Disparities in Lung Cancer Screening Availability: Lessons from Southwest, Virginia

Abstract

Background:

Little is known about health disparities in access to low-dose computed tomography (LDCT) screening. We hypothesized the current capacity for LDCT screening would be exceeded by the number of at-risk individuals in Virginia.

Methods:

Cancer incidence data and adult smoking rates for Virginia were obtained from public sources between 2006-2012. The American College of Radiology website was queried in 2015 to identify lung cancer screening facilities in Virginia, which were surveyed. Spatial exploratory data analysis was used to examine secondary data and descriptive analysis was used to examine primary survey data.

Results:

Rural counties have higher lung cancer death rates and smoking rates than metropolitan counties. Despite tremendous burden for LDCT screening in rural counties, particularly in Southwest, VA, there were only two LDCT facilities. In total, 37 accredited LDCT facilities were identified in Virginia. On average, facilities had been screening for 14.6 months and screened an average of 76 patients.

Conclusions:

At-risk smokers in Virginia, particularly those living in rural areas with high smoking rates, do not have adequate recommended LDCT coverage. More screening centers are needed to care for the high number of rural smokers at-risk for lung cancer.

After the results of the National Lung Screening Trial (NLST) were released in 2011, health economists and policy-makers scrambled to make accurate projections about the potential economic impact of such a screening program.^1-3 The NLST was a multicenter, randomized controlled trial that compared low-dose computed tomography (LDCT) to chest radiography in current and heavy smokers and demonstrated a relative risk reduction in lung cancer-specific mortality of 20%.³ The recently completed NELSON trial also confirmed the importance of LDCT in risk reduction of lung cancer among high risk individuals, in data that has yet to be published. Lung cancer prevention is an area prime for quality improvement in screening and management given the persistently high incidence age-adjusted mortality rates of lung and bronchus cancer.⁴ On February 5, 2015, the Centers for Medicaid and Medicare Services (CMS) approved coverage for lung cancer screening in select high risk patients, most notably for asymptomatic adult patients aged 55-77 years with a 30 pack-year history of smoking.⁵ These coverage recommendations are aligned with those of the American Association for Thoracic Surgery (AATS), which were released in 2012. The age cutoff per AATS guidelines is 79 years, slightly differing from CMS recommendations. ⁶ Unfortunately, very little has been written regarding the health disparities in availability and uptake of LDCT lung cancer screening.

There are critical barriers in accessing screening programs for those in resource-limited settings, such as rural or underserved medical areas. For this study, defined areas at high risk for lung cancer screening-related disparities based on higher than average rates of smoking and/or ever-smokers and simultaneous lower than average rates of healthcare facility access and socioeconomic disadvantage.⁷ Considering these criteria, Southwest Virginia (VA) is an area of significant concern for lung cancer screening access. Adult smoking rates are 30% in Southwest VA, exceeding the national average of 17% and the state of Virginia rate of 19%.^8,9 Importantly, access to American College of Radiology (ACR) certified lung cancer screening centers is limited with only two such centers located in Southwest VA.¹⁰

In 2013, there were 10 million people in the United States eligible for lung cancer screening with LDCT. Unfortunately, even following United States Preventative Services Task Force (USPSTF) recommendations for lung cancer screening, only 3.9% of eligible individuals underwent screening in 2015—this represents only a 0.6% increase in screening rates compared to 2010.¹¹ Academic/urban medical centers likely perform the majority of lung cancer screening, as the CMS requirements for screening require numerous resources, including radiologists, nurses, and practitioners who will engage in a shared decision-making visit, and coordination across disciplines to coordinate the follow-up care for patients who screen positive.

The purpose of the current study is to demonstrate the health disparities related to lung cancer risk in Southwest, VA. Specifically, we hypothesized that the individuals with the highest smoking rates would likely reside in parts of Virginia with less or limited capacity for lung cancer screening. Our primary objective was to assess the current capacity for lung cancer screening in Virginia through a telephone survey. We also sought to estimate and spatially represent county-level incidence rates in geographic relation to current ACR approved lung cancer screening centers.

