OSPAM-C

OSPAM-C predicts five-year survival probabilities for pediatric patients diagnosed with acute myeloid leukemia (AML) using machine learning to provide individualized prognostic assessment.

Key Features:

• Data source: Uses the Surveillance Epidemiology and End Results (SEER) database covering 2000–2011.
• Cohort: Includes pediatric patients aged 0 to 14 years with microscopically confirmed AML.
• Prediction target: Predicts five-year survival outcome.
• Algorithms evaluated: Evaluated logistic regression, support vector machine (SVM), gradient boosting, and K nearest neighbor (KNN).
• Training/test split: Data were partitioned into training and test subsets in an 80/20 ratio.
• Model inputs: Models were trained on patient-specific features.
• Model selection criteria: Selection prioritized discrimination and predictive accuracy.
• Selected model: Gradient boosting was identified as the top-performing algorithm.

Scientific Applications:

• Treatment planning: Provides prognostic information to inform clinical treatment planning for pediatric AML patients.
• Risk stratification: Identifies high-risk patients who may require more aggressive or alternative therapies.
• Outcome assessment: Supports assessment of prognosis to inform patient management and potential outcome improvement.

Methodology:

Development used SEER cases (2000–2011) of pediatric (0–14 years) microscopically confirmed AML; data were split 80/20 into training and test sets, logistic regression, support vector machine (SVM), gradient boosting, and K nearest neighbor (KNN) models were trained to predict five-year survival based on patient features, and the model with the best discrimination and predictive accuracy (gradient boosting) was selected.