Using Data Mining to Differentiate Dengue with Warning Signs from Severe Dengue: A Predictive Model from Oaxaca,...

Area of Science:

• Medical Informatics
• Public Health
• Epidemiology

Background:

• Dengue with Warning Signs (DWS) and severe dengue pose significant global health challenges.
• Accurate differentiation between dengue clinical forms is crucial but often complex.
• Resource-limited settings face particular difficulties in timely diagnosis and management.

Purpose of the Study:

• To enhance diagnostic accuracy for clinical dengue (DWS and severe dengue) using data mining.
• To optimize clinical decision-making and patient management for dengue.
• To identify key predictors for differentiating DWS from severe dengue.

Main Methods:

• Analysis of a dataset comprising 1,341 patients from Oaxaca City, México.
• Training and testing various machine learning algorithms, including linear support vector machine.
• Utilizing performance metrics like precision, recall, F1 score, AUC, and Cohen kappa.
• Employing Shapley values to determine predictor significance.

Main Results:

• The linear support vector machine model achieved precision rates of 62.4% for DWS and 82.9% for severe dengue.
• Other performance metrics included recall/sensitivity (90.8%), F1 score (73.8%), and AUC (71.3%).
• Hematemesis, ascites, and edema were identified as the most significant predictors for differentiating DWS from severe dengue.

Conclusions:

• Data mining and machine learning techniques show strong potential for improving dengue differential diagnosis.
• The models offer valuable tools for clinical decision-making, especially in resource-limited areas.
• Identifying critical risk factors can optimize patient management and resource allocation during dengue outbreaks.