CLASSIFICATION OF CHIEF-COMPLAINTS FOR PATIENTS REPORTING AT EMERGENCY DEPARTMENTS
ABSTRACT: Timely disease detection in case of Infectious Diseases (IDs) can provide sufficient time for disease control activities that might help in controlling the spread of disease to a great extent. Generally, medical diagnostic procedures are time consuming and may take about one to two weeks in case of diagnosis by lab reports; in the meantime, IDs spread may go epidemic. Therefore, it is required to process the complaints presented by patients’ well in time for early warning of ID spread. In this study, we have designed and developed a syndromic classifier, for automated processing of chief-complaints data for patients reporting at emergency departments to corresponding syndromes. METHODS: We have employed Artificial Neural-Networks (ANN) for classification of patients’ chiefcomplaints data into syndromic categories. Network is trained and tested using diagnosed chief-complaints data from leading hospitals of Lahore city. RESULTS: Trained network was able to assign respiratory, gastro-intestinal, hemorrhagic, rash, fever, neurological, shock syndrome with sensitivity 98.1%, 98.9%, 85.4%, 97.7%, 99.1%, 99% and 77.3% respectively. No case was reported for botulinic syndrome. Comparable results were achieved by using International Classification of Disease (ICD-10) codes for classification. CONCLUSION: Using automated syndromic surveillance IDs may be detected two to three weeks in advance as compared to diagnosis based reporting systems. High classification accuracy of the syndromic classifier provides us with the ability to timely detect imminent ID outbreak by processing chiefcomplaints information. Classification results allow monitoring of ID spread and optimum time for disease control activities to mitigate the emerging ID epidemics.
This work is licensed under a Creative Commons Attribution 4.0 International License.