Leveraging undecided cases in chart-reviewed phenotypes to enhance EHR-based association studies.

BackgroundThe annual cost of Alzheimer’s disease and related dementias (ADRD) in the US is $355 billion; that rate is expected to triple by 2050. Individuals ADRD accrue higher healthcare utilization costs than peers without ADRD, including those with comorbid conditions (e.g., cancer, coronary artery disease, diabetes). Although research examining prevalence of ADRD in American Indians/Alaskan Natives (AI/AN) is limited, recent data suggest the prevalence of ADRD in the AI/AN population is 49% higher than in the Non‐Hispanic White population. Only one study has evaluated the healthcare utilization of AI/ANs, which showed greater utilization rates among those with (versus without) ADRD diagnosis and a rapid increase in utilization following ADRD diagnosis. We have yet to comprehensively analyze the additive cost of ADRD within the US among AI/ANs.MethodPost‐paid electronic health record data from Wisconsin Department of Health Services were retrospectively analyzed to compare non‐Tribal and Indian Health Service utilization costs among Medicaid enrolled AI/ANs, ages 40 and older, based on the presence or absence of ADRD throughout data review (7/1/2015 through 7/1/2020). Three separate analyses were completed, separated by payer programs: fee‐for‐service, HMO, and long‐term care. In each analysis, controls (individuals without ADRD) were matched to cases (individuals with ADRD) using exact matching across the subclasses defined by crossings of age group, sex, diabetes status, and emphysema status (Tables 1‐3). Weighted Wilcoxon rank sum tests were then used to compare total payable costs per capita. Tests accounted for both the subclass sampling strata and weights produced through the matching process. Weighted Chi‐square tests of association were also employed to confirm balance was achieved across age groups, sex, diabetes status, and emphysema status in the matching process; no weighted imbalances occurred.ResultAmong all three programs, AI/ANs with dementia had significantly higher healthcare utilization costs than those without dementia, after matching for age, gender, diabetes and emphysema status.ConclusionAI/ANs with ADRD accrued at $25,000 more over 5 years, on average, compared to those without ADRD. The greatest cost disparity was in long‐term care waiver programs ($50,879). Cost‐effective interventions aimed at ADRD prevention and treatment among AI/ANs are greatly needed.

We assessed the suitability of pooled electronic health record (EHR) data from clinical research networks (CRNs) of the patient-centered outcomes research network to conduct studies of the association between tumor necrosis factor inhibitors (TNFi) and infections. EHR data from patients with one of seven autoimmune diseases were obtained from three CRNs and pooled. Person-level linkage of CRN data and Centers for Medicare and Medicaid Services (CMS) fee-for-service claims data was performed where possible. Using filled prescriptions from CMS claims data as the gold standard, we assessed the misclassification of EHR-based new (incident) user definitions. Among new users of TNFi, we assessed subsequent rates of hospitalized infection in EHR and CMS data. The study included 45 483 new users of TNFi, of whom 1416 were successfully linked to their CMS claims. Overall, 44% of new EHR TNFi prescriptions were not associated with medication claims. Our most specific new user definition had a misclassification rate of 3.5%-16.4% for prevalent use, depending on the medication. Greater than 80% of CRN prescriptions had either zero refills or missing refill data. Compared to using EHR data alone, there was a 2- to 8-fold increase in hospitalized infection rates when CMS claims data were added to the analysis. EHR data substantially misclassified TNFi exposure and underestimated the incidence of hospitalized infections compared to claims data. EHR-based new user definitions were reasonably accurate. Overall, using CRN data for pharmacoepidemiology studies is challenging, especially for biologics, and would benefit from supplementation by other sources.

OBJECTIVES:Providers vary in their impact on clinical outcomes, but this is rarely accounted for in healthcare research. By failing to identify the provider responsible for a patient’s care, investigators miss an opportunity to account for nonrandom variation in outcomes. Prior methods of identifying responsible providers have relied on manual chart review, which is time-consuming and expensive, or analysis of claims data, which has been demonstrated to be inaccurate. To address these gaps, we sought to develop an algorithm using electronic health record (EHR) data to identify the responsible provider for each day of a patient’s hospitalization.DESIGN:A multicenter retrospective cohort study.SETTING:Midwest healthcare system.PATIENTS:Hospitalized patients and their providers.INTERVENTIONS:None.MEASUREMENTS AND MAIN RESULTS:We first confirmed high inter-rater reliability of manual chart review to identify the responsible provider. Using manual chart review as the gold standard, we then assessed the accuracy of an automated algorithm in a set of randomly selected patients. The agreement between two independent physicians in their determination of the responsible provider by chart review was 100%. Among 200 randomly selected patients, the algorithm identified the same responsible provider as the physician chart reviewer on 93% (3372/3626; 95% CI, 92–94%) of patient-days.CONCLUSIONS:Readily available EHR data can be used to assign patients to providers daily with a high degree of accuracy. This methodology could be applied in healthcare research to identify sources of variation other than the intervention being studied.