Patients and Methods

Primary and Secondary Data Sources

Both primary and secondary data were utilized for this study. Cancer incidence data for Virginia from 2006 to 2012 were obtained from the Virginia Central Cancer Registry (VCCR). Data for adult smoking rates were acquired from the Behavioral Risk Factor Surveillance System (BRFSS). County-level BRFSS data for the period of 2006-2012 were used. The American College of Radiology website was queried in 2015 to identify names, contact information and locations of all facilities in the State of Virginia currently certified to perform LDCT for lung cancer screening. An initial call was made to all identified facilities using the phone number listed in the registry. If no one at the facility was available to answer questions during the first phone call, contact information was left in the form of a return phone call and/or email address. If the contacted facility did not follow-up with a response, the facility was contacted again within the same business week. This process was repeated until the end of the two-month call period.

Study Outcomes

We sought to estimate incidence rates of lung cancer and smoking rates per county in Virginia. Generalized linear mixed models (GLMMs) were built to estimate county-level smoothed incidence rates. The model incorporated county-level random effects with structured covariance matrix to take into consideration spatial autocorrelation, with the unit of analysis being each county. Implementation of commonly used spatial GLMMs using SAS procedure GLIMMIX are detailed in Rasmussen et al.¹² Multilevel modeling was used to model smoother smoking rates from the BRFSS at the county level. The proposed method using unit-level GLMMs implemented in SAS is detailed in Zhang et al.;¹³ this method takes into consideration unequal probability sampling of individuals in the BRFSS survey.

The potential need for lung cancer screening at the county level was estimated by multiplying four factors: smoking rates × adult population × percentage of heavy smokers × percentage of smokers desiring screening services. Although there is no universal definition of “heavy smoker”, the generally accepted definition is smoking >20 cigarettes per day for 10-20 years (10-20 pack years).¹⁴ We used 10% as a conservative estimate of the percentage of heavy smokers.¹⁵ There is no universally accepted definition of heavy smoker. CMS has used a definition of 30 pack-years, as specified in their decision memo regarding screening recommendations.¹⁶ We hypothesize, based on high rates of smoking in Virginia, that more than 10% smokers >age 55 would meet this definition; however, 10% has been specified as a conservative estimate. Since most people desire screening in the case of breast and colorectal cancers,^17,18 we used 60% as the estimate of the percentage of smokers that want to screen for lung cancer. For the facility survey, questions were asked about number of months screening, number of patients screened in the past year, cost of LDCT, and method of payment for screening.

Statistical Analysis

Spatial exploratory data analysis was used to examine the secondary data and descriptive analysis to examine the primary survey data. The goal was to map, describe, and visually represent the overall spatial structure of lung cancer incidence rates, LDCT facilities, and the potential need for LDCT screening in Virginia. SAS 9.4, ArcGIS, and Google Earth were used to perform the analyses. The University of Virginia Institutional Review Board determined that the data did not involve human subjects and was, therefore, exempt.

Results

LDCT Facilities and Potential Need for Screening

Crude lung cancer incidence rates in Virginia are mapped in Figure 1. Not surprisingly, rural counties, especially those Appalachian counties in the Southwest corner of the state, have higher lung cancer death rates than Northern Virginia, the Richmond metropolitan area, and the Norfolk-Virginia Beach area. Moreover, Figure 2 demonstrates that those counties with high lung cancer death rates had higher adult smoking rates than other counties. Many of these counties have adult smoking rates higher than 30%. In contrast, only 10% of adults are smokers in northern counties with geographic proximity to Washington, DC. Using data from VCCR and BRFSS, were able to estimate the potential need for lung cancer screening in all counties. There is an overwhelming need for LDCT screening services in rural counties, particularly in Southwest VA; however, there were only two LDCT facilities in the region. In total, 37 accredited LDCT facilities were identified in Virginia. These programs were concentrated in areas with very low smoking rates.

