Background Social determinants of health such as residential segregation have been identified as drivers of disparities in health outcomes; however, this has been understudied for out‐of‐hospital cardiac arrest (OHCA). We sought to examine whether there were differences in survival to discharge and survival with good neurological outcome, as well as likelihood of bystander cardiopulmonary resuscitation, using validated measures of racial, ethnic, and economic segregation. Methods We conducted a retrospective observational study using data from the Cardiac Arrest Registry to Enhance Survival data set. The primary predictor for this study was the Index of Concentration at the Extremes. The primary outcomes were survival to discharge and survival with good neurological status. Results During the study period, 626 264 had an out‐of‐hospital cardiac arrest, and patients had a mean age of 62 years (SD 17.2 years). In multivariable models, we observed an increased likelihood of survival to discharge and survival with good neurological outcome for those patients residing in more highly segregated predominately White population and higher‐income census tracts as compared with more highly segregated and lower‐income Black and Hispanic/Latinx population census tracts. We found that the magnitude of this disparity was 24% for the outcome of survival to discharge as compared with reference (relative risk,1.24 [95% CI, 1.20–1.28]). Conclusions This research suggests that areas impacted by residential and economic segregation are important targets for both public policy interventions as well as addressing disparities in care across the chain of survival for out‐of‐hospital cardiac arrest.

Generative Large Language Models (LLMs) hold significant promise in healthcare, demonstrating capabilities such as passing medical licensing exams and providing clinical knowledge. However, their current use as information retrieval tools is limited by challenges like data staleness, resource demands, and occasional generation of incorrect information. This study assessed the potential of LLMs to function as autonomous agents in a simulated tertiary care medical center, using real-world clinical cases across multiple specialties. Both proprietary and open-source LLMs were evaluated, with Retrieval Augmented Generation (RAG) enhancing contextual relevance. Proprietary models, particularly GPT-4, generally outperformed open-source models, showing improved guideline adherence and more accurate responses with RAG. The manual evaluation by expert clinicians was crucial in validating models' outputs, underscoring the importance of human oversight in LLM operation. Further, the study emphasizes Natural Language Programming (NLP) as the appropriate paradigm for modifying model behavior, allowing for precise adjustments through tailored prompts and real-world interactions. This approach highlights the potential of LLMs to significantly enhance and supplement clinical decision-making, while also emphasizing the value of continuous expert involvement and the flexibility of NLP to ensure their reliability and effectiveness in healthcare settings.

Importance. Medical ethics is inherently complex, shaped by a broad spectrum of opinions, experiences, and cultural perspectives. The integration of large language models (LLMs) in healthcare is new and requires an understanding of their consistent adherence to ethical standards. Objective. To compare the agreement rates in answering questions based on ethically ambiguous situations between three frontier LLMs (GPT-4, Gemini-pro-1.5, and Llama-3-70b) and a multi-disciplinary physician group. Methods. In this cross-sectional study, three LLMs generated 1,248 medical ethics questions. These questions were derived based on the principles outlined in the American College of Physicians Ethics Manual. The topics spanned traditional, inclusive, interdisciplinary, and contemporary themes. Each model was then tasked in answering all generated questions. Twelve practicing physicians evaluated and responded to a randomly selected 10% subset of these questions. We compared agreement rates in question answering among the physicians, between the physicians and LLMs, and among LLMs. Results. The models generated a total of 3,744 answers. Despite physicians perceiving the questions' complexity as moderate, with scores between 2 and 3 on a 5-point scale, their agreement rate was only 55.9%. The agreement between physicians and LLMs was also low at 57.9%. In contrast, the agreement rate among LLMs was notably higher at 76.8% (p < 0.001), emphasizing the consistency in LLM responses compared to both physician-physician and physician-LLM agreement. Conclusions. LLMs demonstrate higher agreement rates in ethically complex scenarios compared to physicians, suggesting their potential utility as consultants in ambiguous ethical situations. Future research should explore how LLMs can enhance consistency while adapting to the complexities of real-world ethical dilemmas.

