Background: Lung cancer screening (LCS) reduces lung cancer death, but can also cause harm, especially when applied to patients with co-existing serious health problems. Recognizing these trade-offs, guidelines recommend that persons with ?a health problem that substantially limits life expectancy or ability to have curative lung surgery? should not be screened, and that all patients considering LCS should undergo a shared decision-making (SDM) process to review LCS benefits and harms with their clinicians. Yet these recommendations are difficult to achieve, as there is little evidence to guide clinicians on which health problems or other patient factors tip the balance of LCS from net benefit to net harm, and little is known about Veteran and clinician approaches to and needs for LCS decision-making when anticipated benefit is marginal. Objectives: We propose a sequential explanatory mixed methods study with the following 3 specific aims: 1. Identify factors that predict little LCS benefit due to limited life expectancy or increased LCS harms; 2. Identify clinical patient factors associated with real-world clinician and Veteran LCS decisions; and 3. Characterize approaches to and needs for decision-making when predicted LCS benefit is marginal. Methods:
Aim 1 a. To determine when competing (non-lung cancer) causes of death limit LCS benefit, we will conduct a survival analysis among LCS-eligible but unscreened Veterans, building a competing risks model and applying recursive partitioning to identify clinically meaningful risk groups.
Aim 1 b. We will build a model to identify combinations of patient factors that predict complications of invasive procedures for LCS-detected findings.
Aim 2 : We will compare how well factors associated with actual LCS decisions align with factors that predict little LCS benefit (Aim 1 models), using data from 10 VA sites that tracked rates at which LCS-eligible Veterans were deemed ?too sick? for LCS, were offered LCS, and accepted LCS. We will build mixed effects logistic regression models to complete these subaims: 2a-Identify patient factors associated with clinicians deeming Veterans ?not appropriate? for LCS, characterizing variation across sites in offering LCS, and whether vulnerable groups (minorities, rural, homeless) are disproportionately deemed ?not appropriate? for LCS. 2b- Identify clinical and demographic factors associated with Veteran decisions to decline vs accept LCS.
Aim 3 : We will interview up to 30 clinicians and 30 Veterans (15 who accepted, 15 who declined LCS) for whom predicted LCS benefit is marginal based on our Aim 1 models. For clinicians, we will explore beliefs about, expected outcomes of, and site-level influences on LCS decision-making, presenting vignettes to learn how providing predicted LCS benefit (Aim 1 models) affects LCS decision-making. For patients, we will explore experiences with LCS discussions, health priorities relative to LCS, and other influences on decision-making. For all participants, we will assess informational needs and preferences to support SDM at the point of care. Anticipated Impacts on Veteran's Healthcare: To maximize the life-saving potential of LCS without creating additional harms and wasting VA resources, LCS must be applied appropriately. This proposal will lay the groundwork to inform future development, testing, and implementation of a personalized decision tool to optimize patient selection for LCS and facilitate SDM between Veterans and VA clinicians at the point of care. This work is essential to achieve safe, appropriate, and Veteran-centered care as lung cancer screening is implemented across VA ? a priority for our partners in the National Center for Health Promotion & Disease Prevention, the Office of Primary Care, and the National Program Office for Pulmonary Medicine. This work addresses Secretary Shulkin?s priority to focus resources efficiently on high-impact problems for Veterans.
A top VA priority is to focus resources efficiently on optimizing care for health issues that are particularly salient for Veterans. Lung cancer is the leading cause of cancer death among Veterans, and VA is implementing CT lung cancer screening to reduce these deaths. But clinicians struggle to identify when lung cancer screening is most appropriate, such that some Veterans with little likelihood of benefit are referred for screening, exposing them to harms and wasting resources, while other Veterans who could benefit are not screened, denying them a potentially life-saving intervention. We propose a mixed methods study using real-world VA data to develop clinical prediction models that identify Veterans with anticipated little net benefit from lung cancer screening, and to understand how VA clinicians and Veterans make lung cancer screening decisions at the point of care. Our study will provide critical scientific knowledge to create a personalized decision support tool to optimize patient selection and promote appropriate, safe, and Veteran-centered decisions about lung cancer screening.
