The Impact of Faculty Gender on Resident Evaluations of Faculty Performance in Emergency Medicine.

INTRODUCTION: Gender bias impacts the promotion and tenure of female emergency medicine (EM) physicians and limits their ability to advance in academic rank. Many factors influence the promotion and tenure process including research, evaluations, opportunities for leadership, sponsorship, and mentorship. The goal of this study is to determine if resident evaluations of EM faculty differ by faculty gender. METHODS: A quantitative analysis was used to examine 14,613 teaching evaluations of faculty by residents at a single academic center (The Ohio State University Wexner Medical Center, Columbus) in the years 2017-2019. Anonymized ratings of male and female faculty on a five-point Likert scale were compared using Fischer's exact test and adjusting for multiple comparisons. RESULTS: Male faculty were more likely to hold the rank of Associate Professor or Professor. When taking both faculty gender and rank into account, male Clinical Instructors and Assistant Professors had significantly higher evaluation scores by residents in the domain of resident autonomy than their female counterparts. Regardless of gender or faculty rank, the majority of faculty received scores greater than four. CONCLUSION: A significant gender difference was found in resident evaluation scores of faculty in the domain of resident autonomy at the level of Clinical Instructor and Assistant Professor. Resident autonomy refers to the degree of supervision by faculty which evolves over time and is primarily based on level of training. This is important as it demonstrates a gender difference in scores that could be used to determine faculty compensation and promotion. Evaluation tools used for promotion and tenure of academic faculty should be evaluated for implicit bias and appropriate statistical analysis.

Optimizing Diagnosis and Management of Community-acquired Pneumonia in the Emergency Department.

Pneumonia is split into 3 diagnostic categories: community-acquired pneumonia (CAP), health care-associated pneumonia, and ventilator-associated pneumonia. This classification scheme is driven not only by the location of infection onset but also by the predominant associated causal microorganisms. Pneumonia is diagnosed in over 1.5 million US emergency department visits annually (1.2% of all visits), and most pneumonia diagnosed by emergency physicians is CAP.

Application of an Innovative Data Mining Approach Towards Safe Polypharmacy Practice in Older Adults.

INTRODUCTION: Polypharmacy is common and is associated with higher risk of adverse drug event (ADE) among older adults. Knowledge on the ADE risk level of exposure to different drug combinations is critical for safe polypharmacy practice, while approaches for this type of knowledge discovery are limited. The objective of this study was to apply an innovative data mining approach to discover high-risk and alternative low-risk high-order drug combinations (e.g., three- and four-drug combinations). METHODS: A cohort of older adults (≥ 65 years) who visited an emergency department (ED) were identified from Medicare fee-for-service and MarketScan Medicare supplemental data. We used International Classification of Diseases (ICD) codes to identify ADE cases potentially induced by anticoagulants, antidiabetic drugs, and opioids from ED visit records. We assessed drug exposure data during a 30-day window prior to the ED visit dates. We investigated relationships between exposure of drug combinations and ADEs under the case-control setting. We applied the mixture drug-count response model to identify high-order drug combinations associated with an increased risk of ADE. We conducted therapeutic class-based mining to reveal low-risk alternative drug combinations for high-order drug combinations associated with an increased risk of ADE. RESULTS: We investigated frequent high-order drug combinations from 8.4 million ED visit records (5.1 million from Medicare data and 3.3 million from MarketScan data). We identified 5213 high-order drug combinations associated with an increased risk of ADE by controlling the false discovery rate at 0.01. We identified 1904 high-order, high-risk drug combinations had potential low-risk alternative drug combinations, where each high-order, high-risk drug combination and its corresponding low-risk alternative drug combination(s) have similar therapeutic classes. CONCLUSIONS: We demonstrated the application of a data mining technique to discover high-order drug combinations associated with an increased risk of ADE. We identified high-risk, high-order drug combinations often have low-risk alternative drug combinations in similar therapeutic classes.

Diagnostic Accuracy of the World Health Organization Pediatric Emergency Triage, Assessment and Treatment Tool Plus Among Patients Seeking Care in Nairobi, Kenya: A Cross-sectional Study.

