Lung Cancer Screening Penetration in an Urban Underserved County

To compare residential geography, sex, socioeconomic status (SES), and race/ethnicity of patients screened at Montefiore’s Lung Cancer Screening Program with those of patients diagnosed with lung cancer, assessing whether screening efforts are appropriately focused. This retrospective cohort study involved patients within a multisite urban medical center undergoing lung cancer screening or diagnosed with lung cancer from January 1, 2015 to December 31, 2019. Inclusion criteria were residence within the Bronx, NY and age between 55 and 80 years. Institutional review board approval was obtained. Data were analyzed using the Wilcoxon two-sample t test and χ2. The cohorts comprised 1568 (50.3%) women and 1551 (49.7%) men (mean age 65.6 ± 6.16). The Southeast Bronx had the most diagnosed lung cancers (29.96%) and screenings (31.22%). Sex did not significantly differ (p = 0.053). Cancer and screening cohorts were from impoverished neighborhoods with mean SES of − 3.11 ± 2.78 and − 3.44 ± 2.80 (p < 0.01). The lower tier SES neighborhoods demonstrated more patients in the screening cohort than cancer cohort (p = 0.01). Both cohorts included a majority of Hispanic patients, although race/ethnicity differed significantly (p = 0.01). Lower SES neighborhoods showed no significant difference in race/ethnicity between cancer and screening cohorts (p = 0.262). Though statistically significant differences were found between cohorts, likely due to sample size, few clinically meaningful differences were found, implying our lung cancer screening program was effective in reaching the desired population. Demographics-based programs should be considered in global efforts to screen vulnerable populations.


Introduction
Of all cancers, lung cancer demonstrates the highest mortality worldwide, necessitating early screening methods. It is essential these screens not exclude any demographic, especially those underserved or vulnerable to disease. Lung cancer risk has been shown to vary among race, ethnicity, sex, and socioeconomic status (SES) [1][2][3]. Evidence demonstrates geographic neighborhoods and associated area-based measures of socioeconomic position serve as indicators of lung cancer risk, even while controlling for tobacco exposure [4]. Adie et al. utilized area deprivation index (ADI) to demonstrate higher levels of neighborhood deprivation were associated with higher risks of lung cancer in a cohort of smokers, suggesting that geographic location, which encompasses multiple social determinants of health, should be considered in lung cancer risk models [4].
In meta-analyses, lower SES has been associated with poor coverage of lung cancer screening [5,6]. Furthermore, disparities in cancer screening can be attributed to race and sex. Patient access has been identified as important in screening for lung, breast, and colorectal cancers [7]. Recent studies analyzing screening deficits have focused on mapping geographic areas of the United States with high lung cancer incidences or low access to screens [8]. Other smaller scale studies have demonstrated utility in hospital systems mapping their geographic coverages to assess their efficacies in reaching demographics [9].
Montefiore Medical Center has similarly endeavored to address demographic-based disparities for several disease processes, including atrial fibrillation and ST-elevation myocardial infarctions [10,11]. Specifically, Montefiore-based studies have identified SES as a risk factor for high mortality in atrial fibrillation and breast cancer [12,13]. Montefiore provides care for the majority of the population in the Bronx, an area particularly susceptible to poor health outcomes and with the highest level of poverty of all 62 counties in New York State (NYS) [14]. The Bronx serves as a prototypic impoverished, underserved urban population, with previously identified delays in treatment of head and neck cancers [15]. Moreover, the Bronx has a high prevalence of smokers, ethnicities traditionally underrepresented in studies, and patients with multiple comorbidities [16]. According to the NYS Department of Health, the Bronx reports between 600 and 700 new cases of lung cancer every year [17].

Purpose
The primary aim of the study was to compare residential geography, sex, socioeconomic status (SES), and race/ethnicity of the population who underwent screening at Montefiore's Lung Cancer Screening Program with the population diagnosed with lung cancer. Understanding this relationship will help to assess whether screening efforts are appropriately focused.

