Serious Bleeding in Patients With Atrial Fibrillation Using Diltiazem With Apixaban or Rivaroxaban.

Importance: Diltiazem, a commonly prescribed ventricular rate-control medication for patients with atrial fibrillation, inhibits apixaban and rivaroxaban elimination, possibly causing overanticoagulation. Objective: To compare serious bleeding risk for new users of apixaban or rivaroxaban with atrial fibrillation treated with diltiazem or metoprolol. Design, Setting, and Participants: This retrospective cohort study included Medicare beneficiaries aged 65 years or older with atrial fibrillation who initiated apixaban or rivaroxaban use and also began treatment with diltiazem or metoprolol between January 1, 2012, and November 29, 2020. Patients were followed up to 365 days through November 30, 2020. Data were analyzed from August 2023 to February 2024. Exposures: Diltiazem and metoprolol. Main Outcomes and Measures: The primary outcome was a composite of bleeding-related hospitalization and death with recent evidence of bleeding. Secondary outcomes were ischemic stroke or systemic embolism, major ischemic or hemorrhagic events (ischemic stroke, systemic embolism, intracranial or fatal extracranial bleeding, or death with recent evidence of bleeding), and death without recent evidence of bleeding. Hazard ratios (HRs) and rate differences (RDs) were adjusted for covariate differences with overlap weighting. Results: The study included 204 155 US Medicare beneficiaries, of whom 53 275 received diltiazem and 150 880 received metoprolol. Study patients (mean [SD] age, 76.9 [7.0] years; 52.7% female) had 90 927 person-years (PY) of follow-up (median, 120 [IQR, 59-281] days). Patients receiving diltiazem treatment had increased risk for the primary outcome (RD, 10.6 [95% CI, 7.0-14.2] per 1000 PY; HR, 1.21 [95% CI, 1.13-1.29]) and its components of bleeding-related hospitalization (RD, 8.2 [95% CI, 5.1-11.4] per 1000 PY; HR, 1.22 [95% CI, 1.13-1.31]) and death with recent evidence of bleeding (RD, 2.4 [95% CI, 0.6-4.2] per 1000 PY; HR, 1.19 [95% CI, 1.05-1.34]) compared with patients receiving metoprolol. Risk for the primary outcome with initial diltiazem doses exceeding 120 mg/d (RD, 15.1 [95% CI, 10.2-20.1] per 1000 PY; HR, 1.29 [95% CI, 1.19-1.39]) was greater than that for lower doses (RD, 6.7 [95% CI, 2.0-11.4] per 1000 PY; HR, 1.13 [95% CI, 1.04-1.24]). For doses exceeding 120 mg/d, the risk of major ischemic or hemorrhagic events was increased (HR, 1.14 [95% CI, 1.02-1.27]). Neither dose group had significant changes in the risk for ischemic stroke or systemic embolism or death without recent evidence of bleeding. When patients receiving high- and low-dose diltiazem treatment were directly compared, the HR for the primary outcome was 1.14 (95% CI, 1.02-1.26). Conclusions and Relevance: In Medicare patients with atrial fibrillation receiving apixaban or rivaroxaban, diltiazem was associated with greater risk of serious bleeding than metoprolol, particularly for diltiazem doses exceeding 120 mg/d.

Accuracy and Reliability of Chatbot Responses to Physician Questions.

