Related Experiment Video
Updated: Jan 2, 2026
Author Spotlight: Developing a Point-of-Care Hemoglobin Estimation Method for Anemia Management
Published on: January 19, 2024
A Machine Learning-Based Triage Tool for Children With Acute Infection in a Low Resource Setting
Arthur Kwizera1, Niranjan Kissoon2, Ndidiamaka Musa3
1Department of Anaesthesia and Critical Care, Makerere University College of Health Sciences, Kampala, Uganda.
Insights
Machine learning accurately predicts childhood hospital mortality in low-income countries. The best model uses age, respiratory rate, capillary refill time, and altered mental state for reliable prediction.
Area of Science:
- Pediatric critical care
- Machine learning in healthcare
- Global child health
Background:
- Hospital mortality in children with acute infections remains high in low- and middle-income countries (LMICs).
- Predictive models are crucial for early intervention and resource allocation in pediatric care.
Purpose of the Study:
- To develop and validate a machine learning model for predicting hospital mortality in children with acute infections in LMICs.
- To identify key clinical variables at admission for accurate mortality prediction.
Main Methods:
- A post hoc analysis of a prospective feasibility trial involving 949 children admitted with acute infections in rural Rwanda.
- Random forests, a machine learning algorithm, were employed to build predictive models using variables like age, vital signs, and mental state.
- Five models were tested, comparing different combinations of variables and optimization criteria.
Main Results:
- The overall in-hospital mortality rate was 1.5%.
- All five machine learning models demonstrated good predictive performance, with Area Under the Curve (AUC) ranging from 0.69 to 0.8.
- The optimal model, incorporating age, respiratory rate, capillary refill time, and altered mental state, achieved an AUC of 0.8.
Conclusions:
- Machine learning, utilizing readily available admission data, can reliably predict hospital mortality in pediatric populations in Sub-Saharan Africa.
- The developed model offers a promising tool for improving clinical decision-making and patient outcomes in resource-limited settings.
- Further validation in larger, diverse pediatric cohorts is recommended to strengthen the algorithm's generalizability.
Objectives:
To deploy machine learning tools (random forests) to develop a model that reliably predicts hospital mortality in children with acute infections residing in low- and middle-income countries, using age and other variables collected at hospital admission.
Design:
Post hoc analysis of a single-center, prospective, before-and-after feasibility trial.
Setting:
Rural district hospital in Rwanda, a low-income country in Sub-Sahara Africa.
Patients:
Infants and children greater than 28 days and less than 18 years of life hospitalized because of an acute infection.
Interventions:
None.
Measurements And Main Results:
Age, vital signs (heart rate, respiratory rate, and temperature) capillary refill time, altered mental state collected at hospital admission, as well as survival status at hospital discharge were extracted from the trial database. This information was collected for 1,579 adult and pediatric patients admitted to a regional referral hospital with an acute infection in rural Rwanda. Nine-hundred forty-nine children were included in this analysis. We predicted survival in study subjects using random forests, a machine learning algorithm. Five prediction models, all including age plus two to five other variables, were tested. Three distinct optimization criteria of the algorithm were then compared. The in-hospital mortality was 1.5% (n = 14). All five models could predict in-hospital mortality with an area under the receiver operating characteristic curve ranging between 0.69 and 0.8. The model including age, respiratory rate, capillary refill time, altered mental state exhibited the highest predictive value area under the receiver operating characteristic curve 0.8 (95% CI, 0.78-0.8) with the lowest possible number of variables.
Conclusions:
A machine learning-based algorithm could reliably predict hospital mortality in a Sub-Sahara African population of 949 children with an acute infection using easily collected information at admission which includes age, respiratory rate, capillary refill time, and altered mental state. Future studies need to evaluate and strengthen this algorithm in larger pediatric populations, both in high- and low-/middle-income countries.