Study Design
A retrospective population study was conducted spanning from January 1, 2015 to December 31, 2019. Two cohorts were identified, one consisting of patients who underwent lung cancer screening and the other consisting of patients diagnosed with lung cancer. Patient information was obtained using the electronic medical records (EMR) at Montefiore Medical Center, a multisite academic medical center, and the Montefiore Cancer Registry. Screened patients were identified using the prospective lung cancer screening database. All patients enrolled in our lung cancer screening program met National Lung Screening Trial or Centers for Medicare & Medicaid Services (CMS) eligibility criteria and underwent a low-dose CT with an order type labeled, 'CT chest for lung cancer screening.' The cancer diagnosis group consisted of patients diagnosed with lung cancer and included in the cancer registry.
Inclusion criteria consisted of self-reported residence at one of twenty-five zip codes within the Bronx and age between 55 and 80 years. The zip codes situated patients in one of seven neighborhoods in the Bronx as outlined by the NYS Department of Health [18]. Age criteria corresponded with the 2013 United States Preventive Services Task Force (USPSTF) criteria for lung cancer screening, which recommends screening with low-dose chest CT for those within the age group who have a 30 pack-year smoking history and currently smoke or have quit within the past 15 years [19,20]. Ethnicity and race were self-reported in the medical record and consolidated into one of three categories: non-Hispanic black, non-Hispanic white, and Hispanic. As few patients self-reported as Asian, American Indian and Alaska Native, Native Hawaiian, and Other Pacific Islander, these categories were excluded as outliers. Patients who did not have data for both race and ethnicity were categorized as Other/Unknown.

Institutional Review Board (IRB) Approval
IRB (#2014-3310) and Health Insurance Portability and Accountability (HIPAA) approval were obtained. Informed consent was waived via IRB.

Cohort Population
The cohort population consisted of patients residing in one of seven neighborhoods in the Bronx: Central Bronx, Bronx Park and Fordham, High Bridge and Morrisania, Hunts Point and Mott Haven, Kingsbridge and Riverdale, Northeast Bronx, and Southeast Bronx.

SES Variable
SES score was calculated within the EMR using factor analysis. Using patients' addresses, the EMR linked patients to specific census-block groups, subdivisions of U.S. Census tracts encompassing an average of 1000 people. SES was calculated per census-block group using New York City Department of Health data for each zip code utilizing six variables: (1) log of the median household income, (2) log of the median value of housing units, (3) the percentage of households receiving interest, dividend, or net rental income, (4) the percentage of adults 25 years or older who had completed high school, (5) the percentage of adults 25 years of age or older who had completed college, (6) the percentage of employed persons 16 years of age or older in executive, managerial, or professional specialty occupations. A z-score was generated for each variable, reflecting the deviation of the value from the mean of the population in NYS. The SES score consisted of the sum of the six z-scores, with a score closer to 0 consistent with the mean SES. For example, a score of 1.0 equated a value two standard deviations (SDs) above the mean. Patients from nursing facilities were excluded from scoring as their addresses no longer represented their SES of origin. This standard scoring has been utilized in prior epidemiological studies, including Kargoli et al. and Roux et al. [6,12].

Statistical Analysis
Descriptive analysis was conducted using Statistical Analysis Software (SAS). Descriptive data were analyzed using the Wilcoxon two-sample t test for dichotomous data and χ 2 for other data. Statistical significance was defined by p < 0.05.