Importance: Natural language processing tools, such as ChatGPT (generative pretrained transformer, hereafter referred to as chatbot), have the potential to radically enhance the accessibility of medical information for health professionals and patients. Assessing the safety and efficacy of these tools in answering physician-generated questions is critical to determining their suitability in clinical settings, facilitating complex decision-making, and optimizing health care efficiency. Objective: To assess the accuracy and comprehensiveness of chatbot-generated responses to physician-developed medical queries, highlighting the reliability and limitations of artificial intelligence-generated medical information. Design, Setting, and Participants: Thirty-three physicians across 17 specialties generated 284 medical questions that they subjectively classified as easy, medium, or hard with either binary (yes or no) or descriptive answers. The physicians then graded the chatbot-generated answers to these questions for accuracy (6-point Likert scale with 1 being completely incorrect and 6 being completely correct) and completeness (3-point Likert scale, with 1 being incomplete and 3 being complete plus additional context). Scores were summarized with descriptive statistics and compared using the Mann-Whitney U test or the Kruskal-Wallis test. The study (including data analysis) was conducted from January to May 2023. Main Outcomes and Measures: Accuracy, completeness, and consistency over time and between 2 different versions (GPT-3.5 and GPT-4) of chatbot-generated medical responses. Results: Across all questions (n = 284) generated by 33 physicians (31 faculty members and 2 recent graduates from residency or fellowship programs) across 17 specialties, the median accuracy score was 5.5 (IQR, 4.0-6.0) (between almost completely and complete correct) with a mean (SD) score of 4.8 (1.6) (between mostly and almost completely correct). The median completeness score was 3.0 (IQR, 2.0-3.0) (complete and comprehensive) with a mean (SD) score of 2.5 (0.7). For questions rated easy, medium, and hard, the median accuracy scores were 6.0 (IQR, 5.0-6.0), 5.5 (IQR, 5.0-6.0), and 5.0 (IQR, 4.0-6.0), respectively (mean [SD] scores were 5.0 [1.5], 4.7 [1.7], and 4.6 [1.6], respectively; P = .05). Accuracy scores for binary and descriptive questions were similar (median score, 6.0 [IQR, 4.0-6.0] vs 5.0 [IQR, 3.4-6.0]; mean [SD] score, 4.9 [1.6] vs 4.7 [1.6]; P = .07). Of 36 questions with scores of 1.0 to 2.0, 34 were requeried or regraded 8 to 17 days later with substantial improvement (median score 2.0 [IQR, 1.0-3.0] vs 4.0 [IQR, 2.0-5.3]; P < .01). A subset of questions, regardless of initial scores (version 3.5), were regenerated and rescored using version 4 with improvement (mean accuracy [SD] score, 5.2 [1.5] vs 5.7 [0.8]; median score, 6.0 [IQR, 5.0-6.0] for original and 6.0 [IQR, 6.0-6.0] for rescored; P = .002). Conclusions and Relevance: In this cross-sectional study, chatbot generated largely accurate information to diverse medical queries as judged by academic physician specialists with improvement over time, although it had important limitations. Further research and model development are needed to correct inaccuracies and for validation.

Risk for Bleeding-Related Hospitalizations During Use of Amiodarone With Apixaban or Rivaroxaban in Patients With Atrial Fibrillation : A Retrospective Cohort Study.

