Identification of prognostic factors contributing towards mortality in leptospirosis patients: a statistical and score-based model approach

Abstract Purpose The study aims to identify mortality predictors in patients with leptospirosis using statistical and score-based model approaches. Methods This retrospective study was conducted at a tertiary care hospital in India, involving hospitalized leptospirosis patients. Clinical and biochemical parameters were recorded, and a practical score-based model was developed by calculating risk scores for each attribute. Univariate and multivariate logistic regression analyses were performed to identify significant predictors of mortality. Survival analysis was also conducted to illustrate the interaction among disease parameters affecting survival probabilities. Results Out of 164 patients, the majority were under 45 years old, with a survival rate of 84.8%. Univariate logistic regression indicated that patients over 45 years old, and those with thrombocytopenia, acute kidney injury, total bilirubin, and direct bilirubin, were significant predictors of reduced survival. Multivariate logistic regression confirmed that thrombocytopenia, acute kidney injury, and both total and direct bilirubin were the most significant independent predictors of mortality. Gender, duration of hospital stays, Serum Glutamate Pyruvate Transaminase (SGPT), and Serum Glutamic-Oxaloacetic Transaminase (SGOT) levels were not associated with mortality. The model, developed using Logistic Regression (LR) and Ridge Classifier (RC), showed higher scores for thrombocytopenia, acute kidney injury, leukocytosis, and direct bilirubin features, achieving accuracy rates of 89% and 92% respectively. Conclusion The study suggests that thrombocytopenia, acute kidney injury, leukocytosis, and bilirubin levels are crucial prognostic factors for mortality in leptospirosis patients. The developed model offers an accurate method for early identification of these predictors, enhancing disease prognostics.