Deep Learning-enhanced Evaluation of Quality of Care and Disparities Among Patients with Heart Failure in the Electronic Health Record

Heart failure (HF) is a pervasive, high-risk, and expensive condition that affects over 6.2 million Americans, many of whom endure an excessive burden of hospitalization and reduced life expectancy. This condition, although widely prevalent, disproportionately affects Black individuals who experience a 20-fold higher incidence rate and a 3-fold higher mortality rate in comparison to White individuals.The continuing disparity in HF outcomes among Black individuals, despite advances in HF care, represents a significant challenge that needs urgent attention. The primary concern remains the lack of validated methods to explore and address the underlying reasons for these disparities.
Addressing the challenges, this grant proposal is dedicated to the development of robust models that enhance the assessment and utilization of care-quality process measures in the treatment of HF. We propose to develop and implement robust deep learning models to enhance the evaluation of care quality in HF management. The main objective is to improve the outcome of patients with cardiovascular disease by using deep learning to optimize care management and to identify and reduce systemic care differences in HF leading to disparate care quality in Black populations. Aim 1: Automate the assessment of HF phenotypes to evaluate the non-prescription of evidence-based therapies. The model will use deep learning-based natural language processing (NLP) methods applied to clinical documentation to determine individual HF subtypes and optimize treatment regimens. Aim 2: Automate the identification of social determinants of health and documentation patterns that vary across patient populations. This aim plans to train a deep learning NLP feature extraction model to identify social challenges and clinical language variations, assessing how these features impact care quality across different patient populations.
The outcome of this work will provide an invaluable foundation for advancing data-driven innovations in cardiovascular medicine, promoting data-driven, individualized patient care. This project is anticipated to have a substantial impact on how HF care for across different populations is measured and conceptualized. The goal is to enhance the standardization of care and improvement in health outcomes, thus helping to shape the future of clinical care for one of the most common, high-risk, and high-cost conditions affecting the American population.

Grant Number: 5F30HL176149-02
NIH Institute/Center: NIH
Principal Investigator: Philip Adejumo