Abstract
Introduction: United States Preventive Service Task Force guidelines recommend annual low-dose computed tomography for lung cancer screening (LCS) for people with significant history of cigarette use. While the national prevalence of lung cancer screening remains low, with known racial and ethnic disparities, studies have yet to examine LCS screening disparities by nativity among Latino patients.
Methods: We evaluated the documentation and prevalence of LCS orders in the electronic health record in Latino patients, by place of birth, in a multistate network of community health centers, over a 10-year study period (2013 to 2022), adjusting for patient and clinical demographic factors.
Results: Among patients who reported current or former cigarette use (n = 126,528), the prevalence of a LCS order was 7.3%. Among 62,483 patients with a documented 30+ year pack-history, US-born Latinos had equal odds as non-Latino Whites to have LCS ordered (OR = 0.71, 95% CI = 0.42-1.21), while odds were lower for both foreign-born Latinos (OR = 0.47, 95% CI = 0.29 to 0.75) and Latinos without a place of birth recorded (OR = 0.63, 95% CI = 0.54-0.73).
Conclusion: The prevalence of LCS was low overall in a large sample of Latino and non-Hispanic White patients with cigarette smoking history. Foreign-born Latino and Latino patients without a country of birth noted in the record had significantly lower odds of having LCS ordered; this should be considered in clinical screening workflows. Nativity and pack-year history were not documented in most patients. More complete documentation of nativity and pack-year history is necessary to fully assess LCS need and equity in Latino patients of heterogeneous nativity.
- Cancer
- Cancer Screening
- Carcinoma
- Community Health Centers
- Electronic Health Records
- Latinos
- Lung Cancer
- Prevalence
- Smoking
Introduction
Cancer continues to be a leading cause of death for Latino communities in the United States (US).1 Lung cancer specifically is the most common cause of cancer deaths in Latinos and second most common cause of cancer deaths in Latinas.1 While lung cancer mortality may be lower in Latino patients compared with other racial or ethnic groups,2 lung cancer tends to be diagnosed at later stages and has high mortality in Latinos.3 Beginning in 2013, the United States Presentive Services Task Force (USPSTF) guidelines recommended low-dose computed tomography screening in older patients with significant tobacco history (2013 guidelines specified ≥30 pack-years among those aged 55 to 80 years) and who are healthy enough to undergo curative lung surgery.4 Wide implementation of these guidelines has been low5–8; for instance, one study found that less than 20% of patients eligible for screening were screened.7
While a growing body of research has examined disparities in screening and risk evaluation by race,6,7,9,10 screening differences among Latinos, including Latino subgroups and non-Latino White patients, has been understudied. Latino patients were represented in low numbers in initial lung cancer screening trials,11 are often not considered as primary cohorts in lung cancer equity studies,12,13 and have been omitted from major reviews on lung cancer screening equity.14,15 Existing evidence on lung cancer screening in Latino patients is based on small samples, with mixed results; some studies suggest lower utilization of lung cancer screening in Latinos7,12 and others have shown no difference between Latino and White patients.5 Though Latino patients are less likely to have a history of using cigarettes, and may smoke fewer cigarettes per day16,17 (a prerequisite for screening), those who smoke have also been shown to receive less evidence-based treatment for their tobacco use than non-Hispanic Whites.16,18 It is also uncertain if these tobacco users experience inequity in lung cancer screening as well.
The US Latino population encompasses a large immigrant subpopulation,19 and place of birth in Latino patients has been associated with differential patterns of preventive care.20 Though, data on variation in lung cancer screening offered to Latino patients, by birth country, is largely unknown. Community health centers (CHCs) provide care to millions of Latino patients nationwide,21 and also provide care to populations more likely to use tobacco.18,21 To understand population-wide lung cancer screening in Latino patients, evaluating the services provided in the CHC setting is crucial, and an evaluation of lung cancer screening in Latino patients by nativity is lacking from the literature.
