Leveraging Machine Learning to Aid in the Utilization of Diagnostic Testing in Thrombotic Thrombocytopenic Purpura

Abstract

Artificial intelligence (AI) has the potential to revolutionize the medical field with machine learning utilization, improving patient outcomes. Thrombotic thrombocytopenic purpura (TTP) is a life-threatening, blood clotting disorder which is confirmed by the ADAMTS13 activity assay. The improper usage of ADAMTS13 and constrained resources in laboratories leads to inefficient patient care. This research project will result in a decision tree (DT) algorithm, aiding in efficiently diagnosing TTP. This machine learning (ML) support tool would reduce the over-utilization of ADAMTS13 testing and save lives. In Phase 1, the principal investigator coded the ML algorithm, which was developed by training and testing with preliminary data, producing an overall accuracy of 81%. Phase 2 curates a collection of patient data using the UF Health electronic health record for validation of the algorithm. Phase 3 includes additional testing with new data, while Phase 4 requires review of guidelines for implementation into the laboratory. This knowledge will help close the mortality gap for TTP and provide the framework to advance the development of AI support tools for various diseases. The overarching mission is to create the lab of the future where AI-generated decision support tools guide better diagnostic testing to aid clinicians in improving patient care.