Large language models (LLMs) can optimize clinical workflows; however, the economic and computational challenges of their utilization at the health system scale are underexplored. We evaluated how concatenating queries with multiple clinical notes and tasks simultaneously affects model performance under increasing computational loads. We assessed ten LLMs of different capacities and sizes utilizing real-world patient data. We conducted >300,000 experiments of various task sizes and configurations, measuring accuracy in question-answering and the ability to properly format outputs. Performance deteriorated as the number of questions and notes increased. High-capacity models, like Llama-3-70b, had low failure rates and high accuracies. GPT-4-turbo-128k was similarly resilient across task burdens, but performance deteriorated after 50 tasks at large prompt sizes. After addressing mitigable failures, these two models can concatenate up to 50 simultaneous tasks effectively, with validation on a public medical question-answering dataset. An economic analysis demonstrated up to a 17-fold cost reduction at 50 tasks using concatenation. These results identify the limits of LLMs for effective utilization and highlight avenues for cost-efficiency at the enterprise scale.

INTRODUCTION:Large language models (LLMs) have grown in popularity in recent months and have demonstrated advanced clinical reasoning ability. Given the need to prioritize the sickest patients requesting emergency medical services (EMS), we attempted to identify if an LLM could accurately triage ambulance requests using real-world data from a major metropolitan area. METHODS:An LLM (ChatGPT 4o Mini, Open AI, San Francisco, CA, USA) with no prior task-specific training was given real ambulance requests from a major metropolitan city in the United States. Requests were batched into groups of four, and the LLM was prompted to identify which of the four patients should be prioritized. The same groupings of four requests were then shown to a panel of experienced critical care paramedics who voted on which patient should be prioritized. RESULTS:Across 98 groupings of four ambulance requests (392 total requests), the LLM agreed with the paramedic panel in most cases (76.5 %, n = 75). In groupings where the paramedic panel was unanimous in their decision (n = 48), the LLM agreed with the unanimous panel in 93.8 % of groupings (n = 45). CONCLUSIONS:Our preliminary analysis indicates LLMs may have the potential to become a useful tool for triage and resource allocation in emergency care settings, especially in cases where there is consensus among subject matter experts. Further research is needed to better understand and clarify how they may best be of service.

Background Social determinants of health (SDOH) are critical drivers of health disparities and patient outcomes. However, accessing and collecting patient-level SDOH data can be operationally challenging in the emergency department (ED) clinical setting, requiring innovative approaches. Objective This scoping review examines the potential of AI and data science for modeling, extraction, and incorporation of SDOH data specifically within EDs, further identifying areas for advancement and investigation. Methods We conducted a standardized search for studies published between 2015 and 2022, across Medline (Ovid), Embase (Ovid), CINAHL, Web of Science, and ERIC databases. We focused on identifying studies using AI or data science related to SDOH within emergency care contexts or conditions. Two specialized reviewers in emergency medicine (EM) and clinical informatics independently assessed each article, resolving discrepancies through iterative reviews and discussion. We then extracted data covering study details, methodologies, patient demographics, care settings, and principal outcomes. Results Of the 1047 studies screened, 26 met the inclusion criteria. Notably, 9 out of 26 (35%) studies were solely concentrated on ED patients. Conditions studied spanned broad EM complaints and included sepsis, acute myocardial infarction, and asthma. The majority of studies (n=16) explored multiple SDOH domains, with homelessness/housing insecurity and neighborhood/built environment predominating. Machine learning (ML) techniques were used in 23 of 26 studies, with natural language processing (NLP) being the most commonly used approach (n=11). Rule-based NLP (n=5), deep learning (n=2), and pattern matching (n=4) were the most commonly used NLP techniques. NLP models in the reviewed studies displayed significant predictive performance with outcomes, with F1-scores ranging between 0.40 and 0.75 and specificities nearing 95.9%. Conclusions Although in its infancy, the convergence of AI and data science techniques, especially ML and NLP, with SDOH in EM offers transformative possibilities for better usage and integration of social data into clinical care and research. With a significant focus on the ED and notable NLP model performance, there is an imperative to standardize SDOH data collection, refine algorithms for diverse patient groups, and champion interdisciplinary synergies. These efforts aim to harness SDOH data optimally, enhancing patient care and mitigating health disparities. Our research underscores the vital need for continued investigation in this domain.

