Predicting treatment outcomes using explainable machine learning in
children with asthma
Abstract
Background Asthma in children is a heterogeneous disease manifested by
various phenotypes and endotypes. The level of disease control as well
as the effectiveness of anti-inflammatory treatment is variable and
inadequate in a significant portion of patients. Objectives By applying
machine learning algorithms, we aimed to predict treatment success in a
pediatric asthma cohort and to identify key variables for understanding
underlying mechanisms. Methods We predicted treatment outcomes in
children with mild to severe asthma (N=365), according to changes in
asthma control, lung function (FEV1, MEF50) and FENO values after 6
months of controller medication use, using RandomForest and AdaBoost
classifiers. Results The highest prediction power is achieved for
control- and, to lower extend, for FENO-related treatment outcomes. The
most predictive variables for asthma control are related to asthma
severity and total IgE, which was also predictive for FENO-based
outcomes. MEF50-related treatment outcomes were better predicted than
FEV1-based response and one of the best predictive variables for this
response was hsCRP. Conclusions Our results suggest that asthma control-
and FENO-based outcomes can be more accurately predicted using machine
learning than FEV1 and MEF50. This supports the symptom control-based
asthma management approach and its complementary FENO-guided tool in
children. T2-high asthma seemed to respond best to anti-inflammatory
treatment. The prediction of MEF50-based treatment outcomes emphasizes
the role of the distal airways in childhood asthma. The results of this
study in predicting treatment success will help to enable treatment
optimization and to implement the concept of precision medicine in
pediatric asthma treatment.