Practice as you Teach: Comparing Ordering Practices Between Shared and Physician-Only Visits in Academically Affiliated Community Emergency Departments.

BACKGROUND: Considerable variability exists in emergency physicians' (EPs) rates of resource utilization, which may cluster in distinct patterns. However, previous studies have focused on academic and tertiary care centers, and it is unclear whether similar patterns exist in community practice. OBJECTIVE: Our aim was to examine whether EPs practicing in community emergency departments (EDs) have practice patterns similar to those of academic EDs. Secondarily, we sought to investigate the effects of shared visits with advanced practice professionals and residents. METHODS: This was a retrospective study of two community EDs affiliated with an academic network. There were 62,860 visits among 50 EPs analyzed from October 1, 2018 through January 31, 2020 for rates of advanced imaging, admission, and shared visits. To classify practice patterns, we used a Gaussian Mixture Model (GMM), with groups and covariance determined by Bayesian Information Criteria. RESULTS: Our GMM revealed three groups. The largest had homogeneous patterns of resource use (n = 28; 50% were female; years of experience: 7; interquartile range [IQR] 2-11; advanced imaging: 28%; admission: 19%; shared: 34%), a small group with lower resource use (n = 4; 0% were female; years of experience: 6; IQR 4-10; advanced imaging: 28%; admission: 16%; shared: 8%), and a modest high-resource group (n = 18; 28% female; years of experience: 5; IQR 2-16; advanced imaging: 34%; admission: 23%; shared: 43%). Rates of shared visits had little direct correlation with imaging (r2 = 0.045) or admission (r2 = 0.093), and rates of imaging and admission were weakly correlated (r2 = 0.242). CONCLUSIONS: Our data suggest that community EPs may have multiple patterns of resource use, similar to those in academic EDs.

Revive and Refuse: Capacity, Autonomy, and Refusal of Care After Opioid Overdose.

Physicians generally recommend that patients resuscitated with naloxone after opioid overdose stay in the emergency department for a period of observation in order to prevent harm from delayed sequelae of opioid toxicity. Patients frequently refuse this period of observation despiteenefit to risk. Healthcare providers are thus confronted with the challenge of how best to protect the patient's interests while also respecting autonomy, including assessing whether the patient is making an autonomous choice to refuse care. Previous studies have shown that physicians have widely divergent approaches to navigating these conflicts. This paper reviews what is known about the effects of opioid use disorder on decision-making, and argues that some subset of these refusals are non-autonomous choices, even when patients appear to have decision making capacity. This conclusion has several implications for how physicians assess and respond to patients refusing medical recommendations after naloxone resuscitation.

Race and Ethnicity and Primary Language in Emergency Department Triage.

Importance: Emergency department (ED) triage substantially affects how long patients wait for care but triage scoring relies on few objective criteria. Prior studies suggest that Black and Hispanic patients receive unequal triage scores, paralleled by disparities in the depth of physician evaluations. Objectives: To examine whether racial disparities in triage scores and physician evaluations are present across a multicenter network of academic and community hospitals and evaluate whether patients who do not speak English face similar disparities. Design, Setting, and Participants: This was a cross-sectional, multicenter study examining adults presenting between February 28, 2019, and January 1, 2023, across the Mass General Brigham Integrated Health Care System, encompassing 7 EDs: 2 urban academic hospitals and 5 community hospitals. Analysis included all patients presenting with 1 of 5 common chief symptoms. Exposures: Emergency department nurse-led triage and physician evaluation. Main Outcomes and Measures: Average Triage Emergency Severity Index [ESI] score and average visit work relative value units [wRVUs] were compared across symptoms and between individual minority racial and ethnic groups and White patients. Results: There were 249 829 visits (149 861 female [60%], American Indian or Alaska Native 0.2%, Asian 3.3%, Black 11.8%, Hispanic 18.8%, Native Hawaiian or Other Pacific Islander <0.1%, White 60.8%, and patients identifying as Other race or ethnicity 5.1%). Median age was 48 (IQR, 29-66) years. White patients had more acute ESI scores than Hispanic or Other patients across all symptoms (eg, chest pain: Hispanic, 2.68 [95% CI, 2.67-2.69]; White, 2.55 [95% CI, 2.55-2.56]; Other, 2.66 [95% CI, 2.64-2.68]; P < .001) and Black patients across most symptoms (nausea/vomiting: Black, 2.97 [95% CI, 2.96-2.99]; White: 2.90 [95% CI, 2.89-2.91]; P < .001). These differences were reversed for wRVUs (chest pain: Black, 4.32 [95% CI, 4.25-4.39]; Hispanic, 4.13 [95% CI, 4.08-4.18]; White 3.55 [95% CI, 3.52-3.58]; Other 3.96 [95% CI, 3.84-4.08]; P < .001). Similar patterns were seen for patients whose primary language was not English. Conclusions and Relevance: In this cross-sectional study, patients who identified as Black, Hispanic, and Other race and ethnicity were assigned less acute ESI scores than their White peers despite having received more involved physician workups, suggesting some degree of mistriage. Clinical decision support systems might reduce these disparities but would require careful calibration to avoid replicating bias.