Introduction: Helicopter emergency medical services (HEMS) are used in the United States and globally to respond to patients with critical illness and victims of traumatic injury. Relatively limited research has examined their role in responding to out-of-hospital cardiac arrests (OHCA) in the United States. In this study, we compared OHCA treated by HEMS units with cardiac arrests treated by ground ambulances. Methods: We queried a large national-level database of emergency medical services (EMS) activations in the United States (NEMSIS). Inclusion criteria were OHCA activations between January 1, 2022 and December 31, 2022 treated by either HEMS or ground ambulance. Key arrest data from both groups were then compared. Interfacility transfers and cardiac arrests after EMS arrival were excluded. Results: A total of 1,233 cardiac arrests treated by HEMS and 341,096 cardiac arrests treated by ground ambulances met inclusion criteria. Comparing the two groups, cardiac arrests with HEMS response were more likely to be male (66.7% vs. 62.8%, p < 0.01), White (50.2% vs. 45.7%, p < 0.01), under 18 years old (10.9% vs. 2.7%, p < 0.001), associated with traumatic injury (19.1% vs. 5.7%, p < 0.001), witnessed (72.7% vs. 37.3%, p < 0.001), and initially-shockable (24.7% vs. 11.1%, p < 0.001). Conclusion: Our comparison of cardiac arrests treated by HEMS with cardiac arrests treated by ground ambulance reveals significant differences between the two groups. Further research is needed to better characterize HEMS' ideal role in the response to OHCA as new prehospital resuscitative techniques for non-traumatic and traumatic cardiac arrest are developed.

Witnessed out-of-hospital cardiac arrests (OHCA) are associated with improved outcomes with increased likelihood of rapid activation of the emergency response system thus decreasing no-flow or low-flow times. Current literature has largely treated witnessed cardiac arrest as binary (i.e. was a cardiac arrest witnessed Yes/No). We hypothesize there is a need to further delineate and categorize distinct bystander types; these differences may have important downstream contributions to outcomes and survival and targets for improving training and recognition. For example, a bystander who is a healthcare provider with formal training or has experience performing resuscitation may be associated with better outcomes than a non-medical layperson. We examined a national-level database of emergency medical services (EMS) activations in the United States (NEMSIS). Inclusion criteria were any adult (18+ years) cardiac arrest activation between January 2022 and December 2023. Cardiac arrests taking place after EMS arrival (EMS-witnessed) were excluded. Witnessed status was described as unwitnessed, witnessed by healthcare provider, witnessed by family member, or witnessed by bystander. Across basic demographic variables, we compared overall witnessed rates, family member witnessed rates, and healthcare worker witnessed rates. A total of 791,217 cardiac arrests met inclusion criteria. Overall, 62.5% of arrests were unwitnessed, 22.8% were witnessed by a family member, and 6.3% were witnessed by a healthcare provider. Across sex, cardiac arrests in male patients were more likely to be witnessed, however, cardiac arrests in female patients were more likely to be witnessed by a family member or healthcare provider (p<0.05). Significant differences in witnessed status were also seen across race/ethnicity. For instance, cardiac arrests in Black/African American patients were less likely to be witnessed by family members and cardiac arrests in Hispanic/Latino patients were less likely to be witnessed by healthcare workers (p<0.05). Our analysis of a large and nationally-representative database of cardiac arrest suggests differences in rates of witnessed cardiac arrest and – perhaps, more importantly – the existence of different types of bystanders. Moreover, when considering these differences, there is heterogeneity among demographic variables. Future research is needed to understand whether these differences in bystander type may contribute to differences in outcomes.