BACKGROUND: Amiodarone, the most effective antiarrhythmic drug in atrial fibrillation, inhibits apixaban and rivaroxaban elimination, thus possibly increasing anticoagulant-related risk for bleeding. OBJECTIVE: For patients receiving apixaban or rivaroxaban, to compare risk for bleeding-related hospitalizations during treatment with amiodarone versus flecainide or sotalol, antiarrhythmic drugs that do not inhibit these anticoagulants' elimination. DESIGN: Retrospective cohort study. SETTING: U.S. Medicare beneficiaries aged 65 years or older. PATIENTS: Patients with atrial fibrillation began anticoagulant use between 1 January 2012 and 30 November 2018 and subsequently initiated treatment with study antiarrhythmic drugs. MEASUREMENTS: Time to event for bleeding-related hospitalizations (primary outcome) and ischemic stroke, systemic embolism, and death with or without recent (past 30 days) evidence of bleeding (secondary outcomes), adjusted with propensity score overlap weighting. RESULTS: There were 91 590 patients (mean age, 76.3 years; 52.5% female) initiating use of study anticoagulants and antiarrhythmic drugs, 54 977 with amiodarone and 36 613 with flecainide or sotalol. Risk for bleeding-related hospitalizations increased with amiodarone use (rate difference [RD], 17.5 events [95% CI, 12.0 to 23.0 events] per 1000 person-years; hazard ratio [HR], 1.44 [CI, 1.27 to 1.63]). Incidence of ischemic stroke or systemic embolism did not increase (RD, -2.1 events [CI, -4.7 to 0.4 events] per 1000 person-years; HR, 0.80 [CI, 0.62 to 1.03]). The risk for death with recent evidence of bleeding (RD, 9.1 events [CI, 5.8 to 12.3 events] per 1000 person-years; HR, 1.66 [CI, 1.35 to 2.03]) was greater than that for other deaths (RD, 5.6 events [CI, 0.5 to 10.6 events] per 1000 person-years; HR, 1.15 [CI, 1.00 to 1.31]) (HR comparison: P = 0.003). The increased incidence of bleeding-related hospitalizations for rivaroxaban (RD, 28.0 events [CI, 18.4 to 37.6 events] per 1000 person-years) was greater than that for apixaban (RD, 9.1 events [CI, 2.8 to 15.3 events] per 1000 person-years) (P = 0.001). LIMITATION: Possible residual confounding. CONCLUSION: In this retrospective cohort study, patients aged 65 years or older with atrial fibrillation treated with amiodarone during apixaban or rivaroxaban use had greater risk for bleeding-related hospitalizations than those treated with flecainide or sotalol. PRIMARY FUNDING SOURCE: National Heart, Lung, and Blood Institute.

Assessing the Accuracy and Reliability of AI-Generated Medical Responses: An Evaluation of the Chat-GPT Model.

Background: Natural language processing models such as ChatGPT can generate text-based content and are poised to become a major information source in medicine and beyond. The accuracy and completeness of ChatGPT for medical queries is not known. Methods: Thirty-three physicians across 17 specialties generated 284 medical questions that they subjectively classified as easy, medium, or hard with either binary (yes/no) or descriptive answers. The physicians then graded ChatGPT-generated answers to these questions for accuracy (6-point Likert scale; range 1 - completely incorrect to 6 - completely correct) and completeness (3-point Likert scale; range 1 - incomplete to 3 - complete plus additional context). Scores were summarized with descriptive statistics and compared using Mann-Whitney U or Kruskal-Wallis testing. Results: Across all questions (n=284), median accuracy score was 5.5 (between almost completely and completely correct) with mean score of 4.8 (between mostly and almost completely correct). Median completeness score was 3 (complete and comprehensive) with mean score of 2.5. For questions rated easy, medium, and hard, median accuracy scores were 6, 5.5, and 5 (mean 5.0, 4.7, and 4.6; p=0.05). Accuracy scores for binary and descriptive questions were similar (median 6 vs. 5; mean 4.9 vs. 4.7; p=0.07). Of 36 questions with scores of 1-2, 34 were re-queried/re-graded 8-17 days later with substantial improvement (median 2 vs. 4; p<0.01). Conclusions: ChatGPT generated largely accurate information to diverse medical queries as judged by academic physician specialists although with important limitations. Further research and model development are needed to correct inaccuracies and for validation.

Association of Rivaroxaban vs Apixaban With Major Ischemic or Hemorrhagic Events in Patients With Atrial Fibrillation.

