EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

4049. Peak prediction of pediatric hospitalizations due to respiratory diseases

Invited abstract in session MA-15: Machine learning and game theory in healthcare, stream OR in Health Services (ORAHS).

Monday, 8:30-10:00
Room: 18 (building: 116)

Authors (first author is the speaker)

1. Gloria Henríquez Díaz
Center For Mathematical Modeling (CNRS IRL2807), University of Chile
2. Samuel Marin-Navarro
University of Chile

Abstract

INTRODUCTION
In Chile, every year in fall and winter, respiratory diseases increase in pediatric population (due to different factors) and this produces an increase in the number of hospitalizations, sometimes collapsing the health system. To anticipate, a system has been created to predict when the peak will be.

METHOD
Data from Luis Calvo Mackenna Hospital were obtained from the open access database DEIS of the Ministry of Health. The created system was used to predict the peak date of pediatric hospitalizations from 2017 to 2023, except 2020 and 2021 due to the pandemic. The curve produced were smoothed by moving averages. As the curve ascends, alerts created with machine learning begin to appear. The model was tested using antecedents from 2 to 7 previous years.

RESULTS
Tests performed with 7 and 6 years of historical data had 100% results within the prediction interval (RMSE = 7.0 with respect to the exact day), while with 5 and 4 years were 67% (RMSE = 10.2) and 50% (RMSE = 11.3), respectively. Finally, with 3 years it was 20% (RMSE = 11.4) and with 2 it was 0% (RMSE = 12.6). In all cases, the prediction was approximately one month in advance.

CONCLUSIONS
The system created has promising results if sufficient historical data is added to train the model. A future challenge is to adjust it for other pediatric hospitals.

ACKNOWLEDGMENTS
This work was funded by the ANID FONDEF IDeA I+D 2023 ID23I10423.

Keywords

Status: accepted

Back to the list of papers