Figure 1.

Smoothed crude lung cancer incidence rates per 100,000 Virginia residents, mapped by county.

Figure 2.

Smoothed adult smoking rates including persons age 55 or older for Virginia with corresponding estimated demand for lung cancer screening, mapped by county

LDCT Facility Telephone Survey

Between October 1, 2015 and November 30, 2015, all 37 LDCT programs were contacted. The location of each screening center is not reported here to maintain anonymity. All centers were current ACR accredited centers for lung cancer screening. In total, 25 of 37 accredited centers (68%) provided at least partial answers to the survey questions. On average, centers had been screening for 14.6 months (median = 12 months, SD = 9 months, see Table 1). Two centers were unable to give an exact estimate and responded as “several years.” Conservatively, this was converted to 24 months. Only 15/37 centers (41%) were able to provide an estimate of the number of patients screened. One center was excluded because of an inability to provide aggregate data across all of their radiology centers, some of which were not ACR accredited. On average, screening facilities had screened 76.4 patients (median 25, SD 154). Twenty-one out of 37 centers (56.8%) reported how patients were paying for their imaging studies (Figure 3). There was a combination of multiple payment options (Medicare, out-of-pocket, private insurers) reported with the majority allowing self-pay in combination with other payment methods. Two centers explicitly stated that patients were only screened by physician referral and that self-pay was not allowed. Most centers were aware of the Medicare requirement for shared decision-making prior to reimbursement approval. Several were unaware that a CPT code for LDCT had already been released. Twenty out of 37 centers (54%) reported either an exact amount or estimate for the out-of-pocket cost of LDCT ($262.55, median $252.00, SD $106.26). Several centers were unaware whether or not cost provided included the radiologist read.

Table 1:

Results of Screening Center Surveys (N = 37)

Are you currently screening? (n, %)	25, 67.6%
How many months screened? (average, range)	14.7, 3-36
How many patients screened in last year?	76.4, 0-660
How much did LDCT cost?	$262.55, $106.26-$590

Figure 3.

Survey results for patient payment method for low-dose computed tomography (LDCT) lung cancer screening.

Comment

The USPSTF recommendation for lung cancer screening recognize that smoking history, including amount smoked and length of time smoked, are the most important modifiable risk factors for prevention of lung cancer.^3,19 This study identified an important geographical barrier to accessing lung cancer screening programs for those in resource-limited settings. Virginia counties with the highest smoking rates were found to have insufficient access to lung cancer screening centers. For the present study, we surveyed all 37 certified LDCT screening centers in Virginia, and on average, screening facilities had only screened 76 patients in the past year, totaling far less than the number of at-risk residents. With recent studies demonstrating increased cancer-related incidence and mortality for patients with cancer in non-metropolitan areas.^20,21 Even more concerning are potential differences in access to treatment for cancers that are primarily treated with surgery. A recent large, retrospective study examining urban-rural differences in lung cancer care showed that rural patients with stage 1 non-small cell lung cancer (NSCLC) received fewer surgical interventions compared to urban patients with similar disease stage.²²

According to the Virginia Department of Health (VDH) and the CDC, the state of Virginia has an adult smoking rate of 19%, a number that is above the national average of 17%. This presents a potential problem in that the at-risk screening population is likely higher than the state’s current capacity to provide these annual screening LDCTs. Per our study estimates, lung cancer incidence rates differed widely across the state of Virginia with the highest rates being found in Lee County in far Southwestern, VA (estimated 117-134 cases per 100,000) and Eastern shore (117-160 cases per 100,000 depending on the county), in addition to higher rates in the centrally located Nelson County, VA. In contrast, the VDH reported age-adjusted lung cancer incidence rates for health districts (rather than counties) 2009-2013, which included rates of 95.6 cases per 100,000 for Roanoke in southwestern VA and 88.4 per 100,000 for the Eastern Shore region, rates lower than those found in the present study.²³

The financial impact on institutions that are capable of meeting these requirements are potentially quite large given the population of Virginians that live in medically underserved regions. Southwest Virginia is a medically-underserved region of the state, and many residents have no easily accessible large or academic medical center. Considering the resources needed to provide an effective screening program—including technical and administrative mechanisms, LDCT capabilities, equipment accreditation, and sub-specialty reads—it is not surprising that the smaller medical centers in Southwest VA are ill-equipped to provide screening services, as evidenced by the presence of only two screening centers in the region.

