Changes in Cancer Screening in the US During the COVID-19 Pandemic

Key Points

Question  Did the national prevalence of breast, cervical, and colorectal cancer screening change during the COVID-19 pandemic?

Findings  In this national survey study, between 2018 and 2020, past-year breast and cervical cancer screening prevalence declined by 6% and 11%, respectively. There was no change in past-year colorectal cancer screening prevalence, with a 7% increase in stool testing and a 16% decrease in colonoscopy.

Meaning  These findings suggest that stool testing counterbalanced decreases in colonoscopy during 2020, whereas breast and cervical cancer screening decreased modestly.

Importance  Health care was disrupted in the US during the first quarter of 2020 with the emergence of the COVID-19 pandemic. Early reports in selected samples suggested that cancer screening services decreased greatly, but population-based estimates of cancer screening prevalence during 2020 have not yet been reported.

Objective  To examine changes in breast cancer (BC), cervical cancer (CC), and colorectal cancer (CRC) screening prevalence with contemporary national, population-based Behavioral Risk Factor Surveillance System (BRFSS) data.

Design, Setting, and Participants  This survey study included respondents from the 2014, 2016, 2018, and 2020 BRFSS surveys who were eligible for BC (women aged 50-74 years), CC (women aged 25-64 years), and CRC (women and men aged 50-75 years) screening. Data analysis was performed from September 2021 to February 2022.

Exposures  Calendar year.

Main Outcomes and Measures  Self-reported receipt of a recent (defined as in the past year) BC, CC, and CRC screening test. Adjusted prevalence ratios (aPRs) comparing 2020 vs 2018 prevalence and 95% CIs were computed.

Results  In total, 479 248 individuals were included in the analyses of BC screening, 301 453 individuals were included in CC screening, and 854 210 individuals were included in CRC screening, In 2020, among respondents aged 50 to 75 years, 14 815 (11.4%) were Black, 12 081 (12.6%) were Hispanic, 156 198 (67.3%) were White, and 79 234 (29.9%) graduated from college (all percentages are weighted). After 4 years (2014-2018) of nearly steady prevalence, past-year BC screening decreased by 6% between 2018 and 2020 (from 61.6% in 2018 to 57.8% in 2020; aPR, 0.94; 95% CI, 0.92-0.96), and CC screening decreased by 11% (from 58.3% in 2018 to 51.9% in 2020; aPR, 0.89; 95% CI, 0.87-0.91). The magnitude of these decreases was greater in people with lower educational attainment and Hispanic persons. CRC screening prevalence remained steady; past-year stool testing increased by 7% (aPR, 1.07; 95% CI, 1.02-1.12), offsetting a 16% decrease in colonoscopy (aPR, 0.84; 95% CI, 0.82-0.88) between 2018 and 2020.

Conclusions and Relevance  In this survey study, stool testing increased and counterbalanced a decrease in colonoscopy during 2020, and BC and CC screening modestly decreased. How these findings might be associated with outcomes is not yet known, but they will be important to monitor, especially in populations with lower socioeconomic status, who experienced greater screening decreases during the COVID-19 pandemic.

The emergence of the COVID-19 pandemic in the first quarter of 2020 caused disruptions in health care utilization in the US.^1,2 Stay-at-home orders were issued, and professional societies recommended pausing routine cancer screening tests. Several reports based on commercially or Medicare insured adults note sharp decreases (80%-90%) in breast cancer (BC), cervical cancer (CC), and colorectal cancer (CRC) screening volumes in March and April of 2020.^2-5 After the implementation of safety protocols and reopening of cancer screening services, screening volumes improved to prepandemic levels during the summer of 2020, but perhaps not enough to overcome previous decreases.³

The magnitude of potential deficits and population-based estimates of cancer screening during 2020 are not yet known, because previous studies^3-9 using medical claims and records, from which pandemic-related screening disruptions have been measured, were based on restricted geographical regions or were conducted among people who maintained the same health insurance coverage throughout 2020. Furthermore, those previous studies^3-9 were not able to examine both recent and guideline-concordant screening because medical claims often do not contain a person’s longer-term screening history. Both are important to measure because short-term screening practices may be more closely linked to health care disruptions, whereas adherence to longer screening intervals may be more closely associated with cancer outcomes. It is also unknown whether any changes in population-wide screening were worse among populations who have historically faced barriers to accessing health care and may be especially vulnerable to health care disruptions.¹⁰

In this survey study, we examined recent and guideline-recommended BC, CC, and CRC screening prevalence, with population-based 2020 Behavioral Risk Factor Surveillance System (BRFSS) data and compared it with prior years. We also examined whether potential changes in cancer screening prevalence varied according to race, ethnicity, and socioeconomic status (SES) to assess whether inequities in screening worsened.

