Decreasing length of stay in the emergency department with a split emergency severity index 3 patient flow model

Point-of-care testing improves care timeliness in the emergency department. A multicenter randomized clinical trial (study POCTUR)

OBJECTIVES

Emergency department (ED) crowding is a widespread problem that positions patients at risk. The desire to improve the ED throughput requires novel approaches. Point-of-care testing (POCT) has emerged as useful technology that could contribute to create more efficient patient flow and better timeliness in the ED. The main objective of our study is to demonstrate, in a multicenter study, that POCT benefits care timeliness in the ED.

METHODS

We conducted a multicenter and cluster randomized study. A total of 3,200 patients. We randomly assigned patients to a POCT group or Central Laboratory Group. The primary outcome was the ED time to clinical decision. The secondary outcome included the length of stay and the laboratory turnaround time. Readmission within the seven after discharge was also calculated.

RESULTS

The primary finding of this study is a strategy based on POCT that aims to significantly improve care timeliness in the ED. We found significant reductions in all outcomes regardless of presentation reason, patient disposition or hospital type. Time to clinical decision decreased by 75.2 min (205-129.8), length of stay by 77.5 min (273.1-195.6) and laboratory turnaround time by 56.2 min (82.2-26) in the POCT group. No increase in readmission was found.

CONCLUSIONS

Our strategy represents a good approach to optimize timeliness in the ED. It should be seen as a starting point for further operational research focusing on POCT for improving throughput and reducing crowding in the ED.

Who is coming in? Evaluation of physician performance within multi-physician emergency departments

BACKGROUND

This study aimed to examine how physician performance metrics are affected by the speed of other attendings (co-attendings) concurrently staffing the ED.

METHODS

A retrospective study was conducted using patient data from two EDs between January-2018 and February-2020. Machine learning was used to predict patient length of stay (LOS) conditional on being assigned a physician of average speed, using patient- and departmental-level variables. A physician's patients' actual LOSs were compared to the model's predictions to calculate a measurement of that physician's speed. Linear regression models were employed to assess how physician performance changed based on the measured speed of the concurrent ED co-attendings, on outcomes including patient LOS, patients treated per hour, imaging utilization, admission rates, and 72-h ED revisits.

RESULTS

Eighty physicians and 212,902 ED visits were included. Overall, patients assigned to the fastest physicians have a 17.8 % [13.5 %, 22.0 %] shorter LOS compared to average-speed attendings. When the fastest physicians work alongside the fastest co-attendings, their LOS benefit is reduced to 14.9 %, representing a 2.9 % [0.2 %, 5.6 %] longer LOS than when working without the fastest co-attendings. Similarly, the fastest physicians see 0.21 [0.13, 0.28] more patients per hour compared to average attendings, but this benefit decreases to 0.13 [0.09, 0.17] more patients per hour when the fastest co-attendings are present, reflecting a reduction of 0.08 [0.04, 0.11] patients per hour. The fastest physicians order 0.18 [0.13, 0.23] fewer imaging tests per patient than average-speed attendings; however, this reduction diminishes by 0.05 [0.04, 0.07] imaging tests per patient when the fastest co-attendings are present. Our model found effects of similar magnitudes but in the opposite direction when the slowest co-attendings are present. The speed of co-attendings had no significant association on the attending admission rate or 72-h revisit rate. Additionally, compared to the average attending team speed, slower attending teams, over an 8 h shift, experienced increased waiting room volume by 6.4 % [4.5 %, 8.4 %] while there was no difference when staffed by the fastest attending teams (-1.2 % [-3.2 %,0.7 %]).

CONCLUSION

In this exploratory analysis, physicians have slower throughput and order more imaging when faster co-attendings are present, and faster throughput with less imaging ordered when slower co-attendings are present. Administrators might consider these relationships and balancing attending speeds, particularly at the extremes (slowest and fastest), when designing staffing models as a potential strategy to enhance ED operational efficiency. What is already known on this topic: ED throughput is known to be dependent on multiple factors however physician behavior is commonly modeled as single attendings working in the ED.

WHAT THIS STUDY ADDS

This study examines the association between attending and co-attending speed on physician performance and finds that physicians become faster when a slow co-attending is present and slow down when a fast co-attending is present. How this study might affect research, practice or policy: Physician behavior does not exist in isolation and how an entire ED is staffed may have implications for throughput.

Enhancing trauma triage in low-resource settings using machine learning: a performance comparison with the Kampala Trauma Score

BACKGROUND

Traumatic injuries are a leading cause of morbidity and mortality globally, with a disproportionate impact on populations in low- and middle-income countries (LMICs). The Kampala Trauma Score (KTS) is frequently used for triage in these settings, though its predictive accuracy remains under debate. This study evaluates the effectiveness of machine learning (ML) models in predicting triage decisions and compares their performance to the KTS.

METHODS

Data from 4,109 trauma patients at Soroti Regional Referral Hospital, a rural hospital in Uganda, were used to train and evaluate four ML models: Logistic Regression (LR), Random Forest (RF), Gradient Boosting (GB), and Support Vector Machine (SVM). The models were assessed in regard to accuracy, precision, recall, F1-score, and AUC-ROC (Area Under the Curve of the Receiver Operating Characteristic curve). Additionally, a multinomial logistic regression model using the KTS was developed as a benchmark for the ML models.

RESULTS

All four ML models outperformed the KTS model, with the RF and GB both achieving AUC-ROC values of 0.91, compared to 0.62 (95% CI: 0.61-0.63) for the KTS (p < 0.01). The RF model demonstrated the highest accuracy at 0.69 (95% CI: 0.68-0.70), while the KTS-based model showed an accuracy of 0.54 (95% CI: 0.52-0.55). Sex, hours to hospital, and age were identified as the most significant predictors in both ML models.

CONCLUSION

ML models demonstrated superior predictive capabilities over the KTS in predicting triage decisions, even when utilising a limited set of injury information about the patients. These findings suggest a promising opportunity to advance trauma care in LMICs by integrating ML into triage decision-making. By leveraging basic demographic and clinical data, these models could provide a foundation for improved resource allocation and patient outcomes, addressing the unique challenges of resource-limited settings. However, further validation is essential to ensure their reliability and integration into clinical practice.

Using machine learning and natural language processing in triage for prediction of clinical disposition in the emergency department

BACKGROUND

Accurate triage is required for efficient allocation of resources and to decrease patients' length of stay. Triage decisions are often subjective and vary by provider, leading to patients being over-triaged or under-triaged. This study developed machine learning models that incorporated natural language processing (NLP) to predict patient disposition. The models were assessed by comparing their performance with the judgements of emergency physicians (EPs).

METHOD

This retrospective study obtained data from patients visiting EDs between January 2018 and December 2019. Internal validation data came from China Medical University Hospital (CMUH), while external validation data were obtained from Asia University Hospital (AUH). Nontrauma patients aged ≥ 20 years were included. The models were trained using structured data and unstructured data (free-text notes) processed by NLP. The primary outcome was death in the ED or admission to the intensive care unit, and the secondary outcome was either admission to a general ward or transferal to another hospital. Six machine learning models (CatBoost, Light Gradient Boosting Machine, Logistic Regression, Random Forest, Extremely Randomized Trees, and Gradient Boosting) and one Logistic Regression derived from triage level were developed and evaluated using EPs' predictions as reference.

RESULT

A total of 17,2101 and 41,883 patients were enrolled from CMUH and AUH, respectively. EPs achieved F1 core of 0.361 and 0.498 for the primary and secondary outcomes, respectively. All machine learning models achieved higher F1 scores compared to EPs and Logistic Regression derived from triage level. Random Forest was selected for further evaluation and fine-tuning, because of its robust calibration and predictive performance. In internal validation, it achieved Brier scores of 0.072 and 0.089 for the primary and secondary outcomes, respectively, and 0.076 and 0.095 in external validation. Further analysis revealed that incorporating unstructured data significantly enhanced the model's performance. Threshold adjustments were applied to improve clinical applicability, aiming to balance the trade-off between sensitivity and positive predictive value.

CONCLUSION

This study developed and validated machine learning models that integrate structured and unstructured triage data to predict patient dispositions, distinguishing between general ward and critical conditions like ICU admissions and ED deaths. Integrating both structured and unstructured data significantly improved model performance.

Retrospective Cohort Analysis of the Relationship Between Emergency Department Length of Stay and Timing of First Laboratory Orders

Background The efficiency of patient management in the Emergency Department (ED) is critical for optimizing healthcare delivery. Provider in triage (PIT) and similar ED flow models attempt to expedite throughput by decreasing the amount of time between patient arrival and initial order placement. The exact relationship between ED length of stay (LOS) and the timing of the first laboratory order, however, is unclear. The varying speed at which clinicians of different ages place laboratory orders and move patients through an ED also is understudied.  Methods A retrospective analysis was conducted using SQL from the Clarity data archive to pull all patient encounters in 2023. Linear regression models using Analysis ToolPak in Microsoft Excel were used to create and examine the relationship between LOS and the timing of the first laboratory order. Secondary outcomes using the same models were created to analyze the impact of clinician age on LOS and the relationship between clinician age and the timing of first laboratory orders.  Results Two hundred sixty-nine thousand eight hundred and eight ED visits were reviewed across three academic and 17 community emergency departments. We report a weak but statistically significant positive relationship between the timing of the first laboratory order and LOS (R² = 0.0378, p < 0.001). Secondary outcomes indicated a very weak negative correlation between clinician age and LOS (R² ≈ 0, p < 0.001) and no significant relationship between clinician age and the timing of the first laboratory order (R² ≈ 0, p > 0.05). Conclusion The timing of the first laboratory order is a significant, albeit weak, predictor of LOS in the ED. Clinician age has minimal impact on LOS and does not significantly influence the timing of the first laboratory order.

Identification of avoidable patients at triage in a Paediatric Emergency Department: a decision support system using predictive analytics

BACKGROUND

Crowding has been a longstanding issue in emergency departments. To address this, a fast-track system for avoidable patients is being implemented in the Paediatric Emergency Department where our study is conducted. Our goal is to develop an optimized Decision Support System that helps in directing patients to this fast track. We evaluated various Machine Learning models, focusing on a balance between complexity, predictive performance, and interpretability.

METHODS

This is a retrospective study considering all visits to a university-affiliated metropolitan hospital's PED between 2014 and 2019. Using information available at the time of triage, we trained several models to predict whether a visit is avoidable and should be directed to a fast-track area.

RESULTS

A total of 507,708 visits to the PED were used in the training and testing of the models. Regarding the outcome, 41.6% of the visits were considered avoidable. Except for the classification made by triage rules, i.e. considering levels 1,2, and 3 as non-avoidable and 4 and 5 as avoidable, all models had similar results in model's evaluation metrics, e.g. Area Under the Curve ranging from 74% to 80%.

CONCLUSIONS

Regarding predictive performance, the pruned decision tree had evaluation metrics results that were comparable to the other ML models. Furthermore, it offers a low complexity and easy to implement solution. When considering interpretability, a paramount requisite in healthcare since it relates to the trustworthiness and transparency of the system, the pruned decision tree excels. Overall, this paper contributes to the growing body of research on the use of machine learning in healthcare. It highlights practical benefits for patients and healthcare systems of the use ML-based DSS in emergency medicine. Moreover, the obtained results can potentially help to design patients' flow management strategies in PED settings, which has been sought as a solution for addressing the long-standing problem of overcrowding.

A comparative study of explainable ensemble learning and logistic regression for predicting in-hospital mortality in the emergency department

This study addresses the challenges associated with emergency department (ED) overcrowding and emphasizes the need for efficient risk stratification tools to identify high-risk patients for early intervention. While several scoring systems, often based on logistic regression (LR) models, have been proposed to indicate patient illness severity, this study aims to compare the predictive performance of ensemble learning (EL) models with LR for in-hospital mortality in the ED. A cross-sectional single-center study was conducted at the ED of Imam Reza Hospital in northeast Iran from March 2016 to March 2017. The study included adult patients with one to three levels of emergency severity index. EL models using Bagging, AdaBoost, random forests (RF), Stacking and extreme gradient boosting (XGB) algorithms, along with an LR model, were constructed. The training and validation visits from the ED were randomly divided into 80% and 20%, respectively. After training the proposed models using tenfold cross-validation, their predictive performance was evaluated. Model performance was compared using the Brier score (BS), The area under the receiver operating characteristics curve (AUROC), The area and precision-recall curve (AUCPR), Hosmer-Lemeshow (H-L) goodness-of-fit test, precision, sensitivity, accuracy, F1-score, and Matthews correlation coefficient (MCC). The study included 2025 unique patients admitted to the hospital's ED, with a total percentage of hospital deaths at approximately 19%. In the training group and the validation group, 274 of 1476 (18.6%) and 152 of 728 (20.8%) patients died during hospitalization, respectively. According to the evaluation of the presented framework, EL models, particularly Bagging, predicted in-hospital mortality with the highest AUROC (0.839, CI (0.802-0.875)) and AUCPR = 0.64 comparable in terms of discrimination power with LR (AUROC (0.826, CI (0.787-0.864)) and AUCPR = 0.61). XGB achieved the highest precision (0.83), sensitivity (0.831), accuracy (0.842), F1-score (0.833), and the highest MCC (0.48). Additionally, the most accurate models in the unbalanced dataset belonged to RF with the lowest BS (0.128). Although all studied models overestimate mortality risk and have insufficient calibration (P > 0.05), stacking demonstrated relatively good agreement between predicted and actual mortality. EL models are not superior to LR in predicting in-hospital mortality in the ED. Both EL and LR models can be considered as screening tools to identify patients at risk of mortality.

The Impact of Team Triage Method on Emergency Department Performance Indexes: A quasi-interventional study

Introduction

One of the effective methods of patient triage in the emergency department (ED) is the use of team triage, including physicians and nurses. Considering that there is no conclusive evidence about the effectiveness of team triage, this study aimed to investigate the effect of the team triage method on ED performance indexes.

Methods

The present study is a quasi-interventional study in which 200 patients were referred to the ED in the hospitals of Tabriz in 2020. Participants were randomly assigned to two groups (team triage and conventional triage) and were evaluated. Data were collected by a three-part questionnaire including the participants' demographic characteristics, the five-level triage form, and Press-Ganey satisfaction questionnaire were used. Data were analyzed by SPSS.22 statistical software.

Results

The results showed that the mean score of waiting time for the first physician visit in team triage was statistically significantly lower than the conventional triage (P = 0.001). Furthermore, the mean score of waiting time for receiving the first treatment in team triage was statistically significantly lower than the conventional triage (P = 0.001). Finally, the mean score of patients' satisfaction in team and conventional triage was statistically significantly higher in team triage (P = 0.001).

Conclusion

The study findings revealed that the team triage method, in comparison to conventional triage, decrease the waiting time for receiving the first service and length of stay, but leads to more patient's satisfaction. Therefore, to improve the performance indicators of the ED, it is recommended that hospital managers use the team triage method.

Operations Factors Associated with Emergency Department Length of Stay: Analysis of a National Operations Database

Introduction: Prolonged emergency department (ED) length of stay (LOS) has been shown to adversely affect patient care. We sought to determine factors associated with ED LOS via analysis of a large, national, ED operations database. Methods: We performed retrospective, multivariable, linear regression modeling using the 2019 Emergency Department Benchmarking Alliance survey results to identify associated factors of ED LOS for admitted and discharged patients. Results: A total of 1,052 general and adult-only EDs responded to the survey. Median annual volume was 40,946. The median admit and discharge LOS were 289 minutes and 147 minutes, respectively. R-squared values for the admit and discharge models were 0.63 and 0.56 with out-of-sample R-squared values of 0.54 and 0.59, respectively. Both admit and discharge LOS were associated with academic designation, trauma level designation, annual volume, proportion of ED arrivals occurring via emergency medical services, median boarding, and use of a fast track. Additionally, admit LOS was associated with transfer-out percentage, and discharge LOS was associated with percentage of high Current Procedural Terminology, percentage of patients <18 years old, use of radiographs and computed tomography, and use of an intake physician. Conclusion: Models derived from a large, nationally representative cohort identified diverse associated factors of ED length of stay, several of which were not previously reported. Dominant within the LOS modeling were patient population characteristics and other factors extrinsic to ED operations, including boarding of admitted patients, which was associated with both admitted and discharged LOS. The results of the modeling have significant implications for ED process improvement and appropriate benchmarking.

