Developing a Decision Support Testing Algorithm to Detect Severity Level of Dengue

Abstract

Dengue is a vector borne disease that has become a global threat. In order to reduce the mortality rate early detection of dengue severity level is crucial. This study is an extension of the decision models developed individually for inflammatory mediators and immune parameters. The objective of this study is to improve the individual models by considering their combined effect and to improve the decision making at 96 hours from onset of illness. In order to combine these, three approaches are attempted including, combining together the individual full models on inflammatory mediators and immune parameters, combining the immune parameters based model with decision tree informed cytokines and implementing a decision tree informed model with immune parameters and inflammatory mediators. The decision tree algorithm that is used in model development is Improved ID3 algorithm. The decision tree based model is a two-step decision system with the initial decision being made using the parameters TNF-α, IL-10, dengue NS1 antigen and dengue IgG antibody and, the operator values above 0.4413, are then subjected to the second test including platelet and Platelet Activating Factor. The decision tree based model performed well with an accuracy of 76.19% and 82.3% of DHF patients were correctly classified. Sensitivity analysis indicated the model to be robust.