This was a survey study based on 2014, 2016, 2018, and 2020 BRFSS data, an annual state-based telephone survey overseen by the Centers for Disease Control and Prevention, with respective response rates of 47.0%, 47.0%, 49.9%, and 47.9%.¹¹ The study was based on deidentified publicly available data, which the US Department of Health and Human Services considers as nonhuman research and does not require institutional review board review or informed consent.¹² The National Health Center for Statistics data suppression guidance was followed.¹³

In even-numbered survey years, respondents were asked whether they have ever had a specific screening test and when their most recent test occurred (eTable 1 in the Supplement). BRFSS conducts interviews throughout the year and, in 2020, the national proportions of interviews conducted per month were comparable to those in previous years. However, several states halted interviews beginning in March 2020 and, as such, we focused on national estimates (eTable 2 in the Supplement).

The primary outcomes were self-reported receipt of BC screening (mammography), CC screening (Papanicolaou and human papillomavirus testing [HPV]), and CRC screening (colonoscopy, stool testing, stool DNA, computed tomography colonography, or sigmoidoscopy) in the past year. We focused on past-year screening because health care disruptions may be more closely associated with recent screening testing. We also considered whether participants were up to date (UTD) with US Preventive Services Task Force (USPSTF) recommended cancer screening, which have longer intervals (eg, colonoscopy every 10 years for CRC screening; see eTable 1 in the Supplement), because being UTD is more closely associated with cancer outcomes. Prostate-specific antigen testing for prostate cancer screening was not considered because it is recommended only with shared decision-making.¹⁴

Primary analyses were restricted to the following age ranges to align with USPSTF recommendations: 50 to 74 years for BC, 25 to 64 years for CC, and 50 to 75 years for CRC screening. Respondents with cancer or who were missing cancer screening data (BC, 32 219 individuals [6.3%]; CC, 47 013 individuals [8.1%]; and CRC, 77 832 individuals [8.4%]) were excluded from the analyses (eTable 3 in the Supplement).

Covariates were selected on the basis of previously reported associations with cancer screening and included self-reported race and ethnicity (ie, American Indian or Alaska Native, Asian or Pacific Islander, Hispanic, non-Hispanic Black, non-Hispanic White, and other [ie, not otherwise specified]), age, sex (for CRC only), annual household income, educational attainment, having a usual source of care, and a primary care visit in the past 12 months.^15-17 Employment status and insurance coverage (yes vs no) were considered among people younger than 65 years.

Data analysis was performed from September 2021 to February 2022. Past year and UTD screening prevalence were examined by year. Probabilities calculated from logistic regression models were used to estimate unadjusted and adjusted prevalence ratios (aPRs) and 95% CIs.¹⁸ To determine whether inequities in screening worsened, changes in screening prevalence according to sociodemographic and socioeconomic factors were assessed. In sensitivity analyses, aPRs comparing screening in April to December of 2020, after the onset of the pandemic, vs 2018 were computed. To determine whether changes during 2020 and specifically in April to December were part of ongoing seasonal trends, prevalence according to interview quarters as computed. All estimates used Centers for Disease Control and Prevention–recommended weights to be nationally representative and account for nonresponse. SAS statistical software version 9.4 (SAS Institute) was used for analyses. Significance (P < .05) was determined with 2-sided χ² tests.

In total, 479 248 individuals were included in the analyses of BC, 301 453 individuals were included in the analyses for CC, and 854 210 individuals were included in the analyses for CRC screening. In 2020, among respondents aged 50 to 75 years, 14 815 (11.4%) were Black, 12 081 (12.6%) were Hispanic, 156 198 (67.3%) were White, and 79 234 (29.9%) graduated from college (all percentages are weighted). In terms of health care access and utilization, 10.5% of nonelderly respondents (9890 individuals) were uninsured, and most respondents reported a primary care practitioner visit and usual source of care, similar to 2018 proportions (Table 1). Analogous patterns were observed for persons eligible for BC and CC screening.

Between 2018 and 2020, past-year BC screening decreased by 6% between 2018 and 2020 (from 61.6% in 2018 to 57.8% in 2020; aPR, 0.94; 95% CI, 0.92-0.95), and CC screening decreased by 11% (from 58.3% in 2018 to 51.9% in 2020; aPR, 0.89; 95% CI, 0.87-0.91) after 4 previous years (2014-2018) of mostly stable BC and CC screening prevalence (Figure and Table 2). In 2020, an estimated 2.13 and 4.47 million fewer women reported receiving BC and CC screening, respectively, in the past year than in 2018 (eFigure 1 in the Supplement). Decreases in 2020 past-year BC and CC screening prevalence were largely associated with decreases beginning in the third quarter of 2020, a seasonal pattern that was not observed in previous years (Figure). Past-year screening prevalence during April through December 2020 was 8% (aPR, 0.92; 95% CI, 0.90-0.94) lower for BC and 12% (aPR, 0.88; 95% CI, 0.86-0.90) lower for CC than 2018 estimates (eTable 4 in the Supplement).

