Scoring model for exploring factors influencing mortality in dengue patients at a tertiary care hospital: a retrospective study

This retrospective study took place at a tertiary care hospital involving hospitalized dengue patients in India. Various clinical and biochemical parameters were recorded. A practical score-based model (DENScore) was developed by calculating the risk score for each attribute. Univariate and multivariate logistic regression analyses were carried out to ascertain notable predictors of mortality. The study also conducted a survival analysis test to illustrate the composite interaction among disease parameters that affect survival probabilities. Among 255 patients, most were under 45 years old, with a survival rate of 96.47%. Univariate logistic regression revealed that patients with ages above 45 years showed symptoms associated with acute kidney injury as significant predictors for reduced survival. Multivariate logistic regression analysis also confirmed that age, acute kidney injury, and Leukocytosis remained the most significant independent predictors of mortality. Gender, hospital stay duration, thrombocytopenia, and SGOT (Serum Glutamic-Oxaloacetic Transaminase) levels showed no mortality association. The model was developed using three state-of-the-art algorithms: Logistic Regression (LR), Linear SVM, and Ridge Classifier (RC). The developed model showed higher scores for acute kidney injury, leukocytosis, platelet count, and dengue shock syndrome features, achieving accuracy rates of 95%, 97%, and 91% respectively. The study findings suggest that age, acute kidney injury, Leukocytosis, and dengue shock syndrome are crucial prognostic factors for mortality in dengue fever patients. The developed DENScore model provides an accurate approach to identifying these predictors early, contributing to the enhancement of disease prognostics. Article Highlights Analyzing past cases to identify key risk factors for deaths in dengue patients. Developing a scoring system (DENScore) using advanced data analysis. Aiming to help doctors recognize early warning signs that could lead to severe dengue outcomes.