Given this gap in knowledge of lung cancer screening in Latino patients with a history of cigarette smoking, we evaluated the documentation and prevalence of lung cancer screening in Latino patients, by place of birth, in a multistate network of CHCs with a linked electronic health record (EHR). We hypothesized that, among patients with a history of cigarette use, foreign-born Latino patients will have lower odds of ever having a lung cancer screening, as compared with US-born Latino patients and non-Hispanic Whites.
Methods
Data Source
Data were sourced from the OCHIN Epic electronic health record (EHR) system. As a national nonprofit health IT consultancy, OCHIN provides a fully hosted and shared EHR platform with tailored tools and expertise that are currently used by more than 33,000 providers caring for more than 6 million people in medically underserved communities nationwide. All data, including ethnicity and nativity, were collected during the routine course of clinical care and are used as real world data (secondary) in this observational study.
Study Sample
Inclusion criteria included patients with a current or past history of smoking cigarettes, aged 55 years and over (which was specific to the 2013 guideline criteria, corresponding with the study period) with ≥1 primary care visit to a primary care CHC (546 clinics) across 22 states (AK, AL, CA, CO, CT, GA, ID, IN, LA, MA, MD, MN, MO, MT, NC, NJ, OH, OR, SC, TX, WA, WI) between 1/1/2013-09/30/2022. Past or current cigarette use was defined by the presence of this status in a discrete data field, which is usually updated/reviewed at every clinical encounter. Patients with a diagnosis of lung cancer in an encounter or on the problem list in the period were excluded (n = 314). Pack-years were included when documented in the record and patients with more than 200 pack-years documents were excluded (n = 441). Patient attrition is expected in a clinical setting and variables were operationalized to account for this.
Dependent Variables
The primary outcome was the presence of one or more orders for a lung cancer screening (or other chest computed tomography, as this completion would satisfy screening requirements and diagnostic CTs may be ordered more in this population because of baseline risk) in the study period, as documented in the patient’s EHR. See Appendix Table 1 for the procedural codes used to define the primary outcome.
Independent Variable
The main independent variable combined Hispanic/Latino ethnicity and nativity (place of birth) into 4 mutually exclusive groups1: Foreign-born Latino patients,2 US-born Latino patients,3 Latino patients without a place of birth noted in their record, and4 non-Hispanic Whites. In analyses that subset to Latinos with a country of birth documented, the independent variable of interest was country of birth, and analysis was limited to countries with a suitable sample size for model estimation (ie, country with greater than 100 patients). Ethnicity and nativity are self-reported and documented in discrete EHR fields. Place of birth was obtained from a “country of birth” discrete data field, as previously published.22
Covariates
The following patient-level characteristics were extracted from the EHR and included in regression models as potential confounders: age at first encounter, sex (female or male), household income as a percent of federal poverty level (FPL) (always ≥138%, always <138%, above and below 138%, and never documented to account for missing data), insurance over the study period (always insured or sometimes/never insured), ever having had diagnoses of diabetes, heart disease, obesity, or any other cancer recorded in the EHR, US region (Midwest, Northeast, South, West), and number of primary care encounters (face-to face or telemedicine) per year. In the secondary analysis of all patients, we included the covariate of whether or not they had documented pack-years. These covariates, in this EHR, have been used in previously published research.16,18,23–25
Statistical Analysis
Patient characteristics were described by ethnicity/nativity group, and by specific country of origin for the foreign-born Latino group. We considered 2 analyses to fully understand difference in lung cancer screening orders between ethnicity/nativity groups1: an analysis of study outcomes among patients with ≥30 pack-year history of cigarette use and2 an analysis of study outcomes among patients with current or past cigarette use.
First, because the 2013 guidelines recommended lung cancer screening in patients with a ≥30 pack-year history of cigarette use, we analyzed those patients meeting this criteria. To examine the receipt of at least one lung cancer screening order ever during the study period, generalized estimating equations (GEE) logistic regression26 was utilized including indicator variables for the ethnicity nativity groups and all covariates listed above. Specific country of birth groups had sample sizes too small for meaningful analysis. Estimates of covariate-adjusted prevalence derived from the GEE logistic model and their corresponding 95% confidence intervals are reported.