For CRC screening, past-year colonoscopy prevalence decreased by 16% (from 15.6% to 13.2%; aPR, 0.84; 95% CI, 0.82-0.87), whereas the prevalence of stool testing increased by 7% (from 11.4% to 12.1%; aPR, 1.07; 95% CI, 1.02-1.12) between 2018 and 2020 (Table 2). Sigmoidoscopy was rarely used (2018 prevalence, 0.6%; 2020 prevalence, 1.1%). In 2020, 2.9% of respondents reported a recent sDNA test, and 1.1% reported a recent computed tomography colonography. There was no change in past-year CRC screening between 2018 and 2020 when sDNA and computed tomography colonography were considered (aPR, 1.02; 95% CI, 0.99-1.04) or when these 2 tests were excluded from 2020 estimates (aPR, 0.90; 95% CI, 0.95-1.00).

In 2020, past-year BC and CC screening prevalence rates were lower in people with less than a high school diploma compared with those with college degrees (BC, 50.3% vs 62.2%; CC, 43.6% vs 55.7%) (Table 3). In addition, between 2018 and 2020, past-year BC screening prevalence decreased by 10% (aPR, 0.90; 95% CI, 0.83-0.97) and CC screening prevalence decreased by 17% (aPR, 0.83; 95% CI, 0.75-0.91) among people with less than a high school diploma compared with a 6% decrease for BC (aPR, 0.94; 95% CI, 0.91-0.96) and a 9% decrease for CC (aPR, 0.91; 95% CI, 0.88-0.93) among college-educated respondents (Table 4). Past-year stool testing was an exception to this pattern, where there was no change for college graduates, and people with less than a high school diploma reported a striking 39% increase between 2018 and 2020 (aPR, 1.39; 95% CI, 1.20-1.62) (Table 4). However, overall CRC screening utilization remained lower in people with lower educational attainment (Table 3).

Past-year BC screening prevalence decreased more among Hispanic (aPR, 0.90; 95% CI, 0.83-0.97), Asian (aPR, 0.73; 95% CI, 0.60-0.88), and American Indian or Alaska Native (aPR, 0.83; 95% CI, 0.73-0.95) women compared with White women (aPR, 0.95; 95% CI, 0.93-0.97) (Table 4). Decreases in past-year CC screening prevalence were also greater among Hispanic women (aPR, 0.83; 95% CI, 0.77-0.88) than White women (aPR, 0.89; 95% CI, 0.87-0.91) (Table 4). Among Asian persons, the decrease in colonoscopy prevalence (aPR, 0.64; 95% CI, 0.47-0.88) was greater than those for other racial and ethnic groups, and there was not a concomitant increase in stool testing; past-year CRC prevalence decreased in 2020 among Asian persons (aPR, 0.87; 95% CI, 0.70-1.09) but not significantly from 2018. Changes in self-reported past-year CRC screening prevalence for Black persons were not significantly different between 2018 and 2020 (aPR, 1.01; 95% CI, 0.94-1.08).

Overall, the proportions of women reporting being UTD with BC and CC screening were similar between 2018 and 2020 (eTable 5 in the Supplement), a pattern that was consistent across race and ethnicity, education, and insurance status (eTable 6 in the Supplement). In a post hoc analysis, among women who were UTD with BC and CC screening, the proportion of women who received their screening tests within the past year was lower in 2020 compared with earlier years (eFigure 2 in the Supplement). For example, among women who were UTD with BC screening, 78.4% of 2020 respondents being screened in the past year compared with 78.9% in 2018.

The proportion of adults who were UTD with CRC screening increased by 4% between 2018 and 2020 (aPR, 1.04; 95% CI, 1.03-1.05), with greater increases among adults with lower educational attainment (aPR, 1.21; 95% CI, 1.15-1.26), Hispanic individuals (aPR, 1.16; 95% CI, 1.10-1.22), and uninsured adults (aPR, 1.15; 95% CI, 1.05-1.25) (eTable 5 in the Supplement). Yet, in 2020, people with lower educational attainment, Hispanic adults, and those who were uninsured were still less likely to be UTD with CRC cancer screening vs those with higher education, White adults, and insured persons (eTable 7 in the Supplement). Similar disparities in receipt of BC and CC screening were observed.