BackgroundHealthcare providers are often targeted as research participants, especially for implementation science studies evaluating provider- or system-level issues. Frequently, provider eligibility is based on both provider and patient factors. Manual chart review and self-report are common provider screening strategies but require substantial time, effort, and resources. The automated use of electronic health record (EHR) data may streamline provider identification for implementation science research. Here, we describe an approach to provider screening for a Veterans Health Administration (VHA)-funded study focused on implementing risk-aligned surveillance for bladder cancer patients.MethodsOur goal was to identify providers at 6 pre-specified facilities who performed ≥10 surveillance cystoscopy procedures among bladder cancer patients in the 12 months prior to recruitment start on January 16, 2020, and who were currently practicing at 1 of 6 pre-specified facilities. Using VHA EHR data (using CPT, ICD10 procedure, and ICD10 diagnosis codes), we identified cystoscopy procedures performed after an initial bladder cancer diagnosis (i.e., surveillance procedures). Procedures were linked to VHA staff data to determine the provider of record, the number of cystoscopies they performed, and their current location of practice. To validate this approach, we performed a chart review of 105 procedures performed by a random sample of identified providers. The proportion of correctly identified procedures was calculated (Positive Predictive Value (PPV)), along with binomial 95% confidence intervals (CI).FindingsWe identified 1,917,856 cystoscopies performed on 703,324 patients from October 1, 1999—January 16, 2020, across the nationwide VHA. Of those procedures, 40% were done on patients who had a prior record of bladder cancer and were completed by 15,065 distinct providers. Of those, 61 performed ≥ 10 procedures and were currently practicing at 1 of the 6 facilities of interest in the 1 year prior to study recruitment. The random chart review of 7 providers found 101 of 105 procedures (PPV: 96%; 95% CI: 91% to 99%) were surveillance procedures and were performed by the selected provider on the recorded date.ImplicationsThese results show that EHR data can be used for accurate identification of healthcare providers as research participants when inclusion criteria consist of both patient- (temporal relationship between diagnosis and procedure) and provider-level (frequency of procedure and location of current practice) factors. As administrative codes and provider identifiers are collected in most, if not all, EHRs for billing purposes this approach can be translated from provider recruitment in VHA to other healthcare systems. Implementation studies should consider this method of screening providers.

Herpes zoster (HZ) is a painful condition that is managed with a combination of antivirals, steroids, and pain medications. Nursing home residents with Alzheimer’s disease and related dementias (ADRD) may have difficulty expressing HZ symptoms, thus affecting medication management. This study compared prescription and non-prescription medication use in nursing home residents with HZ by ADRD diagnosis. We conducted a retrospective cohort study using nursing home electronic health record data from 2017-2022. Residents with HZ and ADRD were identified using ICD-10 codes as of the HZ diagnosis date. Administration for pain medications, antivirals, and steroids were measured before and after the HZ diagnosis date in periods of time (e.g., 1-7, 8-14, and 15-30 days before/after diagnosis). To assess relative changes in medication prevalence over time, administrations 15-30 days prior to HZ represented the baseline period. We identified 4,719 residents with HZ (31.5% had ADRD). Among residents without ADRD, oxycodone increased from 1.6% to 3.4% (vs. ADRD=0.5% to 0.8%) and gabapentin from 13.4% to 33.9% (vs. ADRD=12.6% to 25.6%) between baseline and 0-7 days after HZ diagnosis. In the 0-7 days after HZ diagnosis, prednisone use was 1.74 times greater in residents without ADRD (10.3%) vs. with ADRD (5.9%), and acyclovir use was similar between groups (without ADRD=28.5%; with ADRD=27.1%). Differences in medications, especially prednisone, between residents with and without ADRD may reflect undertreatment of HZ symptoms in residents with ADRD.