The FRISK (Fracture Risk)-A New Tool to Indicate the Probability of Fractures

Increasing patient inflow into the emergency department makes it necessary to optimize triage management. The scope of this work was to determine simple factors that could detect fractures in patients without the need for specialized personnel. Between 2014 and 2015, 798 patients were admitted to an orthopedic emergency department and prospectively included in the study. The patients received a questionnaire before contacting the doctor. Objective and subjective data were evaluated to determine fracture risk for the upper and lower extremities. The highest risk for fractures in one region was the hip (73.21%; n = 56), followed by the wrist (60.32%; n = 63) and the femoral shaft (4 of 7, 57.14%; n = 7). The regions with the lowest risk were the knee (8.41%; n = 107), the ankle (18.29%; n = 164), and the forearm shaft (30.00%; n = 10). Age was a predictor for fracture: patients older than 59 years had a risk greater than 59.26%, and patients older than 90 years had a risk greater than 83.33%. The functional questions could exclude fractures. Three factors seem to be able to predict fracture risk: the injured region, the patient's age, and a functional question. They can be used for a probatory heuristic that needs to be proven in a prospective way.

Implementation of Vertical Split Flow Model for Patient Throughput at a Community Hospital Emergency Department

BACKGROUND

Hospitals have implemented innovative strategies to address overcrowding by optimizing patient flow through the emergency department (ED). Vertical split flow refers to the concept of assigning patients to vertical chairs instead of horizontal beds based on patient acuity.

OBJECTIVE

Evaluate the impact of vertical split flow implementation on ED Emergency Severity Index (ESI) level 3, patient length of stay, and throughput at a community hospital.

METHODS

Retrospective cohort study of all ESI level 3 patients presenting to a community hospital ED over a 3-month period prior to and after vertical split flow implementation between 2018 and 2019.

RESULTS

In total, data were collected from 10,638 patient visits: 5262 and 5376 patient visits pre- and postintervention, respectively. There was a significant reduction in mean overall length of stay when ESI-3 patients were triaged with vertical split flow (251 min vs 283 min, p < 0.001).

CONCLUSIONS

Community hospital ED implementation of vertical split flow for ESI level 3 patients was associated with a significant reduction in overall length of stay and improved throughput. This model provides a solution to increase the number of patients that can be simultaneously cared for in the ED without increasing staffing or physical space.

Measuring Potential Visual Exposure of Physicians During Shift-End Handoffs and Its Impact on Interruptions, Privacy, and Collaboration

BACKGROUND

Frequent interruptions, inadequate privacy, and lack of collaboration are barriers to safe and efficient end-of-shift handoffs between emergency physicians. Varying levels of visibility to and from physicians can impact these outcomes. This study quantifies potential visual exposure of physicians in workstations with varying enclosure levels using isovist connectivity (IC) as a measure. Further, this study examines the association of IC with number of interruptions/hour, perceived collaboration, and privacy during handoffs.

METHODS

In-person observations were conducted during 60 handoffs to capture interruptions. Surveys were administered to the incoming and outgoing physicians to garner their perceptions of the extent of interruptions, collaboration, and privacy. Spatial analysis was conducted using DepthmapX.

RESULTS

Findings demonstrate significant differences in IC scores based on (a) physicians location within the workstation during; (b) handoff approach (individual or collaborative); (c) position during handoff (sitting or standing). Documented interruptions were highest in the high IC locations and lowest in the medium and low IC locations. Physicians in low IC locations perceived to have sufficient privacy to conduct handoffs.

LIMITATIONS AND CONCLUSION

It should be noted that the three pods, each housing a physician workstation with different enclosure levels, varied in number of patient rooms, patient acuity, overall size, and the location of workstations. While contextual variables were considered to the extent possible, several other factors could have resulted in differences in number of interruptions and collaboration levels. This study provides design recommendations for handoff locations and a method to test emergency physician workstation designs prior to construction.

Optimizing Pediatric Patient Safety in the Emergency Care Setting

Patient safety is the foundation of high-quality health care and remains a critical priority for all clinicians caring for children. There are numerous aspects of pediatric care that increase the risk of patient harm, including but not limited to risk from medication errors attributable to weight-dependent dosing and need for appropriate equipment and training. Of note, the majority of children who are ill and injured are brought to community hospital emergency departments. It is, therefore, imperative that all emergency departments practice patient safety principles, support a culture of safety, and adopt best practices to improve safety for all children seeking emergency care. This technical report outlined the challenges and resources necessary to minimize pediatric medical errors and to provide safe medical care for children of all ages in emergency care settings.

Optimizing Pediatric Patient Safety in the Emergency Care Setting

This is a revision of the previous American Academy of Pediatrics policy statement titled "Patient Safety in the Emergency Care Setting," and is the first joint policy statement by the American Academy of Pediatrics, the American College of Emergency Physicians, and the Emergency Nurses Association to address pediatric patient safety in the emergency care setting. Caring for children in the emergency setting can be prone to medical errors because of a number of environmental and human factors. The emergency department (ED) has frequent workflow interruptions, multiple care transitions, and barriers to effective communication. In addition, the high volume of patients, high-decision density under time pressure, diagnostic uncertainty, and limited knowledge of patients' history and preexisting conditions make the safe care of critically ill and injured patients even more challenging. It is critical that all EDs, including general EDs who care for the majority of ill and injured children, understand the unique safety issues related to children. Furthermore, it is imperative that all EDs practice patient safety principles, support a culture of safety, and adopt best practices to improve safety for all children seeking emergency care. This policy statement outlines the recommendations necessary for EDs to minimize pediatric medical errors and to provide safe care for children of all ages.

Machine learning-based triage to identify low-severity patients with a short discharge length of stay in emergency department

BACKGROUND

Overcrowding in emergency departments (ED) is a critical problem worldwide, and streaming can alleviate crowding to improve patient flows. Among triage scales, patients labeled as "triage level 3" or "urgent" generally comprise the majority, but there is no uniform criterion for classifying low-severity patients in this diverse population. Our aim is to establish a machine learning model for prediction of low-severity patients with short discharge length of stay (DLOS) in ED.

METHODS

This was a retrospective study in the ED of China Medical University Hospital (CMUH) and Asia University Hospital (AUH) in Taiwan. Adult patients (aged over 20 years) with Taiwan Triage Acuity Scale level 3 were enrolled between 2018 and 2019. We used available information during triage to establish a machine learning model that can predict low-severity patients with short DLOS. To achieve this goal, we trained five models-CatBoost, XGBoost, decision tree, random forest, and logistic regression-by using large ED visit data and examined their performance in internal and external validation.

RESULTS

For internal validation in CMUH, 33,986 patients (75.9%) had a short DLOS (shorter than 4 h), and for external validation in AUH, there were 13,269 (82.7%) patients with short DLOS. The best prediction model was CatBoost in internal validation, and area under the receiver operating cha racteristic curve (AUC) was 0.755 (95% confidence interval (CI): 0.743-0.767). Under the same threshold, XGBoost yielded the best performance, with an AUC value of 0.761 (95% CI: 0.742- 0.765) in external validation.

CONCLUSIONS

This is the first study to establish a machine learning model by applying triage information alone for prediction of short DLOS in ED with both internal and external validation. In future work, the models could be developed as an assisting tool in real-time triage to identify low-severity patients as fast track candidates.

Reduced Emergency Department Length of Stay and Proportion of Patients Who Left Without Being Seen Following Implementation of an Interprofessional Vertical Flow Track With Pivot Triage: A Retrospective Pre-/Postintervention Evaluation

Our objective was to assess change in length of stay and patients who left without being seen following implementation of a pivot triage and interprofessional vertical flow track process at a midwestern academic medical center emergency department. The intervention leveraged an existing interprofessional staffing model including a registered nurse and a paramedic to staff a vertical flow track daily from 1100 to 2300. Pre- and postintervention data were retrospectively abstracted from the electronic charting software. Outcomes included emergency department length of stay and percentage of patients leaving without being seen. Visits for patients during the postintervention period (May 10, 2019, to August 31, 2019) were compared with a corresponding preintervention time period 1 year prior (May 10, 2018, to August 31, 2018). The percentage of patients routed to the vertical flow track increased from 5% to 22% following the process intervention. Median emergency department length of stay decreased from 199 (interquartile range [IQR]: 129-282) to 159 (IQR: 98-232) min. The percentage of patients leaving without being seen decreased from 2.9% to 0.5%; between 1100 and 2300, these changes were more pronounced. Odds of a patient experiencing emergency department length of stay under 180 min increased nearly twofold (odds ratio [OR]: 1.92, 95% confidence interval [CI]: 1.79-2.08) and odds that a patient stayed to be seen by a medical professional increased sixfold (OR: 5.94, 95% CI: 4.08-8.63). Overall, more than 20% of patients were routed through the vertical flow track following the process change. Implementation of an emergency department pivot triage approach with a dedicated interprofessional vertical flow track was associated with significantly shorter emergency department length of stay and reduced patients leaving without being seen.

