Prognostic prediction models for dengue complications in pediatric patients

At the BIPS Institute Seminar, Peter Haddawy presented findings from a multi-site study on prognostic prediction models for dengue hemorrhagic fever in pediatric patients with suspected dengue infection. His talk focused on developing models aimed at identifying the risk of severe disease outcomes at an early stage.

Haddawy explained that the models were developed using 17 years of pediatric patient data from two hospitals in Thailand. The work sought to produce prediction models suitable for use in general hospital settings as well as in primary care environments, including settings with limited laboratory facilities.

Another focus of the presentation was on models designed for typical clinical settings in which patient information becomes available incrementally over the course of a hospital stay. These models account for the stepwise availability of clinical data and are intended to support ongoing risk assessment.

Haddawy also addressed the generalizability of the models. This was examined by training the models on data from one hospital and testing them on data from another hospital.

Peter Haddawy is Global Talent Professor of Computer Science at Mahidol University and Director of the Mahidol-Bremen Medical Informatics Research Unit.