Background:Accurate medical coding is essential for clinical and administrative purposes but complicated, time-consuming, and biased. This study compares Retrieval-Augmented Generation (RAG)-enhanced LLMs to provider-assigned codes in producing ICD-10-CM codes from emergency department (ED) clinical records. Methods:Retrospective cohort study using 500 ED visits randomly selected from the Mount Sinai Health System between January and April 2024. The RAG system integrated past 1,038,066 ED visits data (2021-2023) into the LLMs' predictions to improve coding accuracy. Nine commercial and open-source LLMs were evaluated. The primary outcome was a head-to-head comparison of the ICD-10-CM codes generated by the RAG-enhanced LLMs and those assigned by the original providers. A panel of four physicians and two LLMs blindly reviewed the codes, comparing the RAG-enhanced LLM and provider-assigned codes on accuracy and specificity. Findings:RAG-enhanced LLMs demonstrated superior performance to provider coders in both the accuracy and specificity of code assignments. In a targeted evaluation of 200 cases where discrepancies existed between GPT-4 and provider-assigned codes, human reviewers favored GPT-4 for accuracy in 447 instances, compared to 277 instances where providers' codes were preferred (p<0.001). Similarly, GPT-4 was selected for its superior specificity in 509 cases, whereas human coders were preferred in only 181 cases (p<0.001). Smaller open-access models, such as Llama-3.1-70B, also demonstrated substantial scalability when enhanced with RAG, with 218 instances of accuracy preference compared to 90 for providers' codes. Furthermore, across all models, the exact match rate between LLM-generated and provider-assigned codes significantly improved following RAG integration, with Qwen-2-7B increasing from 0.8% to 17.6% and Gemma-2-9b-it improving from 7.2% to 26.4%. Interpretation:RAG-enhanced LLMs improve medical coding accuracy in EDs, suggesting clinical workflow applications. These findings show that generative AI can improve clinical outcomes and reduce administrative burdens. Funding:This work was supported in part through the computational and data resources and staff expertise provided by Scientific Computing and Data at the Icahn School of Medicine at Mount Sinai and supported by the Clinical and Translational Science Awards (CTSA) grant UL1TR004419 from the National Center for Advancing Translational Sciences. Research reported in this publication was also supported by the Office of Research Infrastructure of the National Institutes of Health under award number S10OD026880 and S10OD030463. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. Twitter Summary:A study showed AI models with retrieval-augmented generation outperformed human doctors in ED diagnostic coding accuracy and specificity. Even smaller AI models perform favorably when using RAG. This suggests potential for reducing administrative burden in healthcare, improving coding efficiency, and enhancing clinical documentation.

Objective Given the consequences of failed intubation, there is great interest in optimizing airway management success rates. Growing evidence suggests that use of a bougie device is associated with improved airway success. Bougies with a flexible tip may increase intubation success by offering operators greater control. In this study, we performed a feasibility assessment of flexible tip bougies involving clinicians who routinely perform emergency airway management. Methods We used a feasibility simulation study design with pre- and postsurveys. Participants first completed a presession survey asking about their intubation experience and preferences. They were then given a brief demonstration of the flexible tip bougie with an opportunity to ask questions. They completed 2 intubations on an airway simulator using video laryngoscopy—one with a flexible tip bougie and one with a standard bougie. During these intubations, the time required to pass the bougie through the vocal cords and the time required to pass a tracheal tube through the vocal cords was assessed. Participants finally completed a postsession survey asking about their experiences with the flexible tip bougie. Results A total of 18 participants took part in the study. All participants rated themselves as familiar with tracheal intubation, and most (72.2%) said that they would consider using a flexible tip bougie on future intubations. Quantitatively, there was no clinically significant difference in the time needed to pass the bougie through the vocal cords (0.38 second) and the time needed to pass a tracheal tube through the vocal cords (8.24 seconds). Conclusion Our simulation-based study comparing a flexible tip bougie with a conventional bougie revealed improved operator feedback with the flexible tip bougie and a clinically insignificant temporal difference in intubation metrics. Further study is needed to better assess the benefits of a flexible tip bougie in emergency and nonemergency airway management.

Background: A growing body of evidence suggests outcomes for cardiac arrest in adults are worse during nights and weekends when compared with daytime and weekdays. Similar research has not yet been carried out in the infant setting. Methods: We examined the National Emergency Medical Services Information System (NEMSIS), a database containing millions of emergency medical services (EMS) runs in the United States. Inclusion criteria were infant out-of-hospital cardiac arrests (patients <1 years old) taking place prior to EMS arrival between January 2021 and December 2022 where EMS documented whether return of spontaneous circulation (ROSC) was achieved. Cardiac arrests were classified as occurring during either the day (defined as 0800-1959) or the night (defined as 2000-0759) and weekends (Saturday/Sunday) or weekdays (Monday-Friday). Rates of ROSC achievement were compared. Results: A total of 8549 infant cardiac arrests met inclusion criteria: 5074 (59.4%) took place during daytime compared with 3475 (40.6%) during nighttime, and 5989 (70.1%) arrests occurred on weekdays compared with 2560 (29.9%) on weekends. Rates of ROSC achievement were significantly lower on weekends versus weekdays (16.8% vs. 14.1%; p = 0.00097). A difference in ROSC rates when comparing daytime and nighttime was seen, but this difference was not statistically significant (16.4% vs. 15.3%; p = 0.08076). Conclusion: ROSC achievement rates for infant out-of-hospital cardiac arrest are significantly lower on weekends when compared with weekdays. Further study and quality improvement work is needed to better understand this. (c) 2024 Published by Elsevier Inc