Second, because many of the patients in our study population did not have pack-years documented, we also analyzed our outcomes among all patients with current and past cigarette use. For the outcome of ever having a lung cancer screening order in this sample, we used a similar GEE logistic model including indicator variables for the ethnicity nativity groups and all covariates. All GEE models used a robust sandwich variance estimator to account for patient clustering within clinics and assumed exchangeable working correlation structure.27 Analyses were conducted in R.v.4.1.2 and Stata v.15.1 and statistical significance was set at P-value < 0.05. Ethical considerations.
This study was approved by the Institutional Review Board of Oregon Health & Science University. Patient consent for data to be used in research was obtained from clinics when care was initiated. Data are available to those who partner with OCHIN for research.
Results
In our overall study population of patients who reported current or former smoking (n = 126,969), the prevalence of a lung cancer screening order was 7.3% (Table 1). Further characteristics of the study population, by independent variable, are included in Table 1. Notable differences in the prevalence of different characteristics by study group include higher prevalence of obesity, diabetes, and being always under 138% of the federal poverty level (a frequent threshold for Medicaid eligibility) in foreign-born-Latinos than the other groups, and higher prevalence of heart disease, lung disease, pack-year documentation, >30-pack year smoking history, and having 0 to 1 encounters per year in non-Hispanic White patients with respect to the Latino groups.
Characteristics of Patients with Current/Past Cigarette Use Ages 55+
In Table 2, we describe the characteristics of patients with and without pack-year documentation. Of note, approximately 50% of patients had pack-years documented. Foreign-born and unknown country of birth Latino patients had a higher proportion of patients without pack-year documentation. Patients with pack-year documentation had a higher prevalence of documented lung disease than those without pack-year documentation. Among patients with ≥30 pack-year history of cigarette use, the prevalence of a lung cancer screening order was 6.3%.
Characteristics of Patients with Current/Past Cigarette Use Aged 55+ by Documented Pack Years
We also described characteristics of the study population by national origin in the 3 countries with the largest representation outside of the United States (Table 3). Of note, Mexican patients in the sample had a higher percentage of patients with pack-years documented than Guatemalan and Salvadoran patients; Mexican patients also had fewer patients in the sometimes/never insured category and had incomes always under 138% of the federal poverty level.
Characteristics of Foreign-Born Latinos with Current/Past Smoking Status by Country
Among those with a documented 30+ year pack-history, compared with non-Hispanic Whites (adjusted prevalence = 7.1%, 95% CI = 6.3 to 7.8), foreign-born Latinos (adjusted prevalence 3.5% 95% CI =1.9-5.1) had lower odds (OR = 0.47, 95% CI = 0.29-0.75) of ever a having lung cancer screening ordered, as did Latino patients without a place of birth documented (adjusted prevalence = 4.6%, 95% CI = 3.9-5.4, OR = 0.63, 95% CI = 0.54-0.73). US-born Latino patients did not differ from non-Hispanic White patients in their odds of lung cancer screening (adjusted prevalence = 5.2%, 95% CI = 2.6 to 7.8, OR = 0.71, 95% CI = 0.42 to 1.21). Figure 1 displays the adjusted odds. Appendix Table 2 displays all regression model results.
Adjusted odds ratios of low dose Computed tomography (CT) orders for patients receiving care in US community health centers from 2013 to 2022 with 30+ documented pack-years. Abbreviation: CI, confidence interval.
We did not have a large enough sample to calculate adjusted odds and prevalences for foreign-born Latino patients with 30+ pack-years documented. The unadjusted model showed no differences in the odds of ever having a lung cancer screening ordered between groups.
