Abstract
Introduction: The COVID-19 pandemic has reduced the number of elective in-person visits to primary care practices. This study examined how the pandemic may have affected cervical cancer (CC) screening rates in primary care settings across the United States.
Methods: We conducted a retrospective cross-sectional study using data from the PRIME Registry of the American Board of Family Medicine from March 15, 2017, to March 14, 2022. We included 2,207,355 women aged 21 to 65 years who had visited a clinician (n = 1,052) from any of 472 primary care practices. We compared CC screening rates among eligible women during in-person visits over the 3 prepandemic years with those during the 2 years of the pandemic.
Results: Screening rates (per 100 eligible patients with in-person visits) decreased from 1.85 to 1.12 in the first quarter of the first year and remained lower throughout both years of the pandemic compared with prepandemic year, had not returned to prepandemic levels by the end of the second year. Hispanic or Latino (odds ratio [OR] = 1.96) and Black or African American (OR = 1.37) women were more likely to be screened, whereas those receiving care from male clinicians (OR = 0.34) were less likely to be screened.
Conclusions: CC screening rates remained below prepandemic levels throughout the 2 years of the pandemic. Policy makers and health care professionals should strategize approaches to enhance CC screening rates, including the exploration of alternative methods, such as home-based CC screening. New screening approaches are needed to ensure preparedness for future health crises.
- Cervical Cancer
- COVID-19
- Cross-Sectional Studies
- Early Detection of Cancer
- Long-Term Effects
- Pandemics
- Primary Health Care
- Screening
- Preventive Care
- Retrospective Studies
Introduction
Approximately 13,820 women may be diagnosed with invasive cervical cancer (CC) and 4,360 women may die from this disease in 2024 in the USA.1 Screening via Papanicolaou (Papanicolaou) smear is an effective intervention to reduce the incidence of CC. Missed or delayed CC screening can result in delayed diagnosis, leading to more advanced stages of cancer detection, poorer prognoses, and increased morbidity and mortality rates.2 Over the last 2 decades in the USA, the annual number of diagnosed CC cases has not decreased, and late-stage diagnoses have increased.3
Primary barriers to CC screening include lack of health insurance, limited access to health care, the need for childcare, and language problems.4 Women with low incomes residing in rural areas demonstrate a lower CC screening uptake, with some individuals never or rarely undergoing screening.5 In addition, the gender of clinicians has been associated cervical cancer (CC) screening rates; specifically, women who have female primary care clinicians are more likely to undergo the recommended screenings.6,7 Similarly, clinician experience may influence screening practices, with more experienced clinicians providing consistent care and recent graduates being more up to date with current guidelines.8,9 Clinician titles may also impact screening practices, as nurse practitioners and physician assistants often focus more on preventive care, potentially increasing screening rates.10
Primary care clinicians, along with gynecologists, play an important role in CC screenings in the USA and as frontline health care professionals who can serve women who have never or rarely been screened.11 Due to stay-at-home orders and concerns about coronavirus disease 2019 (COVID-19) exposure, the recent pandemic was associated with a decline in in-person visits to health care facilities and delays in elective care, such as cancer screenings.12,13 During the peak of the COVID-19 pandemic, the American Cancer Society14 and several other organizations15,16 recommended postponing visits for cancer screening.
Numerous studies have highlighted a sharp decrease in CC screening during the initial months of the pandemic.13,17⇓⇓⇓⇓–22 In June 2020, 2 studies found that CC screening was 35% below the 3-year average17 and 40% below the 5-year average.18 By the end of the first year of the pandemic, CC screening rates remained 21% below historic baselines.19 While most studies have primarily focused on the short-term effects of the pandemic on CC screening, the longer-term impact has been poorly studied. A recent study showed that Papanicolaou testing rates declined in 2022 compared with 2019.23
This study assessed the impact of the COVID-19 pandemic on CC screening over a 2-year period using a nationwide sample from primary care practices and evaluated the potential impact of various patient characteristics, such as race, ethnicity, rurality, and socioeconomic deprivation, along with clinician characteristics, including gender, year of graduation, and titles, on the likelihood of screening.
Methods
Data Source and Study Design
We conducted a retrospective, repeated cross-sectional study using clinical records from the PRIME Registry of the American Board of Family Medicine, a certified outpatient clinical quality data registry that collects electronic health record (EHR) data through automated feeds from participating primary care practices. Currently, PRIME contains data from over 3,000 primary care clinicians who have treated patients in 1,250 practices across 50 states in the USA. This database includes information related to the care of 6.5 million patients across 70 million visits. This study was approved by the ABFM Scientific Review Board and American Academy of Family Physicians Institutional Review Board and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines (Appendix Table 1).