Frequency, compliance, and yield of cardiac testing after high-sensitivity troponin accelerated diagnostic protocol implementation.

BACKGROUND: Among persons presenting to the emergency department with suspected acute myocardial infarction (MI), cardiac troponin (cTn) testing is commonly used to detect acute myocardial injury. Accelerated diagnostic protocols (ADPs) guide clinicians to integrate cTn results with other clinical information to decide whether to order further diagnostic testing. OBJECTIVE: To determine the change in the rate and yield of stress test or coronary CT angiogram following cTn measurement in patients with chest pain presenting to the emergency department pre- and post-transition to a high-sensitivity (hs-cTn) assay in an updated ADP. METHODS: Using electronic health records, we examined visits for chest pain at five emergency departments affiliated with an integrated academic health system 1-year pre- and post-hs-cTn assay transition. Outcomes included stress test or coronary imaging frequency, ADP compliance among those with additional testing, and diagnostic yield (ratio of positive tests to total tests). RESULTS: There were 7564 patient-visits for chest pain, including 3665 in the pre- and 3899 in the post-period. Following the updated ADP using hs-cTn, 862 (23.5 per 100 patient visits) visits led to subsequent testing versus 1085 (27.8 per 100 patient visits) in the pre-hs-cTn period, (P < 0.001). Among those who were tested, the protocol-compliant rate fell from 80.9% to 46.5% (P < 0.001), but the yield of those tests rose from 24.5% to 29.2% (P = 0.07). Among tests that were noncompliant with ADP guidance, yield was similar pre- and post-updated hs-cTn ADP implementation (pre 13.0%, post 15.4% (P = 0.43). CONCLUSION: Implementation of hs-cTn supported by an updated ADP was associated with a lower rate of stress testing and coronary CT angiogram.

Outcomes for Patients Discharged to Involuntary Commitment for Substance Use Disorder Directly from the Hospital.

To evaluate the outcomes of patients discharged to involuntary commitment for substance use disorders directly from the hospital. We performed a retrospective chart review of 22 patients discharged to involuntary commitment for substance use disorder from the hospital between October 2016 and February 2020. We collected demographic data, details about each commitment episode, and healthcare utilization outcomes 1 year following involuntary commitment. Nearly all patients had a primary alcohol use disorder (91%) and had additional medical (82%) and psychiatric comorbidities (71%). One year following involuntary commitment, all patients had relapsed to substance use and had at least one emergency department visit while 78.6% had at least one admission. These findings suggest that patients discharged to involuntary commitment directly from the hospital universally relapsed and experienced significant medical morbidity during the first year following their release. This study adds to a growing literature recognizing the harms of involuntary commitment for substance use disorder.

Machine Learning Methods for Predicting Patient-Level Emergency Department Workload.