INTRODUCTION: The World Health Organization developed Emergency Triage Assessment and Treatment Plus (ETAT+) guidelines to facilitate pediatric care in resource-limited settings. ETAT+ triages patients as nonurgent, priority, or emergency cases, but there is limited research on the performance of ETAT+ regarding patient-oriented outcomes. This study assessed the diagnostic accuracy of ETAT+ in predicting the need for hospital admission in a pediatric emergency unit at Kenyatta National Hospital in Nairobi, Kenya. METHODS: This was a secondary analysis of a cross-sectional study of pediatric emergency unit patients enrolled over a 4-week period using fixed random sampling. Diagnostic accuracy of ETAT+ was evaluated using receiver operating curves (ROCs) and respective 95% confidence intervals (CIs) with associated sensitivity and specificity (reference category: nonurgent). The ROC analysis was performed for the overall population and stratified by age group. RESULTS: A total of 323 patients were studied. The most common reasons for presentation were upper respiratory tract disease (32.8%), gastrointestinal disease (15.5%), and lower respiratory tract disease (12.4%). Two hundred twelve participants were triaged as nonurgent (65.6%), 60 as priority (18.6%), and 51 as emergency (15.8%). In the overall study population, the area under the ROC curve was 0.97 (95% CI, 0.95-0.99). The ETAT+ sensitivity was 93.8% (95% CI, 87.0%-99.0%), and the specificity was 82.0% (95% CI, 77.0%-87.0%) for admission of priority group patients. The sensitivity and specificity for the emergency patients were 66.0% (95% CI, 55.0%-77.0%) and 98.0% (95% CI, 97.0%-100.0%), respectively. CONCLUSIONS: ETAT+ demonstrated diagnostic accuracy for predicting patient need for hospital admission. This finding supports the utility of ETAT+ to inform emergency care practice. Further research on ETAT+ performance in larger populations and additional patient-oriented outcomes would enhance its generalizability and application in resource-limited settings.

Establishment of baseline urinary antimicrobial peptide levels by age: a prospective observational study.

BACKGROUND: Antimicrobial peptides (AMPs) are key effectors of urinary tract innate immunity. Identifying differences in urinary AMP levels between younger and older adults is important in understanding older adults' susceptibility and response to urinary tract infections (UTI) and AMP use as diagnostic biomarkers. We hypothesized that uninfected older adults have higher urinary human neutrophil peptides 1-3 (HNP 1-3), human alpha-defensin-5 (HD-5), and human beta-defensin-2 (hBD-2), but lower urinary cathelicidin (LL-37) than younger adults. METHODS: We conducted a cross-sectional study of patients age ≥18 years completing a family medicine clinic non-acute visit. Enzyme-linked immunosorbent assays (ELISA) were performed for AMPs. We identified associations between age and AMPs using unadjusted and multivariable linear regression models. RESULTS: Of the 308 subjects, 144 (46.8%) were ≥65 years of age. Comparing age ≥65 versus <65 years, there were no significant differences in HNP 1-3 (p=0.371), HD5 (p=0.834) or LL-37 (p=0.348) levels. Values for hBD-2 were lower in older adults versus younger (p <0.001). In multivariable analyses, older males and females had significantly lower hBD-2 levels (p<0.001 and p=0.004). Models also showed urine leukocyte esterase was associated with increased levels of HNP 1-3 and HD5; hematuria with increased hBD-2; and urine cultures with contamination with increased HNP 1-3 and hBD-2. CONCLUSIONS: Baseline urinary HNP 1-3, HD5, and LL-37 did not vary with age. Older adults had lower baseline hBD-2. This finding has implications for the potential use of urinary AMPs as diagnostic markers and will facilitate further investigation into the role of innate immunity in UTI susceptibility in older adults.

Emergency department patient and physician survey accuracy compared to chart abstraction in patients with acute respiratory illness.

BACKGROUND: High-quality research studies in older adults are needed. Unfortunately, the accuracy of chart review data in older adult patients has been called into question by previous studies. Little is known on this topic in patients with suspected pneumonia, a disease with 500,000 annual older adult U.S. emergency department (ED) visits that presents a diagnostic challenge to ED physicians. The study objective was to compare direct interview and chart abstraction as data sources. METHODS: We present a preplanned secondary analysis of a prospective, observational cohort of ED patients ≥65 years of age with suspected pneumonia in two Midwest EDs. We describe the agreement between chart review and a criterion standard of prospective direct patient survey (symptoms) or direct physician survey (examination findings). Data were collected by chart review and from the patient and treating physician by survey. RESULTS: The larger study enrolled 135 older adults; 134 with complete symptom data and 129 with complete examination data were included in this analysis. Pneumonia symptoms (confusion, malaise, rapid breathing, any cough, new/worse cough, any sputum production, change to sputum) had agreement between patient/legally authorized representative survey and chart review ranging from 47.8% (malaise) to 80.6% (confusion). All examination findings (rales, rhonchi, wheeze) had percent agreement between physician survey and chart review of ≥80%. However, all kappas except wheezing were less than 0.60, indicating weak agreement. CONCLUSIONS: Both patient symptoms and examination findings demonstrated discrepancies between chart review and direct survey with larger discrepancies in symptoms reported. Researchers should consider these potential discrepancies during study design and data interpretation.

