Global performance of predictive models for dengue severity, hospitalization and mortality: A systematic review and meta-analysis of 146 studies.
Felipe Mendes Delpino, Igor Tona Peres, Tomoe Gusberti et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis evaluated the performance of predictive models for dengue severity, hospitalization, and mortality across 146 studies. The analysis found that for severity prediction, the pooled sensitivity was 0.85 and the overall AUC was 0.93, with machine learning models slightly outperforming traditional regression methods. High risk of bias was noted in many studies.
Key findings
- Pooled sensitivity for severity prediction was 0.85 (95% CI 0.82-0.87).
- Overall AUC for severity prediction was 0.93 (0.91-0.94).
- Machine learning models had an AUC of 0.93 compared to 0.90 for traditional regression.
- Bleeding, shock, and hypoalbuminemia were consistently associated with adverse outcomes.
Why it matters
Understanding the performance of predictive models for dengue can enhance clinical decision-making and patient management. Accurate predictions can help identify high-risk patients and allocate resources effectively.
What to keep in mind
PROBAST classified 112 of 146 studies as high risk of bias, indicating potential limitations in the reliability of the findings.