MPN-654 Analysis of Cardiovascular Risk in 920 Patients With Myeloproliferative Neoplasms Using Natural Language Processing

• Published in: Clinical lymphoma, myeloma and leukemia myeloma and leukemia, 2024-09, Vol.24, p.S442-S442
• Main Authors: Duminuco, Andrea, Au Yeung, Joshua, Virdee, Sukhraj, Vaghela, Raj, Woodley, Claire, Asirvatham, Susan, Curto-Garcia, Natalia, Sriskandarajah, Priya, O'sullivan, Jennifer, Radia, Deepti, De Lavallade, Hugues, Kordasti, Shahram, Palumbo, Giuseppe, Harrison, Claire, Harrington, Patrick
• Format: Article
• Language: English
• Subjects: Artificial intelligence; Cardiovascular risk factors; Machine learning; MPN; Thrombosis risk
• ISSN: 2152-2650
• DOI: 10.1016/S2152-2650(24)01452-6

Thromboembolic events represent the most common cause of morbidity and mortality in myeloproliferative neoplasms (MPN). Natural language processing is a branch of machine learning involving computational interpretation and human language analysis. CogStack is an information extraction architecture incorporating structured and unstructured electronic health record (EHR) components. We aimed to employ a machine-learning approach to determine the prevalence and impact of cardiovascular risk factors upon thrombotic events during follow-up. Data extracted from CogStack was processed by a medical concept annotation toolkit (MedCAT). We evaluated data from 360 polycythaemia vera (PV) and 560 essential thrombocythaemia (ET) patients seen at Guys’ and St Thomas NHS Foundation Trust (GSTT) between 2005 and April 2023, including 24,155 individual EHR documents. In ET, hypertension (HTN) was the most prevalent comorbidity, identified in 21.3% (119) of patients, followed by hypercholesterolemia (9.6%), while, overall, 20% (112) experienced a thrombotic event. This included thrombosis not otherwise specified (NOS) in 8% (45), cerebrovascular accident (CVA) in 7.7% (43), and myocardial infarction (MI) in 3.6% (20). HTN was also the most identified condition in PV, seen in 23.1% (83) of patients. In PV, thrombosis NOS was observed in 19.4% (70), CVA in 14.2% (51), venous thromboembolism (VTE) in 23.3% (84) and MI in 3.1% (11). Overall, 35% (126) of cases had a thrombotic event. Comparing the two cohorts, a significantly greater frequency of events was observed in PV patients for CVA (p=0.002), portal vein thrombosis, and VTE (both p