Prediction of Incident Heart Failure in TTR Val122Ile Carriers One Year Ahead of Diagnosis in a Multiethnic Biobank

• Published in: The American journal of cardiology 2021-03, Vol.142, p.151-153
• Main Authors: Chaudhary, Kumardeep, Petrazzini, Ben O., Narula, Jagat, Nadkarni, Girish N., Do, Ron
• Format: Article
• Language: English
• Subjects: African Americans; Amyloid Neuropathies, Familial - complications; Amyloid Neuropathies, Familial - genetics; Amyloid Neuropathies, Familial - physiopathology; Angina pectoris; Area Under Curve; Biobanks; Biological Specimen Banks; Cardiomyopathies - complications; Cardiomyopathies - genetics; Cardiomyopathies - physiopathology; Case-Control Studies; Clinical Decision Rules; Congestive heart failure; Diagnosis; Ecosystems; Electronic Health Records; Electronic medical records; Ethnicity; Feature selection; Heart failure; Heart Failure - epidemiology; Heart Failure - etiology; Hispanic Americans; Humans; Incidence; Laboratories; Learning algorithms; Machine Learning; Medical diagnosis; Mutation; Penetrance; Prealbumin - genetics; ROC Curve
• ISSN: 0002-9149; 1879-1913
• DOI: 10.1016/j.amjcard.2020.12.015

[...]we lack external validation; however, we perform 100 iterations of cross-validation on independent datasets to minimize bias in our model performance. [...]we predict heart failure using a 1-year prediction interval only. Importantly, this study shows electronic health records implemented on a machine learning model can predict HF in 82% of high-risk subjects 1 year before diagnosis (Figure 1). [...]notwithstanding limitations, this represents a first effort to address this important unmet need.