Development and Validation of Simple Age-Adjusted Objectified Korean Triage and Acuity Scale for Adult Patients Visiting the Emergency Department

PURPOSE

The study aimed to develop an objectified Korean Triage and Acuity Scale (OTAS) that can objectively and quickly classify severity, as well as a simple age-adjusted OTAS (S-OTAS) that reflects age and evaluate its usefulness.

MATERIALS AND METHODS

A retrospective analysis was performed of all adult patients who had visited the emergency department at three teaching hospitals. Sex, systolic blood pressure, diastolic blood pressure, pulse rate, respiratory rate, body temperature, O₂ saturation, and consciousness level were collected from medical records. The OTAS was developed with objective criterion and minimal OTAS level, and S-OTAS was developed by adding the age variable. For usefulness evaluation, the 30-day mortality, the rates of computed tomography scan and emergency procedures were compared between Korean Triage and Acuity Scale (KTAS) and OTAS.

RESULTS

A total of 44402 patients were analyzed. For 30-day mortality, S-OTAS showed a higher area under the curve (AUC) compared to KTAS (0.751 vs. 0.812 for KTAS and S-OTAS, respectively, p<0.001). Regarding the rates of emergency procedures, AUC was significantly higher in S-OTAS, compared to KTAS (0.807 vs. 0.830, for KTAS and S-OTAS, respectively, p=0.013).

CONCLUSION

S-OTAS showed comparative usefulness for adult patients visiting the emergency department as a triage tool compared to KTAS.

Artificial intelligence and machine learning in emergency medicine: a narrative review

AIM

The emergence and evolution of artificial intelligence (AI) has generated increasing interest in machine learning applications for health care. Specifically, researchers are grasping the potential of machine learning solutions to enhance the quality of care in emergency medicine.

METHODS

We undertook a narrative review of published works on machine learning applications in emergency medicine and provide a synopsis of recent developments.

RESULTS

This review describes fundamental concepts of machine learning and presents clinical applications for triage, risk stratification specific to disease, medical imaging, and emergency department operations. Additionally, we consider how machine learning models could contribute to the improvement of causal inference in medicine, and to conclude, we discuss barriers to safe implementation of AI.

CONCLUSION

We intend that this review serves as an introduction to AI and machine learning in emergency medicine.

Using explainable machine learning to characterise data drift and detect emergent health risks for emergency department admissions during COVID-19

A key task of emergency departments is to promptly identify patients who require hospital admission. Early identification ensures patient safety and aids organisational planning. Supervised machine learning algorithms can use data describing historical episodes to make ahead-of-time predictions of clinical outcomes. Despite this, clinical settings are dynamic environments and the underlying data distributions characterising episodes can change with time (data drift), and so can the relationship between episode characteristics and associated clinical outcomes (concept drift). Practically this means deployed algorithms must be monitored to ensure their safety. We demonstrate how explainable machine learning can be used to monitor data drift, using the COVID-19 pandemic as a severe example. We present a machine learning classifier trained using (pre-COVID-19) data, to identify patients at high risk of admission during an emergency department attendance. We then evaluate our model's performance on attendances occurring pre-pandemic (AUROC of 0.856 with 95%CI [0.852, 0.859]) and during the COVID-19 pandemic (AUROC of 0.826 with 95%CI [0.814, 0.837]). We demonstrate two benefits of explainable machine learning (SHAP) for models deployed in healthcare settings: (1) By tracking the variation in a feature's SHAP value relative to its global importance, a complimentary measure of data drift is found which highlights the need to retrain a predictive model. (2) By observing the relative changes in feature importance emergent health risks can be identified.

Emergency severity level-3 patient flow based on point-of-care testing improves patient outcomes

BACKGROUND

Overcrowding of the Emergency Department is rapidly becoming a global challenge and a major source of concern for emergency physicians. The desire to improve Emergency Department throughput requires novel approaches to patient flow.

MATERIALS AND METHODS

We conducted a prospective and cluster-randomized study, to evaluate the impact in patient outcomes of a new patient flow based on Point-of-Care Testing (POCT). A total of 380 Emergency Severity Level-3 patients were enrolled and studied in two different groups, interventional arm (laboratory analyses performed on POCT analyzers implemented in the Emergency Department) or control arm (central laboratory). The primary outcome was the Emergency Department length of stay. Secondary outcome included the time to first medical intervention, the laboratory turnaround time and the time to disposition decision. Readmission within the 7 days after discharge was also calculated.

RESULTS

Length of stay significantly decreased by 88.50 min (from 247.00 to 158.50), time to disposition decision by 89.00 min (from 192.00 to 103.00) and laboratory turnaround time by 67.11 min (from 89.84 to 22.73) in the POCT group. No increase in readmission was found.

CONCLUSION

Our strategy based on POCT represents a good approach to optimize patient flow in the Emergency Department and it should be seen as a starting point for further studies focusing on improving throughput.

Two-step predictive model for early detection of emergency department patients with prolonged stay and its management implications

OBJECTIVE

To develop a novel model for predicting Emergency Department (ED) prolonged length of stay (LOS) patients upon triage completion, and further investigate the benefit of a targeted intervention for patients with prolonged ED LOS.

MATERIALS AND METHODS

A two-step model to predict patients with prolonged ED LOS (>16 h) was constructed. This model was initially used to predict ED resource usage and was subsequently adapted to predict patient ED LOS based on the number of ED resources using binary logistic regressions and was validated internally with accuracy. Finally, a discrete event simulation was used to move patients with predicted prolonged ED LOS directly to a virtual Clinical Decision Unit (CDU). The changes of ED crowding status (Overcrowding, Crowding, and Not-Crowding) and savings of ED bed-hour equivalents were estimated as the measures of the efficacy of this intervention.

RESULTS

We screened a total of 123,975 patient visits with final enrollment of 110,471 patient visits. The overall accuracy of the final model predicting prolonged patient LOS was 67.8%. The C-index of this model ranges from 0.72 to 0.82. By implementing the proposed intervention, the simulation showed a 12% (1044/8760) reduction of ED overcrowded status - an equivalent savings of 129.3 ED bed-hours per day.

CONCLUSIONS

Early prediction of prolonged ED LOS patients and subsequent (simulated) early CDU transfer could lead to more efficiently utilization of ED resources and improved efficacy of ED operations. This study provides evidence to support the implementation of this novel intervention into real healthcare practice.

Triage Accuracy and Its Association with Patient Factors Using Emergency Severity Index: Findings from United Arab Emirates

Introduction

The Ministry of Health and Prevention of the UAE acquired an electronic medical record system (Wareed) through which they incorporated the Emergency Severity Index as the standard triaging tool. This raised the need to review population dynamics and the accuracy of triage performed by the health-care providers utilizing the tool.

Objective

This research aimed to study demographics and dynamics of the population presenting to emergency departments (EDs) during 2018, evaluate the accuracy of triage assessment using comparative analysis techniques, and determine relationships between patient factors (severity of illness, age-group) and the accuracy of triage.

Methods

This was an observational study that aimed to ascertain findings from ED data over 1 year (January 2018-December 2018) and explore factors associated with reduced accuracy in acuity assignment. We employed comparative analysis to measure the level of agreement between standard guidelines and local findings.

Results

A total of 576,154 patients visited EDs in 2018, of which 54.4% were male. A statistically significant increase in length of stay with increasing severity of illness was observed (Kruskal-Wallis test). Overall triage accuracy was 41.6%, with a positive association with increasing severity of illness. We found a positive association between severity of illness and accuracy of triage (OR 0.14, p=0). We also found on logistic regression that the age-group 11-20 years had the highest probability of accurate triage acuity (R ²=0.41, p=0).