Importance: The comparative effectiveness of rivaroxaban and apixaban, the most frequently prescribed oral anticoagulants for ischemic stroke prevention in patients with atrial fibrillation, is uncertain. Objective: To compare major ischemic and hemorrhagic outcomes in patients with atrial fibrillation treated with rivaroxaban or apixaban. Design, Setting, and Participants: Retrospective cohort study using computerized enrollment and claims files for US Medicare beneficiaries 65 years or older. Between January 1, 2013, and November 30, 2018, a total of 581 451 patients with atrial fibrillation began rivaroxaban or apixaban treatment and were followed up for 4 years, through November 30, 2018. Exposures: Rivaroxaban (n = 227 572) and apixaban (n = 353 879), either standard or reduced dose. Main Outcomes and Measures: The primary outcome was a composite of major ischemic (stroke/systemic embolism) and hemorrhagic (intracerebral hemorrhage/other intracranial bleeding/fatal extracranial bleeding) events. Secondary outcomes were nonfatal extracranial bleeding and total mortality (fatal ischemic/hemorrhagic event or other death during follow-up). Rates, hazard ratios (HRs), and rate differences (RDs) were adjusted for baseline differences in comorbidity with inverse probability of treatment weighting. Results: Study patients (mean age, 77.0 years; 291 966 [50.2%] women; 134 393 [23.1%] receiving reduced dose) had 474 605 person-years of follow-up (median [IQR] of 174 [62-397] days). The adjusted primary outcome rate for rivaroxaban was 16.1 per 1000 person-years vs 13.4 per 1000 person-years for apixaban (RD, 2.7 [95% CI, 1.9-3.5]; HR, 1.18 [95% CI, 1.12-1.24]). The rivaroxaban group had increased risk for both major ischemic events (8.6 vs 7.6 per 1000 person-years; RD, 1.1 [95% CI, 0.5-1.7]; HR, 1.12 [95% CI, 1.04-1.20]) and hemorrhagic events (7.5 vs 5.9 per 1000 person-years; RD, 1.6 [95% CI, 1.1-2.1]; HR, 1.26 [95% CI, 1.16-1.36]), including fatal extracranial bleeding (1.4 vs 1.0 per 1000 person-years; RD, 0.4 [95% CI, 0.2-0.7]; HR, 1.41 [95% CI, 1.18-1.70]). Patients receiving rivaroxaban had increased risk of nonfatal extracranial bleeding (39.7 vs 18.5 per 1000 person-years; RD, 21.1 [95% CI, 20.0-22.3]; HR, 2.07 [95% CI, 1.99-2.15]), fatal ischemic/hemorrhagic events (4.5 vs 3.3 per 1000 person-years; RD, 1.2 [95% CI, 0.8-1.6]; HR, 1.34 [95% CI, 1.21-1.48]), and total mortality (44.2 vs 41.0 per 1000 person-years; RD, 3.1 [95% CI, 1.8-4.5]; HR, 1.06 [95% CI, 1.02-1.09]). The risk of the primary outcome was increased for rivaroxaban in both those receiving the reduced dose (27.4 vs 21.0 per 1000 person-years; RD, 6.4 [95% CI, 4.1-8.7]; HR, 1.28 [95% CI, 1.16-1.40]) and the standard dose (13.2 vs 11.4 per 1000 person-years; RD, 1.8 [95% CI, 1.0-2.6]; HR, 1.13 [95% CI, 1.06-1.21]) groups. Conclusions and Relevance: Among Medicare beneficiaries 65 years or older with atrial fibrillation, treatment with rivaroxaban compared with apixaban was associated with a significantly increased risk of major ischemic or hemorrhagic events.

Brain Death Determination: An Interprofessional Simulation to Determine Brain Death and Communicate with Families Focused on Neurology Residents.

Introduction: Significant variation exists in determining brain death despite an expectation of competence for all neurology residents. In addition, family discussions regarding brain death are challenging and may influence organ donation. Methods: We developed two simulations of increasing complexity for PGY 2 and PGY 3 neurology residents. High-fidelity mannequins were used to simulate patients; standardized actors portrayed family members. In the first simulation, residents determined brain death and shared this information with a grieving family. In the second simulation, residents determined brain death in a more complicated scenario, requiring ancillary testing and accurate result interpretation. Following the determination, residents met with a challenging family. The residents worked with an interdisciplinary team and responded to the family's emotions, used active listening skills, and supported the family through next steps. Results: Twelve residents completed the simulations. Prior to the simulation, three (25%) residents felt comfortable discussing a brain death diagnosis; following the simulation, eight (67%) residents felt comfortable/very comfortable discussing brain death. Prior to the simulation, eight (67%) residents stated they knew prerequisites for performing a brain death examination and seven (58%) agreed they knew indications for ancillary testing; these numbers increased to 100% following the simulation. The number of residents who felt comfortable performing the brain death exam increased from five (42%) to 10 (83%). Discussion: This simulation of determining brain death and leading difficult family meetings was well-received by neurology residents. Further work should focus on the effects of simulation-based education on practice variation and organ donation consent rates.

Teleneurology service provided via tablet technology: 3-year outcomes and physician satisfaction.

INTRODUCTION: This study aimed to demonstrate that teleneurology consultations conducted via tablet technology are an efficient and cost-effective means of managing acute neurologic emergencies at community-based hospitals and that utilizing such technology yields high community physician satisfaction. METHOD: During a 39-month period, Vanderbilt University Medical Center in Tennessee USA, provided teleneurology services to 10 community-based hospitals that lacked adequate neurology coverage. Hospitalists at one community-based hospital were not comfortable treating any patient with a neurologic symptom, resulting in 100% of those patients being transferred. This facility now retains more than 60% of neurology patients. For less than US$1200, these hospitals were able to meet the only capital expenditure required to launch this service: the purchase of handheld tablet computers. Real-time teleneurology consultations were conducted via tablet using two-way video conferencing, radiologic image sharing, and medical record documentation. Community physicians were regularly surveyed to assess satisfaction. RESULTS: From February 2014 to May 2017, 3626 teleneurology consultations were conducted. Community physicians, in partnership with neurologists, successfully managed 87% of patients at the community-based hospital. Only 13% of patients required transfer to another facility for a higher level of care. The most common diagnoses included stroke (34%), seizure (11%), and headache/migraine (6%). The average time for the neurologist to answer a request for consultation page and connect with the community physician was 10.6 minutes. Ninety-one percent of community physicians were satisfied or somewhat satisfied with the overall service. CONCLUSION: In the assessment of neurology patients, tablets are a more cost-effective alternative to traditional telehealth technologies. The devices promote efficiency in consultations through ease of use and low transfer rates, and survey results indicate community physician satisfaction.

A Retrospective Review of Patients With Acute Stroke With and Without Palliative Care Consultations.

Despite advances in stroke care, patients continue to incur significant disability, are at risk for future events, and are inconsistently comanaged with palliative care (PC) specialty teams. The purpose of this study was to review patients with stroke admitted to our institution, comparing patients with and without PC consultation. We retrospectively reviewed medical record data of all patients with stroke admitted to our neurosciences ICU (NICU) in July 2014 to June 2015 with and without PC consultation. Review focused on stroke type, patient demographics, median days to discharge and death, and posthospitalization discharge. Of 463 patients admitted to the NICU with a stroke diagnosis, 27% (125/463) had (PC) consultation. A higher percentage of the patients with PC consult presented with hemorrhagic stroke than those without PC consult (38% vs 21%, P < .001). Patients with PC consult had longer median days to discharge and death ( P < .001) and a higher percentage of mortality (32% vs 11%). Of the 301 patients without PC consult who discharged (89.1%), 36.5% discharged to inpatient rehab while 10% discharged to a skilled nursing facility. In comparison, of the patients with PC consultation who discharged alive (41.1%), 15.7% discharged to inpatient rehab whereas 39% discharged to skilled nursing ( P < .001). The uncertainty of which patients with stroke benefit most from specialty PC is highlighted in that although sicker patients are referred to PC, a substantial portion (41%) of these patients discharge alive, of which 39.2% discharged to skilled nursing. Future research should focus on which patients with stroke would benefit from specialty PC.

Five-Year Experience of an Inpatient Palliative Care Unit at an Academic Referral Center.

BACKGROUND: Palliative care units (PCUs) staffed by specialty-trained physicians and nurses have been established in a number of medical centers. The purpose of this study is to review the 5-year experience of a PCU at a large, urban academic referral center. METHODS: We retrospectively reviewed a prospectively collected database of all admissions to the PCU at Vanderbilt University Medical Center in the first 5 years of its existence, from 2012 through 2017. RESULTS: Over these 5 years, there were 3321 admissions to the PCU. No single underlying disease process accounted for the majority of the patients, but the largest single category of patients were those with malignancy, who accounted for 38% of admissions. Transfers from the intensive care unit accounted for 50% of admissions, with 43% of admissions from a hospital floor and 7% coming from the emergency department or a clinic. Median length of stay in the PCU was 3 days. In hospital deaths occurred for 50% of admitted patients, while 38% of patients were discharged from the PCU to hospice. CONCLUSION: These data show that a successful PCU is enabled by buy in from a wide variety of referring specialists and by a multidisciplinary palliative care team focused on care of the actively dying patient as well as pain and symptom management, advance care planning, and hospice referral since a large proportion of referred patients do not die in house.

Cerebrovascular fibromuscular dysplasia: The MGH cohort and literature review.

BACKGROUND: Fibromuscular dysplasia (FMD) is a rare noninflammatory, nonatherosclerotic arteriopathy of medium-sized arteries affecting up to 7% of the population. The disease can affect any artery but commonly affects renal, extracranial carotid, and vertebral arteries. The epidemiology and natural course of cerebrovascular FMD is unknown and requires further investigation. METHODS: We present demographic and outcomes data on a case series of 81 patients with cerebrovascular FMD from Massachusetts General Hospital presenting between 2011 and 2015 followed by a review of the peer-reviewed literature. RESULTS: Patients were a median age of 53 years (±12 SD) and the majority were women. Approximately 50% had a history of tobacco use and more than two-thirds had hypertension. Most patients were on monoplatelet therapy with aspirin; during follow-up, 7 of 67 had progressive disease or additional symptoms. One of 67 patients had a cerebrovascular event: TIA. There were 5 of 67 who had noncerebrovascular events or disease progression and 1 death of unclear cause. CONCLUSIONS: Cerebrovascular FMD may present with myriad symptoms. Our data support that patients with FMD with symptomatic disease have a low rate of recurrent symptoms or disease progression and can be managed conservatively with stroke risk modification, antiplatelet agents, surveillance imaging, and counseling.

Atypical presentation of cerebral schistosomiasis four years after exposure to Schistosoma mansoni.

Schistosomiasis is the second most socioeconomically devastating parasitic disease worldwide, affecting over 240 million people in 77 countries on 5 continents and killing 300,000 people annually in sub-Saharan Africa alone. Neuroschistosomiasis is caused by granuloma formation around eggs that lodge in the CNS, with Schistosoma mansoni and Schistosoma haematobium usually affecting the spinal cord and Schistosoma japonicum causing most reported cerebral disease. We report a case of a previously healthy 25-year-old woman native to the United States who presented with a single generalized tonic-clonic seizure without other neurologic symptoms four years after spending a semester in Ghana where she went swimming once in a river. Brain MRI showed areas of signal abnormality and mottled nodular linear enhancement in the left temporal and right posterior temporal/parietal lobes and right cerebellum without mass effect. A biopsy of the left temporal lesion showed prominent granulomas with dense mixed inflammatory infiltrates composed of eosinophils, plasma cells, and lymphocytes surrounding refractile egg shells containing characteristic embryonal cells and von Lichtenberg's envelope and displaying the pathognomonic spine shape of S. mansoni. Serum ELISA and antibody immunoblots confirmed exposure to S. mansoni. In summary, we describe the atypical combination of cerebral schistosomiasis due to S. mansoni, after a prolonged interval of four years, from a single known exposure.

Effects of 24 hours of tobacco withdrawal and subsequent tobacco smoking among low and high sensation seekers.