Background The national impact of racial residential segregation on out‐of‐hospital cardiac arrest outcomes after initial resuscitation remains poorly understood. We sought to characterize the association between measures of racial and economic residential segregation at the ZIP code level and long‐term survival and readmissions after out‐of‐hospital cardiac arrest among Medicare beneficiaries. Methods and Results In this retrospective cohort study, using Medicare claims data, our primary predictor was the index of concentration at the extremes, a measure of racial and economic segregation. The primary outcomes were death up to 3 years and readmissions. We estimated hazard ratios (HRs) across all 3 types of index of concentration at the extremes measures for each outcome while adjusting for beneficiary demographics, treating hospital characteristics, and index hospital procedures. In fully adjusted models for long‐term survival, we found a decreased hazard of death and risk of readmission for beneficiaries residing in the more segregated White communities and higher‐income ZIP codes compared with the more segregated Black communities and lower‐income ZIP codes across all 3 indices of concentration at the extremes measures (race: HR, 0.87 [95% CI, 0.81–0.93]; income: HR, 0.75 [95% CI, 0.69–0.78]; and race+income: HR, 0.77 [95% CI, 0.72–0.82]). Conclusions We found a decreased hazard of death and risk for readmission for those residing in the more segregated White communities and higher‐income ZIP codes compared with the more segregated Black communities and lower‐income ZIP codes when using validated measures of racial and economic segregation. Although causal pathways and mechanisms remain unclear, disparities in outcomes after out‐of‐hospital cardiac arrest are associated with the structural components of race and wealth and persist up to 3 years after discharge.

Background: Machine learning clustering offers an unbiased approach to better understand the interactions of complex social and clinical variables via integrative subphenotypes, an approach not studied in out-of-hospital cardiac arrest (OHCA).Objective: We conducted a cluster analysis for a cohort of OHCA survivors to examine the association of clinical and social factors for mortality at 1 year.Methods: We used a retrospective observational OHCA cohort identified from Medicare claims data, including area-level social determinants of health (SDOH) features and hospital-level data sets. We applied k-means clustering algorithms to identify subphenotypes of beneficiaries who had survived an OHCA and examined associations of outcomes by subphenotype.Results: We identified 27,028 unique beneficiaries who survived to discharge after OHCA. We derived 4 distinct subpheno-types. Subphenotype 1 included a distribution of more urban, female, and Black beneficiaries with the least robust area-level SDOH measures and the highest 1-year mortality (2375/4417, 53.8%). Subphenotype 2 was characterized by a greater distribution of male, White beneficiaries and had the strongest zip code-level SDOH measures, with 1-year mortality at 49.9% (4577/9165). Subphenotype 3 had the highest rates of cardiac catheterization at 34.7% (1342/3866) and the greatest distribution with a driving distance to the index OHCA hospital from their primary residence >16.1 km at 85.4% (8179/9580); more were also discharged to a skilled nursing facility after index hospitalization. Subphenotype 4 had moderate median household income at US $51,659.50 (IQR US $41,295 to $67,081) and moderate to high median unemployment at 5.5% (IQR 4.2%-7.1%), with the lowest 1-year mortality (1207/3866, 31.2%). Joint modeling of these features demonstrated an increased hazard of death for subphenotypes 1 to 3 but not for subphenotype 4 when compared to reference.Conclusions: We identified 4 distinct subphenotypes with differences in outcomes by clinical and area-level SDOH features for OHCA. Further work is needed to determine if individual or other SDOH domains are specifically tied to long-term survival after OHCA.

Objective: Guidelines recommend outpatient follow-up after emergency department visits for asthma, but factors related to rates of follow-up among the adult population are understudied. We sought to describe patient and community-level predictors of outpatient follow-up after an index ED visit for asthma and evaluate the association between outpatient follow-up visits and subsequent ED revisits.Methods: We conducted a retrospective observational cohort study of adult patients with emergency departments visits for asthma. The primary predictor was time to outpatient follow-up visit within 30 days of the index ED visit. The primary outcome was all-cause ED revisit within 30 days of the index ED visit. Cox proportional hazards regression was utilized to test the association between time to outpatient follow-up and hazard of ED revisit within 30 days.Results: Time to outpatient follow-up visit within 30 days was not significantly associated with hazard of 30-day ED revisit for asthma (HR 1.05; 95% CI 0.69-1.61). However, male patients (HR 1.45; 95% C 1.11-1.89) and smokers (HR 1.67; 95% CI 1.22-2.29) were significantly more likely to have an ED revisit.Conclusion: Younger, Black patients with Medicaid were less likely to receive follow-up care relative to older patients insured by Medicare. While follow-up visits were not associated with 30-day revisit rates, differences by age, race, and insurance status suggest disproportionate barriers to accessing care. Future research may target these subgroups to improve transitions of care after an ED visit for asthma.

