Jorgensen NW, Budoff M, et al. 10-Year Coronary Heart Disease Risk Prediction Using Coronary Artery Calcium and Traditional Risk Factors: Derivation in the MESA (Multi-Ethnic Study of Atherosclerosis) With Validation in the HNR (Heinz Nixdorf Recall) Study and the DHS (Dallas Heart Study). J Am Coll Cardiol 2015;66(15):1643\u0026ndash;53.\u003c/li\u003e\n\u003cli\u003eYano Y, Vongpatanasin W, Ayers C, et al. Regional Fat Distribution and Blood Pressure Level and Variability: The Dallas Heart Study. Hypertension 2016;68(3):576\u0026ndash;83.\u003c/li\u003e\n\u003cli\u003eCarlson SM, Kim J, Khan DA, et al. Hippocampal volume in patients with asthma: Results from the Dallas Heart Study. J Asthma 2017;54(1):9\u0026ndash;16.\u003c/li\u003e\n\u003cli\u003eCeasar JN, Ayers C, Andrews MR, et al. Unfavorable perceived neighborhood environment associates with less routine healthcare utilization: Data from the Dallas Heart Study. PLoS One 2020;15(3):e0230041.\u003c/li\u003e\n\u003cli\u003eTejani S, McCoy C, Ayers CR, et al. Cardiometabolic Health Outcomes Associated With Discordant Visceral and Liver Fat Phenotypes: Insights From the Dallas Heart Study and UK Biobank. Mayo Clin Proc 2022;97(2):225\u0026ndash;237.\u003c/li\u003e\n\u003cli\u003eYuan C, Palka JM, Rohatgi A, et al. The Relationship Between Coronary Artery Calcification and Carotid Intima Media Thickness and Hippocampal Volume: An Analysis From the Dallas Heart Study. J Acad Consult Liaison Psychiatry 2023;64(3):218\u0026ndash;225.\u003c/li\u003e\n\u003cli\u003eRossetti HC, Weiner M, Hynan LS, Cullum CM, Khera A, Lacritz LH. Subclinical atherosclerosis and subsequent cognitive function. Atherosclerosis 2015;241(1):36\u0026ndash;41. (In eng).\u003c/li\u003e\n\u003cli\u003eNasreddine ZS, Phillips NA, B\u0026eacute;dirian V, et al. The Montreal Cognitive Assessment, MoCA: A Brief Screening Tool For Mild Cognitive Impairment. Journal of the American Geriatrics Society 2005;53(4):695\u0026ndash;699.\u003c/li\u003e\n\u003cli\u003eRoalf DR, Moberg PJ, Xie SX, Wolk DA, Moelter ST, Arnold SE. Comparative accuracies of two common screening instruments for classification of Alzheimer\u0026apos;s disease, mild cognitive impairment, and healthy aging. Alzheimers Dement 2013;9(5):529\u0026ndash;37. (In eng).\u003c/li\u003e\n\u003cli\u003eFalkowski JA, Hynan LS, Krishnan K, et al. Conversion of MoCA to MMSE scores. Alzheimers Dement (Amst) 2015;1(1):125. (In eng).\u003c/li\u003e\n\u003cli\u003eBlagus R, Lusa L. SMOTE for high-dimensional class-imbalanced data. BMC Bioinformatics 2013;14:106.\u003c/li\u003e\n\u003cli\u003eRobin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011;12:77.\u003c/li\u003e\n\u003cli\u003eBensken WP, McGrath BM, Gold R, Cottrell EK. Area-level social determinants of health and individual-level social risks: Assessing predictive ability and biases in social risk screening. Journal of Clinical and Translational Science 2023;7(1):e257.\u003c/li\u003e\n\u003cli\u003eLokar K, Zagar T, Zadnik V. Estimation of the Ecological Fallacy in the Geographical Analysis of the Association of Socio-Economic Deprivation and Cancer Incidence. Int J Environ Res Public Health 2019;16(3) (In eng).\u003c/li\u003e\n\u003cli\u003eLundberg SM, Lee S-I. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems; 2017.\u003c/li\u003e\n\u003cli\u003eKind AJH, Buckingham WR. Making Neighborhood-Disadvantage Metrics Accessible \u0026mdash; The Neighborhood Atlas. New England Journal of Medicine 2018;378(26):2456\u0026ndash;2458.\u003c/li\u003e\n\u003cli\u003eFlanagan BE, Gregory EW, Hallisey EJ, Heitgerd JL, Lewis B. A Social Vulnerability Index for Disaster Management. Journal of Homeland Security and Emergency Management;8(1).\u003c/li\u003e\n\u003cli\u003eTay JK, Narasimhan B, Hastie T. Elastic Net Regularization Paths for All Generalized Linear Models. Journal of Statistical Software 2023;106(1):1 \u0026ndash; 31.\u003c/li\u003e\n\u003cli\u003eFriedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 2010;33(1):1\u0026ndash;22.\u003c/li\u003e\n\u003cli\u003eBreiman L. Random Forests. Machine Learning 2001;45(1):5\u0026ndash;32.\u003c/li\u003e\n\u003cli\u003eMeyer D. e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. 2024.\u003c/li\u003e\n\u003cli\u003eKrzywinski M, Altman N. Classification and regression trees. Nature Methods 2017;14(8):757\u0026ndash;758.\u003c/li\u003e\n\u003cli\u003eZahodne LB, Kraal AZ, Zaheed AB, others. Neighborhood disadvantage reduces cognitive reserve independent of ethnicity, education, and occupation. Neurology 2024;102(5):e209047.\u003c/li\u003e\n\u003cli\u003eMeyer OL, Tsai J, Marcum ZA, others. Neighborhood disadvantage and the incidence of dementia in US older adults. Alzheimer\u0026apos;s \u0026amp; Dementia 2025.\u003c/li\u003e\n\u003cli\u003eKind AJH, others. Area Deprivation Index and Progression to Dementia. Neurology 2022;98(18 Supplement):1244.\u003c/li\u003e\n\u003cli\u003eAlzheimer\u0026apos;s A. Social Determinants of Health and Dementia Risk - Compiled Report. 2023.\u003c/li\u003e\n\u003cli\u003eLivingston G, Huntley J, Liu XY, others. Dementia prevention, intervention, and care: 2024 report of the Lancet Commission. The Lancet 2024;404(10452):572\u0026ndash;628.\u003c/li\u003e\n\u003cli\u003eCenters for Disease C, Prevention. Non-Medical Factors that Affect Alzheimer\u0026apos;s Disease and Related Dementias. 2025.\u003c/li\u003e\n\u003cli\u003eSmith ER, Broughton M, Baker R, others. Social determinants of dementia: A scoping review. Alzheimer\u0026apos;s \u0026amp; Dementia 2025.\u003c/li\u003e\n\u003cli\u003eWang Y, Wang D, Ye X, others. Practical guide to SHAP analysis: Explaining supervised machine learning models in healthcare. Communications Medicine 2024;4(1):206.\u003c/li\u003e\n\u003cli\u003eSavage SA. Characterizing Performance Gaps of A Code-Based Dementia Algorithm. The University of Texas Health Science Center at Houston; 2022.\u003c/li\u003e\n\u003cli\u003eZhu Y, Wang L, Yin P, others. Automated phenotyping of mild cognitive impairment and dementias using electronic health records. International Journal of Medical Informatics 2025;193:105649.\u003c/li\u003e\n\u003cli\u003eMayeda ER, Gilsanz P, Haneuse S, others. Regional variation in diagnostic intensity of dementia among older Medicare beneficiaries: 2016-2019. Alzheimer\u0026apos;s \u0026amp; Dementia 2024.\u003c/li\u003e\n\u003cli\u003eRichardson S, Montfort C. Estimation of the Ecological Fallacy in the Geographical Analysis of the Association of Socio-economic Deprivation and Cancer Incidence. Spatial and Spatio-temporal Epidemiology 2019;28:1\u0026ndash;9.