Improving the liver transplant evaluation process: a data science-focused and team-based approach

Johns Hopkins University, Baltimore MD

Investigators

Linked publications, trials & patents

Abstract

In the US, 4.5 million adults have liver disease, and liver transplantation (LT) is the only curative treatment for those with cirrhosis; transplant centers are charged with determining recipients for a life-saving organ. LT centers assess each patient’s appropriateness for transplant, culminating in a decision to list for transplant or decline. If listed, patients are prioritized based on disease severity and will either receive a liver or be de-listed for a variety of reasons, such as death. While prior research has targeted factors affecting post-listing outcomes (e.g., waitlist dropout, post-LT survival), an upstream focus on factors impacting listing status has not been well studied. LT listing decision-making is variable. Objective clinical measures are utilized, but complex data wrangling requirements and subjectivity permeate data gathering, clinical observations, and psychosocial assessments. A data-driven approach to LT listing has yet to be described. Predictive analytics (supervised machine learning) can be harnessed to strengthen objectivity in complex decision-making. Preliminary data from Dr. Strauss’s qualitative work are the first to comprehensively outline potential pathways related to listing status and reveal that transplant center providers are cautiously optimistic for machine learning-based clinical decision support tools in LT evaluation. The hypothesis is that timely access to summarized, objective data can improve provider decision-making and patient listing. Using a multi-disciplinary approach, Dr. Strauss will leverage her strong relationships with experts from Johns Hopkins Medical Center: experienced transplant team, transplant research lab, Malone Center for Engineering in Healthcare, and School of Public Health. The overarching project goal is to improve LT decision-making using a data-driven and team-based intervention; the overarching training goal is to gain skills in machine learning and implementation science. AIM 1: Develop and internally validate a machine learning-based model to assist LT listing decision-making. AIM 2: Create a data-driven intervention for team decision-making in LT evaluation. AIM 3: Design a multicenter pilot implementation trial of a data-driven intervention for LT evaluation. Impact: Through this project, Dr. Strauss will develop a data-driven intervention that will improve LT listing. This mentored award will develop Dr. Strauss into an R01-funded, independent physician-scientist with advanced skills in machine learning and implementation science.