Background: Although many emergency departments (EDs) employ some form of vertical patient flow, there is limited published literature describing variations of vertical pathways. Objectives: We sought to describe our ED's emergency physician (EP)-driven vertical model and to characterize patients selected by EPs to be seen in our vertical space. Methods: We retrospectively reviewed all vertical ED encounters in the study period, separately analyzing those who eventually received an ED bed (“ED Bed”) and those who were discharged without being roomed (“Vertical only”). We report patient demographics, ESI, vital signs, oxygen use, chief complaints, resource utilization, ED LOS, disposition, and 72 ​h return rates. Results: Physicians elected to perform initial evaluations of a variety of patients in the vertical space. The two most common classes of complaints evaluated in the vertical space were extremity issues (21.4%) and skin complaints (13.5%). Patients presenting with abdominal pain and chest pain initially assessed in vertical were significantly more likely to later receive an ED bed (Standardized Difference of 38% and 21.4% respectively), and patients with skin complaints were more frequently discharged from the vertical space and were less likely to receive an ED bed (Standardized Difference of 32.2%). Most (56.2%) Vertical only patients were ESI 3, although EPs also discharged ESI 2, 4, and 5 patients from Vertical. Conclusions: EP-driven patient selection for a vertical pathway allowed EPs to discharge some patients without bed placement while simultaneously functioning as their own triage physicians for higher-acuity patients who would go on to receive an ED bed.

Introduction:The clinical diversity of patients presenting to the emergency department (ED) allows emergency medicine (EM) and non-EM residents to sharpen their clinical skills. In most EDs, residents self-assign patients at their discretion. Our institution transitioned from a self-assignment-system to an automated-system, after which we sought to determine the productivity of our non-EM residents compared to the previous system.Methods:In this retrospective cross-sectional study, resident productivity was measured as number of patient visits per hour and per 8.5-hour shift before and after the implementation of an automated patient assignment system in emergency department. The automated-system assigns one patient at the start of the shift, another 30 minutes later, and one patient every hour thereafter, throughout the shift.Results:28 residents performed 406 total shifts prior to implementation and 14 residents performed 252 total shifts post-implementation. The average number of patient visits per hour significantly increased from 0.52 ± 0.18 (95% CI 0.45-0.59, IQR 0.43-0.60) to 0.82 ± 0.11 (95% CI 0.75-0.88, IQR 0.74-0.89) after implementation of our assignment system (p<0.00001; figure 1). Additionally, the average number of patient visits per 8.5-hour shift significantly increased from 4.46 ± 1.53 (CI 3.86-5.05, IQR 3.66-5.08) to 6.52 ± 0.86 (CI 6.02-7.02, IQR 5.90-7.09) after the implementation of our system (p<0.00001; figure 1).Conclusion:These findings warrant further evaluation of the impact of patient assignment systems on trainee education.

As a major public health crisis, the opioid epidemic caused over 556,000 deaths in the U.S. between 2000 and 2020. To control the epidemic, the Centers for Disease Control and Prevention (CDC) has developed some general guidelines, encouraging physicians to use opioid medications only when their benefits outweigh their risks. The CDC’s 2016 guidelines mainly left it to physicians to decide when the benefits outweigh the risks. A few years later (in 2022), the CDC made some modifications to make its recommendations a bit less reliant on each individual physician’s perception of benefits versus risks. In complex and high stake decision-making environments such as those pertaining the use of opioid medications, it is not clear whether and how human-based perceptions might differ from algorithmic-based ones. In this study, we first develop some longitudinal machine learning algorithms (e.g., historical random forest, recurrent neural networks, and long short-term memory networks) and train them on clinical evidence of more than 3 million patients. We then feed the best machine learning algorithm to a mathematical model that enables determining cost-effective treatments for each patient in a personalized manner. Through extensive numerical experiments, we compare the treatment options and recommendations from our algorithmic-based approach with human-based ones that are currently followed in the medical practice. Compared to the human-based approach, our results show that the average saving in quality-adjusted life years and costs obtained by following our algorithmic-based treatments are about 2.82 days and $ 461.46 per patient per year. Finally, we make use of our findings and generate insights for policymakers as well as individual physicians into better ways of managing opioid prescriptions (and hence, the opioid epidemic) by incorporating and interacting with our algorithmic-based approach.

