Dataset Information

Creating an automated trigger for sepsis clinical decision support at emergency department triage using machine learning.

ABSTRACT:

Objective

To demonstrate the incremental benefit of using free text data in addition to vital sign and demographic data to identify patients with suspected infection in the emergency department.

Methods

This was a retrospective, observational cohort study performed at a tertiary academic teaching hospital. All consecutive ED patient visits between 12/17/08 and 2/17/13 were included. No patients were excluded. The primary outcome measure was infection diagnosed in the emergency department defined as a patient having an infection related ED ICD-9-CM discharge diagnosis. Patients were randomly allocated to train (64%), validate (20%), and test (16%) data sets. After preprocessing the free text using bigram and negation detection, we built four models to predict infection, incrementally adding vital signs, chief complaint, and free text nursing assessment. We used two different methods to represent free text: a bag of words model and a topic model. We then used a support vector machine to build the prediction model. We calculated the area under the receiver operating characteristic curve to compare the discriminatory power of each model.

Results

A total of 230,936 patient visits were included in the study. Approximately 14% of patients had the primary outcome of diagnosed infection. The area under the ROC curve (AUC) for the vitals model, which used only vital signs and demographic data, was 0.67 for the training data set, 0.67 for the validation data set, and 0.67 (95% CI 0.65-0.69) for the test data set. The AUC for the chief complaint model which also included demographic and vital sign data was 0.84 for the training data set, 0.83 for the validation data set, and 0.83 (95% CI 0.81-0.84) for the test data set. The best performing methods made use of all of the free text. In particular, the AUC for the bag-of-words model was 0.89 for training data set, 0.86 for the validation data set, and 0.86 (95% CI 0.85-0.87) for the test data set. The AUC for the topic model was 0.86 for the training data set, 0.86 for the validation data set, and 0.85 (95% CI 0.84-0.86) for the test data set.

Conclusion

Compared to previous work that only used structured data such as vital signs and demographic information, utilizing free text drastically improves the discriminatory ability (increase in AUC from 0.67 to 0.86) of identifying infection.

SUBMITTER: Horng S

PROVIDER: S-EPMC5383046 | biostudies-literature | 2017

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Creating an automated trigger for sepsis clinical decision support at emergency department triage using machine learning.

Horng Steven S Sontag David A DA Halpern Yoni Y Jernite Yacine Y Shapiro Nathan I NI Nathanson Larry A LA

PloS one 20170406 4

<h4>Objective</h4>To demonstrate the incremental benefit of using free text data in addition to vital sign and demographic data to identify patients with suspected infection in the emergency department.<h4>Methods</h4>This was a retrospective, observational cohort study performed at a tertiary academic teaching hospital. All consecutive ED patient visits between 12/17/08 and 2/17/13 were included. No patients were excluded. The primary outcome measure was infection diagnosed in the emergency dep ...[more]

PMID: 28384212

Similar Datasets

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.

Project description:BackgroundDevelopment 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).MethodsUsing 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.ResultsOf 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.ConclusionsCompared 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.

Project description:Background. While often first treated in the emergency department (ED), identification of sepsis is difficult. Electronic medical record (EMR) clinical decision tools offer a novel strategy for identifying patients with sepsis. The objective of this study was to test the accuracy of an EMR-based, automated sepsis identification system. Methods. We tested an EMR-based sepsis identification tool at a major academic, urban ED with 64,000 annual visits. The EMR system collected vital sign and laboratory test information on all ED patients, triggering a "sepsis alert" for those with ?2 SIRS (systemic inflammatory response syndrome) criteria (fever, tachycardia, tachypnea, leukocytosis) plus ?1 major organ dysfunction (SBP ? 90 mm Hg, lactic acid ?2.0 mg/dL). We confirmed the presence of sepsis through manual review of physician, nursing, and laboratory records. We also reviewed a random selection of ED cases that did not trigger a sepsis alert. We evaluated the diagnostic accuracy of the sepsis identification tool. Results. From January 1 through March 31, 2012, there were 795 automated sepsis alerts. We randomly selected 300 cases without a sepsis alert from the same period. The true prevalence of sepsis was 355/795 (44.7%) among alerts and 0/300 (0%) among non-alerts. The positive predictive value of the sepsis alert was 44.7% (95% CI [41.2-48.2%]). Pneumonia and respiratory infections (38%) and urinary tract infection (32.7%) were the most common infections among the 355 patients with true sepsis (true positives). Among false-positive sepsis alerts, the most common medical conditions were gastrointestinal (26.1%), traumatic (25.7%), and cardiovascular (20.0%) conditions. Rates of hospital admission were: true-positive sepsis alert 91.0%, false-positive alert 83.0%, no sepsis alert 5.7%. Conclusions. This ED EMR-based automated sepsis identification system was able to detect cases with sepsis. Automated EMR-based detection may provide a viable strategy for identifying sepsis in the ED.

Project description:ImportanceWhile machine learning approaches may enhance prediction ability, little is known about their utility in emergency department (ED) triage.ObjectivesTo 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 participantsPrognostic 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 measuresThe 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.ResultsOf 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 relevanceMachine 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.

Project description:We undertook a process improvement initiative to expedite rapid identification of potential sepsis patients based on triage chief complaint, vital signs, and initial lactate level.DesignProspective cohort study.SettingSeven hundred-bed tertiary care hospital with ≅65,000 patient visits/yr.PatientsPatients presenting to emergency department (ED) triage who met the following criteria: greater than or equal to two of the three systemic inflammatory response syndrome criteria assessable in triage, a chief complaint suggestive of infection, emergency severity index 2 or 3, and ambulatory to ED.InterventionsA computer-generated lactate order was created, staff education and resources increased, and point-of-care lactate testing was introduced.Measurements and main resultsPrimary endpoints include the following: percent of patients having a lactate level drawn, percent of lactate samples resulting before room placement, and time intervals from triage to lactate blood draw and to lactate result. Secondary endpoints were percentage of patients admitted to the hospital, percentage admitted to the ICU, and in-hospital mortality. Six thousand nine hundred six patients were included: 226 historic controls (HCs) and 6,680 intervention group patients. The mean serum lactate level was 1.77 ± 1.18 mmol/L. The percentage of patients having a lactate resulted increased from 27.4% in the HC period to 79.6%. The percentage of these lactate results available while the patient was still in the waiting room increased from 0.4% during the HC period to 33.7% during Phase 5 (p < 0.0001). In the intervention period, time from triage to lactate result decreased (78.1-63.4 min; p < 0.0001) and time to treatment room decreased (59.3-39.6 min; p < 0.0001).ConclusionsImplementation of a computerized lactate order using readily available data obtained during ED triage, combined with point-of-care lactate testing, improves time to lactate blood draw and lactate result in patients at risk for severe sepsis. Initial lactate levels correlated with admission to the hospital, admission to the ICU, and in-hospital mortality.

Project description:BackgroundComputerized clinical decision support systems (CDSSs) are increasingly adopted in health care to optimize resources and streamline patient flow. However, they often lack scientific validation against standard medical care.ObjectiveThe purpose of this study was to assess the performance, safety, and usability of a CDSS in a university hospital emergency department setting in Kuopio, Finland.MethodsPatients entering the emergency department were asked to voluntarily participate in this study. Patients aged 17 years or younger, patients with cognitive impairments, and patients who entered the unit in an ambulance or with the need for immediate care were excluded. Patients completed the CDSS web-based form and usability questionnaire when waiting for the triage nurse's evaluation. The CDSS data were anonymized and did not affect the patients' usual evaluation or treatment. Retrospectively, 2 medical doctors evaluated the urgency of each patient's condition by using the triage nurse's information, and urgent and nonurgent groups were created. The International Statistical Classification of Diseases, Tenth Revision diagnoses were collected from the electronic health records. Usability was assessed by using a positive version of the System Usability Scale questionnaire.ResultsIn total, our analyses included 248 patients. Regarding urgency, the mean sensitivities were 85% and 19%, respectively, for urgent and nonurgent cases when assessing the performance of CDSS evaluations in comparison to that of physicians. The mean sensitivities were 85% and 35%, respectively, when comparing the evaluations between the two physicians. Our CDSS did not miss any cases that were evaluated to be emergencies by physicians; thus, all emergency cases evaluated by physicians were evaluated as either urgent cases or emergency cases by the CDSS. In differential diagnosis, the CDSS had an exact match accuracy of 45.5% (97/213). The usability was good, with a mean System Usability Scale score of 78.2 (SD 16.8).ConclusionsIn a university hospital emergency department setting with a large real-world population, our CDSS was found to be equally as sensitive in urgent patient cases as physicians and was found to have an acceptable differential diagnosis accuracy, with good usability. These results suggest that this CDSS can be safely assessed further in a real-world setting. A CDSS could accelerate triage by providing patient-provided data in advance of patients' initial consultations and categorize patient cases as urgent and nonurgent cases upon patients' arrival to the emergency department.

Dataset Information

Creating an automated trigger for sepsis clinical decision support at emergency department triage using machine learning.

Objective

Methods

Results

Conclusion

Publications

Creating an automated trigger for sepsis clinical decision support at emergency department triage using machine learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets