Dataset Information

Predicting emergency department orders with multilabel machine learning techniques and simulating effects on length of stay.

ABSTRACT: OBJECTIVE:Emergency departments (EDs) continue to pursue optimal patient flow without sacrificing quality of care. The speed with which a healthcare provider receives pertinent information, such as results from clinical orders, can impact flow. We seek to determine if clinical ordering behavior can be predicted at triage during an ED visit. MATERIALS AND METHODS:Using data available during triage, we trained multilabel machine learning classifiers to predict clinical orders placed during an ED visit. We benchmarked 4 classifiers with 2 multilabel learning frameworks that predict orders independently (binary relevance) or simultaneously (random k-labelsets). We evaluated algorithm performance, calculated variable importance, and conducted a simple simulation study to examine the effects of algorithm implementation on length of stay and cost. RESULTS:Aggregate performance across orders was highest when predicting orders independently with a multilayer perceptron (median F1 score = 0.56), but prediction frameworks that simultaneously predict orders for a visit enhanced predictive performance for correlated orders. Visit acuity was the most important predictor for most orders. Simulation results indicated that direct implementation of the model would increase ordering costs (from $21 to $45 per visit) but reduce length of stay (from 158 minutes to 151 minutes) over all visits. DISCUSSION:Simulated implementations of the predictive algorithm decreased length of stay but increased ordering costs. Optimal implementation of these predictions to reduce patient length of stay without incurring additional costs requires more exploration. CONCLUSIONS:It is possible to predict common clinical orders placed during an ED visit with data available at triage.

SUBMITTER: Hunter-Zinck HS

PROVIDER: S-EPMC7647143 | biostudies-literature | 2019 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Predicting emergency department orders with multilabel machine learning techniques and simulating effects on length of stay.

Hunter-Zinck Haley S HS Peck Jordan S JS Strout Tania D TD Gaehde Stephan A SA

Journal of the American Medical Informatics Association : JAMIA 20191201 12

<h4>Objective</h4>Emergency departments (EDs) continue to pursue optimal patient flow without sacrificing quality of care. The speed with which a healthcare provider receives pertinent information, such as results from clinical orders, can impact flow. We seek to determine if clinical ordering behavior can be predicted at triage during an ED visit.<h4>Materials and methods</h4>Using data available during triage, we trained multilabel machine learning classifiers to predict clinical orders placed ...[more]

PMID: 31578568

Similar Datasets

Project description:Performance measures, particularly pay for performance, may have unintended consequences for safety-net institutions caring for disproportionate shares of Medicaid or uninsured patients.To describe emergency department (ED) compliance with proposed length-of-stay measures for admissions (8 hours or 480 minutes) and discharges, transfers, and observations (4 hours or 240 minutes) by safety-net status.The 2008 National Hospital Ambulatory Medical Care Survey (NHAMCS) ED data were stratified by safety-net status (Centers for Disease Control and Prevention definition) and disposition (admission, discharge, observation, transfer). The 2008 NHAMCS is a national probability sample of 396 hospitals (90.2% unweighted response rate) and 34 134 patient records. Visits were excluded for patients younger than 18 years, missing length-of-stay data or dispositions of missing, other, left against medical advice, or dead on arrival. Median and 90th percentile ED lengths of stay were calculated for each disposition and admission/discharge subcategories (critical care, psychiatric, routine) stratified by safety-net status. Multivariable analyses determined associations with length-of-stay measure compliance.Emergency Department length-of-stay measure compliance by disposition and safety-net status.Of the 72.1% ED visits (N = 24 719) included in the analysis, 42.3% were to safety-net EDs and 57.7% were to non-safety-net EDs. The median length of stay for safety-net was 269 minutes (interquartile range [IQR], 178-397 minutes) for admission vs 281 minutes (IQR, 178-401 minutes) for non-safety-net EDs; 156 minutes (IQR, 95-239 minutes) for discharge vs 148 minutes (IQR, 88-238 minutes); 355 minutes (IQR, 221-675 minutes) for observations vs 298 minutes (IQR, 195-440 minutes); and 235 minutes (IQR, 155-378 minutes) for transfers vs 239 minutes (IQR, 142-368 minutes). Safety-net status was not independently associated with compliance with ED length-of-stay measures; the odds ratio was 0.83 for admissions (95% CI, 0.52-1.34); 1.03 for discharges (95% CI, 0.83-1.27); 1.05 for observations (95% CI, 0.57-1.95), 1.30 for transfers (95% CI, 0.70-2.45]); or subcategories except for psychiatric discharges (1.67, [95% CI, 1.02-2.74]).Compliance with proposed ED length-of-stay measures for admissions, discharges, transfers, and observations did not differ significantly between safety-net and non-safety-net hospitals.

