Using Big Data to Estimate the Effects of Complex Cost-Sharing Rules on Colorectal Cancer Screening and Patient Health

点击翻译按钮获取中文摘要

Colorectal cancer is the third-leading cause of cancer-related mortality and approximately 5% of individuals will develop colorectal cancer during their lifetimes. However, colorectal cancer is highly preventable if detected early. Colonoscopies are the most common form of colorectal cancer screenings and all adults between the ages of 50 and 74 are recommended to receive a colonoscopy at least once every 10 years. This study will examine how two recent policies, coverage mandates and high-deductible health plans, that change consumer cost-sharing for colonoscopies have changed patient adherence to these guidelines among the commercially insured population. This study will also examine how changes in patient utilization of colonoscopies have led to changes in colorectal cancer detection and mortality. These policies have created a rapidly changing cost- sharing environment for patients and so fully understanding the patient health effects of these policies is will help inform policy makers on the patient health effects of these changes. Recent years have also seen a rapid increase in data, computing power, and analytical methodologies. This study will apply recent advances in data mining techniques to one of the largest sources of data available to researchers. The machine learning approaches used in this study will two sources of bias that are potentially present in traditional approaches—multiple hypothesis testing and selective reporting of results. The machine learning model approaches will also be used to estimate heterogeneity in treatment effects, which will help inform policy makers of how to tailor cost-sharing policies for colonoscopy services.