Performance of the FIND-FH machine learning algorithm for the identification of individuals with suspected familial hypercholesterolemia.

BackgroundDespite the growing number and proportion of the immigrant population in the US, there is a lack of knowledge on healthcare expenditures among immigrants with cognitive limitations or Alzheimer’s disease and related dementias (ADRD). This study estimated differences in total healthcare expenditures between US‐born and foreign‐born individuals by the presence of ADRD and/or cognitive limitations. This study also examined the differences in healthcare expenditures by service types, including inpatient stays, office‐based visits, emergency department (ED) visits, home healthcare, and prescription drugs.MethodThe study used nationally representative data from the 2007‐2020 Medical Expenditure Panel Survey (MEPS). Generalized linear regression models were used to estimate differences in healthcare expenditures between US‐born and foreign‐born individuals with ADRD, cognitive limitations without ADRD, and without ADRD or cognitive limitations. Analyses were also implemented by the immigrants’ duration of residence in the US, classified as foreign‐born living less than 10 years in the US and more than 10 years in the US. The study sample included all community‐dwelling adult Americans and had separate subpopulation analyses on the elderly population aged 65 and above. Survey weights were applied to all estimates.ResultForeign‐born individuals with cognitive limitations or ADRD had substantially higher total healthcare expenditures than their counterparts. Compared to US‐born individuals with ADRD, foreign‐born individuals with ADRD spent significantly more on home healthcare ($4,456.1, p = 0.029). Among individuals without ADRD or cognitive limitations, foreign‐born individuals spent significantly less than their US‐born counterparts in most healthcare services. Compared to their US‐born counterparts, immigrants with ADRD who had lived in the US for more than 10 years encountered significantly higher home healthcare costs ($4,245.6, p = 0.031), whereas those who had lived in the US for less than 10 years incurred significantly lower costs on office‐based visits (‐$1,833.8, p<0.001) and ED visits (‐$514.4, p<0.001). In contrast, immigrants with no ADRD and cognitive limitations spent significantly less than their US‐born counterparts on healthcare regardless of residence duration in the US.ConclusionForeign‐born individuals with cognitive limitations or ADRD had substantially higher total healthcare expenditures than their US‐born counterparts, while foreign‐born individuals without ADRD or cognitive limitations spent significantly less than their US‐born counterparts.

Alzheimer’s Disease and Related Dementias (ADRD) afflict almost 7 million people in the USA alone. The majority of research in ADRD is conducted using post-mortem samples of brain tissue or carefully recruited clinical trial patients. While these resources are excellent, they suffer from lack of sex/gender, and racial/ethnic inclusiveness. Electronic Health Records (EHR) data has the potential to bridge this gap by including real-world ADRD patients treated during routine clinical care. In this study, we utilize EHR data from a cohort of 70,420 ADRD patients diagnosed and treated at Penn Medicine. Our goal is to uncover important risk features leading to three types of Neuro-Degenerative Disorders (NDD), including Alzheimer’s Disease (AD), Parkinson’s Disease (PD) and Other Dementias (OD). We employ a variety of Machine Learning (ML) Methods, including uni-variate and multivariate ML approaches and compare accuracies across the ML methods. We also investigate the types of features identified by each method, the overlapping features and the unique features to highlight important advantages and disadvantages of each approach specific for certain NDD types. Our study is important for those interested in studying ADRD and NDD in EHRs as it highlights the strengths and limitations of popular approaches employed in the ML community. We found that the uni-variate approach was able to uncover features that were important and rare for specific types of NDD (AD, PD, OD), which is important from a clinical perspective. Features that were found across all methods represent features that are the most robust.

Alzheimer's Disease and Related Dementias (ADRD) afflict almost 7 million people in the USA alone. The majority of research in ADRD is conducted using post-mortem samples of brain tissue or carefully recruited clinical trial patients. While these resources are excellent, they suffer from lack of sex/gender, and racial/ethnic inclusiveness. Electronic Health Records (EHR) data has the potential to bridge this gap by including real-world ADRD patients treated during routine clinical care. In this study, we utilize EHR data from a cohort of 70,420 ADRD patients diagnosed and treated at Penn Medicine. Our goal is to uncover important risk features leading to three types of Neuro-Degenerative Disorders (NDD), including Alzheimer's Disease (AD), Parkinson's Disease (PD) and Other Dementias (OD). We employ a variety of Machine Learning (ML) Methods, including uni-variate and multivariate ML approaches and compare accuracies across the ML methods. We also investigate the types of features identified by each method, the overlapping features and the unique features to highlight important advantages and disadvantages of each approach specific for certain NDD types. Our study is important for those interested in studying ADRD and NDD in EHRs as it highlights the strengths and limitations of popular approaches employed in the ML community. We found that the uni-variate approach was able to uncover features that were important and rare for specific types of NDD (AD, PD, OD), which is important from a clinical perspective. Features that were found across all methods represent features that are the most robust.