Results
The cohorts consisted of 3,119 ethnically diverse patients meeting inclusion criteria, with 1568 (50.3%) women and 1551 (49.7%) men, mean age of 65.6 ± 6.16 years. There was an overlap of 50 patients between groups. The cancer diagnosis cohort (64.7 ± 6.48) was slightly younger than the screening cohort (66.4 ± 5.79) (p < 0.01) ( Table 1). Sex between cohorts did not significantly differ (p = 0.053). Both cohorts comprised predominantly ethnic minorities with a plurality of patients identifying as Hispanic (30.4% in the diagnosis cohort and 38.9% in the screening cohort) (p = 0.01). The fewest patients identified as non-Hispanic white, with 18.4% in the cancer cohort and 17.9% in the screening cohort. Both cancer and screening cohorts were from impoverished neighborhoods with mean SES of − 3.11 ± 2.78 and − 3.44 ± 2.80, respectively (p < 0.01) ( Table 1). Table 2 demonstrates the geographic locations of the cohorts. The Southeast Bronx had both the highest rate of diagnosed lung cancers (423, 29.96%) and lung cancer screenings (533, 31.22%) ( Table 2). However, there was overall difference in geography of lung cancer diagnoses and lung cancer screens (p = 0.0015).
To account for the differences between the neighborhoods, the frequencies of race/ethnicity, SES, and age with respective to individual neighborhoods were also studied. For both the lung cancer and screening cohorts, ethnicity varied significantly between Bronx neighborhoods (p = < 0.0001 and p = < 0.0001, respectively). Furthermore, in the cancer cohort, SES varied by neighborhood (p < 0.0001) as did age (p = 0.0359) (Table 3). Similarly, in the screening cohort, both SES and age varied by neighborhood p < 0.0001 and p = 0.0002, respectively. For both cohorts, the neighborhood with the lowest SES in the screening cohort was Highbridge  (Table 3). To further demonstrate this fact, neighborhoods were divided into two groups of similar size: a group with neighborhoods of overall lower SES scores (Central Bronx, Bronx Park and Fordham, High Bridge and Morrisania, Hunts Point and Mott Haven) and a group with neighborhoods of higher SES scores (Kingsbridge and Riverdale, Northeast Bronx, Southeast Bronx). The lower SES neighborhoods tended to have more patients in the screening cohort than cancer cohort (809 versus 604) (p = 0.010) ( Table 4).
The distribution of ethno-racial groups was also analyzed based on higher and lower SES neighborhoods. χ 2 analysis was performed to assess for differences without including the Other/Unknown groups, to focus on differences among the ethno-racial groups of interest. Among the higher SES neighborhoods, there was a significant difference (p = 0.045) in the distribution of ethno-racial groups, with 100 more Hispanic and 42 more non-Hispanic black patients reflected in the screening group (Table 5). Among the lower SES neighborhoods, there was no significant difference in the distribution of ethno-racial groups (p = 0.262) ( Table 6).

