Machine Learning-Integrated Private Blockchain Framework for Bias Detection in clinical trials data

Contemporary clinical trials face major operational challenges. These include high attrition rates, rising costs, and a risk to data integrity and transparency. Many current blockchain models in healthcare do not chain biased data and lack real-time auditing. This shows the need for a better approach to integrating predictive analytics with decentralized security. The framework proposed in this paper combines machine learning (ML) algorithms with a privacy-preserving blockchain platform. This combination aims to improve the safety, transparency, and accuracy of clinical trial outcome predictions and to identify biases in reported trials. This system introduces a majority-voting bias detection mechanism that highlights necessary biases in demographics and reporting before they are recorded permanently. The system architecture relies on Hyperledger Fabric, which provides a decentralized basis for immutable ledger services. Six different ML algorithms were used: Decision Tree, Linear Regression, Logistic Regression, XG-Boost, Random Forest, and Support Vector Machine (SVM). These algorithms predicted trial outcomes and assessed biases based on patient demographics, dosage levels, and study duration. The best-performing model was XG-Boost, achieving the highest accuracy (0.91) and precision (0.90). The results of this research indicate that this hybrid solution leads to a safer, clearer, and ethically sound clinical trial management process. Blockchain, Clinical Trials, Hyperledger Fabric, IPFS, Smart Contracts, Machine Learning, Bias Detection, Data Integrity, ZKP.