Machine Learning-Based Forecasting of Dengue Severity from Initial Clinical Indicators

Research Article

Machine Learning-Based Forecasting of Dengue Severity from Initial Clinical Indicators

Anika Voss^1*, Elijah Carter¹, Clara Moreau²

¹Department of Viral Pathogenesis, Berlin Institute of Virology and Immunology, Berlin, Germany

²Division of Immunovirology, Melbourne Institute of Infectious Disease Research, Melbourne, Australia

Available online: 23 May 2010

Abstract

Dengue fever remains a significant public health concern, particularly in tropical and subtropical regions. Predicting the severity of dengue in the early stages of infection is crucial for timely clinical management and resource allocation. This study aimed to develop a robust forecast model for dengue severity using readily available clinical parameters at the time of initial presentation. We retrospectively analyzed clinical data from a cohort of laboratory-confirmed dengue patients, employing various machine learning algorithms, including logistic regression, support vector machines, random forests, and gradient boosting. The performance of these models in predicting severe dengue (SD) was evaluated based on accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Our results demonstrate that a gradient boosting model, utilizing a combination of early clinical features such as platelet count, hematocrit, abdominal pain, and bleeding manifestations, achieved high predictive accuracy for SD. This model offers a promising tool for early risk stratification and improved clinical decision-making in dengue management.

Keywords

Dengue Fever; Severity Prediction; Machine Learning; Early Diagnosis