BackgroundThe natural language processing (NLP) algorithm for predetermined asthma criteria (NLP-PAC) was successfully developed and validated for automatically ascertaining pediatric asthma from electronic health record (EHRs) systems. A scalable, efficient, and automated tool for ascertaining adult asthma status from EHRs remains nonexistent.ObjectiveWe validated NLP-PAC enabling ascertainment and early identification of adult asthma status in their EHRs.MethodsWe applied the validated NLP-PAC to EHRs of a convenient sample (adult cohorts who participated in our previous population-based studies) in which a reference standard (ie, asthma status defined by manual chart review) is available. The performance of NLP-PAC was assessed by determining criterion validity against manual chart review and construct validity before and after the new EHR (Epic) system was implemented in 2018.ResultsThe cohort consisted of 1,898 subjects, with 43% male and a median age at time of last follow-up of 65 years (interquartile range, 55-76). Manual chart review and NLP-PAC identified 97 (5.1%) and 98 (5.1%) subjects with asthma, respectively, with 89 subjects commonly identified by both methods. The sensitivity, specificity, positive predictive value, and negative predictive value of NLP-PAC were 92%, 99%, 91%, and 99%, respectively, before the new EHR system was implement, which remained similar after introducing the system (95%, 88%, 96%, and 85%, respectively). The risk factors for asthma identified either by NLP-PAC or manual chart review were similar.ConclusionAutomatic asthma ascertainment for adults based on EHR data is feasible with our NLP algorithm, offering immense scientific and clinical value for large-scale clinical research and population management for adult asthma care.

Objective: To evaluate the continuity and completeness of electronic health record (EHR) data, and the concordance of select clinical outcomes and baseline comorbidities between EHR and linked claims data, from three healthcare delivery systems in Taiwan. Methods: We identified oral hypoglycemic agent (OHA) users from the Integrated Medical Database of National Taiwan University Hospital (NTUH-iMD), which was linked to the National Health Insurance Research Database (NHIRD), from June 2011 to December 2016. A secondary evaluation involved two additional EHR databases. We created consecutive 90-day periods before and after the first recorded OHA prescription and defined patients as having continuous EHR data if there was at least one encounter or prescription in a 90-day interval. EHR data completeness was measured by dividing the number of encounters in the NTUH-iMD by the number of encounters in the NHIRD. We assessed the concordance between EHR and claims data on three clinical outcomes (cardiovascular events, nephropathy-related events, and heart failure admission). We used individual comorbidities that comprised the Charlson comorbidity index to examine the concordance of select baseline comorbidities between EHRs and claims. Results: We identified 39,268 OHA users in the NTUH-iMD. Thirty-one percent (n = 12,296) of these users contributed to the analysis that examined data continuity during the 6-month baseline and 24-month follow-up period; 31% (n = 3,845) of the 12,296 users had continuous data during this 30-month period and EHR data completeness was 52%. The concordance of major cardiovascular events, nephropathy-related events, and heart failure admission was moderate, with the NTU-iMD capturing 49–55% of the outcome events recorded in the NHIRD. The concordance of comorbidities was considerably different between the NTUH-iMD and NHIRD, with an absolute standardized difference >0.1 for most comorbidities examined. Across the three EHR databases studied, 29–55% of the OHA users had continuous records during the 6-month baseline and 24-month follow-up period. Conclusion: EHR data continuity and data completeness may be suboptimal. A thorough evaluation of data continuity and completeness is recommended before conducting clinical and translational research using EHR data in Taiwan.