Descriptive Characteristics of All Patients in the PRIME Registry from 2017 to 2022 by Screening Status, Including a Comparison of Screening Rates Between Pre-Pandemic and Pandemic Years
We defined March 15, 2020, as the onset of the COVID-19 pandemic, aligned with the nationwide emergency declaration of COVID-19 in the USA. To assess the potential impact of the COVID-19 pandemic on CC screening, we defined a 5-year study period from March 15, 2017, to March 14, 2022, and divided this into 5 distinct time periods starting from March 15th of each year and ending on March 14th of the subsequent year: prepandemic year-3 (Pre-3): March 15, 2017, to March 14, 2018; prepandemic year-2 (Pre-2): March 15, 2018, to March 14, 2019; prepandemic year-1 (Pre-1): March 15, 2019, to March 14, 2020; pandemic year-1 (PY-1): March 15, 2020, to March 14, 2021; and pandemic year-2 (PY-2): March 15, 2021, to March 14, 2022. Our analysis primarily compared the prepandemic years (Pre-1 through Pre-3) with the first 2 pandemic years (PY-1 through PY-2). We included data from 3 years before the start of the pandemic to capture any trends that may have occurred during this period fully.
Study Population
To establish a comparable study population across the different time periods, we applied several exclusion criteria. We initially identified clinicians who did not provide the minimum level of service throughout the study period. Clinicians with no visits for a continuous 6-month period and those with fewer than 250 total visits in each year between 2017 and 2022 were excluded from the study database. After these exclusions, our database included 1,052 primary care clinicians from 472 primary care clinics. We then identified female patients aged 21 to 65 years who visited these clinicians at least once during the study period. In addition, we excluded women with a history of hysterectomy, precancer status, or CC, identified through pertinent diagnostic and procedural codes. Finally, telehealth visits were excluded from the sample because our focus was screenings that require in-person visits.
Outcome Measures and Study Variables
The primary outcome measure was the CC screening rate represents the number of screenings per 100 eligible patients during each period. CC screening was identified as any cervical cytology (Papanicolaou test) or human papillomavirus test, based on the recommendations of the US Preventive Services Task Force (USPSTF).24 To align our populations with USPSTF screening recommendations, we classified our sample into 2 age categories: 21 to 29 years and 30 to 64 years. Screenings were determined using the International Classification of Diseases and Related Health Problems, Tenth Revision, and the Current Procedural Terminology/Health care Common Procedure Coding System codes (Appendix Table 2). In cases of conflicting information, procedural codes (CPT/HCPCS) were prioritized for accurately identifying the screening event.
Descriptive Characteristics of Clinicians in the PRIME Registry Included in the Study from 2017 to 2022
To account for patient characteristics, the independent variables were age, sex, race/ethnicity, rurality, and residence in a historically marginalized area. We used the 2010 Rural–urban commuting area (RUCA) codes to classify patient zip codes as urban or rural.25 Residence in a historically marginalized area was assessed using 5-digit zip code level Social Deprivation Index (SDI), which is derived from socioeconomic indicators and reflects an individual's exposure to historic disinvestment in resources and marginalization.26 The SDI scores range from 0 to 100 and were categorized into 2 groups: ≤50 (low SDI) and >50 (high SDI). Clinician sex, titles (medical doctor, doctor of osteopathic medicine, nurse practitioner, physician assistant, and others), and medical school graduation year were extracted from the doctor and clinician national downloadable file of the Centers for Medicare & Medicaid Services (CMS). Graduation years were categorized into 4 groups, based on the time since graduation from medical school: <10 years, 10 to 19 years, 20 to 29 years, and >30 years.
Statistical Analyses
Quarterly practice-level screening rates were calculated by dividing the number of eligible patients who underwent CC screening in a quarter of the practice period by the total number of eligible patients for screening who had an in-person office visit in that quarter. To align with our annual observation periods, the quarters were defined as follows: first quarter (March 15 to June 14), second quarter (June 15 to September 14), third quarter (September 15 to December 14), and fourth quarter (December 15 to March 14 of the following year). We combined the 3 prepandemic years and the 2 pandemic years to compare CC screening rates for each descriptive characteristic using a Chi-Square test. To evaluate the association between the likelihood of screening in each period and independent variables, we developed a multivariate logistic regression model with errors clustered at the practice level to account for unmeasured confounders. This model predicted the likelihood of undergoing CC screening using patient demographics, clinician sex, titles, and graduation year as predictors. Given that approximately 75% of our sample was White, we developed separate models for White and non-White patients to control for race/ethnicity while still comparing the variables of interest.