BACKGROUND: Work Relative Value Units (wRVUs) are a component of many compensation models, and a proxy for the effort required to care for a patient. Accurate prediction of wRVUs generated per patient at triage could facilitate real-time load balancing between physicians and provide many practical operational and clinical benefits. OBJECTIVE: We examined whether deep-learning approaches could predict the wRVUs generated by a patient's visit using data commonly available at triage. METHODS: Adult patients presenting to an urban, academic emergency department from July 1, 2016-March 1, 2020 were included. Deidentified triage information included structured data (age, sex, vital signs, Emergency Severity Index score, language, race, standardized chief complaint) and unstructured data (free-text chief complaint) with wRVUs as outcome. Five models were examined: average wRVUs per chief complaint, linear regression, neural network and gradient-boosted tree on structured data, and neural network on unstructured textual data. Models were evaluated using mean absolute error. RESULTS: We analyzed 204,064 visits between July 1, 2016 and March 1, 2020. The median wRVUs were 3.80 (interquartile range 2.56-4.21), with significant effects of age, gender, and race. Models demonstrated lower error as complexity increased. Predictions using averages from chief complaints alone demonstrated a mean error of 2.17 predicted wRVUs per visit (95% confidence interval [CI] 2.07-2.27), the linear regression model: 1.00 wRVUs (95% CI 0.97-1.04), gradient-boosted tree: 0.85 wRVUs (95% CI 0.84-0.86), neural network with structured data: 0.86 wRVUs (95% CI 0.85-0.87), and neural network with unstructured data: 0.78 wRVUs (95% CI 0.76-0.80). CONCLUSIONS: Chief complaints are a poor predictor of the effort needed to evaluate a patient; however, deep-learning techniques show promise. These algorithms have the potential to provide many practical applications, including balancing workloads and compensation between emergency physicians, quantify crowding and mobilizing resources, and reducing bias in the triage process.

The effect of batched patient-physician assignment on patient length of stay in the emergency department.

Objectives: Queuing theory suggests that signing up for multiple patients at once (batching) can negatively affect patients' length of stay (LOS). At academic centers, resident assignment adds a second layer to this effect. In this study, we measured the rate of batched patient assignment by resident physicians, examined the effect on patient in-room LOS, and surveyed residents on underlying drivers and perceptions of batching. Methods: This was a retrospective study of discharged patients from August 1, 2020 to October 27, 2020, supplemented with survey data conducted at a large, urban, academic hospital with an emergency medicine training program in which residents self-assign to patients. Time stamps were extracted from the electronic health record and a definition of batching was set based on findings of a published time and motion study. Results: A total of 3794 patients were seen by 28 residents and ultimately discharged during the study period. Overall, residents batched 23.7% of patients, with a greater rate of batching associated with increasing resident seniority and during the first hour of resident shifts. In-room LOS for batched assignment patients was 15.9 minutes longer than single assignment patients (P value < 0.01). Residents' predictions of their rates of batching closely approximated actual rates; however, they underestimated the effect of batching on LOS. Conclusions: Emergency residents often batch patients during signup with negative consequences to LOS. Moreover, residents significantly underestimate this negative effect.

Time on shift in the emergency department and decision to prescribe opioids to patients without chronic opioid use.

OBJECTIVE: To study the effect of time on shift on the opioid prescribing practices of emergency physicians among patients without chronic opioid use. METHODS: We analyzed pain-related visits for five painful conditions from 2010 to 2017 at a single academic hospital in Boston. Visits were categorized according to national guidelines as conditions for which opioids are "sometimes indicated" (fracture and renal colic) or "usually not indicated" (headache, low back pain, and fibromyalgia). Using conditional logistic regression with fixed effects for clinicians, we estimated the probability of opioid prescribing for pain-related visits as a function of shift hour at discharge, time of day, and patient-level confounders (age, sex, and pain score). RESULTS: Among 16,115 visits for which opioids were sometimes indicated, opioid prescribing increased over the course of a shift (28% in the first hour compared with 40% in the last hour; adjusted odds ratio, 1.06; 95% confidence interval, 1.02-1.10; adjusted P-trend <0.01). However, among visits for which opioids are usually not indicated, relative to the first hour, opioid prescriptions progressively fell (40% in the first hour compared with 23% in the last hour; adjusted odds ratio, 0.93; 95% confidence interval, 0.91-0.96; adjusted P-trend <0.01). CONCLUSION: As shift hour progressed, emergency physicians became more likely to prescribe opioids for conditions that are sometimes indicated, and less likely to prescribe opioids for nonindicated conditions. Our study suggests that clinical decision making in the emergency department can be substantially influenced by external factors such as clinician shift hour.