The burden of screening and referrals for lung cancer will likely be left on the shoulders of primary care providers (PCPs), who often have low-quality information regarding how to appropriately refer patients to specialty care.²⁵ Furthermore, PCPs are already potentially overburdened by their current workload, which involves making age-appropriate screening recommendations and referrals for other cancer types. Given the lack of nearby screening centers, identified in the current studies, PCPs may have a difficult time finding a geographically convenient location to refer their at-risk patients for screening.

Unlike other cancers that are screened for, lung cancer screening approval requires a shared decision-making visit—a potential additional barrier to screening for low-income populations, as this requires another trip to a doctor or screening facility. Although the shared decision-making visit requires extra time and money spent on the part of both the patient and provider, research has shown that the visit does improve patient knowledge regarding the need for lung cancer screening.²⁶ It is not clear from currently available data that improved knowledge results in increased screening uptake among this patient population and may just be an added burden of cost and time. Future studies that look at patient-centered perspectives and outcomes should assess for the potential additional financial burden imposed by the shared decision making visit and should look at its potential impact on decreasing uptake of lung cancer screening.

Compounding this access issue is the mismatch between smoking prevalence and density of screening centers throughout the state, where rates of smoking in Southwest VA approach 30%.⁹ Eberth and colleagues utilized ArcGIS data, number of LDCT screening centers and state and county population totals to create measures of lung cancer screening capacity based on geographic proximity and population density. In their study, Virginia was identified as a state with one of the best capacities for screening at 0.8 per 100,000 persons aged 55-79, however they demonstrate a clear concentration of screening centers in northern and eastern VA, where smoking rates are lowest.²⁷ Furthermore, although Centers for Medicaid and Medicare Services (CMS) recently approved coverage for lung cancer screening, LDCT screening scans cost approximately $200 and are not yet covered by all insurance companies. Given the higher prevalence of smoking among uninsured individuals, many uninsured and underinsured screening candidates will find yearly screening cost-prohibitive. With potential changes to key pieces of the Affordable Care Act in the future, there are concerns that gains made in decreasing the rate of uninsured will be lost if alterations in coverage for smoking-related cancers, such as lung adenocarcinoma, are made.²⁸

This present analysis represents a novel assessment of the readiness for implementation of lung cancer screening recommendations for patients in underserved parts of the US. There are multiple limitations of the current study, most notably a lack of patient-centered perspectives, which the current study methodology does not account for given that patient-level data was not collected. Another limitation is lack of availability of data regarding screening centers that are nearby Virginia, but not technically in Virginia, such as southeastern Tennessee/Kentucky or Washington, DC. Predictably, there were challenges associated with the collection of data via survey response, with 68% of Virginia screening centers providing responses. As several centers were affiliated with one overarching entity, if the single responsible employee could not be reached after multiple attempts, more than one screening center could not be included. Additionally, at least one screening center did not wish to answer questions as program protocols were undergoing revision causing concern that survey responses could be misinterpreted. There were also challenges associated with reporting both the duration of screening and number of patients screened. The lack of a designated contact person at multiple centers limited the availability of these data. There was also variability in the precision of these estimates, as some centers did not seem to be accurately recording these data while others were preparing to report to official databases, such as the ACR.

Conclusion

This study identified barriers to lung cancer screening for Virginia residents, related primarily to geographic location. We believe the current inability of Virginia to provide adequate lung cancer screening for at-risk patients is a major public health problem that needs to be addressed at the State and Federal levels.