Background: Guidelines recommend an in-haled corticosteroid (ICS) prescription on emergency de-partment (ED) discharge after acute asthma exacerbations.Objective: We sought to identify rates and predictors of ICS prescription at ED discharge. Secondary outcomes included ICS prescription rates in a high-risk subgroup, outpatient follow-up rates within 30 days, and variation in ICS pre-scriptions among attending emergency physicians.Methods: This was a retrospective cohort study of adult asthma ED discharges for acute asthma exacerbation across 5 urban academic hospitals. We used multivariable logistic regres-sion to evaluate predictors of ICS prescription after adjust-ing for patient characteristics and hospital-level clustering.Results: Among 3948 adult ED visits, an ICS was prescribed in 6% (n = 238) of visits. Only 14% (n = 552) completed an outpatient visit within 30 days. Among patients with 2 or more ED visits in 12 months, the ICS prescription rate was 6.7%. ICS administration in the ED (odds ratio [OR] 9.91; 95% CI 7.99-12.28) and prescribing a /3-agonist on dis-charge (OR 2.67; 95% CI 2.08-3.44) were associated with higher odds of ICS prescription. Decreased odds of ICS prescription were associated with Hispanic ethnicity (OR 0.71; 95% CI 0.51-0.99) relative to Black race, and private (OR 0.75; 95% CI 0.62-0.91) or no insurance (OR 0.54; 95% CI 0.35-0.84) relative to Medicaid. One-third (36%, n = 66) of ED attendings prescribed 0 ICS prescriptions during the study period.Conclusions: An ICS is infrequently prescribed on ED asthma discharge, and most patients do not have an outpatient follow-up within 30 days. Future stud-ies should examine the extent to which ED ICS prescriptions improve outcomes for patients with barriers to accessing pri-mary care.(c) 2023 Elsevier Inc. All rights reserved.

Introduction: Racial and economic segregation is associated with poor health outcomes, including out-of-hospital cardiac arrest (OHCA) survival. Prior work identified disparities in OHCA survival for Medicare beneficiaries associated with residential segregation. We sought to use registry data to evaluate this association of racial and economic segregation with OHCA survival to discharge. Methods: We conducted a retrospective analysis of the 2013-2021 Cardiac Arrest Registry toEnhance Survival OHCA dataset. Arrests with missing data, pediatric patients, or EMS-witnessed were excluded. Our primary predictor, the index of concentration at the extremes(ICE), is a validated measure of racial and economic segregation. We examined three ICE measures (race, income, and race + income) calculated from US Census data and binned into tract-level quintiles. The primary outcome was survival to discharge. We performed multi-levelPoisson regression, with hospital random effect, and adjusted for age, gender, race and ethnicity, witness status, location, and bystander CPR to predict relative risk of survival to discharge and good neurological outcome. Results: We identified 624,626 adult OHCA patients with median age 64 years (SD 17), 38%female, 22% Black/African American, 50% White, 6.8% Hispanic/Latino. Overall, 156,712 (25%)survived to hospital admission, 10% survived to hospital discharge and 6.3% had good neurological outcome. Patients residing in the most segregated White and higher income tracts(Q5) had a 7-10% higher likelihood of survival to discharge compared to the most segregatedBlack and lower income communities (Q1) across all three measures (race RR: 1.09, CI 1.05-1.12; income RR: 1.07, CI 1.04-1.10; race + income RR: 1.10, CI 1.06-1.13). Similarly, neurological recovery was higher in Q5 (race RR: 1.20, CI 1.15-1.25; income RR: 1.12, CI 1.08-1.16; race + income RR: 1.19, CI 1.15-1.23) compared to Q1. Conclusion: We identified outcome disparities across three ICE measures for residents of the most segregated and lowest income Black census tracts compared to those in most segregated higher income White census tracts. Future work should focus on interventions for these disparities to improve survival and address health inequities.