Explainable artificial intelligence for predicting length of stay and treatment response in pediatric asthma and allergic rhinitis: An irregular fuzzy cellular automata approach across pre- and during-COVID-19 periods.

OBJECTIVE

The COVID-19 pandemic significantly altered treatment strategies and healthcare utilization patterns in pediatric respiratory diseases. Given the need for interpretable predictive tools in clinical decision-making, this study aimed to investigate changes in treatment approaches and to predict hospital length of stay and treatment response in children with asthma and allergic rhinitis across pre- and during-COVID-19 periods.

METHODS

This retrospective study analyzed medical records of 450 hospitalized children under 12 years diagnosed with asthma and/or allergic rhinitis at Abuazar Hospital, Ahvaz, between January 2018 and December 2021, according to GINA and ARIA criteria. An Irregular Fuzzy Cellular Automata (IFCA) model was implemented using Python. Data preprocessing included one-hot encoding and balancing with SMOTE. Model performance was evaluated using 5-fold cross-validation and compared with Random Forest and SVM models.

RESULTS

Antibiotic use significantly decreased during COVID-19 (38.2% vs. 58.7%), while corticosteroid use increased (72.3% vs. 57.2%) (P < 0.001). The IFCA model achieved 86.2% accuracy and an AUC-ROC of 0.89, outperforming baseline models. Disease severity (SHAP = 0.35) and treatment type (SHAP = 0.28) were the most influential predictors.

CONCLUSION

The IFCA model provides accurate and interpretable predictions for hospitalization outcomes in pediatric asthma and allergic rhinitis, supporting clinical decision-making and resource optimization during pandemic conditions.