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. (C) 2022 Elsevier Inc. (C)) 2021 Published by Elsevier Inc.

INTRODUCTION:Emergency departments (ED) are rapidly replacing conventional troponin assays with high-sensitivity troponin tests. We sought to evaluate emergency physician utilization of troponin tests before and after high-sensitivity troponin introduction in our ED.METHODS:We retrospectively examined 9,477 ED encounters, identifying the percentage in which physicians ordered a serum troponin both before and after our institution adopted a high-sensitivity troponin test.RESULTS:After introduction of high-sensitivity troponin testing, the percentage of ED encounters in which physicians ordered troponin studies decreased (28.3% before vs 22% after; P <.001), with the drop most pronounced in admitted patients (decrease of 10.9% [95% confidence interval [CI]: 7.3%-14.5%] in admitted patients vs decrease of 3.6% [95% CI: 1.7%-5.4%] in discharged patients; P<.001) CONCLUSION: Introduction of high-sensitivity troponin testing was associated with a decrease in troponin ordering. While the reasons for this are unclear, it is possible that physicians became more selective in their ordering behavior because of the lower specificity of high-sensitivity troponin.

Objectives We characterize physician workflow in two distinctive emergency departments (ED). Physician practices mediated by electronic health records (EHR) are explored within the context of organizational complexity for the delivery of care. Methods Two urban clinical sites, including an academic teaching ED, were selected. Fourteen physicians were recruited. Overall, 62hours of direct clinical observations were conducted characterizing clinical activities (EHR use, team communication, and patient care). Data were analyzed using qualitative open-coding techniques and descriptive statistics. Timeline belts were used to represent temporal events. Results At site 1, physicians, engaged in more team communication, followed by direct patient care. Although physicians spent 61% of their clinical time at workstations, only 25% was spent on the EHR, primarily for clinical documentation and review. Site 2 physicians engaged primarily in direct patient care spending 52% of their time at a workstation, and 31% dedicated to EHRs, focused on chart review. At site 1, physicians showed nonlinear complex workflow patterns with a greater frequency of multitasking and interruptions, resulting in workflow fragmentation. In comparison, at site 2, a less complex environment with a unique patient assignment system, resulting in a more linear workflow pattern. Conclusion The nature of the clinical practice and EHR-mediated workflow reflects the ED work practices. Physicians in more complex organizations may be less efficient because of the fragmented workflow. However, these effects can be mitigated by effort distribution through team communication, which affords inherent safety checks.

Introduction Daily patient volume in emergency departments (ED) varies considerably between days and sites. Although studies have attempted to define “high-volume” days, no standard definition exists. Furthermore, it is not clear whether the frequency of high-volume days, by any definition, is related to the size of an ED. We aimed to determine the correlation between ED size and the frequency of high-volume days for various volume thresholds, and to develop a measure to identify high-volume days. Methods We queried retrospective patient arrival data including 1,682,374 patient visits from 32 EDs in 12 states between July 1, 2018–June 30, 2019 and developed linear regression models to determine the correlation between ED size and volume variability. In addition, we performed a regression analysis and applied the Pearson correlation test to investigate the significance of median daily volumes with respect to the percent of days that crossed four volume thresholds ranging from 5–20% (in 5% increments) greater than each site’s median daily volume. Results We found a strong negative correlation between ED median daily volume and volume variability (R2 = 81.0%; P < 0.0001). In addition, the four regression models for the percent of days exceeding specified thresholds greater than their daily median volumes had R2 values of 49.4%, 61.2%, 70.0%, and 71.8%, respectively, all with P < 0.0001. Conclusion We sought to determine whether smaller EDs experience high-volume days more frequently than larger EDs. We found that high-volume days, when defined as days with a count of arrivals at or above certain median-based thresholds, are significantly more likely to occur in lower-volume EDs than in higher-volume EDs. To the extent that EDs allocate resources and plan to staff based on median volumes, these results suggest that smaller EDs are more likely to experience unpredictable, volume-based staffing challenges and operational costs. Given the lack of a standard measure to define a high-volume day in an ED, we recommend 10% above the median daily volume as a metric, for its relevance, generalizability across a broad range of EDs, and computational simplicity.

Early assignment of patients to specific treatment teams improves length of stay, rate of patients leaving without being seen, patient satisfaction, and resident education. Multiple variations of patient assignment systems exist, including provider-in-triage/team triage, fast-tracks/vertical pathways, and rotational patient assignment. The authors discuss the theory behind patient assignment systems and review potential benefits of specific models of patient assignment found in the current literature.

Objectives: Patients admitted to a medical-surgical unit infrequently require early transfer to higher level care, although how their inpatient length of stay compares to untransferred patients, or those directly admitted to intermediate care, is unknown. We sought to compare the inpatient length of stay of these groups. Design: Single-site retrospective analysis. Setting: An academic hospital specializing in complex care. Patients: We evaluated 23,694 patients admitted to the Hospital Internal Medicine service over a 4-year period (January 1, 2013, to December 31, 2016). Interventions: None. Measurements and Main Results: Using 6- and 24-hour definitions of early transfer, we categorized patients as admitted to medical-surgical unit without early transfer (medical-surgical unit), transferred (TX) early to higher level care, or initially admitted to an intermediate care unit. We report patient characteristics and inpatient length of stay adjusted for patient demographics (age and sex) and initial acuity (measured by Emergency Severity Index). There were significant increases in both unadjusted inpatient length of stay (6 hr: medical-surgical unit = 73.4 hr, TX = 137.9 hr, intermediate care unit = 101.1 hr; 24 hr: medical-surgical unit = 72.4 hr, TX = 141.9 hr, intermediate care unit = 98.2 hr; p < 0.01 for all groups) and adjusted inpatient length of stay (6-hr definition: medical-surgical unit = 50.9 hr [95% CI, 50.3–51.6 hr], TX = 100.4 hr [90.4–112.0 hr], intermediate care unit = 72.3 hr [70.6–74.0 hr]; 24-hr definition: medical-surgical unit = 50.3 hr [49.7–50.9 hr], TX = 108.3 hr [101.5–116.0 hr], intermediate care unit = 70.7 hr [69.0–72.3 hr]; p < 0.0001 for comparison of TX to medical-surgical unit and intermediate care unit in both groups). The increases in inpatient length of stay for the TX groups were not explained by differences in demographics or acuity. Conclusions: In a single facility study, patients admitted to a medical/surgical unit who require early transfer to intermediate care unit have a significant and unexplained increase in inpatient length of stay. This unexplained increased inpatient length of stay suggests that triage to the appropriate inpatient unit significantly affects inpatient length of stay.

Since the turn of the millennium, numerous healthcare venues all over the world have made a standard of communication transition from the classic telephone call to a sophisticated online patient portal system. More and more, a majority of patients prefer using portal-style communication for clinician contact, checking lab results, and other informational transactions, in which hundreds of thousands of patient portal messages (PPMs) are daily generated as free-text data with multiple requests often buried in one single message. Thus, there is a pressing need to design and implement artificial intelligence (AI) algorithms to accurately organize this wealth of data in a timely fashion. With the present contribution, an attempt was made to first develop an ensemble deep learning text classification component and then integrate it with rule-based named entity recognition to categorize free-text PPMs submitted under the "Non-Urgent Medical Question" subject in the patient portal as either containing active symptom descriptions or logistical requests (e.g., appointment rescheduling).

