SCREENING OF PATIENTS AT RISK FOR WILD TYPE ATTR-CM USING A COMPUTATIONAL MACHINE LEARNING ALGORITHM

Abstract

BACKGROUND Wild-type transthyretin amyloid cardiomyopathy (ATTRwt-CM) is a progressive, life-threatening disease that is an underdiagnosed cause of heart failure (HF). To enable the identification of patients with suspected ATTRwt-CM, a machine learning algorithm (MLA) was developed using U.S. medical claims data sourced from IQVIA and Optum. Our goal was to test high-yield clinical combinations derived from the MLA in undiagnosed and diagnosed patients with ATTRwt-CM. METHODS AND RESULTS 21 combinations of clinical characteristics associated with ATTRwt-CM were tested within University of Utah Health. The undiagnosed cohort included patients 50 years of age or older with at least 1 office visit within the past 18 months, diagnosis of heart failure (HF), and no prior diagnosis of amyloidosis or end stage renal disease. High-yield combinations and the clinical profile of patients at risk for ATTRwt-CM were identified. A second cohort of diagnosed patients was included to analyze the sensitivity of the combinations. 9,051 HF patients met inclusion criteria, of whom 1,091 (12%) were identified as at risk for ATTRwt-CM by the clinical combinations (mean age 74 years, 54% female). Higher yielding clinical combinations included 1) systolic or combined HF and diastolic HF (n = 537, 49%), 2) carpal tunnel (n = 340, 31%), and 3) atrial flutter and/or atrial fibrillation (AFl/AF) and joint disorders and diastolic HF (n = 196, 18%). Common clinical features of patients at risk included osteroarthrosis (60%), AFl/AF (56%), and chronic kidney disease (36%). Over the previous 18 months, patients identified as ‘at risk’ visited cardiology (77%), family practice (48%), and orthopedics (30%) clinics. In cohort 2, 1 of the 21 clinical combinations was observed in 52 (75%) patients with confirmed ATTRwt-CM (mean age 80 years, 6% female). CONCLUSION The MLA combinations identified patients at risk for ATTRwt-CM using clinical characteristics. Integration of the clinical combinations within electronic health records may improve the identification of patients with ATTRwt-CM. Wild-type transthyretin amyloid cardiomyopathy (ATTRwt-CM) is a progressive, life-threatening disease that is an underdiagnosed cause of heart failure (HF). To enable the identification of patients with suspected ATTRwt-CM, a machine learning algorithm (MLA) was developed using U.S. medical claims data sourced from IQVIA and Optum. Our goal was to test high-yield clinical combinations derived from the MLA in undiagnosed and diagnosed patients with ATTRwt-CM. 21 combinations of clinical characteristics associated with ATTRwt-CM were tested within University of Utah Health. The undiagnosed cohort included patients 50 years of age or older with at least 1 office visit within the past 18 months, diagnosis of heart failure (HF), and no prior diagnosis of amyloidosis or end stage renal disease. High-yield combinations and the clinical profile of patients at risk for ATTRwt-CM were identified. A second cohort of diagnosed patients was included to analyze the sensitivity of the combinations. 9,051 HF patients met inclusion criteria, of whom 1,091 (12%) were identified as at risk for ATTRwt-CM by the clinical combinations (mean age 74 years, 54% female). Higher yielding clinical combinations included 1) systolic or combined HF and diastolic HF (n = 537, 49%), 2) carpal tunnel (n = 340, 31%), and 3) atrial flutter and/or atrial fibrillation (AFl/AF) and joint disorders and diastolic HF (n = 196, 18%). Common clinical features of patients at risk included osteroarthrosis (60%), AFl/AF (56%), and chronic kidney disease (36%). Over the previous 18 months, patients identified as ‘at risk’ visited cardiology (77%), family practice (48%), and orthopedics (30%) clinics. In cohort 2, 1 of the 21 clinical combinations was observed in 52 (75%) patients with confirmed ATTRwt-CM (mean age 80 years, 6% female). The MLA combinations identified patients at risk for ATTRwt-CM using clinical characteristics. Integration of the clinical combinations within electronic health records may improve the identification of patients with ATTRwt-CM.