Improving colorectal cancer screening decisions through consideration of life expectancy

Project Summary Colorectal cancer (CRC) is a leading cause of cancer-related deaths in the US, and screening is the most effective way to reduce colon cancer-specific mortality. For average risk adults, the United States Preventive Services Task Force (USPSTF) recommends routine CRC screening from 45 to 75 years. Between age 76 and 85, both the USPSTF and the American Cancer Society recommend physicians select appropriate patients for screening, considering life expectancy, comorbidities, patient preference, and screening history; however, there are no recommendations to guide clinicians on how to do this. Since the benefits of CRC screening accrue over a 10- to 15-year period due to the interval required for the progression of adenomas to CRC, screening patients with a life expectancy of <10 years offers little benefit. Older adults have a substantially higher risk of adverse events associated with colonoscopy and a low likelihood of actionable findings, making the balance of potential benefits versus harms unfavorable in patients with a life expectancy <10 years. Conversely, many patients over 75 years in good health can expect to live 10 years or longer, making colonoscopy potentially beneficial. Yet under current age-based guidelines, they are unlikely to be offered it by their physician. Most physicians believe that life expectancy is a sufficient criterion for colorectal cancer screening discontinuation, but making such predictions is difficult. As a result, many patients with life expectancy <10 years are being unnecessarily exposed to the harms of screening, while others with a life expectancy of ≥10 years are missing out on the benefits. To address this, we propose a two-phased investigation. Multiple life expectancy prediction models exist, but these have not been compared against each other to determine which is most appropriate for use in cancer screening decisions. In Aim 1 we will develop a life expectancy prediction model using data from our electronic health record, comparing our results against two existing models to determine which is most appropriate for routine use. Life expectancy predictions aren't useful unless physicians feel comfortable using them to guide decision making. To maximize acceptability, we will develop and test the life expectancy prediction model with input from primary care physicians throughout the process via interviews and meetings with physician leaders. Through these interactions, we will determine physicians' preferred presentation style for life expectancy information in the electronic health record. Our study team has extensive experience creating, implementing, and studying clinical decision support. In Aim 2 we will conduct a cluster randomized trial of a clinical decision support-delivered life expectancy notification on physician colonoscopy ordering for patients aged >75 years. We hypothesize this intervention will reduce both under- and over-screening of older patients based on their predicted life expectancy. Together, this series of investigations will generate essential information to optimize colorectal cancer screening for older adults, ensuring benefits are maximized and harms are minimized.