An artificial intelligence-driven predictive model for pediatric allogeneic hematopoietic stem cell transplantation...

Area of Science:

• Pediatric Oncology
• Hematology
• Biostatistics

Background:

• Hematopoietic stem cell transplantation (HSCT) is associated with significant morbidity and mortality.
• Accurate risk assessment is vital for optimizing patient selection and outcomes in HSCT.
• Existing predictive models for adult HSCT have limitations when applied to pediatric populations.

Purpose of the Study:

• To develop an automated machine learning algorithm for predicting survival in pediatric patients with malignant disorders undergoing HSCT.
• To create a predictive tool that addresses the limitations of existing models in the pediatric population.

Main Methods:

• Analysis of allogeneic HSCTs in children with malignant disorders (1991-2021).
• Survival analysis using Kaplan-Meier, log-rank test, and Cox regression.
• Development of a prognostic index and a random forest predictive model for 1-year survival.

Main Results:

• Older age, later period of HSCT, and mismatched donor were significant factors in multivariate analysis.
• The developed prognostic index correlated with 3-year overall survival.
• The random forest model achieved 72% accuracy in predicting 1-year survival using key clinical variables.

Conclusions:

• The developed prognostic index and random forest model are effective for predicting 1-year survival post-HSCT in children.
• Further validation in diverse pediatric populations is required to ensure generalizability of the predictive models.