In this national survey study, past-year BC and CC screening prevalence decreased by 6% and 11%, respectively, between 2018 and 2020, and these decreases were consistently greater among people with lower educational attainment and Hispanic persons. Overall, there were an estimated 2.13 fewer million women who reported being recently screened for BC and 4.47 fewer million women screened for CC screening in 2020 than 2018. However, past-year CRC screening prevalence remained steady because decreases in colonoscopy use were offset by increases in stool testing.

Our finding that increased stool testing counterbalanced decreases in colonoscopy is important because it shows the realized potential of at-home testing to maintain population-wide screening rates during a major health care disruption. Commentaries^19,20 hypothesized that increased stool tests could buffer against reported decreases in colonoscopy use during 2020, and our findings suggest that this appears to be the case. This finding also has implications for CC screening, where self-collected HPV sampling has been shown to improve screening uptake, although self-HPV sampling is not yet recommended because questions about its efficacy remain.^21,22 The striking increases in recent stool testing among people with lower SES from 2018 to 2020 is an encouraging finding that may stem from growing use of stool testing outreach in health systems, including federally qualified health clinics (FQHCs) that serve lower-income populations.^23,24 This is a strategy shown to boost CRC screening utilization.^23,24 In our study, UTD CRC screening improved the most among people who had the least education, but despite these improvements, inequities in screening remain.

The recent increase in stool-based CRC screening heightens the importance of implementing evidence-based strategies (eg, patient navigation and tracking) to ensure that positive stool tests are promptly followed by colonoscopy, as stool testing is effective only when this recommended screening process is complete.^25-28 This may be especially important in the populations (ie, people with the lower SES) for whom stool testing is commonly used. Before the pandemic, receipt of follow-up colonoscopy among people with positive stool tests was suboptimal, with reported rates of less than 50% among FQHCs.²⁹ This finding also underscores the timeliness of a federal rule issued in early 2022 addressing patient out-of-pocket costs for a follow-up colonoscopy after a positive stool test.³⁰ Previously, health insurers and Medicare were not required to fully pay for follow-up colonoscopy after a positive stool test because it was considered diagnostic, even though a follow-up colonoscopy is a necessary component of the screening process and all preventive services with A or B USPSTF recommendation are supposed to be completely cost-free under the Patient Protection and Affordable Care Act.

Our observation of population-wide decreases in recent BC and CC screening prevalence in 2020 compared with 2018, alongside studies of medical records and claims,^6,8,9 may signal that the rebounds in screening volume beginning in the summer of 2020 may not have fully compensated for the near halt in screening during March and April. This may be especially true for Hispanic persons and people with lower SES as these groups experienced greater decreases in BC and CC screening,^5,9 reflecting newly emerging and exacerbation of long-standing barriers to screening. This is concerning because these populations are less likely to be UTD with cancer screening and have higher death rates for some screen-detectable cancers.³¹ Early in the pandemic when routine procedures were paused, BC screening volumes decreased more sharply among Hispanic, Asian, and American Indian and Alaska Native persons, and recovery was also weaker in these groups.^32,33 At least 1 FQHC experienced greater delays in fully resuming BC screening services compared with nearby academic centers,³⁴ and other studies of FQHCs note that staff were diverted away from screening navigation and outreach during the pandemic.³⁵ Less obvious structural factors may have created barriers; for example, health care systems may have prioritized those at higher risk of developing cancer, and risk-based models are often based on White populations.³⁴ In addition, people cited fear of COVID-19 as a reason to avoid medical care, and people’s assessment of the benefits of preventive medical care vs COVID-19 risk may be influenced by COVID-19 burden in their community.¹

Broader social and economic issues may have disproportionately affected cancer screening utilization during the pandemic. People with lower SES were more likely to lose employment, have reduced hours or wages, and have a worsening financial situation during the pandemic.^36,37 Health insurance rates were unexpectedly steady during 2020 amid increasing unemployment rates. This may be partly because job losses were concentrated among low-wage workers who did not have employer-sponsored insurance to begin with, and, in 2020, there was enhanced Medicaid funding, and Marketplace and Medicaid Expansion enrollment periods were opened.^38-40