The Ethics of Procedural Education Under Pandemic Conditions.

BACKGROUND: The COVID-19 pandemic significantly disrupted emergency medicine residents' education. Early in the pandemic, many facilities lacked adequate personal protective equipment (PPE), and intubation was considered particularly high risk for transmission to physicians, leading hospitals to limit the number of individuals present during the procedure. This posed difficulties for residents and academic faculty, as opportunities to perform endotracheal intubation during residency are limited, but patients with COVID-19 requiring intubation are unstable and have difficult airways. Case Scenario: When PPE is being rationed, who should be the one to perform an intubation on a patient with respiratory failure from severe COVID-19? DISCUSSION: We examined this case scenario using the ethical frameworks of bioethical principles and virtue ethics. Bioethical principles include justice, beneficence, nonmalfeasance, and autonomy, and virtue ethics emphasizes the provision of moral exemplars and opportunities to exercise practical wisdom. Arguments for an attending-only strategy include the role of the attending as a truly autonomous decision maker and the importance of providing residents with a moral exemplar. A resident-only strategy benefits a resident's future patients and provides opportunities for residents to exercise character. Strategies preserving the dyad of attending and resident maintain these advantages and mitigate some drawbacks, while intubation teams may provide the most parsimonious use of PPE, but may elide resident involvement. CONCLUSIONS: There exist compelling motivations for involving senior residents and attendings in high-risk intubations during the COVID-19 pandemic. A just strategy will preserve residents' role whenever possible, while maximizing supervision and providing alternative routes for intubation practice.

Not All Testers are Admitters: An Analysis of Emergency Physician Resource Utilization and Consultation Rates.

BACKGROUND: Variability exists in emergency physician (EP) resource utilization as measured by ordering practices, rate of consultation, and propensity to admit patients. OBJECTIVE: To validate and expand upon previous data showing that resource utilization as measured by EP ordering patterns is positively correlated with admission rates. METHODS: This is a retrospective study of routinely gathered operational data from the ED of an urban academic tertiary care hospital. We collected individual EP data on advanced imaging, consultation, and admission rates per patient encounter. To investigate whether there might be distinct groups of practice patterns relating these 3 resources, we used a Gaussian mixture model, a classification method used to determine the likelihood of distinct subgroups within a larger population. RESULTS: Our Gaussian mixture model revealed 3 distinct groups of EPs based on their ordering practices. The largest group is characterized by a homogenous pattern of neither high or low resource utilization (n = 37, 27% female, median years' experience: 6 [interquartile ratio {IQR} 3-18]; rates of advanced imaging, 38.9%; consultation, 45.1%; and admission 39.3%), with a modest group of low-resource users (n = 15, 60% female, median years' experience: 6 [IQR 5-14]; rates of advanced imaging, 37%; consultation, 42.6%; and admission 37.3%), and far fewer members of a high-resource use group (n = 6, 0% female, median years' experience: 6 [IQR 4-16]; rates of advanced imaging, 42.2%; consultation, 45.8%; and admission 40.6%). This variation suggests that not "all testers are admitters," but that there exist wider practice variations among EPs. CONCLUSIONS: At our academic tertiary center, 3 distinct subgroups of EP ordering practices exist based on consultation rates, advanced imaging use, and propensity to admit a patient. These data validate previous work showing that resource utilization and admission rates are related, while demonstrating that more nuanced patterns of EP ordering practices exist. Further investigation is needed to understand the impact of EP characteristics and behavior on throughput and quality of care. © 2022 Elsevier Inc.

Analysis of time-to-disposition intervals during early and late parts of an emergency department shift.