Background: Collecting a predefined set of blood tubes (the "rainbow draw") is a common but controversial practice in many emergency departments (EDs), with limited data to support it. We determined the actual utilization of rainbow draw tubes at a single facility and evaluated the perceptions of ED staff regarding the utility of rainbow draws. Methods: We analyzed 2 weeks of ED visits (1326 visits by 1240 unique patients) to determine blood tube utilization for initial and add-on testing, as well as the incidence of additional venipunctures. We also surveyed ED staff regarding aspects of ED phlebotomy and test ordering. Utilization data analysis was structured to satisfy specific concerns addressed in the ED staff survey. Results: Observed tube utilization data showed that fluoride/oxalate, citrate, and serum separator tubes were frequently discarded unused, and that the actual utility of the rainbow draw for add-on testing and avoiding additional venipunctures was low. ED staff perceived that the rainbow draw was highly valuable, both to expedite add-on testing and to avoid additional venipunctures. Contrasting the objective (utilization data) and subjective (survey results) to drive changes in the standard ED blood collection reduced the estimated waste blood by 175 L/year. Conclusions: Comparison of perceptions and objective utilization data drove process changes that were mutually agreeable to ED and laboratory staff. Although specifics of ED and laboratory work flows vary between institutions, the principles and strategy of this study are widely applicable.

Introduction: We sought to determine the association of abnormal vital signs with emergency department (ED) process outcomes in both discharged and admitted patients. Methods: We performed a retrospective review of five years of operational data at a single site. We identified all visits for patients 18 and older who were discharged home without ancillary services, and separately identified all visits for patients admitted to a floor (ward) bed. We assessed two process outcomes for discharged visits (returns to the ED within 72 hours and returns to the ED within 72 hours resulting in admission) and two process outcomes for admitted patients (transfer to a higher level of care [intermediate care or intensive care] within either six hours or 24 hours of arrival to floor). Last-recorded ED vital signs were obtained for all patients. We report rates of abnormal vital signs in each group, as well as the relative risk of meeting a process outcome for each individual vital sign abnormality. Results: Patients with tachycardia, tachypnea, or fever more commonly experienced all measured process outcomes compared to patients without these abnormal vitals; admitted hypotensive patients more frequently required transfer to a higher level of care within 24 hours. Conclusion: In a single facility, patients with abnormal last-recorded ED vital signs experienced more undesirable process outcomes than patients with normal vitals. Vital sign abnormalities may serve as a useful signal in outcome forecasting.

Purpose: (1) To investigate empirically the effects of meaningful use (MU) criteria on efficiency and quality metrics regarding clinical workflow at two urban emergency departments (ED) with different Electronic Health Records (EHRs); (2) To develop contextually-plausible and relevant patient safety and quality guidelines for EHR improvements Scope: Advances in health information technology (HIT), such as EHRs, can reduce the burden on clinicians, potentially improving quality, safety, and efficiency of their care activities. EHR acceptability and impact are of increasing importance. We ask how we can leverage these advances in HIT to increase efficiency (time and cost) without compromising safety. Method: A mixed-methods approach utilizing semi-structured interviews, ethnographic observations, clinician shadowing, EHR data logs, and sensor-based technology was used for data collection. To generate a composite picture of ED workflow, we used a combination of qualitative and novel data analytic and visualization approaches. Results: Changes in clinical workflow with EHR implementation are influenced by the nature of clinicians’ tasks, as dictated by specific organizational contexts (e.g., whether it is a teaching environment). These changes reflect an intricate relationship between EHR-related clinical activities (e.g., documentation, chart review, medication ordering). We identified generalizable mechanisms for characterizing aspects of workflow, where use of specific modern technology increases efficiency, and suggest ways to manage actions that have more chances of generating errors, affecting the quality of care.

