Cardiovascular Data Science (CarDS) Lab at the American Heart Association Scientific Sessions 2021
The CarDS Lab congratulates its members for its 11 abstracts selected for presentation at AHA 2021. We hope to see you virtually on November 13-15. Our presenters describe their work below:
National Trends and County-Level Patterns of Prescription for PCSK-9 Inhibitors Among Medicare Beneficiaries in the United States
Arash A. Nargesi will share his findings on the prescription patterns of PCSK-9 inhibitors among Medicare beneficiaries. Using the most recent data, he will report a steady but slow uptrend of PCSK-9 inhibitors prescriptions in recent years, with persistent limited prescription in small and less affluent counties.
Patterns of Prescribing Sodium-Glucose Cotransporter-2 Inhibitors for Medicare Beneficiaries in the United States
Veer Sangha will present on patterns of prescribing SGLT-2 Inhibitors for Medicare beneficiaries in the United States. He will report on a baseline pattern of underuse relative to other second line medications without proven cardiovascular benefits, highlighting an opportunity to proactively identify challenges against the wider use of SGLT2i to realize their potential public health benefits.
Prescribing Patterns of Cardioprotective Anti-hyperglycemics Among Medicare Beneficiaries Across United States Counties by their Socioeconomic Disadvantageousness
Jonathan Hanna will report his study on prescription patterns of SGLT-2 inhibitors and GLP-1RAs in Medicare beneficiaries. He will describe disparities of use for these drug classes at the county level, with lower prescription rates in communities with predominantly African-Americans.
Racial and Gender Differences in Lifetime Healthcare Costs Across Cardiovascular Risk Factors
In a large cohort study of 2184 individuals in the Dallas Heart Study followed across all regional hospitals for over 18 years, Rohan Khera and Mengni Liu found that the Black individuals, particularly those with cardiovascular risk factors had substantial higher lifetime healthcare expenses, defined as all-cause healthcare spending between 40 and 80 years of age. There was significantly cost inflection in the 60s, suggestive of higher healthcare utilization among Black individuals in the Medicare age group
Natural Language Processing of Hospitalization Discharge Summary to Predict 1-year Post-Discharge Mortality Among Patients With Acute Heart Failure
Benjamin Rosand has developed a highly accurate model for prediction of patient mortality post-hospital-discharge using clinical notes from patients hospitalized for heart failure in the Yale New Haven Hospital network. He will describe the model accuracy and show how the model picks up on key clinical features within the text.
Individualizing the Cardiovascular Benefits of Intensive Blood Pressure Reduction Using Computational Phenomaps of the SPRINT and ACCORD Trials
&
Individualizing Cardiovascular Benefits of Canagliflozin Based on Multidimensional Computational Phenomaps of the CANVAS Trials
Evangelos Oikonomou will discuss how a novel machine learning-based method of creating computational phenomaps can personalize the interpretation of clinical trials. In an analysis of the SPRINT and ACCORD-BP trials, which will be presented as part of the Lifestyle and Cardiometabolic Health Early Career Investigator Award Competition, Evangelos will describe a tool that individualizes the cardiovascular benefits of intensive versus standard blood pressure reduction. In a separate session, this method will be expanded to novel cardiometabolic therapies, reviewing how computational phenomaps of the CANVAS trials can inform clinicians and patients on the personalized cardiovascular benefits of canagliflozin use based on each patient’s unique profile.
ECG-DualNet: A MultiLabel Model for Automated Diagnosis on Both Electrocardiographic Signals and Printed 12-lead Electrocardiograms
Veer Sangha will present on ECG-DualNet, a multi-label model for automated diagnosis on both electrocardiographic signals and printed 12-lead EKGs. Using data from Brazil validated across continents, he will report on a versatile, interpretable, and generalizable model that has the potential to broaden the application of artificial intelligence to clinical electrocardiography.
Out-of-Pocket Health Spending and Subjective Financial Hardship in Atherosclerotic Cardiovascular Disease and Heart Failure in the United States
In an oral presentation, Stephen Y Wang will describe his study comparing subjective and objective financial hardship among patients with atherosclerotic cardiovascular disease or heart failure in the United States. Using the nationally representative Medical Expenditure Panel Survey, he examines deferred and foregone care and conducts subgroup analyses by type of financial burden, in order to demonstrate that studying both forms of financial burden is necessary to elucidate the full extent of financial toxicity in cardiovascular diseases.
Predicting In-Hospital Mortality Among Patients Hospitalized With Heart Failure Using a Convolutional Neural Network and Transfer Learning on a Single Admission Chest X-Ray
David Hidalgo-Gato will share his study on transfer learning in patients with heart failure. Using MIMIC-CXR dataset, he will report a new method to improve prediction of mortality in patients with heart failure by a model trained to determine the presence of pleural effusion.
Variation In Time In Therapeutic Range For Cooling For Out-of-hospital Cardiac Arrest
Kevin Wheelock will present his findings on the association of time-in-therapeutic range and outcomes in out-of-hospital cardiac arrest patients undergoing therapeutic hypothermia. Using data from the ROC-CCC trial, he will show that a greater time-in-therapeutic range at target temperature 32-36C was not associated with improved survival or neurologic disability at hospital discharge.