INTRODUCTION: Time-to-disposition is an important metric for emergency department throughput. We hypothesized that providers view the shift end as a key timepoint and attempt to leave as few dispositions as possible to the oncoming team, thereby making quicker decisions later in the shift. This study evaluates disposition distribution relative to when patients are assigned a provider during the course of a shift. METHODS: 50,802 cases were analyzed over the one-year study interval. 31,869 patients were seen in the early half of a shift (hours 1-4) and 18,933 were seen in the later half (hours 5+). We ran a linear mixed model that adjusted for age, gender, emergency severity index score, time of day, weekend arrivals, quarter of arrival and shift type. RESULTS: Median time-to-disposition for the early group was 3.25 h (IQR 1.90-5.04), and 2.62 h (IQR 1.51-4.31) for the late group. From our mixed model, we conclude that in the later parts of the shift, providers take on average 15.1% less time to make a disposition decision than in the earlier parts of the shift. CONCLUSION: Patients seen during the latter half of a shift were more likely to have a shorter time-to-disposition than similar patients seen in the first half of a shift. This may be influenced by many factors, such as providers spending the early hours of a shift seeing new patients which generate new tasks and delay dispositions, and viewing the end of shift as a landmark with a goal to maximize dispositions prior to sign-out.

A retrospective multistate analysis: Do regional football games impact emergency department patient volume?

OBJECTIVE: We sought to assess the effect of National Football League (NFL) games played by a regional sports team, the New England Patriots, on emergency department (ED) patient volume. METHODS: We conducted a multicenter, retrospective chart review at the following 3 tertiary centers in New England from 2012 to 2019: Beth Israel Deaconess Medical Center, Boston, MA; Dartmouth Hitchcock Medical Center, Lebanon, NH; and Maine Medical Center, Portland, ME. RESULTS: Within the NFL season, we observed a 2.6% overall decrease (-10.4 patients) in average total daily volume across the study sites on Sundays when Patriots games were played compared with Sundays when games were not played (P = 0.07; 95% confidence interval [CI], -22.37 to 1.62). We observed a 4.3% reduction (-19.0 patients) in average total daily volume across the study sites on Mondays during which Patriots games were played compared with Mondays without games (P = 0.15; 95% CI, -43.51 to 5.47). Subanalyses on the 5-hour period corresponding with each Patriots game showed reductions in mean patient volume per hour. Although our primary and subanalyses showed reductions in patient volume during Patriots games, these results were not statistically significant. CONCLUSIONS: Our data support prior studies that showed a minimal impact of major sporting events on ED patient volume at tertiary centers. These results add to the limited data on this topic and can inform administrators whether staffing adjustments are necessary during similar types of sporting events.

Understanding Demand and Capacity Mismatch in an Academic Emergency Department Using a Staircase Model of Provider Capacity and Staggered Shift Start Times.

BACKGROUND: Staffing and provider capacity are essential components of emergency department (ED) throughput. Patient flow is dependent on matching patient arrivals with provider capacity. Current models assume a static rate of patients per hour for providers; however, this metric has been shown to decrease throughout a shift in a pattern we describe as a staircase. OBJECTIVE: We sought to analyze the demand capacity mismatch based on both a static and staircase model of resident productivity. We then suggest a new staggered staffing model that would improve flow in the ED. METHODS: This was a retrospective analysis of patient demand and productivity, analyzing both static and staircase models of provider capacity. An alternative staggered shift model was then suggested, and a 2-sample t test was performed to assess if a new model reduces the amount of demand/capacity mismatch. RESULTS: Seventeen thousand five hundred twenty data points were analyzed over the 2-year interval, comparing the difference between actual patients placed into a treatment space at each hour and projected resident capacity based on the staircase model, using both the existing schedule and a new staggered schedule. Mean absolute values for the disparity in coverage was 2.69 (95% confidence interval 2.65-2.72) for the staircase scheduling model, and 2.14 (95% confidence interval 2.12-2.17) when staggering provider start times. The mean difference between these data sets was 0.54 (95% confidence interval 0.52-0.57; p < 0.0001). CONCLUSIONS: Academic EDs may find value in using a staircase model to analyze provider capacity because it is more reflective of actual capacity. EDs may benefit from visualizing their capacity curves to identify mismatches and staggering resident shifts to improve throughput and flow.

Association of resident characteristics with patterns of patient self-assignment.