Implementation of geriatric screening in the emergency department using the Consolidated Framework for Implementation Research.

OBJECTIVE: Implementation of evidence-based care processes (EBP) into the emergency department (ED) is challenging and there are only a few studies of real-world use of theory-based implementation frameworks. We report final implementation results and sustainability of an EBP geriatric screening program in the ED using the Consolidated Framework for Implementation Research (CFIR). METHODS: The EBP involved nurses screening older patients for delirium (Delirium Triage Screen), fall risk (4-Stage Balance Test), and vulnerability (Identification of Seniors at Risk score) with subsequent appropriate referrals to physicians, therapy specialists, or social workers. The proportions of screened adults ≥65 years old were tracked monthly. Outcomes are reported January 2021-December 2022. Barriers encountered were classified according to CFIR. Implementation strategies were classified according to the CFIR-Expert Recommendations for Implementing Change (ERIC). RESULTS: Implementation strategies increased geriatric screening from 5% to 68%. This did not meet our prespecified goal of 80%. Change was sustained through several COVID-19 waves. Inner setting barriers included culture and implementation climate. Initially, the ED was treated as a single inner setting, but we found different cultures and uptake between ED units, including night versus day shifts. Characteristics of individuals barriers included high levels of staff turnover in both clinical and administrative roles and very low self-efficacy from stress and staff turnover. Initial attempts with individualized audit and feedback were not successful in improving self-efficacy and may have caused moral injury. Adjusting feedback to a team/unit level approach with unitwide stretch goals worked better. Identifying early adopters and conducting on-shift education increased uptake. Lessons learned regarding ED culture, implementation in interconnected health systems, and rapid cycle process improvement are reported. CONCLUSIONS: The pandemic exacerbated barriers to implementation in the ED. Cognizance of a large ED as a sum of smaller units and using the CFIR model resulted in improvements.

Estimating treatment effects for time-to-treatment antibiotic stewardship in sepsis.

Sepsis is a life-threatening condition with a high in-hospital mortality rate. The timing of antibiotic administration poses a critical problem for sepsis management. Existing work studying antibiotic timing either ignores the temporality of the observational data or the heterogeneity of the treatment effects. Here we propose a novel method (called T4) to estimate treatment effects for time-to-treatment antibiotic stewardship in sepsis. T4 estimates individual treatment effects by recurrently encoding temporal and static variables as potential confounders, and then decoding the outcomes under different treatment sequences. We propose mini-batch balancing matching that mimics the randomized controlled trial process to adjust the confounding. The model achieves interpretability through a global-level attention mechanism and a variable-level importance examination. Meanwhile, we equip T4 with an uncertainty quantification to help prevent overconfident recommendations. We demonstrate that T4 can identify effective treatment timing with estimated individual treatment effects for antibiotic stewardship on two real-world datasets. Moreover, comprehensive experiments on a synthetic dataset exhibit the outstanding performance of T4 compared with the state-of-the-art models on estimation of individual treatment effect.

A comprehensive assessment of statin discontinuation among patients who concurrently initiate statins and CYP3A4-inhibitor drugs; a multistate transition model.

AIMS: The aim of this study was to describe the 1-year direct and indirect transition probabilities to premature discontinuation of statin therapy after concurrently initiating statins and CYP3A4-inhibitor drugs. METHODS: A retrospective new-user cohort study design was used to identify (N = 160 828) patients who concurrently initiated CYP3A4 inhibitors (diltiazem, ketoconazole, clarithromycin, others) and CYP3A4-metabolized statins (statin DDI exposed, n = 104 774) vs. other statins (unexposed to statin DDI, n = 56 054) from the MarketScan commercial claims database (2012-2017). The statin DDI exposed and unexposed groups were matched (2:1) through propensity score matching techniques. We applied a multistate transition model to compare the 1-year transition probabilities involving four distinct states (start, adverse drug events [ADEs], discontinuation of CYP3A4-inhibitor drugs, and discontinuation of statin therapy) between those exposed to statin DDIs vs. those unexposed. Statistically significant differences were assessed by comparing the 95% confidence intervals (CIs) of probabilities. RESULTS: After concurrently starting stains and CYP3A, patients exposed to statin DDIs, vs. unexposed, were significantly less likely to discontinue statin therapy (71.4% [95% CI: 71.1, 71.6] vs. 73.3% [95% CI: 72.9, 73.6]) but more likely to experience an ADE (3.4% [95% CI: 3.3, 3.5] vs. 3.2% [95% CI: 3.1, 3.3]) and discontinue with CYP3A4-inhibitor therapy (21.0% [95% CI: 20.8, 21.3] vs. 19.5% [95% CI: 19.2, 19.8]). ADEs did not change these associations because those exposed to statin DDIs, vs. unexposed, were still less likely to discontinue statin therapy but more likely to discontinue CYP3A4-inhibitor therapy after experiencing an ADE. CONCLUSION: We did not observe any meaningful clinical differences in the probability of premature statin discontinuation between statin users exposed to statin DDIs and those unexposed.

