Ebola CARE (Computational Assignment of Risk Estimates) predicts mortality risk associated with clinical symptoms in Ebola patients
This app aggregates several Neural Network predictors of mortality risk for Ebola Virus Disease (EVD). These predictors were trained on the only publicly available clinical dataset from the West African EVD outbreak, described by Schieffelin et al. (Schieffelin JS, Shaffer JG, Goba A, et al. Clinical illness and outcomes in patients with Ebola in Sierra Leone. N Engl J Med 2014;371:2092-100).
This dataset includes epidemiological, clinical, and laboratory records of 106 patients treated at Kenema Government Hospital in Sierra Leone during the initial stages of the outbreak, and it is available at http://fathom.info/mirador/ebola/datarelease
The app allows the user to enter clinical chart symptoms, metabolic panel results, and virus load levels (PCR) of an EVD patient, and selects the best predictor for the available data. The variables that can be entered into the app are the following: age, gender, temperature, hearth rate, weakness, vomit, diarrhea, aspartate aminotransferase (AST), alkaline phosphatase (ALK), creatinine (CRE), total carbon dioxide (tCO2), and blood urea nitrogen (BUN).
After entering the patient data, the app runs a Neural Network predictor on it, and outputs a mortality risk estimate, as a number between 0 and 1. Values above 0.5 indicate high risk of death.