Use of structural equation models to predict dengue illness phenotype

Area of Science:

• Infectious Diseases
• Epidemiology
• Biostatistics

Background:

• Early recognition of dengue is crucial for effective clinical management, especially for patients at risk of plasma leakage.
• Dengue hemorrhagic fever (DHF) and dengue shock syndrome (DSS) represent severe forms requiring prompt identification.
• Structural Equation Modelling (SEM) offers a statistical approach to analyze complex mechanistic pathways in disease progression.

Purpose of the Study:

• To develop predictive models for dengue, DHF, and DSS using SEM.
• To identify key clinical and laboratory predictors for these dengue classifications.
• To assess the predictive performance of the models in a validation cohort.

Main Methods:

• Structural Equation Modelling (SEM) was applied to data from 257 Thai children with febrile illnesses.
• Predictive models were constructed using data from fever days -3 and -1 relative to fever resolution.
• Model validation was performed on an independent dataset of 897 subjects.

Main Results:

• Predictors for dengue and DSS included age, tourniquet test, AST, WBC, lymphocyte percentage, and platelet counts.
• Predictors for DHF included age, AST, hematocrit, tourniquet test, WBC, and platelet counts.
• Models demonstrated good predictive performance in the validation set, with AUCs at fever day -3 ranging from 0.67 to 0.84, improving closer to the critical phase.

Conclusions:

• SEM-based predictive models show promise for guiding the clinical management of suspected dengue cases.
• These models can aid in identifying high-risk patients prior to the critical phase of illness.
• Early prediction facilitates timely intervention and potentially reduces disease severity.