Objectives To (i) investigate alterations in homotopic functional connectivity (hfc) in concussed patients relative to healthy controls (HC) and to (ii) interrogate whether hfc in concussed patients normalized during the recovery process. The relationship between symptom recovery and change in hfc was assessed using post-hoc analyses. Methods This study included 15 concussed patients (mean age = 39.1, SD = 10.1; sex: 13 females, 2 males) and 15 HC (mean age = 39.1, SD = 11.7; sex: 13 females, 2 males). Hfc patterns were interrogated using resting-state magnetic resonance imaging (rs-MRI) for 29 a priori selected pain-processing regions. Concussed patients underwent imaging at two time-points; at 1-month post-concussion (mean time following concussion: 28 days, SD = 9.5) and again at 5-months post-concussion (mean time following concussion: 121 days, SD = 13). At both time-points, symptoms associated with concussion were assessed using the Sports Concussion Assessment Tool (SCAT-3). Results Concussed patients had significantly weaker hfc in the following six regions 1-month post-concussion compared to HC: middle cingulate, posterior insula, middle occipital, spinal trigeminal nucleus, precentral and the pulvinar. There were no regions of significantly stronger hfc in concussed patients relative to HC. Longitudinally, patients showed significant symptom recovery 5-months post-concussion and had significant strengthening of hfc patterns in seven homotopic ROIs: middle cingulate, posterior insula, middle occipital, secondary somatosensory area, spinal trigeminal nucleus, precentral, and the pulvinar. Post-hoc analyses indicated a significant negative correlation between somatosensory functional connectivity strengthening and symptom severity. Conclusion At 1-month post-concussion, patients had significantly weaker hfc in a number of pain-processing regions relative to HC. However, over a period of 5-months, region-pair connectivity showed significant recovery and normalization. Those patients with more successful symptom recovery at 5-months post-concussion had more functional somatosensory strengthening, suggesting an association between functional strengthening and post-concussion symptom recovery.

Understanding potential ways through which physicians impact each other's performance can yield new insights into better management of hospitals' operations. We use evidence from Emergency Medicine to study whether and how physicians who work alongside each other during same shifts affect each other's performance. We find strong empirical evidence that physicians affect each other's speed and quality, and scheduling diverse peers during the same shift could have a positive net impact on the operations of a hospital Emergency Department (ED). Specifically, our results show that a faster (slower) peer decreases (increases) the average speed of a focal physician compared to a same-speed peer. Similarly, a higher- (lower-) quality peer decreases (increases) a focal physician's average quality. Furthermore, the presence of a less-experienced peer improves a focal physician's average speed. However, in contrast to the conventional wisdom, we do not find any evidence that more-experienced physicians can affect the performance of their less-experienced peers. We investigate various mechanisms that might be the driving force behind our findings, including psychological channels such as learning, social influence, and homophily as well as resource spillover. We identify resource spillover as the main driver of the effects we observe and show that, under high ED volumes (i.e., when the shared resources are most constrained), the magnitude of the observed effects increases. While some of these observed effects tend to be long-lived, we find that their magnitudes are fairly heterogeneous among physicians. In particular, our results show that newly-hired and/or high-performing physicians are typically more influenced than others by their peers. Finally, we draw conclusions from our results and discuss how they can be utilized by hospital administrators to improve the overall performance of physicians via better scheduling patterns and/or training programs that require physicians to work during same shifts.

The pursuit of increased efficiency and quality of clinical care based on the analysis of workflow has seen the introduction of several modern technologies into medical environments. Electronic health records (EHRs) remain central to analysis of workflow, owing to their wide-ranging impact on clinical processes. The two most common interventions to facilitate EHR-related workflow analysis are automated location tracking using sensor-based technologies and EHR usage data logs. However, to maximize the potential of these technologies, and especially to facilitate workflow redesign, it is necessary to overlay these quantitative findings on the contextual data from qualitative methods such as ethnography. Such a complementary approach promises to yield more precise measures of clinical workflow that provide insights into how redesign could address inefficiencies. In this paper, we categorize clinical workflow in the Emergency Department (ED) into three types (perceived, real and ideal) to create a structured approach to workflow redesign using the available data. We use diverse data sources: sensor-based location tracking through Radio-Frequency Identification (RFID), summary EHR usage data logs, and data from physician interviews augmented by direct observations (through clinician shadowing). Our goal is to discover inefficiencies and bottlenecks that can be addressed to achieve a more ideal workflow state relative to its real and perceived state. We thereby seek to demonstrate a novel data-driven approach toward iterative workflow redesign that generalizes for use in a variety of settings. We also propose types of targeted support or adjustments to offset some of the inefficiencies we noted.