Among all patients with a current or past history of cigarette use, compared with non-Hispanic Whites (adjusted prevalence = 7.9% 95% CI = 7.3 to 8.4), US-born Latinos (adjusted prevalence 5.3% 95% CI = 3.7–6.9) had a lower odds (OR = 0.6, 95% CI = 0.5 to 0.9) of ever having lung cancer screening order, as did Latino patients without a place of birth documented (adjusted prevalence = 5.3% 95% CI = 4.7–5.9, OR = 0.6, 95% CI = 0.6 to 0.7). Foreign-born Latino patients also had lower odds of screening orders (adjusted prevalence = 3.8% 95% CI = 2.7–4.8, OR = 0.4, 95% CI = 0.3 to 0.6). Figure 2 displays the adjusted odds.
Adjusted odds of low dose Computed tomography (CT) orders for all patients receiving care in US community health centers from 2013 to 2022. Abbreviation: CI, confidence interval.
Discussion
We performed a novel analysis of a large sample of Latino and non-Hispanic White patients with smoking history over a nearly 10 year study period to assess the prevalence of low-dose computed tomography screening orders for lung cancer screening. Overall, the prevalence of screening was low. While this finding is consistent with other settings and survey-based analyses, the low numbers in this study, among such a large population, are still striking. Our study population, across all groups, has relatively high levels of comorbidities, uninsurance, and low-income status compared with the general population,21 possibly contributing to barriers to receipt of these tests; although, it is unclear from our data (see Limitations section below) whether or not these barriers enter into conversations with providers and any resulting orders for the tests in the first place. Our capture of pack-years, while still low, is and may be higher than other reports because of Epic-related tools and specific national CHC initiatives.28
In those with documented 30+ pack-year history, foreign-born Latino patients and Latino patients without documented country of birth had significantly lower odds of lung cancer screening being ordered. In other cancers, some groups of Latino patients have higher screening prevalence,25,23,29 but we saw a lower prevalence in these specific nativity groups in our analysis. The absolute differences between these groups were small, although in this large, multistate sample, smaller absolute differences may be clinically significant, depending on completion of the ordered service. Barriers to care, including language barriers, access to transportation, access to imaging/hospital services could all play a role in this disparity, which requires accessing care outside of the community health center. Findings may also be due related to the prevention associated with this specific cancer type (eg, lung), in contrast to breast cancer for example,30 where there is less societal wide knowledge and/or concern. Regardless, in the setting of low screening overall, it is crucial to realize that some groups may be even less likely to receive screening.
While clinics and providers should design lung cancer screening workflows to improve overall screening prevalence, they should consider workflow improvements tailored for foreign-born Latino patients. It is notable, however, that US-born Latinos did not differ from non-Hispanic Whites in their receipt of a lung cancer screening order. This suggests some measure of equity in ordering of lung cancer screening among different ethnic groups in community health centers, albeit, again, with overall low prevalence of orders for screening. Our findings may also provide a rationale for systematic collection of place of birth in Latino and other patient populations. While challenging to interpret the screening odds of the “unknown” birth country group and those of Latinos from specific country groups, which had low screening overall, these groups’ comparisons would have benefitted from larger sample sizes as a result of a systematic collection of place of birth. In many Latino populations, who may have a significant proportion of foreign-born members, collecting and understanding place of birth should be a persistent priority for cancer screening equity.
Our findings underscore the importance of collecting pack-year (or smoking amount) data. EHRs have been shown to underreport pack-years,31 and in our analysis Latino groups met the 30-pack year criteria less often. And since ∼50% of patients with cigarette smoking history in the EHR had a documented assessment of amount in these data fields overall, improving the collection of this data in all patients should be paramount, to properly assess the uptake of screening and to evaluate its effectiveness in multiple populations.
