Detection of Patients with Influenza Syndrome Using Machine-Learning Models Learned from Emergency Department Reports

Authors

  • Arturo Lopez Pineda University of Pittsburgh, Department of Biomedical Informatics
  • Fu-Chiang Tsui University of Pittsburgh, Department of Biomedical Informatics
  • Shyam Visweswaran University of Pittsburgh, Department of Biomedical Informatics
  • Gregory F. Cooper University of Pittsburgh, Department of Biomedical Informatics

DOI:

https://doi.org/10.5210/ojphi.v5i1.4446

Abstract

Information available in ED reports has the potential to improve detection of syndromic diseases. Our goal is to provide a machine-learning model characterized by improved predictive accuracy of influenza syndrome. Seven machine-learning algorithms (K2-BN, NB, EBMC, SVM, LR, ANN, RF) for the construction of models were used. Our dataset correspond to 40853 ED cases (67% training, 33% testing). The measurements used were AUROC, calibration and statistical significance testing. The results show high AUROCs with no significant difference between the algorithms and the expert model. EBMC is the most general algorithms.

Author Biography

Arturo Lopez Pineda, University of Pittsburgh, Department of Biomedical Informatics

Arturo Lopez Pineda is a graduate fellow in the Department of Biomedical Informatics at the University of Pittsburgh. Arturo holds an MS in Intelligent Systems and a BS in Computer Science from Tecnologico de Monterrey. His research interest is in the application of Artificial Intelligence in Medicine, with a focus on personalized medicine. Arturo is the recipient of the International S&T Fulbright Award and the CONACyT Scholarship.

Downloads

Published

2013-03-24

How to Cite

Lopez Pineda, A., Tsui, F.-C., Visweswaran, S., & Cooper, G. F. (2013). Detection of Patients with Influenza Syndrome Using Machine-Learning Models Learned from Emergency Department Reports. Online Journal of Public Health Informatics, 5(1). https://doi.org/10.5210/ojphi.v5i1.4446

Issue

Section

Oral Presentations: Influenza Surveillance Methods - Research