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1. Characteristics of Study Population Used for the Development of the Brain Vulnerability Index\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSource: Authors\u0026apos; own analyses.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eOverall\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eEHR-Evidence for Cognitive Impairment= No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eEHR-Evidence for Cognitive Impairment=Yes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN=231,630\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN=192,060\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN=39,570\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep-value*\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e61.87 (14.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e66.70 (18.61)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e60.87 (13.82)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSex\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eFemale\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e126,894 (54.78%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e105,605 (54.99%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e21,289 (53.80%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eMale\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e104,736 (45.22%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e86,455 (45.01%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e18,281 (46.20%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFirst Race\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eAmerican Indian\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e643 (0.28%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e528 (0.27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e115 (0.29%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eAsian\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e7,740 (3.34%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e6,347 (3.30%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e1,393 (3.52%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eBlack or African American\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e72,028 (31.10%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e57,640 (30.01%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e14,388 (36.36%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eOther Pacific Islander\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e727 (0.31%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e605 (0.32%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e122 (0.31%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eUnkown\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e21,158 (9.13%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e20,291 (10.56%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e867 (2.19%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eWhite\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e129,334 (55.84%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e106,649 (55.53%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e22,685 (57.33%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEthnicity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eHispanic\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e90,241 (38.96%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e77,060 (40.12%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e13,181 (33.31%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eNon-Hispanic\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e118,723 (51.26%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e93,232 (48.54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e25,491 (64,42%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e22,666 (9.79%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e21,768 (11.33%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e898 (2.27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSmoking Status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eEver\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e67,571 (29.17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e50,613 (26.35%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e16,958 (42.86%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eNever\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e164,059 (70.83%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e141,447 (73.65%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e22,612 (57.14%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMarital Status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eDivorced or Legally Separated\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e27,371 (11.82%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e21,389 (11.14%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e5,982 (15.12%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eMarried, Common Law or Significant Other\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e78,834 (34.03%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e68,250 (35.54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e10,584 (26.75%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eSingle\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e81,640 (35.17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e65,084 (33.89%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e16,376 (31.38%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e28,161 (12.16%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e27,415 (14.27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e746 (1.89%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eWidowed\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e15,804 (6.82%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e9,922 (5.17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e5,882 (14.86%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eCharlson Comorbidity Index (Age-Adjusted)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e0.61 (1.58)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e2.00 (2.56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.32 (1.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSDOH (Average Index/Indicator Value) **\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eCommunity Vulnerability Index)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e0.55 (0.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.54 (0.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.59 (0.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eHousehold Essentials Index\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e0.54 (0.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.53 (0.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.58 (0.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eEmpowered People Index\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e0.51 (0.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.5 (0.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.55 (0.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eEquitable Communities Index\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e0.61 (0.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.61 (0.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.64 (0.23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eGood Health Index\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e0.48 (0.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.47 (0.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e0.52 (0.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHealthcare Utilization, 1 year prior to diagnosis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003ePrimary Care Visits\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e136085 (0.59\u0026plusmn;3.29 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e86758 (0.45\u0026plusmn;2.64 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e49327 (1.25\u0026plusmn;5.37 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eED Visits\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e1505 (0.01\u0026plusmn;0.08 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e868 (0\u0026plusmn;0.07 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e637 (0.02\u0026plusmn;0.13 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eHospital Admissions\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e276 (0\u0026plusmn;0.03 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e50 (0\u0026plusmn;0.02 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e226 (0.01\u0026plusmn;0.08 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eMental Health Encounters\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e1250 (0.01\u0026plusmn;0.24 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e370 (0\u0026plusmn;0.13 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e880 (0.02\u0026plusmn;0.5 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eNeurology Encounters\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e5732 (0.02\u0026plusmn;0.51 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e1951 (0.01\u0026plusmn;0.29 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e3781 (0.1\u0026plusmn;1.04 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eRheumatology Encounter\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e1464 (0.01\u0026plusmn;0.21 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e911 (0\u0026plusmn;0.18 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e553 (0.01\u0026plusmn;0.3 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003eGastroenterology Encounter\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e3928 (0.02\u0026plusmn;0.36 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e2136 (0.01\u0026plusmn;0.27 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e1792 (0.05\u0026plusmn;0.64 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003e\u003cem\u003ePulmonology Encounter\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cem\u003e1921 (0.01\u0026plusmn;0.23 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e959 (0\u0026plusmn;0.18 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e\u003cem\u003e962 (0.02\u0026plusmn;0.42 per patient)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd nowrap=\"\" valign=\"top\" style=\"width: 5px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026lt;0.001\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003e* Student t-test performed for mean difference between Alzheimer\u0026apos;s Dementia Status\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e** Higher the value, higher the vulnerability\u003c/em\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2. Key Characteristics of External Validation Study Participants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSource: Authors\u0026apos; own analyses.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCharacteristic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eADC (N=863)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDHS (N=532)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCombined (N=1,395)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDemographics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAge, years, mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e70.1 (8.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e60.3 (10.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e66.4 (10.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eFemale, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e440 (51.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e322 (60.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e762 (54.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eRace/Ethnicity, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eWhite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e713 (82.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e161 (30.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e874 (62.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAfrican American\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e139 (16.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e310 (58.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e449 (32.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e11 (1.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e61 (11.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e72 (5.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCognitive Measures\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMoCA score, mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e22.3 (5.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e22.1 (4.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e22.2 (5.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCognitive impairment, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e551 (63.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e156 (29.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e707 (50.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eImpairment Category, n (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eNormal cognition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e312 (36.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e376 (70.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e688 (49.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMild cognitive impairment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e218 (25.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e153 (28.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e371 (26.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDementia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e333 (38.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e3 (0.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e336 (24.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNeighborhood Indices\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eBVI (block group), mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.45 (0.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.48 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.46 (0.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSVI, mean (SD)ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.37 (0.28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.54 (0.30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.44 (0.30)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eADI national percentile, mean (SD)ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e38.7 (25.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e56.5 (26.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e45.7 (27.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eAbbreviations: ADC = Alzheimer\u0026apos;s Disease Center; DHS = Dallas Heart Study; SD = standard deviation; MoCA = Montreal Cognitive Assessment; BVI = Brain Vulnerability Index; SVI = Social Vulnerability Index; ADI = Area Deprivation Index.\u003c/p\u003e\n\u003cp\u003eᵃ SVI available for N=772 (ADC) and N=496 (DHS); ADI available for N=771 (ADC) and N=495 (DHS).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3. Cross-sectional Association of Neighborhood Vulnerability Indices with Cognitive Outcomes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSource: Authors\u0026apos; own analyses.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCharacteristic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eADC (N=863)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDHS (N=532)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCombined (N=1,395)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMoCA scores\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUnadjusted, \u0026beta; (95% CI)ᵃ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.14 (\u0026minus;0.25, 0.54)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;1.36 (\u0026minus;1.72, \u0026minus;1.01)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;0.47 (\u0026minus;0.76, \u0026minus;0.19)**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAdjusted, \u0026beta; (95% CI)ᵇ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.14 (\u0026minus;0.26, 0.53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;1.06 (\u0026minus;1.41, \u0026minus;0.71)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;0.35 (\u0026minus;0.63, \u0026minus;0.07)*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eR\u0026sup2;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.028\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.191\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.061\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eRisk Stratification (BVI \u0026ge;0.5282)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHigh risk, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e119 (13.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e159 (29.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e278 (19.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMoCA difference, unadjusted\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;0.63 (\u0026minus;1.73, 0.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;2.73 (\u0026minus;3.52, \u0026minus;1.93)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;1.72 (\u0026minus;2.42, \u0026minus;1.02)*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMoCA difference, adjustedᵇ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;0.82 (\u0026minus;1.91, 0.28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;2.21 (\u0026minus;2.97, \u0026minus;1.45)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;1.55 (\u0026minus;2.23, \u0026minus;0.87)***\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCohen\u0026apos;s d\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;0.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026minus;0.33\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCognitive Impairment risk for top quintile\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eBVI (\u0026ge;0.5282)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHigh risk, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e122 (14.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e159 (29.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e281 (20.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUnadjusted OR (95% CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.92 (0.62\u0026ndash;1.37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.61 (1.08\u0026ndash;2.40)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.22 (0.92\u0026ndash;1.63)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAdjusted OR (95% CI)ᵇ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.07 (0.71\u0026ndash;1.60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.55 (1.04\u0026ndash;2.33)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.27 (0.95\u0026ndash;1.70)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSVI (\u0026ge;0.60)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHigh risk, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e186 (24.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e231 (46.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e417 (32.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUnadjusted OR (95% CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.03 (0.73\u0026ndash;1.44)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.39 (0.94\u0026ndash;2.06)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.17 (0.91\u0026ndash;1.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAdjusted OR (95% CI)ᵇ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.10 (0.77\u0026ndash;1.56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.39 (0.93\u0026ndash;2.06)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.20 (0.93\u0026ndash;1.