OBJECTIVE: We hypothesized that resident characteristics impact patterns of patient self-assignment in the emergency department (ED). Our goal was to determine if male residents would be less likely than their female colleagues to see patients with sensitive (e.g. breast-related or gynecologic) chief complaints (CCs). We also investigated whether resident specialty was associated with preferentially choosing patients with more familiar chief complaints. METHODS: We performed a retrospective cross-sectional study at a tertiary academic medical center using data from all adult patients presenting to the ED between 2010 and 2019 with one of six CC categories (vaginal bleeding, breast-related concerns, male genitourinary [GU] concerns, gastrointestinal bleeding, epistaxis, and laceration). These CCs were chosen as they each require either an invasive medical exam or procedure, and cannot easily be evaluated with an exam in a hallway bed. We used logistic regression to assess the likelihood of being treated by a male resident compared to a female resident for each CC, adjusting for candidate variables of patient age, race, primary language, ESI score, bed location, time of day, day of week, calendar month, and resident specialty. We also similarly analyzed patterns of patient self-assignment according to resident specialty. RESULTS: Male residents were significantly less likely than female residents to treat patients with breast-related CCs (adjusted OR 0.67, 95% CI 0.54-0.83, p < 0.001) or vaginal bleeding (adjusted OR 0.73, 95% CI 0.63-0.84, p < 0.001, reference group: epistaxis). Off-service residents were more likely to assign themselves to familiar chief complaints, for example surgery residents were more likely to see patients with lacerations (adjusted OR 2.11, 95% CI 1.71-2.61, p < 0.001) and OB/GYN residents were less likely to see patients with male GU concerns (adjusted OR 0.21, 95% CI 0.05-0.85, p = 0.029), compared to emergency medicine residents. CONCLUSION: In a single facility, resident characteristics were associated with preferential patient self-assignment. Further work is necessary to determine the underlying reasons for patient avoidance, and to create work environments in which preferentially choosing patients is discouraged.

Estimating maximum patient throughput at an ED scanner: Is three patients per hour the magic number?

.

Decreased hospital admissions through emergency departments during the COVID-19 pandemic.

STUDY OBJECTIVE: Emergency Department (ED) visits decreased significantly in the United States during the COVID-19 pandemic. A troubling proportion of this decrease was among patients who typically would have been admitted to the hospital, suggesting substantial deferment of care. We sought to describe and characterize the impact of COVID-19 on hospital admissions through EDs, with a specific focus on diagnosis group, age, gender, and insurance coverage. METHODS: We conducted a retrospective, observational study of aggregated third-party, anonymized ED patient data. This data included 501,369 patient visits from twelve EDs in Massachusetts from 1/1/2019-9/9/2019, and 1/1/2020-9/8/2020. We analyzed the total arrivals and hospital admissions and calculated confidence intervals for the change in admissions for each characteristic. We then developed a Poisson regression model to estimate the relative contribution of each characteristic to the decrease in admissions after the statewide lockdown, corresponding to weeks 11 through 36 (3/11/2020-9/8/2020). RESULTS: We observed a 32% decrease in admissions during weeks 11 to 36 in 2020, with significant decreases in admissions for chronic respiratory conditions and non-orthopedic needs. Decreases were particularly acute among women and children, as well as patients with Medicare or without insurance. The most common diagnosis during this time was SARS-CoV-2. CONCLUSION: Our findings demonstrate decreased hospital admissions through EDs during the pandemic and suggest that several patient populations may have deferred necessary care. Further research is needed to determine the clinical and operational consequences of this delay.

Emergency physicians' active patient queues over the course of a shift.

OBJECTIVES: When emergency physicians see new patients in an ad libitum system, they see fewer patients as the shift progresses. However, it is unclear if this reflects a decreasing workload, as patient assessments often span many hours. We sought to investigate whether the size of a physician's queue of active patients similarly declines over a shift. METHODS: Retrospective cohort study, conducted over two years in three community hospitals in the Northeastern United States, with 8 and 9-h shifts. Timestamps of all encounters were recorded electronically. Generalized estimating equations were constructed to predict the number of active patients a physician concurrently managed per hour. RESULTS: We evaluated 64 physicians over a two-year period, with 9822 physician-shifts. Across all sites, physicians managed an increasing queue of active patients in the first several hours. This queue plateaued in the middle of the shift, declining in the final hours, independently of other factors. Physicians' queues of active patients increased slightly with greater volume and acuity, but did not affect the overall pattern of work. Similarly, working alone or with colleagues had little effect on the number of active patients managed. CONCLUSIONS: Emergency physicians in an ad libitum system tend to see new patients until reaching a stable roster of active patients. This pattern may help explain why physicians see fewer new patients over the course of a shift, should be factored into models of throughput, and suggests new avenues for evaluating relationships between physician workload, patient safety, physicians' well-being, and the quality of care.

