Collaboratively Learning Optimal Patient Outcomes Using Smart Contracts in Limited Data Settings

The widespread presence of electronic health records and digitizing patient health data has led to various Clinical Decision Support (CDS) systems, leveraging novel analysis tools such as Machine Learning (ML). While current ML-based CDS systems are continually developed, they are frequently limited by the data they are trained on. We, therefore, propose a novel architecture called "BlockAI," which is designed to employ blockchain and smart contracts to continuously receive data from participating healthcare infrastructures to aid ML processes. With the help of smart contract based transactions, and incentive mechanisms, BlockAI allows sustainable machine learning upon a much more considerable set of healthcare data collected from healthcare institutions worldwide. As a motivating example, we have implemented BlockAI for sepsis mortality prediction and have compared the key functionalities of our system with existing comparable state-of-the-art systems available in healthcare research. We have also evaluated our system’s accuracy and performance costs compared to guideline and ML-based sepsis mortality prediction algorithms, which shows promising results. The future widespread application of BlockAI would ensure healthcare providers obtain accurate and timely information for complicated patients with multiple comorbid conditions that are rare and emerging diseases that they do not have much information about, which will significantly assist their diagnosis process and ensure better care for the patients.