Conclusion

Conducted on a very large data set from the UAE, our study reflects upon population dynamics and triage accuracy distribution among different variables. This study paves the way for further in-depth analysis of factors that may impact triage accuracy within EDs, and utilizing a similar approach it can be replicated in other settings as well.

Effect of a split-flow physician in triage model on abdominal CT ordering rate and yield

OBJECTIVE

The objective of this study was to compare the rate and clinical yield of computed tomography (CT) imaging between patients presenting with abdominal pain initially seen by a physician in triage (PIT) versus those seen only by physicians working in the main emergency department (ED).

METHODS

A retrospective study was conducted of all self-arrivals >18 years old presenting to a single ED with abdominal pain. Nine-hundred patients were randomly selected from both the PIT and traditional patient flow groups and rates and yields of CT imaging were compared, both alone and in a model controlling for potential confounders. Predetermined criteria for CT significance included need for admission, consult, or targeted medications.

RESULTS

The overall rate of CT imaging (unadjusted) did not differ between the PIT and traditional groups, 48.7% (95% CI 45.4-51.9) vs. 45.1% (95% CI 41.8-48.4), respectively (p = .13). The CT yield for patients seen in in the PIT group was also similar to that of the traditional group: 49.1% (95% CI 44.4-53.8) vs. 50.5% (95% CI 45.6-55.4) (p = .68). In the logistic regression model, when controlling for age, gender, ESI-acuity, race and insurance payor, PIT vs. traditional was not a predictor of CT ordering (OR 1.14, 95% CI 0.94-1.38).

CONCLUSIONS

For patients with abdominal pain, we found no significant differences in rates of CT ordering or CT yield for patients seen in a PIT vs. traditional models, suggesting the increased efficiencies offered by PIT models do not come at the cost of increased or decreased imaging utilization.

Efficacy of deep learning methods for predicting under-five mortality in 34 low-income and middle-income countries

OBJECTIVES

To explore the efficacy of machine learning (ML) techniques in predicting under-five mortality (U5M) in low-income and middle-income countries (LMICs) and to identify significant predictors of U5M.

DESIGN

This is a cross-sectional, proof-of-concept study.

SETTINGS AND PARTICIPANTS

We analysed data from the Demographic and Health Survey. The data were drawn from 34 LMICs, comprising a total of n=1 520 018 children drawn from 956 995 unique households.

PRIMARY AND SECONDARY OUTCOME MEASURES

The primary outcome measure was U5M; secondary outcome was comparing the efficacy of deep learning algorithms: deep neural network (DNN); convolution neural network (CNN); hybrid CNN-DNN with logistic regression (LR) for the prediction of child's survival.

RESULTS

We found that duration of breast feeding, number of antenatal visits, household wealth index, postnatal care and the level of maternal education are some of the most important predictors of U5M. We found that deep learning techniques are superior to LR for the classification of child survival: LR sensitivity=0.47, specificity=0.53; DNN sensitivity=0.69, specificity=0.83; CNN sensitivity=0.68, specificity=0.83; CNN-DNN sensitivity=0.71, specificity=0.83.

CONCLUSION

Our findings provide an understanding of determinants of U5M in LMICs. It also demonstrates that deep learning models are more efficacious than traditional analytical approach.

Design of the Academic Emergency Department

This article introduces a clinical audience to the process of emergency department (ED) design, particularly relating to academic EDs. It explains some of the major terms, processes, and key decisions that clinical staff will experience as participants in the design process. Topics covered include an overview of the planning and design process, issues related to determining needed patient capacity, the impact of patient flow models on design, and a description of several common ED design types and their advantages and disadvantages.

Factors Affecting Emergency Department Crowding

Emergency department crowding is a multifactorial issue with causes intrinsic to the emergency department and to the health care system. Understanding that the causes of emergency department crowding span this continuum allows for a more accurate analysis of its effects and a more global consideration of potential solutions. Within the emergency department, boarding of inpatients is the most appreciable effect of hospital-wide crowding, and leads to further emergency department crowding. We explore the concept of emergency department crowding, and its causes, effects, and potential strategies to overcome this problem.

Adjusting Daily Inpatient Bed Allocation to Smooth Emergency Department Occupancy Variation

Study Objective: Overcrowding in emergency departments (ED) is an increasingly common problem in Taiwanese hospitals, and strategies to improve efficiency are in demand. We propose a bed resource allocation strategy to overcome the overcrowding problem. Method: We investigated ED occupancy using discrete-event simulation and evaluated the effects of suppressing day-to-day variations in ED occupancy by adjusting the number of empty beds per day. Administrative data recorded at the ED of Taichung Veterans General Hospital (TCVGH) in Taiwan with 1500 beds and an annual ED volume of 66,000 visits were analyzed. Key indices of ED quality in the analysis were the length of stay and the time in waiting for outward transfers to in-patient beds. The model is able to analyze and compare several scenarios for finding a feasible allocation strategy. Results: We compared several scenarios, and the results showed that by reducing the allocated beds for the ED by 20% on weekdays, the variance of daily ED occupancy was reduced by 36.25% (i.e., the percentage of reduction in standard deviation). Conclusions: This new allocation strategy was able to both reduce the average ED occupancy and maintain the ED quality indices.

The Association Between Patient Outcomes and the Initial Emergency Severity Index Triage Score in Patients With Suspected Acute Coronary Syndrome

BACKGROUND

The Emergency Severity Index (ESI) is a widely used tool to triage patients in emergency departments. The ESI tool is used to assess all complaints and has significant limitation for accurately triaging patients with suspected acute coronary syndrome (ACS).

OBJECTIVE

We evaluated the accuracy of ESI in predicting serious outcomes in suspected ACS and aimed to assess the incremental reclassification performance if ESI is supplemented with a clinically validated tool used to risk-stratify suspected ACS.

METHODS

We used existing data from an observational cohort study of patients with chest pain. We extracted ESI scores documented by triage nurses during routine medical care. Two independent reviewers adjudicated the primary outcome, incidence of 30-day major adverse cardiac events. We compared ESI with the well-established modified HEAR/T (patient History, Electrocardiogram, Age, Risk factors, but without Troponin) score.

RESULTS

Our sample included 750 patients (age, 59 ± 17 years; 43% female; 40% black). A total of 145 patients (19%) experienced major adverse cardiac event. The area under the receiver operating characteristic curve for ESI score for predicting major adverse cardiac event was 0.656, compared with 0.796 for the modified HEAR/T score. Using the modified HEAR/T score, 181 of the 391 false positives (46%) and 16 of the 19 false negatives (84%) assigned by ESI could be reclassified correctly.

CONCLUSION

The ESI score is poorly associated with serious outcomes in patients with suspected ACS. Supplementing the ESI tool with input from other validated clinical tools can greatly improve the accuracy of triage in patients with suspected ACS.

A Gradient Boosting Machine Learning Model for Predicting Early Mortality in the Emergency Department Triage: Devising a Nine-Point Triage Score

BACKGROUND

Emergency departments (ED) are becoming increasingly overwhelmed, increasing poor outcomes. Triage scores aim to optimize the waiting time and prioritize the resource usage. Artificial intelligence (AI) algorithms offer advantages for creating predictive clinical applications.

OBJECTIVE

Evaluate a state-of-the-art machine learning model for predicting mortality at the triage level and, by validating this automatic tool, improve the categorization of patients in the ED.

DESIGN

An institutional review board (IRB) approval was granted for this retrospective study. Information of consecutive adult patients (ages 18-100) admitted at the emergency department (ED) of one hospital were retrieved (January 1, 2012-December 31, 2018). Features included the following: demographics, admission date, arrival mode, referral code, chief complaint, previous ED visits, previous hospitalizations, comorbidities, home medications, vital signs, and Emergency Severity Index (ESI). The following outcomes were evaluated: early mortality (up to 2 days post ED registration) and short-term mortality (2-30 days post ED registration). A gradient boosting model was trained on data from years 2012-2017 and examined on data from the final year (2018). The area under the curve (AUC) for mortality prediction was used as an outcome metric. Single-variable analysis was conducted to develop a nine-point triage score for early mortality.