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eADI (\u0026ge;60)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHigh risk, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e174 (22.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e236 (47.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e410 (32.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eUnadjusted OR (95% CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.98 (0.69\u0026ndash;1.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.45 (0.98\u0026ndash;2.14)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.17 (0.90\u0026ndash;1.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAdjusted OR (95% CI)ᵇ\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.12 (0.78\u0026ndash;1.60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.48 (0.99\u0026ndash;2.20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.25 (0.96\u0026ndash;1.63)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eAbbreviations: BVI = Brain Vulnerability Index; SVI = Social Vulnerability Index; ADI = Area Deprivation Index; OR = odds ratio; CI = confidence interval.\u003c/p\u003e\n\u003cp\u003eᵃ Values are regression coefficients from the general linear models (GLM) with MoCA score modeled as a continuous dependent variable; BVI entered as a continuous standardized BVI along with study indicator in combined analysis.\u003c/p\u003e\n\u003cp\u003eᵇ Adjusted for age and sex (study indicator added in combined analysis).\u003c/p\u003e\n\u003cp\u003e*p\u0026lt;0.05, **p\u0026lt;0.01, ***p\u0026lt;0.001\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9381770/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9381770/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eSocial determinants of health (SDOH) are increasingly recognized as important contributors to cognitive impairment, including Alzheimer\u0026rsquo;s disease and related dementias. Existing indices are heavily weighted toward financial indicators and are not validated against cognitive outcomes. We aimed to develop and validate a novel brain-specific SDOH index that identifies high-risk communities for cognitive impairment.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThe Brain Vulnerability Index (BVI) integrates patient-level electronic health record data with neighborhood-level SDOH from the Community Vulnerability Compass, a population health analytics framework. In the model development phase, electronic health record data were obtained from patients seen at Parkland Health (Dallas, TX). Cases were defined as individuals with at least one ICD-10 diagnosis suggestive of cognitive impairment from 2015 to 2023 and controls were age-matched with no ICD-10 diagnosis. External validation against serial cognitive performance measured by the Montreal Cognitive Assessment (MoCA) and clinically adjudicated consensus diagnosis of cognitive impairment of any etiology was performed in community-based (Dallas Heart Study) and clinically based (Alzheimer\u0026rsquo;s Disease Research Center) cohorts.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe model development sample included 39,570 cases and 192,060 controls. Derived BVI at the block group level achieved detection of electronic health documentation suggestive of cognitive impairment at a balanced accuracy of 54.6%. In external validation (n\u0026thinsp;=\u0026thinsp;1,395), higher BVI values were associated with lower MoCA scores (β = \u0026minus;0.35; p\u0026thinsp;=\u0026thinsp;0.009) and with faster cognitive decline (\u0026minus;\u0026thinsp;1.04 vs\u0026thinsp;\u0026minus;\u0026thinsp;0.42 MoCA points/year in high- vs low-risk groups; p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). BVI was also associated with adjudicated cognitive impairment in the community-based cohort only (p\u0026thinsp;=\u0026thinsp;0.04). Across analyses, BVI correlated with but outperformed existing neighborhood indices.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eBVI, a novel multidimensional brain-specific SDOH index, is clinically anchored and is externally validated. It may assist in earlier identification of at-risk communities with accelerated cognitive decline and aid in resource allocation for targeted prevention strategies.\u003c/p\u003e","manuscriptTitle":"The Brain Vulnerability Index: Development and Validation of a Machine Learning–Derived, Community-Informed Geospatial Risk Score for Cognitive Impairment","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-27 14:18:51","doi":"10.21203/rs.3.rs-9381770/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-02T10:54:13+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-01T14:02:42+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-24T14:59:05+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-23T18:34:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"210946899345752099751900337915866365783","date":"2026-04-17T17:32:27+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"254742194171042703361266192790510141860","date":"2026-04-17T16:46:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"223775721013313718698122191827554160796","date":"2026-04-17T16:43:05+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"165971590774889133021095812628645132542","date":"2026-04-17T16:42:35+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-17T16:40:47+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-17T00:25:09+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-16T01:32:21+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Digital Medicine","date":"2026-04-10T16:33:35+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1fecf405-a216-48d3-bcc9-a52b33cea9f0","owner":[],"postedDate":"April 27th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-02T10:54:13+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-01T14:02:42+00:00","index":24,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[{"id":66963031,"name":"Health sciences/Diseases"},{"id":66963032,"name":"Health sciences/Health care"},{"id":66963033,"name":"Health sciences/Medical research"},{"id":66963034,"name":"Health sciences/Neurology"},{"id":66963035,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2026-05-02T11:08:29+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-27 14:18:51","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9381770","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9381770","identity":"rs-9381770","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}