Project description:Emergency Departments (ED) are trying to alleviate crowding using various interventions. We assessed the effect of an alternative model of care, the Medical Team Evaluation (MTE) concept, encompassing team triage, quick registration, redesign of triage rooms and electronic medical records (EMR) on door-to-doctor (waiting) time and ED length of stay (LOS). We conducted an observational, before-and-after study at an urban academic tertiary care centre. On July 17th 2014, MTE was initiated from 9:00 a.m. to 10 p.m., 7 days a week. A registered triage nurse was teamed with an additional senior ED physician. Data of the 5-month pre-MTE and the 5-month MTE period were analysed. A matched comparison of waiting times and ED LOS of discharged and admitted patients pertaining to various Emergency Severity Index (ESI) triage categories was performed based on propensity scores. With MTE, the median waiting times improved from 41.2 (24.8-66.6) to 10.2 (5.7-18.1) minutes (min; P < 0.01). Though being beneficial for all strata, the improvement was somewhat greater for discharged, than for admitted patients. With a reduction from 54.3 (34.2-84.7) to 10.5 (5.9-18.4) min (P < 0.01), in terms of waiting times, MTE was most advantageous for ESI4 patients. The overall median ED LOS increased for about 15 min (P < 0.01), increasing from 3.4 (2.1-5.3) to 3.7 (2.3-5.6) hours. A significant increase was observed for all the strata, except for ESI5 patients. Their median ED LOS dropped by 73% from 1.2 (0.8-1.8) to 0.3 (0.2-0.5) hours (P < 0.01). In the same period the total orders for diagnostic radiology increased by 1,178 (11%) from 10,924 to 12,102 orders, with more imaging tests being ordered for ESI 2, 3 and 4 patients. Despite improved waiting times a decrease of ED LOS was only seen in ESI level 5 patients, whereas in all the other strata ED LOS increased. We speculate that this was brought about by the tendency of triage physicians to order more diagnostic radiology, anticipating that it may be better for the downstream physician to have more information rather than less.

Project description:BackgroundA prolonged length of stay in an emergency department is related to lower quality of care and adverse outcomes, which are often linked with overcrowding.ObjectiveExamine the influence of demographic factors on prolonged length of stay in the emergency department.MethodsThis study used a cross-sectional design. It used secondary data for all patients admitted during the specific duration at the emergency department of a governmental hospital in Saudi Arabia. The independent variables were gender, age, disposition status, shift time, and clinical acuity (CTAS) level while the dependent variable was prolonged length of stay.ResultsThe study shows that 30% of patients stay at the emergency department for four hours or more. The results also show a significant association between demographic factors which are age, gender, disposition status, shift time, clinical acuity (CTAS) level and prolonged length of stay in an emergency department. Based on the results males are more likely to stay in the emergency department than females (OR = 1.20; 95% CI = 1.04 to 1.38). Patients aged 60 and older are less likely to stay in the emergency department than patients aged 29 or smaller (OR = 0.58; 95% CI = 0.39 to 0.84). According to disposition status discharged patients after examination stays in the emergency department more than admitted patients after the examination (OR = 2.78; 95% CI = 1.67 to 4.99). Patients who come to the night shift are less likely to stay in the emergency department than patients who come in the morning shift (OR = 0.67; 95% CI = 0.56 to 0.81). Patients who are classified in level three of CTAS are less likely to stay in the emergency department than patients who are classified in level one (OR = 0.28; 95% CI = 0.88 to 0.023).ConclusionDemographic factors such as age, gender, shift time, disposition status and clinical acuity (CTAS) were important factors that needed to be considered to reduce the length of stay of patients in the emergency department. it is possible to formulate a machine learning model to predict the anticipated length of stay in the hospital for each patient. This prediction with an accepted margin of uncertainty will help the clinicians to communicate the evidence-based anticipated length of stay with the patient's caregivers. In addition, hospital managers need to provide the emergency department with enough staff and materials to reduce the length of stay of patients.

Project description:ObjectiveTo predict hospital admission at the time of ED triage using patient history in addition to information collected at triage.MethodsThis 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.ResultsA 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).ConclusionMachine 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.

Dataset Information

Predicting emergency department orders with multilabel machine learning techniques and simulating effects on length of stay.

Publications

Predicting emergency department orders with multilabel machine learning techniques and simulating effects on length of stay.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets