ABSTRACT
OBJECTIVE: To apply and compare common machine learning techniques with an expert-built Bayesian Network to determine eligibility for asthma guidelines in pediatric emergency department patients. POPULATION: All patients 2-18 years of age presenting to a pediatric emergency department during a 2-month study period. METHODS: We created an artificial neural network, a support vector machine, a Gaussian process, and a learned Bayesian network to compare each method's ability to detect patients eligible for asthma guidelines. Our outcome measures included the area under the receiver operating characteristic curves, sensitivity, specificity, predictive values, and likelihood ratios. RESULTS: The data were randomly split into a training set (n=3017) and test set (n=1006) for analysis. The systems performed equally well. The area under the receiver operating characteristic curve was 0.959 for the expert-built Bayesian network, 0.962 for the automatically constructed Bayesian network, 0.956 for the Gaussian Process, and 0.937 for the artificial neural network. DISCUSSION: All four evaluated machine learning methods achieved high accuracy. The automatically created Bayesian network performed similarly to the expert-built network. These methods could be applied to create a realtime detection system for identifying asthma patients.
Subject(s)
Artificial Intelligence , Asthma/diagnosis , Decision Support Techniques , Adolescent , Bayes Theorem , Child , Child, Preschool , Decision Support Systems, Clinical , Emergency Service, Hospital , Humans , Likelihood Functions , Neural Networks, Computer , Patient Selection , ROC Curve , Sensitivity and SpecificityABSTRACT
Predicting hospital admission for Emergency Department (ED) patients at the time of triage may improve throughput. To predict admission we created and validated a Bayesian Network from 47,993 encounters (training: n=23,996, validation: n=9,599, test: n=14,398). The area under the receiver operator characteristic curve was 0.833 (0.8260.840) for the network and 0.790 (0.7810.799) for the control variable (acuity only). Predicting hospital admission early during an encounter may help anticipate ED workload and potential overcrowding.
Subject(s)
Neural Networks, Computer , Patient Admission , Triage , Bayes Theorem , Emergency Service, Hospital/organization & administration , Humans , ROC CurveABSTRACT
When the Emergency Department (ED) reaches a critical level of overcrowding, it diverts ambulances to other hospitals. We evaluated the accuracy of a Gaussian process for prediction of ambulance diversion using March 1, 2005 November 30, 2005 data. The area under the receiver operating curve (AUC) for 120 minutes in advance was 0.93 (SE: 0.19). The instrument demonstrated a high AUC and may be used to alert ED managers earlier of a diversion episode.
Subject(s)
Ambulances , Emergency Service, Hospital/organization & administration , Models, Statistical , Patient Transfer/organization & administration , Academic Medical Centers/organization & administration , Bayes Theorem , Bed Occupancy , Humans , Neural Networks, ComputerABSTRACT
INTRODUCTION: The healthcare environment is constantly changing. Probabilistic clinical decision support systems need to recognize and incorporate the changing patterns and adjust the decision model to maintain high levels of accuracy. METHODS: Using data from >75,000 ED patients during a 19-month study period we examined the impact of various static and dynamic training strategies on a decision support system designed to predict hospital admission status for ED patients. Training durations ranged from 1 to 12 weeks. During the study period major institutional changes occurred that affected the system's performance level. RESULTS: The average area under the receiver operating characteristic curve was higher and more stable when longer training periods were used. The system showed higher accuracy when retrained an updated with more recent data as compared to static training period. DISCUSSION: To adjust for temporal trends the accuracy of decision support systems can benefit from longer training periods and retraining with more recent data.
Subject(s)
Bayes Theorem , Decision Support Systems, Clinical , Hospitalization , Neural Networks, Computer , Adult , Artificial Intelligence , Emergency Service, Hospital , Expert Systems , Female , Humans , Male , Probability , ROC CurveABSTRACT
Hospital admission delays in the Emergency Department (ED) reduce capacity and contribute to the ED's diversion problem. We evaluated the accuracy of an Artificial Neural Network for the early prediction of hospital admission using data from 43,077 pediatric ED encounters. We used 9 variables commonly available in the ED setting. The area under the receiver operating characteristic curve was 0.897 (95% CI: 0.887-0.896). The instrument demonstrated high accuracy and may be used to alert clinicians to initiate admission processes earlier during a patient's ED encounter.
Subject(s)
Emergency Service, Hospital , Neural Networks, Computer , Patient Admission , Child , Emergency Service, Hospital/organization & administration , Humans , ROC CurveABSTRACT
Hospital admission delays in the Emergency Department (ED) reduce volume capacity and contribute to the nation's ED diversion problem. This study evaluated the accuracy of a Bayesian network for the early prediction of hospital admission status using data from 16,900 ED encounters. The final model included nine variables that are commonly available in many ED settings. The area under the receiver operating characteristic curve was 0.894 (95% CI: 0.887-0.902) for the validation set. The system had high accuracy an may be used to alert clinicians to initiate admission processes earlier during a patient's ED encounter.