Related Experiment Video
Updated: Jul 6, 2025
10:30
Author Spotlight: Developing a Microfluidic Lung-on-Chip Model for In-Depth Study of Human Immune Response and Infection Mechanisms
Published on: May 31, 2024
1.4K
Machine learning in predicting severe acute respiratory infection outbreaks
Amauri Duarte da Silva1, Marcelo Ferreira da Costa Gomes2, Tatiana Schäffer Gregianini3
1Universidade Federal de Ciências da Saúde de Porto Alegre, Porto Alegre, Brasil.
Cadernos De Saude Publica
|January 10, 2024
Summary
Machine learning models accurately predict severe acute respiratory infection (SARI) outbreaks, including peak timing and case volume. These tools aid healthcare resource planning for SARI seasons.
Area of Science:
- Epidemiology
- Machine Learning
- Public Health
Background:
- Severe acute respiratory infection (SARI) outbreaks are a recurring public health concern with variable seasonal peaks.
- Effective healthcare resource planning is crucial for managing SARI seasons and patient hospitalization demands.
Purpose of the Study:
- To develop and validate machine learning models for predicting SARI outbreaks in Brazil.
- To provide health managers with tools for anticipating SARI peak incidence, case volume, and pre-epidemic periods.
Main Methods:
- Utilized SARI hospitalization data from Brazil (2013-2020), excluding COVID-19 cases.
- Employed a neural network with a time series pipeline to generate predictive models for five Brazilian regions.
- Validated models against historical SARI outbreak data.
Main Results:
- Neural network models accurately predicted SARI peaks, seasonal case volumes, and the onset of pre-epidemic periods.
- Achieved high weekly incidence correlation (R2 = 0.97) in Southeastern Brazil for the 2019 season.
- Demonstrated strong predictive accuracy for case volume, with a median prediction of 9,405 cases versus 9,936 observed in Southern Brazil in 2019.
Conclusions:
- Machine learning, specifically neural networks and time series algorithms, can effectively predict SARI outbreaks.
- Predictive models offer valuable insights for optimizing healthcare resource allocation during SARI seasons.
