Using Artificial Intelligence to Develop Clinical Decision Support Systems-The Evolving Road of Personalized...

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Artificial intelligence (AI) offers potential to enhance clinical decision-making in oncology.
• Managing risks associated with targeted therapies like Bevacizumab is crucial for patient outcomes.

Purpose of the Study:

• To develop and validate a machine learning-based clinical decision support system (CDSS) for predicting complications from Bevacizumab or its biosimilars.
• To translate the predictive model into a clinically applicable tool for personalized cancer care.

Main Methods:

• A prospective observational study involved 395 patients treated with Bevacizumab or biosimilars for solid tumors.
• Machine learning models (Random Forest, logistic regression, XGBoost) were trained and validated using pretherapeutic variables.
• The best model informed a logistic risk score and an interactive HTML form for risk stratification.

Main Results:

• The optimized Random Forest model achieved 70.63% accuracy and an AUC-ROC of 0.75.
• The derived logistic risk score demonstrated good performance (AUC-ROC = 0.720) and identified key predictors like age, anemia, and tumor stage.
• The tool effectively stratifies patients into low-, intermediate-, and high-risk categories for complications.

Conclusions:

• This study demonstrates the feasibility of AI-driven predictive tools in oncology using real-world data.
• The developed risk score and interactive form can aid clinicians in tailoring targeted therapy decisions, promoting personalized oncology care.
• The tool serves as a decision support, complementing rather than replacing clinical judgment.