Reducing Emergency Department Transfers from Skilled Nursing Facilities Through an Emergency Physician Telemedicine Service.

INTRODUCTION: Transfers of skilled nursing facility (SNF) residents to emergency departments (ED) are linked to morbidity, mortality and significant cost, especially when transfers result in hospital admissions. This study investigated an alternative approach for emergency care delivery comprised of SNF-based telemedicine services provided by emergency physicians (EP). We compared this on-site emergency care option to traditional ED-based care, evaluating hospital admission rates following care by an EP. METHODS: We conducted a retrospective, observational study of SNF residents who underwent emergency evaluation between January 1, 2017-January 1, 2018. The intervention group was comprised of residents at six urban SNFs in the Northeastern United States, who received an on-demand telemedicine service provided by an EP. The comparison group consisted of residents of SNFs that did not offer on-demand services and were transferred via ambulance to the ED. Using electronic health record data from both the telemedicine and ambulance transfers, our primary outcome was the odds ratio (OR) of a hospital admission. We also conducted a subanalysis examining the same OR for the three most common chronic disease-related presentations found among the telemedicine study population. RESULTS: A total of 4,606 patients were evaluated in both the SNF-based intervention and ED-based comparison groups (n=2,311 for SNF based group and 2,295 controls). Patients who received the SNF-based acute care were less likely to be admitted to the hospital compared to patients who were transferred to the ED in our primary and subgroup analyses. Overall, only 27% of the intervention group was transported to the ED for additional care and presumed admission, whereas 71% of the comparison group was admitted (OR for admission = 0.15 [9% confidence interval, 0.13-0.17]). CONCLUSION: The use of an EP-staffed telemedicine service provided to SNF residents was associated with a significantly lower rate of hospital admissions compared to the usual ED-based care for a similarly aged population of SNF residents. Providing SNF-based care by EPs could decrease costs associated with hospital-based care and risks associated with hospitalization, including cognitive and functional decline, nosocomial infections, and falls.

Deep-learning approaches to identify critically Ill patients at emergency department triage using limited information.

STUDY OBJECTIVE: Triage quickly identifies critically ill patients, facilitating timely interventions. Many emergency departments (EDs) use emergency severity index (ESI) or abnormal vital sign triggers to guide triage. However, both use fixed thresholds, and false activations are costly. Prior approaches using machinelearning have relied on information that is often unavailable during the triage process. We examined whether deep-learning approaches could identify critically ill patients only using data immediately available at triage. METHODS: We conducted a retrospective, cross-sectional study at an urban tertiary care center, from January 1, 2012-January 1, 2020. De-identified triage information included structured (age, sex, initial vital signs) and textual (chief complaint) data, with critical illness (mortality or ICU admission within 24 hours) as the outcome. Four progressively complex deep-learning models were trained and applied to triage information from all patients. We compared the accuracy of the models against ESI as the standard diagnostic test, using area under the receiver-operator curve (AUC). RESULTS: A total of 445,925 patients were included, with 60,901 (13.7%) critically ill. Vital sign thresholds identified critically ill patients with AUC 0.521 (95% confidence interval [CI] = 0.519-0.522), and ESI <3 demonstrated AUC 0.672 (95% CI = 0.671-0.674), logistic regression classified patients with AUC 0.803 (95% CI = 0.802-0.804), 2-layer neural network with structured data with AUC 0.811 (95% CI = 0.807-0.815), gradient tree boosting with AUC 0.820 (95% CI = 0.818-0.821), and the neural network model with textual data with AUC 0.851 (95% CI = 0.849-0.852). All successive increases in AUC were statistically significant. CONCLUSION: Deep-learning techniques represent a promising method of augmenting triage, even with limited information. Further research is needed to determine if improved predictions yield clinical and operational benefits.