KEY RESULTS

Overall, 799,522 ED visits were available for analysis. The early and short-term mortality rates were 0.6% and 2.5%, respectively. Models trained on the full set of features yielded an AUC of 0.962 for early mortality and 0.923 for short-term mortality. A model that utilized the nine features with the highest single-variable AUC scores (age, arrival mode, chief complaint, five primary vital signs, and ESI) yielded an AUC of 0.962 for early mortality.

CONCLUSION

The gradient boosting model shows high predictive ability for screening patients at risk of early mortality utilizing data available at the time of triage in the ED.

The Effect of a Rapid Assessment Zone on Emergency Department Operations and Throughput

STUDY OBJECTIVE

We examine the effects of a front-end flow model designated the rapid assessment zone on multiple emergency department (ED) operational metrics.

METHODS

This was a retrospective, before-after study of consecutive patient visits at an urban community ED. Six-month periods were compared before and after an intervention in 2017 that changed patient flow and the intake process. A lead nurse role splits patient flow immediately on patient arrival according to only age and chief complaint, allowing direct bedding without the bottlenecks of vital sign measurement, full triage assessment, or Emergency Severity Index assignment. A new patient care area (designated rapid assessment zone) preferentially expedites treatment of patients likely to remain ambulatory and serves as flexible acute care space when needed by individual cases and the ED. The outcomes measured were ED length of stay, arrival-to-provider time, the rate of leaving before treatment completion, and the rate of leaving before being seen. Data were analyzed with nonparametric testing, χ² analysis, and multiple linear regression, controlling for patient visit characteristics, ED daily census volumes, and measurements of boarding patients.

RESULTS

We analyzed 43,847 visits in the preintervention and 44,792 visits in the postintervention periods. The intervention was associated with the following changes: median ED length of stay from 203 to 171 minutes (-15.8%), median arrival-to-provider time from 28 to 13 minutes (-53.6%), leaving before treatment completion from 1.0% to 0.8% (-20%), and leaving before being seen from 3.1% to 0.5% (-84%). Regression analysis accounting for multiple confounders demonstrated that the reduced length of stay after rapid assessment zone implementation persisted across Emergency Severity Index levels 2 to 5 and all ED daily census levels.

CONCLUSION

The rapid assessment zone model aims to decrease front-end bottlenecks and minimize serial intake assessments at a high-volume, urban ED. It was associated with improved patient throughput and decreased early patient departure. It may represent a useful model for similar centers.

Emergency department triage prediction of clinical outcomes using machine learning models

Background

Development of emergency department (ED) triage systems that accurately differentiate and prioritize critically ill from stable patients remains challenging. We used machine learning models to predict clinical outcomes, and then compared their performance with that of a conventional approach—the Emergency Severity Index (ESI).

Methods

Using National Hospital and Ambulatory Medical Care Survey (NHAMCS) ED data, from 2007 through 2015, we identified all adult patients (aged ≥ 18 years). In the randomly sampled training set (70%), using routinely available triage data as predictors (e.g., demographics, triage vital signs, chief complaints, comorbidities), we developed four machine learning models: Lasso regression, random forest, gradient boosted decision tree, and deep neural network. As the reference model, we constructed a logistic regression model using the five-level ESI data. The clinical outcomes were critical care (admission to intensive care unit or in-hospital death) and hospitalization (direct hospital admission or transfer). In the test set (the remaining 30%), we measured the predictive performance, including area under the receiver-operating-characteristics curve (AUC) and net benefit (decision curves) for each model.

Results

Of 135,470 eligible ED visits, 2.1% had critical care outcome and 16.2% had hospitalization outcome. In the critical care outcome prediction, all four machine learning models outperformed the reference model (e.g., AUC, 0.86 [95%CI 0.85–0.87] in the deep neural network vs 0.74 [95%CI 0.72–0.75] in the reference model), with less under-triaged patients in ESI triage levels 3 to 5 (urgent to non-urgent). Likewise, in the hospitalization outcome prediction, all machine learning models outperformed the reference model (e.g., AUC, 0.82 [95%CI 0.82–0.83] in the deep neural network vs 0.69 [95%CI 0.68–0.69] in the reference model) with less over-triages in ESI triage levels 1 to 3 (immediate to urgent). In the decision curve analysis, all machine learning models consistently achieved a greater net benefit—a larger number of appropriate triages considering a trade-off with over-triages—across the range of clinical thresholds.

Conclusions

Compared to the conventional approach, the machine learning models demonstrated a superior performance to predict critical care and hospitalization outcomes. The application of modern machine learning models may enhance clinicians’ triage decision making, thereby achieving better clinical care and optimal resource utilization.

Electronic supplementary material

The online version of this article (10.1186/s13054-019-2351-7) contains supplementary material, which is available to authorized users.

Machine Learning-Based Prediction of Clinical Outcomes for Children During Emergency Department Triage

IMPORTANCE

While machine learning approaches may enhance prediction ability, little is known about their utility in emergency department (ED) triage.

OBJECTIVES

To examine the performance of machine learning approaches to predict clinical outcomes and disposition in children in the ED and to compare their performance with conventional triage approaches.

DESIGN, SETTING, AND PARTICIPANTS

Prognostic study of ED data from the National Hospital Ambulatory Medical Care Survey from January 1, 2007, through December 31, 2015. A nationally representative sample of 52 037 children aged 18 years or younger who presented to the ED were included. Data analysis was performed in August 2018.

MAIN OUTCOMES AND MEASURES

The outcomes were critical care (admission to an intensive care unit and/or in-hospital death) and hospitalization (direct hospital admission or transfer). In the training set (70% random sample), using routinely available triage data as predictors (eg, demographic characteristics and vital signs), we derived 4 machine learning-based models: lasso regression, random forest, gradient-boosted decision tree, and deep neural network. In the test set (the remaining 30% of the sample), we measured the models' prediction performance by computing C statistics, prospective prediction results, and decision curves. These machine learning models were built for each outcome and compared with the reference model using the conventional triage classification information.

RESULTS

Of 52 037 eligible ED visits by children (median [interquartile range] age, 6 [2-14] years; 24 929 [48.0%] female), 163 (0.3%) had the critical care outcome and 2352 (4.5%) had the hospitalization outcome. For the critical care prediction, all machine learning approaches had higher discriminative ability compared with the reference model, although the difference was not statistically significant (eg, C statistics of 0.85 [95% CI, 0.78-0.92] for the deep neural network vs 0.78 [95% CI, 0.71-0.85] for the reference; P = .16), and lower number of undertriaged critically ill children in the conventional triage levels 3 to 5 (urgent to nonurgent). For the hospitalization prediction, all machine learning approaches had significantly higher discrimination ability (eg, C statistic, 0.80 [95% CI, 0.78-0.81] for the deep neural network vs 0.73 [95% CI, 0.71-0.75] for the reference; P < .001) and fewer overtriaged children who did not require inpatient management in the conventional triage levels 1 to 3 (immediate to urgent). The decision curve analysis demonstrated a greater net benefit of machine learning models over ranges of clinical thresholds.

CONCLUSIONS AND RELEVANCE

Machine learning-based triage had better discrimination ability to predict clinical outcomes and disposition, with reduction in undertriaging critically ill children and overtriaging children who are less ill.

Emergency department models of care in Queensland: a multisite cross-sectional study

Objective The acuity and number of presentations being made to emergency departments (EDs) is increasing. In an effort to safely and efficiently manage this increase and optimise patient outcomes, innovative models of care (MOC) have been implemented. What is not clear is how these MOC reflect the needs of patients or relate to each other or to ED performance. The aim of this study was to describe ED MOC in Queensland, Australia. Methods Situated within a larger mixed-methods study, the present study was a cross-sectional study. In early 2015, leaders (medical directors and nurse managers) from public hospital EDs in Queensland were invited to complete a survey detailing ED activity, staffing profiles, treatment space, MOC and National Emergency Access Target (NEAT) performance. Routinely collected ED information system data was also used. Results Twenty of the 27 EDs invited participated in the study (response rate 74%). An extensive array of MOC were identified that were categorised into those that facilitate input, throughput and output from the ED. There was no consistent evidence as to the relative effectiveness of these MOC in achieving ED performance benchmarks, such as NEAT performance. Conclusion There is considerable variability in the MOC used throughout EDs in Queensland. A more complete analysis of the relative effectiveness of different MOC either in isolation or as part of a comprehensive approach would help inform more consistent MOC in Queensland EDs. What is known about the topic? MOC in any given ED are implemented in response to factors such as the geographical location of the hospital, hospital-specific characteristics and service profile, staffing profile and patient demographic profile. In the era of time-based targets, they may also serve to address a particular aspect of flow in the face of rising ED demand. Although many of the MOC attempt to deal with flow in a linear fashion, target specific phases of the ED journey or address particular patient cohorts, what is clear is that not all EDs are shaped and formed the same. What does this paper add? The study provides a comprehensive description of the varied models of care operating within Queensland public hospital EDs and how they relate to ED performance. A basic taxonomy of contemporary ED MOC is necessary to allow comparison between departments and inform decisions regarding safety, efficiency and cost-effectiveness. What are the implications to practitioners? A contemporary understanding of the presence and profile of ED MOC that currently exist within a network of hospitals and health services is important for managers, clinicians and patients to inform decision-making regarding the safety, clinical effectiveness and cost-effectiveness of these models. This understanding can also inform where and how further improvements in care delivery can progress.

Emergency department crowding: A systematic review of causes, consequences and solutions

BACKGROUND

Emergency department crowding is a major global healthcare issue. There is much debate as to the causes of the phenomenon, leading to difficulties in developing successful, targeted solutions.

AIM

The aim of this systematic review was to critically analyse and summarise the findings of peer-reviewed research studies investigating the causes and consequences of, and solutions to, emergency department crowding.

METHOD

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed. A structured search of four databases (Medline, CINAHL, EMBASE and Web of Science) was undertaken to identify peer-reviewed research publications aimed at investigating the causes or consequences of, or solutions to, emergency department crowding, published between January 2000 and June 2018. Two reviewers used validated critical appraisal tools to independently assess the quality of the studies. The study protocol was registered with the International prospective register of systematic reviews (PROSPERO 2017: CRD42017073439).

RESULTS

From 4,131 identified studies and 162 full text reviews, 102 studies met the inclusion criteria. The majority were retrospective cohort studies, with the greatest proportion (51%) trialling or modelling potential solutions to emergency department crowding. Fourteen studies examined causes and 40 investigated consequences. Two studies looked at both causes and consequences, and two investigated causes and solutions.

CONCLUSIONS

The negative consequences of ED crowding are well established, including poorer patient outcomes and the inability of staff to adhere to guideline-recommended treatment. This review identified a mismatch between causes and solutions. The majority of identified causes related to the number and type of people attending ED and timely discharge from ED, while reported solutions focused on efficient patient flow within the ED. Solutions aimed at the introduction of whole-of-system initiatives to meet timed patient disposition targets, as well as extended hours of primary care, demonstrated promising outcomes. While the review identified increased presentations by the elderly with complex and chronic conditions as an emerging and widespread driver of crowding, more research is required to isolate the precise local factors leading to ED crowding, with system-wide solutions tailored to address identified causes.

NHAMCS Validation of Emergency Severity Index as an Indicator of Emergency Department Resource Utilization

Introduction

Triage systems play a vital role in emergency department (ED) operations and can determine how well a given ED serves its local population. We sought to describe ED utilization patterns for different triage levels using the National Hospital Ambulatory Medical Care Survey (NHAMCS) database.

Methods

We conducted a multi-year secondary analysis of the NHAMCS database from 2009-2011. National visit estimates were made using standard methods in Analytics Software and Solutions (SAS, Cary, NC). We compared patients in the mid-urgency range in regard to ED lengths of stay, hospital admission rates, and numbers of tests and procedures in comparison to lower or higher acuity levels.

Results

We analyzed 100,962 emergency visits (representing 402,211,907 emergency visits nationwide). In 2011, patients classified as triage levels 1–3 had a higher number of diagnoses (5.5, 5.6 and 4.2, respectively) when compared to those classified as levels 4 and 5 (1.61 and 1.25). This group also underwent a higher number of procedures (1.0, 0.8 and 0.7, versus 0.4 and 0.4), had a higher ED length of stay (220, 280 and 237, vs. 157 and 135), and admission rates (32.2%, 32.3% and 15.5%, vs. 3.1% and 3.6%).

Conclusion

Patients classified as mid-level (3) triage urgency require more resources and have higher indicators of acuity as those in triage levels 4 and 5. These patients’ indicators are more similar to those classified as triage levels 1 and 2.

Predicting hospital admission at emergency department triage using machine learning

OBJECTIVE

To predict hospital admission at the time of ED triage using patient history in addition to information collected at triage.

METHODS

This retrospective study included all adult ED visits between March 2014 and July 2017 from one academic and two community emergency rooms that resulted in either admission or discharge. A total of 972 variables were extracted per patient visit. Samples were randomly partitioned into training (80%), validation (10%), and test (10%) sets. We trained a series of nine binary classifiers using logistic regression (LR), gradient boosting (XGBoost), and deep neural networks (DNN) on three dataset types: one using only triage information, one using only patient history, and one using the full set of variables. Next, we tested the potential benefit of additional training samples by training models on increasing fractions of our data. Lastly, variables of importance were identified using information gain as a metric to create a low-dimensional model.

RESULTS

A total of 560,486 patient visits were included in the study, with an overall admission risk of 29.7%. Models trained on triage information yielded a test AUC of 0.87 for LR (95% CI 0.86-0.87), 0.87 for XGBoost (95% CI 0.87-0.88) and 0.87 for DNN (95% CI 0.87-0.88). Models trained on patient history yielded an AUC of 0.86 for LR (95% CI 0.86-0.87), 0.87 for XGBoost (95% CI 0.87-0.87) and 0.87 for DNN (95% CI 0.87-0.88). Models trained on the full set of variables yielded an AUC of 0.91 for LR (95% CI 0.91-0.91), 0.92 for XGBoost (95% CI 0.92-0.93) and 0.92 for DNN (95% CI 0.92-0.92). All algorithms reached maximum performance at 50% of the training set or less. A low-dimensional XGBoost model built on ESI level, outpatient medication counts, demographics, and hospital usage statistics yielded an AUC of 0.91 (95% CI 0.91-0.91).

CONCLUSION

Machine learning can robustly predict hospital admission using triage information and patient history. The addition of historical information improves predictive performance significantly compared to using triage information alone, highlighting the need to incorporate these variables into prediction models.

Machine learning approaches for predicting disposition of asthma and COPD exacerbations in the ED

OBJECTIVE

The prediction of emergency department (ED) disposition at triage remains challenging. Machine learning approaches may enhance prediction. We compared the performance of several machine learning approaches for predicting two clinical outcomes (critical care and hospitalization) among ED patients with asthma or COPD exacerbation.

METHODS

Using the 2007-2015 National Hospital and Ambulatory Medical Care Survey (NHAMCS) ED data, we identified adults with asthma or COPD exacerbation. In the training set (70% random sample), using routinely-available triage data as predictors (e.g., demographics, arrival mode, vital signs, chief complaint, comorbidities), we derived four machine learning-based models: Lasso regression, random forest, boosting, and deep neural network. In the test set (the remaining 30% of sample), we compared their prediction ability against traditional logistic regression with Emergency Severity Index (ESI, reference model).

RESULTS

Of 3206 eligible ED visits, corresponding to weighted estimates of 13.9 million visits, 4% had critical care outcome and 26% had hospitalization outcome. For the critical care prediction, the best performing approach- boosting - achieved the highest discriminative ability (C-statistics 0.80 vs. 0.68), reclassification improvement (net reclassification improvement [NRI] 53%, P = 0.002), and sensitivity (0.79 vs. 0.53) over the reference model. For the hospitalization prediction, random forest provided the highest discriminative ability (C-statistics 0.83 vs. 0.64) reclassification improvement (NRI 92%, P < 0.001), and sensitivity (0.75 vs. 0.33). Results were generally consistent across the asthma and COPD subgroups.

CONCLUSIONS

Based on nationally-representative ED data, machine learning approaches improved the ability to predict disposition of patients with asthma or COPD exacerbation.

Grow the Pie: Interdepartmental Cooperation as a Method for Achieving Operational Efficiency in an Emergency Department

BACKGROUND

Despite sufficient literature analyzing macroscopic and microscopic methods of addressing emergency department (ED) operations, there is a paucity of studies that analyze methods between these extremes.