Barriers to and recommendations for integrating the age-friendly 4-Ms framework into electronic health records.

BACKGROUND: The Institute for Healthcare Improvement's 4-Ms framework of care for older adults recommends a multidisciplinary assessment of a patient's Medications, Mentation, Mobility, and What Matters Most. Electronic health record (EHR) systems were developed prior to this emphasis on the 4-Ms. We sought to understand how healthcare providers across the healthcare system perceive their EHRs and to identify any current best practices and ideas for improvement regarding integration of the 4-Ms. METHODS: Anonymous survey of healthcare providers who care for older adults. The survey aimed to evaluate efficiency, error tolerance, and satisfaction (usefulness and likeability). The survey was distributed through organizational list serves that focus on the care of older adults and through social media. RESULTS: Sixty-six respondents from all geographic segments of the U.S. (n = 62) and non-U.S. practices (n = 4) responded. Most (82%) were physicians. Respondents used a range of EHRs and 82% had >5 years of experience with their current EHR. Over half of respondents agreed that their EHR had easy to find contact information (56%) and advance directives. Finding a patient's prior cognitive status (26% agreement), goals of care (24%), functional status (14%), and multidisciplinary geriatric assessments (27%) was more difficult. Only 3% were satisfied with how their EHR handles geriatric syndromes. In free text responses, respondents (79%) described three areas that the EHR assists in the care of older adults: screening tied to actions or orders; advance care planning, and medication alerts or review. Common suggestions on how to improve the EHR included incorporating geriatric assessments in notes, establishing a unified place to review the 4-Ms, and creating age-specific best practice alerts. CONCLUSIONS: The majority of healthcare providers were not satisfied with how their EHR handles multidisciplinary geriatric assessment and geriatric care. EHR modifications would aide in reporting, communicating, and tracking the 4-Ms in EHRs.

Building a RAFFT: Impact of a professional development program for women faculty and residents in emergency medicine.

Background: Women comprise 28% of faculty in academic departments of emergency medicine (EM) and 11% of academic chairs. Professional development programs for women are key to career success and to prevent pipeline attrition. Within emergency medicine, there is a paucity of outcomes-level data for such programs. Objectives: We aim to measure the impact of a novel structured professional development curriculum and mentorship group (Resident and Faculty Female Tribe, or RAFFT) within an academic department of EM. Methods: This prospective single-center curriculum implementation and evaluation was conducted in the academic year 2020-2021. A planning group identified potential curricular topics using an iterative Delphi process. We developed a 10-session longitudinal curriculum; a postcurriculum survey was conducted to assess the perceived benefit of the program in four domains. Results: A total of 76% of 51 eligible women attended at least one session; for this project we analyzed the 24 participants (47%) who attended at least one session and completed both the pre- and the postsurvey. The majority of participants reported a positive benefit, which aligned with their expectations in the following areas: professional development (79.2%), job satisfaction (83.3%), professional well-being (70.8%), and personal well-being (79.2%). Resident physicians more often reported less benefit than expected compared to fellow/faculty physicians. Median perceived impact on career choice and trajectory was positive for all respondents. Conclusions: Success of this professional development program was measured through a perceived benefit aligning with participant expectations, a positive impact on career choice and career trajectory for participants in each career stage, and a high level of engagement in this voluntary program. Recommendations for the successful implementation of professional development programs include early engagement of stakeholders, the application of data from a program-specific needs assessment, early dissemination of session dates to allow for protected time off, and structured discussions with appropriate identification of presession resources.

Novel technique for continuous pressure monitoring of esophageal balloon in balloon tamponade device for acute variceal bleed.