BackgroundAmerican Indian and Alaska Native (AI/AN) peoples face a rising burden of Alzheimer's disease and related dementia (ADRD). The 2024 Lancet Commission identified modifiable risk factors for dementia for the general population; however, a comprehensive understanding of risk factors associated with ADRD among AI/AN peoples is missing. This study utilizes a national database to estimate the prevalence of health conditions associated with ADRD and examine disparities between older AI/AN and White populations.MethodWe analyzed 2019 Medicare Master Beneficiary Summary File data for Medicare beneficiaries aged 68 and older, including all AI/AN beneficiaries and a 5% random sample of White beneficiaries. Two types of health conditions were examined: Lancet risk factors available in the Medicare data and other conditions that have been associated with ADRD (Table 1). Prevalence of ADRD and each condition was reported. Bivariate and multivariate logistic regressions assessed the associations between health conditions and ADRD.ResultAmong 112,280 AI/AN and 1,010,862 White beneficiaries, the former had higher age‐adjusted prevalence of ADRD (15.6 % vs. 13.3%), and substantially higher prevalence of 5 out of 9 Lancet risk factors, including diabetes, alcohol use disorder (AUD), tobacco use disorder, visual impairment, and hearing loss; similar prevalence of hypertension, depression, and traumatic brain injury (TBI); but a lower prevalence of hyperlipidemia (Table 1). Bivariate analyses revealed that all conditions were associated with higher ADRD odds in both populations, with TBI, depression, and AUD showing the strongest association with ADRD (Table 2). In multivariate regressions, the associations for TBI (odds ratio (OR) = 8.00), AUD (OR=3.30), visual impairment (OR=2.38), and hearing loss (OR = 1.34) were stronger in AI/AN beneficiaries, while depression, diabetes, and hypertension had stronger associations in Whites (Figure 1).ConclusionOlder AI/AN Medicare beneficiaries compared to their White counterparts, exhibit a higher prevalence of ADRD and 5 of the 9 Lancet risk factors. Furthermore, several conditions had stronger associations with ADRD among AI/AN beneficiaries, underscoring the importance of addressing disparities in ADRD risk factors to mitigate ADRD risk among AI/AN peoples. Further research using longitudinal data with social and behavioral determinants is essential to better understand and reduce these disparities.

Aging populations in low‐ and middle‐income countries (LMICs) are growing more rapidly than those in high income countries (HICs). Thus, globally, the burden of Alzheimer’s disease and related dementias (ADRD) profoundly impacts LMICs, with greater burden and risk for late‐life women than men. LMICs are unable to meet diagnostic challenges with underfunded healthcare infrastructure, low numbers of clinicians with training and expertise in ADRD, and a lack of culturally and linguistically sensitive screening assessments. Emerging social and societal factors that are important risk and protective factors for later life ADRD and are differentially experienced by gender/sex are relatively understudied, yet essential in LMICs. Typical samples in neuroscience studies have been small and highly selected. They have been biased toward Western, Educated, Industrialized, Rich, and Democratic (WEIRD) samples and do not apply well to target populations in LMICs. In addition, levels‐of‐analysis across social/societal and biological determinants of disease are not typically integrated, and social/societal factors such as sociocultural (e.g., culture, religion, gender norms, access to education, discrimination, social aspects of pandemics, disability), political (e.g., governments and policies; migrants and displaced persons; war and conflict; reproductive health, justice, and agency), environmental (e.g., climate change and pollution), and economic factors (e.g., poverty), and their co‐occurrence, have been minimally considered for their role and impact on ADRD. To better understand gender/sex differences in ADRD in LMICs, we review how these social and societal factors may be important to ADRD using an integrated Population Neuroscience and Syndemics framework. Population neuroscience considers the brain in context using multiple levels of analyses, a life course perspective, and tools to enhance internal and external validity, while Syndemics suggest that diseases and social conditions may cluster and interact in populations with syndemic risk factors—sociocultural, political, economic, and environmental factors which promote stress pathways. We lay out a model in which syndemic risk and protective factors contribute to or protect from stress and biological aging and ultimately ADRD risk, both directly and indirectly through physical and mental health conditions. We close with recommendations for future aging and ADRD research based on this integrated Population Neuroscience‐Syndemics of ADRD framework.

