Evaluation of dengue severity scores for shock prediction: A pediatric cohort in the Mekong Delta, Vietnam

Abstract

Background

Dengue shock syndrome (DSS) is a life-threatening complication of dengue infection, particularly in children. Early identification of those at risk of progression to DSS is essential for improving outcomes, especially in resource-limited settings such as Vietnam. This study aimed to determine the utility of three dengue scoring systems—Pongpan 2013 (Score I), Srisuphanunt 2022 (Score II), and a modified Bhaskar 2022 (Score III)—in predicting DSS in pediatric patients by systematically assessing their discrimination and calibration.

Methods

This cohort study included children aged ≤ 16 years with dengue infection, admitted to a central hospital in the Mekong Delta region between December 2022 and October 2024. The predictive utility of three scoring systems was evaluated using measures of discrimination and calibration. Discrimination was evaluated using area under the receiver operating characteristic curve (AUC), and calibration was assessed using the Hosmer-Lemeshow test, calibration plots, mean absolute error (MAE), and 90th percentile of absolute calibration error (P90).

Results

Among the studied patients, 25.0 % progressed to DSS. The median day of illness at admission was 4 (interquartile range: 2–4). All three scores demonstrated excellent discrimination: Score (I), AUC = 0.916 (95 % CI: 0.865–0.967); Score (II), AUC = 0.897 (95 % CI: 0.842–0.951); Score (III), AUC = 0.849 (95 % CI: 0.781–0.918); all p < 0.001. Calibration was best in Score (III), with χ²(df) = 2.24(3), p = 0.524, MAE = 0.036, and P90 = 0.094. Score (I) and Score (II) showed poorer calibration, with significant Hosmer-Lemeshow p-values and higher error metrics.

Conclusion

Score (III) demonstrates balanced predictive performance and may aid early DSS risk stratification in pediatric dengue. Further validation in diverse settings is needed before routine use. Scoring systems should complement, not replace, continuous clinical monitoring.