Improving early prediction for adverse dengue outcomes using data from the IDAMS prospective multicentre study.

Early identification of individuals at risk of complications is important to improve dengue case-management and promote appropriate use of limited resources in high burden settings. In this prospective observational cohort, 7428 febrile outpatients were enrolled across 8 countries in Asia and Latin America and followed daily; 2230 individuals with confirmed dengue were included in the analysis, among whom 304 (14%) progressed to moderate/severe dengue and 38 (2%) to severe dengue during follow-up. We demonstrate evolving relationships between common clinical features, WHO clinical warning signs, and simple haematological parameters, with progression. Key predictors of moderate/severe dengue include low lymphocyte percentage, low total white blood cell count, low platelet count, and persistent vomiting and significant abdominal pain/tenderness. Multivariable models incorporating these variables predict the risk of moderate/severe dengue, initially with modest performance (AUC ~ 0.68), but improving subsequently as new daily data is included into the predictions (AUC ~ 0.73-0.85). We also demonstrate that using these models in combination with clinical decision-making pathways could facilitate earlier admission of high-risk patients without markedly increasing admissions of uncomplicated cases. These results should prove instrumental for updating guidelines on dengue triage and management, potentially providing major public health benefits in resource poor settings.