BackgroundEscalating drug costs are likely to disproportionately burden individuals with Alzheimer’s disease and related dementias (ADRD) as they often have multiple chronic conditions (MCC). This study aimed to estimate the incremental total and out‐of‐pocket (OOP) drug costs associated with ADRD and MCC across the drug cost distributions.MethodThe study used a nationally representative cross‐sectional dataset from 2003 to 2021 Medical Expenditure Panel Survey. The study sample comprised of 83,667 elderly community dwelling individuals aged 65 and above, among which 3,046 individuals had ADRD diagnosis and 2,710 had both ADRD and MCC. To address the skewed distribution of the drug costs, quantile regression models were used to examine the associations between prescription drug costs and having ADRD/MCC across the drug cost distributions at 25th, 50th, 75th, and 90th percentiles, after adjustment for other sociodemographic factors. MCC was defined as having two or more chronic conditions. Survey weights were applied to all estimates.ResultADRD patients and those with ADRD and MCC had significantly higher incremental costs at the upper end of the total and OOP drug cost distributions than at the lower end of the distributions, compared to those who had no ADRD diagnosis. The incremental OOP drug costs of ADRD at the 90th percentile were over 14 times higher than that at the 25th percentile ($624.43, p<0.001, 90th percentile; $42.66, p<0.001, 25th percentile), while the incremental total drug costs of ADRD at the 90th percentile was about 9 times higher than that of ADRD at the 25th percentile ($3,372.27, p<0.001, 90th percentile; $381.31, p<0.001, 25th percentile). Similarly, the incremental OOP drug costs among individuals with ADRD and MCC at the 90th percentile were 11 times more than that at the 25th percentile ($756.91 vs. $68.55, respectively), while the incremental cost difference for the total drug costs was 7 times higher among individuals with ADRD and MCC at the 90th percentile than those at the 25th percentile ($3,940.53 vs. $548.80, respectively).ConclusionThe results of the study highlighted that ADRD patients with and without MCC who are heavy prescription drug users face a disproportionately greater burden in OOP payments.

Research ObjectiveTo explore racial differences in the nursing home (NH) placement among Medicare‐Medicaid dual‐eligible home and community‐based service (HCBS) users with Alzheimer's disease and related dementias (ADRD), and determine the extent to which HCBS generosity may be influential.Study DesignThe following 2010‐2013 national data were linked at the individual level: Medicaid Analytic eXtract (MAX) Personal Summary (PS) file and Other Therapy (OT) file, the Minimum Data Set (MDS) 2.0/3.0 and Medicare Beneficiary Summary File and MedPAR.Dually eligible fee‐for‐service Medicare and Medicaid beneficiaries were identified using MBSF. Diagnosis of ADRD was based on the MBSF chronic condition file. New users were defined as those who did not have HCBS episodes in a prior 30‐day look‐back period, based on the OT records.The outcome variable was any NH admission within one year after the HBCS use, defined based on the MDS data. The main variables of interest were race (ie, white, black, other), the intensity of HCBS among ADRD population at the county level (the average monthly spending per HCBS user with ADRD in a county, based on MAX PS file), and the interaction terms between the intensity of ADRD and race. A linear probability model with county fixed‐effect and robust standard errors were estimated, accounting for individual covariates (eg, demographic characteristics, chronic conditions, prior use of institutional care).Population Studied1 527 974 new users of HCBS with ADRD during CY2010 to 2013.Principal FindingsThe intensity of HCBS among individuals with ADRD varies widely across counties, the 25th percentile, median, and the 75th percentile was $654, $827, and $1099, respectively. The overall NH admission rate among HCBS users with ADRD was 67%, 55%, and 36% for whites, blacks, and others, respectively. After accounting for individual covariates, whites continued to exhibit a higher NH admission rate, as compared with two other race groups. However, while a higher intensity of HCBS reduced the likelihood of NH placement for whites, it increased the risk among blacks. NH admission rate for blacks was 7.2 percentage points (P < 0.01) lower than for whites in a county with a median HCBS monthly spending. A one hundred dollars increase in monthly HCBS spending was associated with a 0.3 percentage points (P < 0.01) decrease in admission rate among whites, but a 0.3 percentage points (P < 0.01) increase among blacks.ConclusionsOur findings suggest that among dual‐eligible HCBS users with ADRD, whites were more likely to have NH admission within 1 year than non‐whites. Moreover, higher intensity of HCBS in a county was associated with lower a NH admission rate among whites, but a higher rate among blacks.Implications for Policy or PracticeStates have been promoting the use of HCBS as the means of rebalancing the cost of traditional institutional care while allowing more people to remain living in the community. However, the expansion of HCBS may have differential impact on NH admission risk for whites and blacks. Better understanding of racial differences in care needs is critical to inform the State Medicaid program design.Primary Funding SourceNational Institutes of Health.