Objective: To describe the relationship between emergency department resource utilization and admission rate at the level of the individual physician. Methods: Retrospective observational study of physician resource utilization and admitting data at two emergency departments. We calculated observed to expected (O/E) ratios for four measures of resource utilization (intravenous medications and fluids, laboratory testing, plain radiographs, and advanced imaging studies) as well as for admission rate. Expected values reflect adjustment for patient- and time-based variables. We compared O/E ratios for each type of resource utilization to the O/E ratio for admission for each provider. We report degree of correlation (slope of the trendline) and strength of correlation (adjusted R-2 value) for each association, as well as categorical results after clustering physicians based on the relationship of resource utilization to admission rate. Results: There were statistically significant positive correlations between resource utilization and physician admission rate. Physicians with lower resource utilization rates were more likely to have lower admission rates, and those with higher resource utilization rates were more likely to have higher admission rates. Conclusions: In a two-facility study, emergency physician resource utilization and admission rate were positively correlated: those who used more ED resources also tended to admit more patients. These results add to a growing understanding of emergency physician variability. (C) 2018 Elsevier Inc. All rights reserved.

Objectives Emergency physician productivity, often defined as new patients evaluated per hour, is essential to planning clinical operations. Prior research in this area considered this a static quantity; however, our group's study of resident physicians demonstrated significant decreases in hourly productivity throughout shifts. We now examine attending physicians' productivity to determine if it is also dynamic. Methods This is a retrospective cohort study, conducted from 2014 to 2016 across three community hospitals in the north-eastern USA, with different schedules and coverage. Timestamps of all patient encounters were automatically logged by the sites' electronic health record. Generalised estimating equations were constructed to predict productivity in terms of new patients per shift hour. Results 207 169 patients were seen by 64 physicians over 2 years, comprising 9822 physician shifts. Physicians saw an average of 15.0 (SD 4.7), 20.9 (SD 6.4) and 13.2 (SD 3.8) patients per shift at the three sites, with 2.97 (SD 0.22), 2.95 (SD 0.24) and 2.17 (SD 0.09) in the first hour. Across all sites, physicians saw significantly fewer new patients after the first hour, with more gradual decreases subsequently. Additional patient arrivals were associated with greater productivity; however, this attenuates substantially late in the shift. The presence of other physicians was also associated with slightly decreased productivity. Conclusions Physician productivity over a single shift follows a predictable pattern that decreases significantly on an hourly basis, even if there are new patients to be seen. Estimating productivity as a simple average substantially underestimates physicians' capacity early in a shift and overestimates it later. This pattern of productivity should be factored into hospitals' staffing plans, with shifts aligned to start with the greatest volumes of patient arrivals.

Background: Emergency physicians differ in many ways with respect to practice. One area in which interphysician practice differences are not well characterized is emergency department (ED) length of stay (LOS). Objective: To describe how ED LOS differs among physicians. Methods: We performed a 3-year, five-ED retrospective study of non-fast-track visits evaluated primarily by physicians. We report each provider's observed LOS, as well as each provider's ratio of observed LOS/expected LOS (LOSO/E); we determined expected LOS based on site average adjusted for the patient characteristics of age, gender, acuity, and disposition status, as well as the time characteristics of shift, day of week, season, and calendar year. Results: Three hundred twenty-seven thousand, seven hundred fifty-three visits seen by 92 physicians were eligible for analysis. For the five sites, the average shortest observed LOS was 151 min (range 106-184 min), and the average longest observed LOS was 232 min (range 196-270 min); the average difference was 81 min (range 69-90 min). For LOSO/E, the average lowest LOSO/E was 0.801 (range 0.702-0.887), and the average highest LOSO/E was 1.210 (range 1.186-1.275); the average difference between the lowest LOSO/E and the highest LOSO/E was 0.409 (range 0.305-0.493). Conclusion: There are significant differences in EDLOS at the level of the individual physician, even after accounting for multiple confounders. We found that the LOSO/E for physicians with the lowest LOSO/E at each site averaged approximately 20% less than predicted, and that the LOSO/E for physicians with the highest LOSO/E at each site averaged approximately 20% more than predicted. (C) 2018 Elsevier Inc. All rights reserved.