Predicting outcomes in older ED patients with influenza in real time using a big data-driven and machine learning approach to the hospital information system

Abstract

Background Predicting outcomes in older patients with influenza in the emergency department (ED) by machine learning (ML) has never been implemented. Therefore, we conducted this study to clarify the clinical utility of implementing ML. Methods We recruited 5508 older ED patients (≥65 years old) in three hospitals between 2009 and 2018. Patients were randomized into a 70%/30% split for model training and testing. Using 10 clinical variables from their electronic health records, a prediction model using the synthetic minority oversampling technique preprocessing algorithm was constructed to predict five outcomes. Results The best areas under the curves of predicting outcomes were: random forest model for hospitalization (0.840), pneumonia (0.765), and sepsis or septic shock (0.857), XGBoost for intensive care unit admission (0.902), and logistic regression for in-hospital mortality (0.889) in the testing data. The predictive model was further applied in the hospital information system to assist physicians’ decisions in real time. Conclusions ML is a promising way to assist physicians in predicting outcomes in older ED patients with influenza in real time. Evaluations of the effectiveness and impact are needed in the future.

Use of Artificial Intelligence to Manage Patient Flow in Emergency Department during the COVID-19 Pandemic: A Prospective, Single-Center Study

“…Hong et al chose their threshold and compared their models by setting the specificity to 85%. Tan et al [ 26 ] built a model that predicted the patient’s outcome on the basis of the triage data, the medical history, and the laboratory data. The highest AUROC for admission prediction was 84%, using a random forest model.…”

Design and Implementation of a Comprehensive AI Dashboard for Real-Time Prediction of Adverse Prognosis of ED Patients

“…Each type of patient has several outcome prediction models. For instance, the AI for chest pain has two prediction models (AMI, death) [ 10 ]; the AI for influenza in the elderly has five prediction models (hospitalization, pneumonia, sepsis/septic shock, ICU admission, and death) [ 8 ]; the AI for pneumonia has three prediction models (respiratory failure, sepsis/septic shock, and death) [ 9 ]; and the AI for brain trauma has three prediction models (ICU admission, hospitalization, and death) [ 7 ]. All the AI prediction models and their launch times are shown in Table 1 .…”

“…For example, when a patient has just been admitted to the ED and is coded as elderly with influenza by the physician (ICD9 codes as 487 or 488), the dashboard will automatically capture the patients’ 10 feature values of influenza in the elderly models (tachypnea, GCS, history of hypertension, history of CAD, history of malignancy, bedridden, leukocytosis, bacteremia, anemia, and elevated CRP) from the HIS to calculate the risk probabilities of the five outcomes (hospitalization, pneumonia, sepsis or septic shock, ICU admission, and in-hospital death) [ 8 ].…”

“…Therefore, focusing on practicalities, the ED in Chi Mei Medical Center (CMMC) has developed and implemented eight artificial intelligence (AI) disease predictive systems and integrated them into the HIS, including influenza in old age, chest pain, pancreatitis, hyperglycemic crisis, dengue, pneumonia, brain trauma, and fever. These predictive systems have been investigated and implemented, and the relevant results were published [ 7 , 8 , 9 , 10 ].…”

Real‐time artificial intelligence predicts adverse outcomes in acute pancreatitis in the emergency department: Comparison with clinical decision rule

“…Artificial intelligence (AI) has seen widespread application in the ED for outcome prediction in various diseases in recent years. [10][11][12][13] Compared to conventional CDRs, AI-based predictions are often more precise and user-friendly, particularly in the fast-paced environment of the ED. [10][11][12][13] Despite AI being explored for improving diagnosis, treatment, and severity prediction in AP for the past two decades, 14…”