Acute variceal bleeding is a life-threatening emergency associated with high mortality. Balloon tamponade is required for refractory bleeding to allow stabilization for definitive therapy. Unfortunately, these devices are associated with iatrogenic complications such as esophageal necrosis and perforation. It is imperative to accurately measure the esophageal balloon pressure to limit these complications. We describe a novel technique for both initial and continuous pressure monitoring of the esophageal balloon.

Emergency medicine resident clinical experience vs. in-training examination content: A national database study.

Objectives: Emergency medicine (EM) residents take the In-Training Examination (ITE) annually to assess medical knowledge. Question content is derived from the Model of Clinical Practice of Emergency Medicine (EM Model), but it is unknown how well clinical encounters reflect the EM Model. The objective of this study was to compare the content of resident patient encounters from 2016-2018 to the content of the EM Model represented by the ITE Blueprint. Methods: This was a retrospective cross-sectional study utilizing the National Hospital Ambulatory Medical Care Survey (NHAMCS). Reason for visit (RFV) codes were matched to the 20 categories of the American Board of Emergency Medicine (ABEM) ITE Blueprint. All analyses were done with weighted methodology. The proportion of visits in each of the 20 content categories and 5 acuity levels were compared to the proportion in the ITE Blueprint using 95% confidence intervals (CIs). Results: Both resident and nonresident patient visits demonstrated content differences from the ITE Blueprint. The most common EM Model category were visits with only RFV codes related to signs, symptoms, and presentations regardless of resident involvement. Musculoskeletal disorders (nontraumatic), psychobehavioral disorders, and traumatic disorders categories were overrepresented in resident encounters. Cardiovascular disorders and systemic infectious diseases were underrepresented. When residents were involved with patient care, visits had a higher proportion of RFV codes in the emergent and urgent acuity categories compared to those without a resident. Conclusions: Resident physicians see higher acuity patients with varied patient presentations, but the distribution of encounters differ in content category than those represented by the ITE Blueprint.

Pediatric Medical Emergencies and Injury Prevention Practices in the Pediatric Emergency Unit of Kenyatta National Hospital, Nairobi, Kenya.

INTRODUCTION: The epidemiology and presence of pediatric medical emergencies and injury prevention practices in Kenya and resource-limited settings are not well understood. This is a barrier to planning and providing quality emergency care within the local health systems. We performed a prospective, cross-sectional study to describe the epidemiology of case encounters to the pediatric emergency unit (PEU) at Kenyatta National Hospital in Nairobi, Kenya; and to explore injury prevention measures used in the population. METHODS: Patients were enrolled prospectively using systematic sampling over four weeks in the Kenyatta National Hospital PEU. Demographic data, PEU visit data and lifestyle practices associated with pediatric injury prevention were collected directly from patients or guardians and through chart review. Data were analyzed with descriptive statistics with stratification based on pediatric age groups. RESULTS: Of the 332 patients included, the majority were female (56%) and 76% were under 5 years of age. The most common presenting complaints were cough (40%) fever (34%), and nausea/vomiting (19%). The most common PEU diagnoses were upper respiratory tract infections (27%), gastroenteritis (11%), and pneumonia (8%). The majority of patients (77%) were discharged from the PEU, while 22% were admitted. Regarding injury prevention practices, the majority (68%) of guardians reported their child never used seatbelts or car seats. Of 68 patients that rode bicycles/motorbikes, one reported helmet use. More than half of caregivers cook at potentially dangerous heights; 59% use ground/low level stoves. CONCLUSIONS: Chief complaints and diagnoses in the PEU population were congruent with communicable disease burdens seen globally. Measures for primary injury prevention were reported as rarely used in the sample studied. The epidemiology described by this study provides a framework for improving public health education and provider training in resource-limited settings.

Inclusion of older adults in emergency department clinical research: Strategies to achieve a critical goal.

Medical research across all fields has historically excluded older adults (aged 65 years and older). Because older adults have a higher burden of chronic illness, respond differently to treatment, and are more prone to medication side effects, the results of current research may not be applicable to this important population. To address this major research deficiency, the National Institutes of Health established the Inclusion Across the Lifespan policy, effective January 2019. We present important considerations and proven strategies for successful inclusion of older adults in emergency care research relating to study design, participant recruitment and retention, and sources of support for investigators.

An implementation science approach to geriatric screening in an emergency department.

