Lung cancer is the leading cause of cancer-related death for Americans [1]. The National Lung Screening Trial (NLST) demonstrated that among individuals with a high risk of lung cancer, a 20% relative reduction in lung cancer mortality was observed with low-dose computed tomography for lung cancer screening (LDCT-LCS) compared to chest X-ray [2]. Those results formed the basis of the current screening recommendations adopted by almost all major organizations [3–7]. In December 2013, the US Preventive Services Task Force (USPSTF) released a recommendation for LDCT-LCS for adults 55–80 years of age who have a 30 pack-year smoking history and currently smoke, or those who have quit within the past 15 years [3]. Beginning February 2015, LDCT-LCS became a Medicare-covered preventive service for eligible beneficiaries [8]. Under Section 2713 of the Affordable Care Act (ACA) [9], private health plans must provide coverage for annual lung cancer screening for eligible adults aged 55–80 years and may not impose cost-sharing (such as copayments, deductibles, or co-insurance) on patients. New recommendations issued by the USPSTF are required to be covered without cost-sharing beginning in the plan year that begins on or after exactly 1 year from the latest issue date. As the only procedure proven to reduce lung cancer mortality in this high-risk population, implementation of LDCT-LCS could reduce mortality.

While the recent policy changes provide guidelines for screening that should lead to identifying early-stage lung cancer, it is unknown how screening is implemented in real-world clinical care. Data are being collected via the national Lung Cancer Screening Registry, a CMS-approved clinical practice registry gathering information from all patients who undergo LDCT-LCS, and these data will ultimately provide the opportunity to learn about lung cancer screening. Development of the registry, however, is still underway and may take several years to become publicly available [10]. Importantly, the registry tracks patients who are screened only, not those who might be eligible and are not screened. One recent Veterans Health Administration (VHA) study reported low screening rates overall, and several challenges in implementing LDCT-LCS including difficulties identifying patients who were eligible for screening and coordinating the screening and post-screening follow-up processes [11].

Research exploring factors associated with lung cancer screening is still nascent, with early qualitative work exploring perceptions of lung cancer screening among smokers [12–15] and health care providers [16, 17] and focusing on intention to screen rather than actual screening behavior. The clinical trials such as the NLST enrolled a relatively healthy population (e.g., younger, more likely to be former smokers) whose adherence was high [18], and with those declining trial participation described being too old to benefit [15]. Studies examining relationship between smoking status and participation in lung cancer screening showed conflicting results. One large study of nonrandomized lung cancer screening of smokers [19] reported, the self-selected screening participants tended to be current smokers, have smoked a greater number of pack-years, and more likely to have a family history of lung cancer than nonparticipants, whereas another clinical trial identified current smoking as a barrier to trial uptake among high-risk individuals declining participation [20]. Furthermore, the severity of comorbidities among elderly patients may influence LDCT-LCS referrals as potential benefit of screening is minimal for people with major health problem that substantially limits life expectancy or the ability to have curative lung surgery [2]. A positive patient-provider relationship also has been identified as potentially influential in the shared decision making process in cancer screening [21, 22]. Our aim is to characterize and understand the current uptake of lung cancer screening after the policy changes, and describe existing factors associated with LDCT-LCS referrals.

Within a large community healthcare setting in northern California serving a diverse patient population with varying insurance types, we examined the implementation of the USPSTF recommendations for LDCT-LCS following a series of guideline and reimbursement policy changes. The purpose of the study is twofold: a) to assess the changes in documentation of smoking history, number of eligible patients, and medical orders for LDCT-LCS between 2010 and 2016, and (b) to explore patient- and provider-level factors associated with medical orders for LDCT-LCS between 2014 and 2016 (up to 3 years after the 2013 USPSTF recommendation). To our knowledge, this is the first study of these factors in this type of setting.

Using EHR data from a large community healthcare system, we assessed changes in documentation of smoking history and orders for LDCT for lung cancer after the USPSTF released its recommendations. We found substantial variation across providers in referral for LDCT-LCS. Lower medical order rates were found for male providers and foreign medical school graduates, but even after taking into account these factors, we observed large, unexplained variation across providers. Our findings are consistent with a previous population-based study looking at PCP barriers to LDCT-LCS which found that only about half indicated that they knew LCS was recommended by the USPSTF [24]. Reported barriers to LCS among providers included the potential harms associated with LDCT-LCS, uncertainty about the benefit of LDCT-LCS, and questions about insurance coverage for LCS.

At the patient level, we found that younger age, Asian race, being a current smoker, having minimal severe comorbidities, and seeing one’s own PCP were positive predictors of getting LDCT-LCS orders. Previous studies exploring smokers’ attitudes and beliefs about LDCT-LCS in both US and European suggest that being too old to benefit, concerns about cost, inconvenience, perceived smoking-related stigma, fatalism, radiation exposure fears, and distrust of the healthcare system may constitute barriers to LDCT-LCS [12, 14, 15]. The maximum age limit of 77 set by CMS for reimbursement of annual LDCT for lung cancer screening [8] may in part explain the lower referral rates among the oldest age group (78–80 years old), as well as the decreasing relative benefit of screening and early detection with increasing patient age. We found a lower rate (5.8%) of referrals for patients with the most pack-years (≥60 pack-years) but smoking history did not retain a significant relationship with getting LDCT-LCS orders after controlling for all other independent variables.

We found that LDCT-LCS orders are much more likely to occur if the patient has seen his or her own PCP. This makes sense given the counseling necessary before LDCT-LCS. On an interpersonal level, trusting relationships shape whether physicians and patients are able to engage in good discussions on cancer screening [25]. Relationships between patients and their own PCP will increase the likelihood that LDCT-LCS counseling will take place.

In our data, orders for LDCT-LCS have increased since the USPSTF recommendations, but overall remain very low. This is consistent with findings from the National Health Interview Surveys where a significant but still low uptake of CT scans for lung cancer screening was observed in high-risk smokers (5.8% in 2015 vs. 2.9% in 2010, P < .001) [26]. A potential explanation for the low medical order rate in the NHIS data is the lack of documentation of smoking history enabling appropriate selection of high-risk individuals for screening [11]. Recommending and covering LDCT in a targeted, high-risk population is designed to minimize the burden of screening, including the need for confirmatory tests in those with false-positive results and the anxiety related to testing [27]. In our study, although documentation of smoking history improved substantially from 2010 to 2016 (from 59.2 to 77.8%), incomplete documentation of smoking history remains an area for improvement.

Even when smoking status is known, determining screening eligibility for those who are at a high risk of lung cancer may not be straightforward. Many PCPs may not be fully aware of the selection criteria for lung cancer screening and may not be prepared to refer patients for appropriate screening [28, 29]. Further, eligibility based on USPSTF guideline may be debatable. Recent studies have shown that selection for screening using an individualized risk assessment tool is superior (in terms of higher sensitivity and specificity) to the current eligibility criteria based on age and cumulative smoking exposure alone [30]. The feasibility of assessing individual risk during routine primary care visits is challenging but an important area to build additional knowledge.

Studies have shown that, from a physician’s perspective, implementation of the new guidelines is hindered by the lack of available staff time, financial factors and the complexity of the information presented [31]. PCPs practicing in New Mexico clinics reported concerns about the feasibility and appropriateness of implementing LDCT screening, including insufficient infrastructure, access barriers, and financial burdens for patients [16]. A 3-year demonstration project at the VHA contacted all patients deemed eligible for LCS to discuss screening and set up an appointment, but this required devoted staff members including a coordinator, significant time resources, a rigorous and clear implementation guide for healthcare providers, and electronic tools and database [11]. This resource-intensive approach may not reflect what happens when implementing LDCT-LCS in health care systems other than VHA. In the community healthcare system of this study, there had been minimal effort promoting or advertising lung cancer screening at the system level between 2014 and 2016. For example, there was an introduction of the new recommendations for lung cancer screening by pulmonologists at one primary care provider meeting at the largest medical foundation in this healthcare system.

Several limitations of this analysis must be noted. We relied on structured data from billing, procedure, ordering, and administrative records, but many important constructs remain unexamined. For example, we did not examine guideline non-adherence (i.e., referrals for LDCT screening in persons not meeting the USPSTF criteria). We lack information on specific factors influencing patient and provider decision-making including physicians’ verbal recommendations and patients’ preferences, which may be in part available in unstructured, free-text EHR notes. Similarly, the completion of LDCT-LCS warrants further investigation but the organization is not an HMO, it is difficult to know whether a patient receiving a referral obtained a LDCT-LCS from a provider outside the system. Furthermore, using data from a single healthcare organization with a generally highly-insured patient population may limit the generalizability of our findings. By focusing on the adoption and implementation of the guidelines among those who have access to a healthcare system, we reduce the influence of disparities in access in the patterns we observed. In addition, smoking rates in California are well below the national average [32].

Three years after the inception of USPSTF recommendations and Medicare coverage, uptake of LDCT-LCS remains very low. Wider adherence to the guidelines will first require better documentation of smoking history to enable providers to determine patients’ eligibility for lung cancer screening. There is a wide variation across PCPs in a healthcare organization in the uptake of guideline-based practice, suggesting needs for system-wide efforts to facilitate appropriate LCS. Barriers and facilitators specific to lung cancer screening at the patient, provider, and system level need to be better understood to inform targeted interventions to improve the entire lung cancer screening process. The higher rates of orders for current smokers suggests that either providers or patients may see less need for LDCT-LCS if the patient had quit smoking many years before. This is an area worthy of further exploration. Organizational factors should also be explored including the support available to PCPs, workflows, and other “standard practices” that can facilitate appropriate use of LDCT-LCS. Future interventions to improve lung cancer screening should pay particular attention to reaching former smokers, as well as those who are not visiting his/her own PCP.