INTRODUCTION: Previous studies have indicated that high sensation seekers are more sensitive to the reinforcing effects of nicotine, initiate smoking at an earlier age, and smoke greater amounts of cigarettes. This study examined the influence of sensation-seeking status on tobacco smoking following deprivation in regular tobacco users. METHODS: Twenty healthy tobacco-smoking volunteers with low or high impulsive sensation-seeking subscale scores completed 2 consecutive test days per week for 3 consecutive weeks. Each week, a range of self-report, performance, and cardiovascular assessments were completed during ad libitum smoking on Day 1 and before and after the paced smoking of a tobacco cigarette containing 0.05, 0.6, or 0.9 mg of nicotine following 24 hr of tobacco deprivation on Day 2. In addition, self-administration behavior was analyzed during a 2-hr free access period after the initial tobacco administration. RESULTS: In high sensation seekers, tobacco smoking independent of nicotine yield ameliorated deprivation effects, whereas amelioration of deprivation effects was dependent on nicotine yield among low sensation seekers. However, this effect was limited to a small subset of measures. Subsequent cigarette self-administration increased in a nicotine-dependent manner for high sensation seekers only. CONCLUSIONS: Compared with low sensation seekers, high sensation seekers were more sensitive to the withdrawal relieving effects of nonnicotine components of smoking following 24 hr of deprivation on selective measures and more sensitive to nicotine yield during subsequent tobacco self-administration. These results are consistent with studies suggesting that factors driving tobacco dependence may vary as a function of sensation-seeking status.

Delirium in older emergency department patients: recognition, risk factors, and psychomotor subtypes.

OBJECTIVES: Missing delirium in the emergency department (ED) has been described as a medical error, yet this diagnosis is frequently unrecognized by emergency physicians (EPs). Identifying a subset of patients at high risk for delirium may improve delirium screening compliance by EPs. The authors sought to determine how often delirium is missed in the ED and how often these missed cases are detected by admitting hospital physicians at the time of admission, to identify delirium risk factors in older ED patients, and to characterize delirium by psychomotor subtypes in the ED setting. METHODS: This cross-sectional study was a convenience sample of patients conducted at a tertiary care, academic ED. English-speaking patients who were 65 years and older and present in the ED for less than 12 hours at the time of enrollment were included. Patients were excluded if they refused consent, were previously enrolled, had severe dementia, were unarousable to verbal stimuli for all delirium assessments, or had incomplete data. Delirium status was determined by using the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) administered by trained research assistants (RAs). Recognition of delirium by emergency and hospital physicians was determined from the medical record, blinded to CAM-ICU status. Multivariable logistic regression was used to identify independent delirium risk factors. The Richmond Agitation and Sedation Scale was used to classify delirium by its psychomotor subtypes. RESULTS: Inclusion and exclusion criteria were met in 303 patients, and 25 (8.3%) presented to the ED with delirium. The vast majority (92.0%, 95% confidence interval [CI] = 74.0% to 99.0%) of delirious patients had the hypoactive psychomotor subtype. Of the 25 patients with delirium, 19 (76.0%, 95% CI = 54.9% to 90.6%) were not recognized to be delirious by the EP. Of the 16 admitted delirious patients who were undiagnosed by the EPs, 15 (93.8%, 95% CI = 69.8% to 99.8%) remained unrecognized by the hospital physician at the time of admission. Dementia, a Katz Activities of Daily Living (ADL) < or = 4, and hearing impairment were independently associated with presenting with delirium in the ED. Based on the multivariable model, a delirium risk score was constructed. Dementia, Katz ADL < or = 4, and hearing impairment were weighed equally. Patients with higher risk scores were more likely to be CAM-ICU positive (area under the receiver operating characteristic [ROC] curve = 0.82). If older ED patients with one or more delirium risk factors were screened for delirium, 165 (54.5%, 95% CI = 48.7% to 60.2%) would have required a delirium assessment at the expense of missing 1 patient with delirium, while screening 141 patients without delirium. CONCLUSIONS: Delirium was a common occurrence in the ED, and the vast majority of delirium in the ED was of the hypoactive subtype. EPs missed delirium in 76% of the cases. Delirium that was missed in the ED was nearly always missed by hospital physicians at the time of admission. Using a delirium risk score has the potential to improve delirium screening efficiency in the ED setting.