Factors associated with prolonged hospitalization in children with dengue infection: a retrospective cohort study from Indonesia

Dengue infection remains a major cause of pediatric hospitalization in tropical countries, including Indonesia. Length of stay (LOS) is an important indicator of disease severity, healthcare utilization, and hospital resource allocation. Although platelet count has traditionally been used as a marker of dengue severity, the predictive value of combining nutritional status, platelet indices, and warning signs for prolonged hospitalization remains limited and heterogenous. To evaluate the predictive ability of nutritional status, platelet count, platelet indices, and warning signs for prolonged hospitalization among children with dengue infection, and to compare the performance of a platelet-only, nutrition-only, and full models . A retrospective cohort study was conducted among pediatric patients diagnosed with dengue infection who were admitted to Dr. Sardjito General Hospital, Yogyakarta, Indonesia, between 2019 and 2024. The primary outcome was prolonged hospitalization, defined as LOS > 4 days. Nutritional status, platelet count, platelet indices including mean platelet volume (MPV), platelet distribution width (PDW), plateletcrit (PCT), platelet large cell ratio (P-LCR), and warning signs were analyzed. Bivariate analysis and multivariable logistic regression were performed along with 5-fold cross-validation for internal validation. Predictive performance was evaluated using receiver operating characteristic (ROC) curve analysis. A total of 104 children fulfilled the inclusion criteria, whereas 102 survivors entered prolonged LOS analyses. Prolonged hospitalization occurred in 32/102 patients (31.4%). Nutritional status was significantly associated with prolonged hospitalizatio, whereas platelet count, platelet indices, and warning signs were not. In the full model multivariable model, undernutrition and overweight/obesity remained independently associated with prolonged hospitalization, while P-LCR showed only a weak, borderline inverse association that requires confirmation. The platelet-only model showed poor discrimination (AUC = 0.533, 5-fold cross-validated AUC 0.494). In contrast, the nutrition-only and full models demonstrated good discriminations (AUC = 0.777 and 0.855, respectively), which decreased to 0.729 and 0.754 after 5-fold cross-validation. Therefore, the incremental gain in discrimination from adding platelet parameters and warning signs to nutritional status was small after internal validation (ΔAUC = 0.025).The nutrition-only model also retained higher cross-validated sensitivity (84.4% vs. 65.6%) and NPV (90.7% vs. 82.3%) than the full model. The full model showed better apparent discrimination for prolonged hospitalization than the platelet-only and nutrition-only models, among children with dengue infection. The models retained acceptable discrimination on internal cross-validation. Nutritional status showed the strongest independent association with prolonged hospitalization, the borderline inverse association of P-LCR should be considered exploratory. External validation and further adjustment for disease severity are needed before clinical application. Not applicable.