1
|
Husband G, D"amico A, Hasnie U, Batra N, Cochrun S, Gann A, Li E, Nguyen D, Philip George A, Soto M, Rogers C, Ahmed M, Andrikopoulou E. Machine learning analysis including social determinants of health for predication of mortality following transcatheter aortic valve implantation: a single center experience. Eur Heart J Cardiovasc Imaging 2022. [DOI: 10.1093/ehjci/jeab289.244] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Indexed: 11/12/2022] Open
Abstract
Abstract
Funding Acknowledgements
Type of funding sources: None.
Introduction
Social determinants of health (SDOH) are increasingly being recognized as critical, independent prognosticators in cardiovascular disease. Despite this, little is known about the role of SDOH in predicting outcomes following transcatheter aortic valve implantation (TAVI).
Purpose
To assess the value of adding census-derived SDOH in developing machine learning (ML) models for prediction of all-cause mortality in patients following TAVI.
Methods
A total of 398 patients, who underwent TAVI in 2019, were studied. Clinical, demographic, echocardiographic (echo) and census-derived SDOH data were collected. All-cause mortality at 1 year was the endpoint. A general linear ML model was fit with 100 iterations and a 70:30 training-test split. We compared the predictive performance of the model with and without adding SDOH. The SDOH included in the ML model were race (white vs. non-white), % zip code population as female, and zip code average yearly income less than $45,000.
Results
Baseline SDOH, demographic, clinical, and echo data are shown in Table 1. Following univariate and multivariate predictor analysis, the following input data were used for the ML model without the SDOH: post TAVI all-cause hospitalizations, history of outpatient hemodialysis, atrial fibrillation, heart failure with reduced ejection fraction, myocardial infarction, coronary artery disease and beta-blockers. The ML model with SDOH used the same input as well as the SDOH variables. The model with vs. without SDOH had a median AUC of 0.75 vs. 0.73 (p = 0.9957).
Conclusions
Despite not reaching statistical significance, our ML model provides a holistic picture of mortality predictors. Larger studies are needed to more assess the predictive value of SDOH post TAVI. Abstract Figure. Baseline patient characteristics Abstract Figure. ML Model: Area Under Curve
Collapse
Affiliation(s)
- G Husband
- UAB Hospital, Birmingham, United States of America
| | - A D"amico
- UAB Hospital, Birmingham, United States of America
| | - U Hasnie
- UAB Hospital, Birmingham, United States of America
| | - N Batra
- UAB Hospital, Birmingham, United States of America
| | - S Cochrun
- UAB Hospital, Birmingham, United States of America
| | - A Gann
- University of South Alabama, Mobile, United States of America
| | - E Li
- UAB Hospital, Birmingham, United States of America
| | - D Nguyen
- The Johns Hopkins Hospital, Baltimore, United States of America
| | | | - M Soto
- UAB Hospital, Birmingham, United States of America
| | - C Rogers
- UAB Hospital, Birmingham, United States of America
| | - M Ahmed
- UAB Hospital, Birmingham, United States of America
| | | |
Collapse
|