Discussion
As efforts are made to improve early diagnosis of deadly cancers, it is essential to identify discrepancies between populations undergoing screening compared with those diagnosed with cancer. Our study evaluated the reach of lung cancer screening at an urban institution. There were no differences in sex between the cohorts, which is particularly significant, as women have traditionally been underrepresented in cancer screens [21]. Overall, we found statistically significant but probably not clinically meaningful differences in age (64.7 ± 6.48 years versus 66.4 ± 5.79 years), and SES (-3.11 versus -3.44). Lower SES neighborhoods had significantly more patients in the screening than cancer cohort, implying efforts are being made to reach lower SES individuals. The results for SES were especially important, as low SES has been shown to negatively impact care in breast cancer, making patients less likely to communicate with clinical teams and follow up on treatments [22]. It is vital to identify deficits in screening these populations, as there exist strategies to attenuate treatment-related risks by providing health care literacy resources such as pictorial health information [22]. Though both cancer and screening cohorts demonstrated the ethnic diversity characteristic of the Bronx, there were statistical discrepancies in the breakdown of ethnicity between cohorts, of uncertain clinical significance. Importantly, within the lower SES neighborhoods, no differences were identified based on race/ethnicity, implying no discrimination against low-income patients of underserved ethno-racial groups. As the Bronx has a high predominance of Hispanic patients, it concurs that Hispanic patients would be in the majority for both cohorts. Although Hispanic populations in the United States demonstrate lower incidences of cancers, they are more prone to health care disparities related to sociodemographic factors, with one third of the demographic lacking health insurance in 2013 [23]. Recent institution-based studies have identified Hispanic ethnicity as high risk for morbidities including high-mortality STEMIs, incidence of acute critical illness, and metabolic syndrome [11,24,25]. The predominance of Hispanics in the screening cohort suggests our hospital system is reaching this traditionally underserved population.
Although there was a significant difference in the breakdown of neighborhood between cohorts, both cohorts demonstrated the highest number of patients from the Southeast Bronx. The highest rates of screening were in the Southeast Bronx and Bronx Park and Fordham, the neighborhoods where Montefiore has two major campuses. Our analysis identified the Northeast Bronx, which demonstrated the second highest caseload of lung cancer in the borough, as an area toward which more screening efforts should be made. Recognizing a geographic area in need of screening has allowed institutions to distribute patient-facing resources and clinician-led education as well as facilitate primary care referrals [26]. If the number of clinics proves to be a limiting factor, efforts toward establishing mobile clinics should be considered, as mobile low-dose whole body CT screening has emerged as a screening mechanism with a comparable detection rate to that described in the National Lung Screening Trial [27].
The neighborhoods were heterogeneous regarding race/ ethnicity and SES. Interestingly, the data demonstrated variable ranges in SES based on neighborhood. For example, in Kingsbridge and Riverdale, which demonstrated the highest average SES, the standard deviation was 3.652, whereas in the Northeast Bronx, the standard deviation in SES was 1.906. Future studies may investigate whether living in a neighborhood with a high average SES confers health benefits despite an individual falling into a low SES bracket. Previous literature has identified a smaller incidence of nonsmall cell lung cancer in areas with higher income [28]. Moreover, lung cancer incidence is positively associated with tobacco exposure, including secondhand smoke [29]. As individuals within higher SES brackets tend smoke less [30], residence within neighborhoods with higher average SES where carcinogenic exposure is more likely to be minimal may confer a protective effect.
Demographics-based personalization of lung cancer screening programs should be considered in global efforts to achieve better screening. Since our study was conducted, the 2021 USPSTF guidelines have expanded criteria for lung cancer screening to reduce unintended race and gender disparities, which will likely reflect beneficially on our patient population [31]. Future studies expanding on sex in screening should consider the role of gender identity in lung cancer screening, as non-gender conforming individuals have demonstrated lower rates of disease detection, especially when individuals are in underserved areas [32]. These studies will take place in a continued effort to address psychosocial barriers to health and maintain equal access to care among all.

Limitations
Patients were not stratified by smoking behavior when SES was evaluated in neighborhoods within the cancer diagnosis cohort; smoking has previously been identified as a confounding variable in the relationship between SES and lung cancer. That said, recent evidence has limited the significance of this confounding relationship, suggesting that controlling for smoking history attenuates but does not eliminate the significant indirect relationship between SES and lung cancer incidence [5]. There were additional limitations in the method by which the EMR calculated SES, which did not include variables such as the number of domestic occupants supported per household. Some selection bias may have been introduced, as the study cohorts comprised patients who willingly reported to the hospital. Furthermore, demographics were not stratified by age, which was shown to affect the risk profile of ethnicity and medical comorbidities for STEMI patients at Montefiore [11]. A future avenue of study may involve ascertaining the number of patients eligible for Montefiore's lung cancer screening program and comparing the number of patients who present for screening CTs. Another future study may involve evaluating the actual stages of diagnosed lung cancers and positive screening tests, studying whether the discrepancies between these groups vary by demographics.

Conclusion
Overall, the demographics of the lung cancer screening and cancer diagnosis populations at our urban academic medical center were highly similar, indicating that our lung cancer screening program is effective in reaching the most at-risk populations. Demographics-based programs should be considered in global efforts to screen vulnerable populations.