OBJECTIVE

We conducted a quasi-experimental study incorporating a pre/post-intervention comparison to determine whether interdepartmental cooperation is effective at improving ED operations by combining microscopic and macroscopic concepts.

METHODS

We performed an analysis of operational and financial data from a cooperative investment in imaging transport personnel between the emergency and radiology departments. Our primary outcome, order to table time (OTT), measured imaging times by modality (computed tomography [CT], ultrasound [US], magnetic resonance imaging [MRI]). These were compared for statistically significant change before and after the intervention. Our secondary outcome, gross profit, was calculated using the revenue generated from gained outpatient studies minus the associated direct personnel costs.

RESULTS

Transporters improved OTTs by decreasing median imaging times from 132 min to 116 min (p < 0.0005). Efficiency improved for CT scans with median time decreasing from 142 min to 114 min (p < 0.0005). Transport hires had adverse effects on US, with an increase in median OTT from 91 min to 99 min (p < 0.018). MRI experienced a similar trend in OTT, as median times worsened from 215 min to 235 min (p < 0.225). The investment in transporters generated a gross profit of $1.03 million for the radiology department over 9 months.

CONCLUSIONS

Interdepartmental cooperation is a broadly applicable macroscopic method that is effective at achieving microscopic, site-specific gains in ED efficiency. Transporters provided operational gains for the ED and financial gains for the radiology department.

An Evolutionary Computation Approach for Optimizing Multilevel Data to Predict Patient Outcomes

Widespread adoption of electronic health records (EHR) and objectives for meaningful use have increased opportunities for data-driven predictive applications in healthcare. These decision support applications are often fueled by large-scale, heterogeneous, and multilevel (i.e., defined at hierarchical levels of specificity) patient data that challenge the development of predictive models. Our objective is to develop and evaluate an approach for optimally specifying multilevel patient data for prediction problems. We present a general evolutionary computational framework to optimally specify multilevel data to predict individual patient outcomes. We evaluate this method for both flattening (single level) and retaining the hierarchical predictor structure (multiple levels) using data collected to predict critical outcomes for emergency department patients across five populations. We find that the performance of both the flattened and hierarchical predictor structures in predicting critical outcomes for emergency department patients improve upon the baseline models for which only a single level of predictor—either more general or more specific—is used (p < 0.001). Our framework for optimizing the specificity of multilevel data improves upon more traditional single-level predictor structures and can readily be adapted to similar problems in healthcare and other domains.

Two Hour Evaluation and Referral Model for Shorter Turnaround Times in the emergency department

OBJECTIVE

The objective of this study was to assess the implementation of a novel ED model of care, which combines clinical streaming, team-based assessment and early senior consultation to reduce length of stay.

METHODS

A pre-post-intervention study was used to compare ED performance following an extensive clinical redesign programme. Clinical teams and work sequences were reconfigured to promote the role of the staff specialist, with a focus on earlier decisions regarding disposition. Primary outcome measures were ED length of stay and National Emergency Access Target (NEAT) compliance. Secondary outcomes included referral and workup times, wait times by triage category, ambulance offload times, ward discharges and unit transfers within 24 h of admission, representation within 48 h, and Medical Emergency Response Team (MERT) calls within 24 h of admission.

RESULTS

Two seasonally matched 26 week intervals were compared with adjustment for demographics, triage category and arrival by ambulance. Overall, there was an 18.4% rise in NEAT performance (95% confidence interval (CI): 17.7-19.1) while ED length of stay decreased by a total of 86.8 min (95% CI: 83.6-90.1). Time series analysis did not suggest any preexisting trends to explain these results. The average time to referral decreased by 74.7 min (95% CI: 69.8-79.6) and waiting times decreased across all triage categories. Rates of MERT activation and unplanned representation were unchanged.

CONCLUSION

A facilitated team leader role for senior doctors can help to reduce length of stay by via early disposition, without significant risks to the patient.

Using Lean to Rapidly and Sustainably Transform a Behavioral Health Crisis Program: Impact on Throughput and Safety

BACKGROUND

Lean has been increasingly applied in health care to reduce waste and improve quality, particularly in fast-paced and high-acuity clinical settings such as emergency departments. In addition, Lean's focus on engagement of frontline staff in problem solving can be a catalyst for organizational change. In this study, ConnectionsAZ demonstrates how they applied Lean principles to rapidly and sustainably transform clinical operations in a behavioral health crisis facility.

METHODS

A multidisciplinary team of management and frontline staff defined values-based outcome measures, mapped the current and ideal processes, and developed new processes to achieve the ideal. Phase I was implemented within three months of assuming management of the facility and involved a redesign of flow, space utilization, and clinical protocols. Phase II was implemented three months later and improved the provider staffing model. Organizational changes such as the development of shift leads and daily huddles were implemented to sustain change and create an environment supportive of future improvements.

RESULTS

Post-Phase I, there were significant decreases (pre vs. post and one-year post) in median door-to-door dwell time (343 min vs. 118 and 99), calls to security for behavioral emergencies (13.5 per month vs. 4.3 and 4.8), and staff injuries (3.3 per month vs. 1.2 and 1.2). Post-Phase II, there were decreases in median door-to-doctor time (8.2 hours vs. 1.6 and 1.4) and hours on diversion (90% vs. 17% and 34%).

CONCLUSIONS

Lean methods can positively affect safety and throughput and are complementary to patient-centered clinical goals in a behavioral health setting.

Increased door to admission time is associated with prolonged throughput for ED patients discharged home

BACKGROUND

Emergency Department (ED) service evaluations are typically based on surveys of discharged patients. Physicians/administrators benefit from data that quantifies system-based factors that adversely impact the experience of those who represent the survey cohort.

OBJECTIVE

While investigators have established that admitted patient boarding impacts overall ED throughput times, we sought to specifically quantify the relationship between throughput times for patients admitted (EDLOS) versus discharged home from the ED (DCLOS).

METHODS

We performed a prospective analysis of consecutive patient encounters at an inner-city ED. Variables collected: median daily DCLOS for ED patients, ED daily census, left without being seen (LWBS), median door to doctor, median room to doctor, and daily number admitted. Admitted patients divided into 2 groups based on daily median EDLOS for admits (<6 hours, ≥6 hours). Continuous variables analyzed by t-tests. Multivariate regression utilized to identify independent effects of the co-variants on median daily DCLOS.

RESULTS

We analyzed 24,127 patient visits. ED patient DCLOS was longer for patients seen on days with prolonged EDLOS (193.7 minutes, 95%CI 186.7-200.7 vs. 152.8, 144.9-160.5, P< .0001). Variables that were associated with increased daily median EDLOS for admits included: daily admits (P= 0.01), room to doctor time (P< .01), number of patients that left without being seen (P< .01). When controlling for the covariate daily census, differences in DCLOS remained significant for the ≥6 hours group (189.4 minutes, 95%CI 184.1-194.7 vs. 164.8, 155.7-173.9 (P< .0001).

CONCLUSION

Prolonged ED stays for admitted patients were associated with prolonged throughput times for patients discharged home from the ED.

Are Split Flow and Provider in Triage Models in the Emergency Department Effective in Reducing Discharge Length of Stay?

A quality improvement (QI) project was completed early in 2015 to evaluate the split flow model of care delivery and a provider in triage model within a newly constructed emergency department. The QI project compared 2 emergency departments of similar volumes, one that splits the patient flow and employs a provider in triage model and the other that blends the patient flow and employs a traditional nurse triage model. A total of 68,603 patients were included in this project. The purpose of the split flow model is to create a second flow stream of patients through the emergency department, parallel to the regular acute/critical care flow stream, for patients with problems that are not complex. Specific patient outcomes that were evaluated for the purpose of this QI project were door to discharge or discharge length of stay (DLOS) for all ED patients. The provider in triage model enhances patient triage assessment, as well as patient flow within the emergency department, by allowing patients to be evaluated by an ED provider immediately at the point of triage when the patient first presents to the emergency department. The QI project demonstrated that the split flow model alone reduced DLOS for all ED patients, and when coupled with the provider in triage model, a greater reduction in DLOS, as well as an improvement in front-end throughput metrics, was realized.