Table 1.
Baseline demographics and clinical characteristics.
Fig 1.
Study population demographics.
Panel A: Age distribution histogram by hospital outcome. Died (n = 75): median 54 years. Survived (n = 1,409): median 44 years. Panel B: Outcome by sex. Males: 44/831 (5.3%) mortality. Females: 31/653 (4.7%) mortality.
Fig 2.
Dengue severity classification and outcomes.
Panel A: WHO 2009 classification: Classical 977 (65.8%), Severe 470 (31.7%), Shock 37 (2.5%). Panel B: Mortality: Classical 24/977 (2.5%), Severe 28/470 (6.0%), Shock 23/37 (62.2%). Chi-square p < 0.001.
Fig 3.
Kaplan-Meier survival curves stratified by dengue severity and ICU admission status.
Panel A: Survival probability by dengue severity classification. Patients with dengue shock syndrome exhibited markedly worse survival compared to severe dengue without shock and classical dengue (CLASSICAL: 24/977 deaths, 2.5%; SEVERE: 28/470 deaths, 6.0%; SHOCK: 23/37 deaths, 62.2%; log-rank p < 0.001). Panel B: Survival probability by ICU admission status. ICU patients (admitted at presentation or transferred during hospitalization) had significantly higher mortality than non-ICU patients (ICU: 33/194 deaths, 17.0%; non-ICU: 42/1,290 deaths, 3.3%; log-rank p < 0.001). Shaded areas represent 95% confidence intervals. Total cohort: n = 1,484; 75 deaths (5.1% overall mortality).
Table 2.
Univariate [A] analysis – Risk factors for mortality and multivariable [B] logistic regression – Predictors of mortality.
Table 3.
Multivariable logistic regression – Predictors of ICU admission.
Fig 4.
Panel A: Distribution: Paediatric (<18y) 159 (10.7%), Young adult (18-39y) 465 (31.3%), Middle-aged (40-59y) 521 (35.1%), Elderly (≥60y) 339 (22.8%). Panel B: Mortality/ICU rates: Paediatric 4.4%/13.8%, Young adult 4.7%/4.9%, Middle-aged 2.9%/11.9%, Elderly 9.1%/16.8%. Chi-square p < 0.001.
Fig 5.
Panel A: Platelet trajectories. Survived: admission median 120, nadir 75 × 10³/μL. Died: admission 103, nadir 55 × 10³/μL. Panel B: Mortality by thrombocytopenia severity: Normal (≥100) 20/538 (3.7%), Mild (50-100) 22/486 (4.5%), Moderate (20-50) 20/319 (6.3%), Severe (<20) 13/141 (9.2%). Chi-square p < 0.001.
Table 4.
ROC analysis – Biomarkers for mortality prediction.
Fig 6.
Albumin as a prognostic marker.
Panel A: Kernel density plots. Survived (n = 1,347): median 4.00 g/dL. Died (n = 72): median 3.70 g/dL. Cutoff 3.8 g/dL marked. Panel B: Mortality by category: < 2.5 g/dL 5/8 (62.5%), 2.5-3.0 10/24 (41.7%), 3.0-3.5 12/119 (10.1%), 3.5-3.8 16/295 (5.4%), 3.8-4.5 26/868 (3.0%), > 4.5 3/105 (2.9%). Chi-square p < 0.001.
Fig 7.
Machine learning analysis for mortality prediction in dengue fever (n = 1,416 with complete data; 72 deaths, 5.1%).
(A) Random Forest feature importance (Gini importance) with 95% confidence intervals from 1,000 bootstrap iterations. AST (0.161, 95% CI 0.134–0.195), NLR (0.157, 95% CI 0.128–0.187), and albumin (0.152, 95% CI 0.111–0.201) were identified as top predictors. (B) Model comparison using 5-fold stratified cross-validation. Logistic regression achieved the highest discriminative performance (AUC 0.718 ± 0.101), followed by Random Forest (0.661 ± 0.114) and Gradient Boosting (0.648 ± 0.072). Black dots represent individual fold AUC values; error bars represent standard deviation.
Fig 8.
Note: Analysis restricted to patients without pre-existing chronic liver disease (n = 1,450) to avoid confounding. Panel A: No liver involvement 479 (33.0%), Abnormal LFT 563 (38.8%), Acute Hepatitis 389 (26.8%), Acute Hepatitis with Jaundice 5 (0.3%), Multi-Organ Failure 14 (1.0%). Panel B: Mortality: No involvement 23/479 (4.8%), Abnormal LFT 21/563 (3.7%), Acute Hepatitis 14/389 (3.6%), Acute Hepatitis with Jaundice 0/5 (0.0%), MOF 14/14 (100.0%). Chi-square p < 0.001.
Fig 9.
Secondary HLH and critical care complications.
Panel A: HLH impact (n = 57 with HLH, n = 1,426 without): Mortality 8.8% vs 4.9%, ICU admission 45.6% vs 9.7%, Ventilator 1.8% vs 1.1%. Panel B: Complications: ICU 164 (11.1%), Haemorrhage 202 (13.6%), Secondary infection 24 (1.6%), Shock 20 (1.3%), Ventilator 17 (1.1%), Dialysis 8 (0.5%).
Fig 10.
Panel A: Length of stay. Survived (n = 1,409): median 4 days (IQR 3-6). Died (n = 75): median 2 days (IQR 1-5). Panel B: Factors in prolonged stay (>7 days, n = 113 vs ≤ 7 days, n = 1,296): ICU 23.9% vs 8.2%, Severe dengue 50.4% vs 30.8%, Secondary infection 10.6% vs 0.4%, Haemorrhage 16.8% vs 12.7%.
Fig 11.
Panel A: Lower triangular correlation matrix. Notable: AST-ALT r = 0.91, Platelets admission-nadir r = 0.80. Panel B: Spearman correlations with mortality (sorted): AST r = 0.266, Creatinine r = 0.242, ALT r = 0.223, CRP r = 0.200, NLR r = 0.122, Age r = 0.080, Platelets admission r = −0.036, Platelets nadir r = −0.057, Hemoglobin r = −0.072, Albumin r = −0.203.
Fig 12.
Panel A: Annual cases: 2021 (n = 63, 1.6% mortality), 2022 (n = 155, 2.6%), 2023 (n = 763, 4.7%), 2024 (n = 503, 6.8%). Panel B: Monthly distribution. Monsoon months (June-September): 914/1,484 (61.6%).
Fig 13.
ROC curves for mortality prediction.
Panel A: Biomarker AUCs: Albumin 0.666, NLR 0.637, Platelet nadir 0.588, Hemoglobin 0.571. Panel B: SAPS-3 in ICU patients (n = 188): AUC 0.852, optimal cutoff 62, sensitivity 71.0%, specificity 93.9%.