Modeling and Estimating a Threshold Effect: An Application to Improving Cardiac Surgery Practices

• Activity Number: 193 - Section on Medical Devices and Diagnostics: Student Paper Competition
• Type: Topic Contributed
• Date/Time: Monday, August 8, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Guangyu Yang and Baqun Zhang and Jonathan Haft and Robert Hawkins and David Sturmer and Donald Likosky and Min Zhang*
• Companies: University of Michigan and Shanghai University of Finance and Economics and University of Michigan and University of Michigan and University of Michigan and University of Michigan and University of Michigan
• Keywords: Broken-stick Model; Change-Point; Linear Spline

Abstract

Estimating the threshold has important applications in biomedical research. However, the existing method commonly used in biomedical literature may lead to a biased estimate. For patients undergoing coronary artery bypass grafting (CABG), it is thought that exposure to low oxygen delivery (DO2) leads to an increased risk of avoidable acute kidney injury. This research is motivated by estimating the threshold of nadir DO2 for CABG patients to help develop an evidence-based guideline for improving cardiac surgery practices. We review several models that can be adapted to estimate the threshold and discuss modeling assumptions, scientific plausibility, and implications in estimating the threshold. Under each model, various estimation methods are studied and compared. We propose an algorithm under the constrained broken-stick model and show that the constrained broken-stick model is flexible, robust, and able to incorporate scientific knowledge to improve efficiency. Through simulations and a real data application, we show that the proposed algorithm has good computational performances and enjoys nice properties in terms of bias, robustness and efficiency relative to existing methods.