BACKGROUND: The Geriatric Emergency Department (ED) Guidelines recommend screening older adults during their ED visit for delirium, fall risk/safe mobility, and home safety needs. We used the Consolidated Framework for Implementation Research (CFIR) and the Expert Recommendations for Implementation Change (ERIC) tool for preimplementation planning. METHODS: The cross-sectional survey was conducted among ED nurses at an academic medical center. The survey was adapted from the CFIR Interview Guide Tool and consisted of 21 Likert scale questions based on four CFIR domains. Potential barriers identified by the survey were mapped to identify recommended implementation strategies using ERIC. RESULTS: Forty-six of 160 potential participants (29%) responded. Intervention Characteristics: Nurses felt geriatric screening should be standard practice for all EDs (76.1% agreed some/very much) and that there was good evidence (67.4% agreed some/very much). Outer setting: The national and regional practices such as the existence of guidelines or similar practices in other hospitals were unknown to many (20.0%). Nurses did agree some/very much (64.4%) that the intervention was good for the hospital/health system. Inner Setting: 67.4% felt more staff or infrastructure and 63.0% felt more equipment were needed for the intervention. When asked to pick from a list of potential barriers, the most commonly chosen were motivational (I often do not remember (n = 27, 58.7%) and It is not a priority (n = 14, 30.4%)). The identified barriers were mapped using the ERIC tool to rate potential implementation strategies. Strategies to target culture change were identifying champions, improve adaptability, facilitate the nurses performing the intervention, and increase demand for the intervention. CONCLUSION: CFIR domains and ERIC tools are applicable to an ED intervention for older adults. This preimplementation process could be replicated in other EDs considering implementing geriatric screening.

Diagnostic Uncertainty in Dyspneic Patients with Cancer in the Emergency Department.

OBJECTIVE: Dyspnea is the second most common symptom experienced by the approximately 4.5 million patients with cancer presenting to emergency departments (ED) each year. Distinguishing pneumonia, the most common reason for presentation, from other causes of dyspnea is challenging. This report characterizes the diagnostic uncertainty in patients with dyspnea and pneumonia presenting to an ED by establishing the rates of co-diagnosis, co-treatment, and misdiagnosis. METHODS: Visits by individuals ≥18 years old with cancer who presented with a complaint of dyspnea were identified using the National Hospital Ambulatory Medical Care Survey between 2012-2014 and analyzed for rates of co-diagnosis, co-treatment (treatment or diagnosis for >1 of pneumonia, chronic obstructive pulmonary disease [COPD], and heart failure), and misdiagnosis of pneumonia. Additionally, we assessed rates of diagnostic uncertainty (co-diagnosis, co-treatment, or a lone diagnosis of dyspnea not otherwise specified [NOS]). RESULTS: Among dyspneic cancer visits (1,593,930), 15.2% (95% confidence interval [CI], 11.1-20.5%) were diagnosed with pneumonia, 22.5% (95% CI, 16.7-29.7%) with COPD, and 7.4% (95% CI 4.7-11.4%) with heart failure. Dyspnea NOS was diagnosed in 32.3% (95% CI, 25.7-39.7%) of visits and as the only diagnosis in 23.1% (95% CI, 16.3-31.6%) of all visits. Co-diagnosis occurred in 4.0% (95% CI, 2.0-7.6%) of dyspneic adults with cancer and co-treatment in 12.1% (95% CI, 7.5-18.9%). Agreement between emergency physician and inpatient documentation for presence of pneumonia was 57.7% (95% CI, 37.0-76.1%). CONCLUSION: Diagnostic uncertainty remains a significant concern in patients with cancer presenting to the ED with dyspnea. Clinical uncertainty among dyspneic patients results in both misdiagnosis and under-treatment of patients with pneumonia and cancer.

An interpretable deep-learning model for early prediction of sepsis in the emergency department.

Sepsis is a life-threatening condition with high mortality rates and expensive treatment costs. Early prediction of sepsis improves survival in septic patients. In this paper, we report our top-performing method in the 2019 DII National Data Science Challenge to predict onset of sepsis 4 h before its diagnosis on electronic health records of over 100,000 unique patients in emergency departments. A long short-term memory (LSTM)-based model with event embedding and time encoding is leveraged to model clinical time series and boost prediction performance. Attention mechanism and global max pooling techniques are utilized to enable interpretation for the deep-learning model. Our model achieved an average area under the curve of 0.892 and was selected as one of the winners of the challenge for both prediction accuracy and clinical interpretability. This study paves the way for future intelligent clinical decision support, helping to deliver early, life-saving care to the bedside of septic patients.