In our study, decreases in recent CC and BC screening have not yet influenced the proportion of women UTD with these cancer screenings. Women screened more frequently than recommended may have maintained their UTD BC and CC screening status in 2020 despite not having a recent test. Previous studies^41,42 have shown that physicians frequently recommend, and women often receive, BC and CC screening on an annual basis, even though the USPSTF, American Cancer Society, and professional organizations lengthened BC and CC screening intervals many years ago. It is possible that the pandemic may have reduced overuse of screening. Among women in our study who were UTD for BC and CC screening, the proportion who received their cancer screening tests within the past year was lower in 2020 compared with earlier years. What the potential lengthening of screening intervals and declines in recent BC and CC screening means for outcomes is not yet known. Modeling studies⁴³ assuming a 6-month pause in mammography projected a relatively small number of excess BC deaths through 2030, and a 6-month pause in CC screening may lead to a slight stage shift, with minimal impacts on CC outcomes.⁴⁴ A recent study⁴⁵ of electronic pathology reports showed a 9% decrease in the number of BCs in 2020 compared with 2019, although the stage and outcomes of cancers diagnosed in 2020 and beyond is not yet known. Previous observational studies examining outcomes after the 2009 USPSTF recommendation to lengthen BC screening intervals for women aged 50 to 74 years to every 2 years showed no immediate changes in stage distribution,⁴⁶ although screening intervals and practices also did not change.⁴² Ongoing study of dynamics of screening utilization throughout the pandemic and outcomes are needed.

The magnitude and timing of population-wide decreases in past-year BC and CC screening as well as colonoscopy use in our study were different than medical claims or regional health system data. In those studies,^3-9,47 mammography, Papanicolaou and HPV testing, and colonoscopy volumes decreased by 70% to 90% in March and April of 2020 compared with expected volumes, and then approached expected volumes by summer or fall of 2020. In our study, decreases in self-reported screening behavior began in the third quarter (July-September) of 2020. This is likely due to how screening was ascertained because we examined self-reported past-year screening rates. For example, people asked about their most recent screening experience in April of 2020 could have been screened in 2019, whereas people interviewed in the latter part of 2020 would have a greater chance of reporting their 2020 experience. In addition, the BRFSS sample was national and not limited geographical area, health care network, or specific type of health insurance coverage.

Our study has several limitations. Overall BRFSS response rates were similar in 2018 and 2020, yet some states paused interviews during shutdowns, and the raw proportion of respondents with lower educational attainment decreased in 2020. Although we used population-based weights that accounted for nonresponse, biases may remain and pandemic-related changes in cancer screening prevalence may be underestimated, especially among people with lower educational attainment. Bias analyses of US Census surveys have noted greater nonresponse among people with lower incomes during 2020 than previous years, and ongoing efforts to improve data collection in vulnerable populations are needed.⁴⁸ Furthermore, survey respondents tend be more healthy than the general population and it is unclear whether this worsened during the pandemic.⁴⁹ Our data were self-reported and prone to recall biases, although a previous meta-analysis⁵⁰ suggests that the overall accuracy of self-reported screening is good. An additional limitation is that we did not examine the reason for a test (eg, for routine screening vs a problem or diagnostic reasons).

In a national survey study, past-year CRC screening prevalence remained level during 2020 because decreases in colonoscopy use were offset by an increase in stool testing, showing the promise of at-home testing to maintain population-wide screening rates during a major health care disruption. Meanwhile, past-year BC and CC screening prevalence decreased, especially among people with lower educational attainment and Hispanic persons, perhaps because of both newly emerging and existing barriers to health care. What these decreases in recent BC and CC screening mean for immediate and long-term outcomes is not yet known but will be important to monitor, especially among people with lower SES.

Accepted for Publication: March 28, 2022.

Published: June 3, 2022. doi:10.1001/jamanetworkopen.2022.15490

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Fedewa SA et al. JAMA Network Open.

Corresponding Author: Stacey A. Fedewa, PhD, Department of Hematology and Oncology, Emory University, 550 Peachtree St, Ste 1075, Atlanta, GA 30308 (sfedewa@emory.edu).

Author Contributions: Dr Fedewa had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Fedewa, Star, Minihan, Han, Yabroff, Jemal.

Acquisition, analysis, or interpretation of data: Fedewa, Star, Bandi, Han.

Drafting of the manuscript: Fedewa.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Fedewa, Star, Bandi.

Administrative, technical, or material support: Fedewa.

Supervision: Fedewa, Jemal.

Conflict of Interest Disclosures: All authors were employed by the American Cancer Society, which receives grants from private and corporate foundations, including foundations associated with companies in the health sector for research outside of the submitted work. The authors are not funded by or key personnel for any of these grants, and their salaries are solely funded through American Cancer Society funds. Dr Han reported receiving grants from AstraZeneca outside the submitted work. Dr Yabroff reported serving on the Flatiron Health Equity Advisory Board; all honoraria are donated to the American Cancer Society. No other disclosures were reported.