Limitations
We performed our analysis using data from patients in CHCs, and these findings may not generalize to other settings. However, this remains an evaluation of cancer screening orders across a large population in a setting representative of where > 30 million people receive their care.21 As described above, pack-year history and nativity were not collected in many study patients, limiting our conclusions. However, given the paucity of large-scale studies of lung cancer screening in low income populations, this analysis makes a novel contribution to the scientific landscape on lung cancer prevention. The calculation of pack-years may be imprecise in this data; however this imprecision matches the imprecision in the real-world delivery of care, because our data matches that used to counsel patients. Further initiatives will need to improve the capture of this data and the surrounding counseling discussions. Our data also contain a low prevalence of those with lung cancer on their problem list, suggesting that additional lung cancer related data may be missing, which could bias results. In addition, importantly, this analysis of discrete EHR data fields was not able to consider declinations, provider-patient conversations, provider rationale not documented in the EHR or documented only in free text fields, or the time elapsed between quit date and a given clinical encounter, as the guidelines specify that “screening should be discontinued once a person has not smoked for 15 years or develops a health problem that substantially limits life expectancy or the ability or willingness to have curative lung surgery.” Further work can examine these encounters in CHCs and other settings and in other groups. Our decision to include all lung CT orders (because if completed, they would satisfy recommended guidelines) may bias results toward higher prevalence of screening orders. However, excluding patients with diagnostic tests could a) bias against toward a lower screening order prevalence in a population more likely to get a diagnostic lung CT because of baseline risk. We were also unable to assess receipt of tests, which are not conducted in CHCs. This is a necessary future step to understand lung cancer screening rates in this CHC population and whether this varies by ethnicity and/or country of birth. The Center for Medicare and Medicaid Services decided to cover this expense in 2015.32 Finally, in 2021, the USPTSF guidelines were revised to specify that individuals aged 50 to 80 years with a 20 or more pack-year history undergo screening. Analyses have demonstrated that these guidelines may have increased eligibility in Latino patients.33,34 We were unable to analyze the period between March 2021 and September 2022 separately due to inadequate sample size. However, we included this several month period at the very end of our study period in our main analyses, for several reasons: 1) those meeting old criteria are still eligible in the new period, 2) clinicians are unlikely to change their practice patterns rapidly,35 and 3) attempting to evaluate changing practice patterns in this limited time frame so close to the guideline change is likely unrealistic and biased. However, it remains a limitation of this work, and future research is warranted to better understand how the 2021 USPSTF guideline revision may have differentially affected Latino patients with a cigarette smoking history in real-world settings, as we accumulate more years of data.
Conclusion
We performed a novel analysis of Latino and non-Hispanic White patients with current or former smoking history over a nearly 10-year study period to assess the prevalence of orders for low-dose computed tomography screening for lung cancer. The prevalence of screening orders was low overall in community health centers, and focused efforts should be made to increase orders among those who are eligible for screening. In those with documented ≥30 pack-year history, foreign-born Latino patients and Latino patients without a country of birth noted in the record had significantly lower odds of having a lung cancer screening ordered. Nativity and pack-year history were not documented in all patients. More complete documentation of these elements is necessary to fully assess lung cancer screening equity in Latino patients of heterogeneous nativity, and clinics and providers should be aware that certain groups of Latino patients may be at risk for underutilization of screening even more than broader populations.
Acknowledgments
The research reported in this work was powered by PCORnet®. PCORnet has been developed with funding from the Patient-Centered Outcomes Research Institute® (PCORI®) and conducted with the Accelerating Data Value Across a National Community Health Center Network (ADVANCE) Clinical Research Network (CRN). ADVANCE is a Clinical Research Network in PCORnet® led by OCHIN in partnership with Health Choice Network, Fenway Health, University of Washington, and Oregon Health & Science University. ADVANCE’s participation in PCORnet® is funded through the PCORI Award RI-OCHIN-01-MC. This study was approved by the Institutional Review Board of Oregon Health & Science University. Raw data underlying this article were generated from multiple health systems across the OCHIN network; restrictions apply to the availability and rerelease of data under organizational agreements.
Appendix
Adjusted Logistic Regression Model for Patients with 30+ Pack Years Recorded
Adjusted Logistic Regression Model for Everyone in the Sample
Notes
This article was externally peer reviewed.
Funding: National Cancer Institute (Grant no. R01CA258464), Silver Family Innovation Fund.
Conflict of interest: The authors have no conflicts of interest to declare.
To see this article online, please go to: http://jabfm.org/content/38/3/451.full.
- Received for publication October 15, 2024.
- Revision received December 27, 2024.
- Accepted for publication February 21, 2025.