Statistical analyses were performed using Stata 17.0 (StataCorp, College Station, TX, USA). A 2-sided P-value <0.05 indicated statistical significance.
Results
Among 2,207,355 patients who met the eligibility criteria, 87,859 (4.0%) underwent CC screening in the included practices during the study period (Table 1). The sample descriptive characteristics remained consistent across all study periods, suggesting comparability across study periods. Of the screened individuals, 65.8% were White, 18.3% were Hispanic or Latino, 41.3% having an SDI ≤50, 76.3% residing in urban areas and 16.6% were aged between 21 to 29 years. Screening rates significantly decreased during the pandemic across all groups, except for Black or African American individuals and those in the Other race/ethnicity category.
Of the 1,052 clinicians, 45.2% were female, 66.0% held MD degrees, and 35.5% had graduated from medical school 20 to 29 years ago (Table 2).
Figure 1 shows the screening rates during the study period. Overall, in both PY1 and PY2, quarterly CC screening rates (range: 1.12 to 1.77) were lower than those in the 3 prepandemic years (range: 1.64 to 2.00). Before the pandemic, the quarterly CC trends were similar across quarters; however, they decreased sharply by 39.5% from 1.85 to 1.12 in the first quarter of the PY-1. Subsequently, rates recovered during the remainder of PY-1, with screening rates reaching 1.66 in the third quarter of PY-1. Our data indicated that the rates did not rebound to prepandemic levels during PY-2. The highest CC screening rate was 1.77 (15 June through 14 September in PY-2) during the pandemic, which was still 10% lower than the highest rates observed in the same quarters in the prepandemic years (range: 1.96 to 2.00).
Cervical cancer screening rates (per 100 eligible patients with in-person visits) in the PRIME registry from 2017 to 2022, quarterly.
The distribution of screenings within a clinician’s practice remained consistent in each annual period, with approximately 74.0% of the screenings performed by female clinicians (Figure 2). While the proportion of screenings by male clinicians remained stable, there was a slight, but statistically nonsignificant (P = .987) increase during the pandemic years. The quarterly percentages of clinicians who performed at least one screening during the pandemic ranged from 37.6% to 49.3% (Appendix Figure 1). These rates were consistently lower than those in the prepandemic years, except in the third quarter of PY-1.
Distribution of cervical cancer screenings within a practice in the PRIME registry from 2017 to 2022, by clinician sex during the study period.
In the logistic regression model, patients who were identified as Hispanic/Latino, Black/African American, or those within the 21 to 29-year age-group had the highest odds of undergoing CC screening (Table 3). Compared with Pre-3 (reference year), PY-1 was associated with a significantly lower likelihood of screening (odds ratio [OR] 0.86; 95% confidence interval [CI]: 0.79 to 0.95; P = .003). The odds of CC screening in PY-2 were lower than those in Pre-3 (the comparison year); however, this difference was not statistically significant. In addition, patients seeing a male clinician had significantly lower odds of undergoing CC screening (OR 0.34; 95%CI: 0.26 to 0.45; P < .001). The observed trends were consistent across both White and non-White participants (Appendix Table 3).
Logistic Regression Model Estimating the Odds Ratio for Cervical Cancer Screening by Patient and Clinician Characteristics in the PRIME Registry from 2017 to 2022
Discussion
In this study, conducted using data from a national sample of primary care practices, we observed lower CC screening rates during the first 2 years of the COVID-19 pandemic than during the 3 prepandemic years. We noted a sharp decline in CC screening rates in the first quarter of the pandemic, followed by a rebound in subsequent quarters. However, despite this partial recovery, screening rates remained below the levels observed in the same quarters during the prepandemic period. To the best of our knowledge, this study is one of the first to examine the CC screening rates at the end of the second year of the COVID-19 pandemic. Extending the observation period is critical for understanding the longer-term impact of the COVID-19 pandemic on preventive service use.
Our results are consistent with those of other studies, indicating a decline in CC screening rates in the USA during the pandemic, with decreases ranging from 36% to 94% during the early months of the pandemic.13,17⇓–19,21,27,28 A study19 explored an extended timeline, and its findings indicated similar trends to ours, highlighting that CC screening remained below prepandemic levels by the end of 2021. Another study found a 30% decrease in Papanicolaou smear testing odds in 2022 compared with 2019.23 Our study covered 5-year period and assessed the potential impact of patient demographics and clinician characteristics on CC screening in primary care.
A decline in the initial months of the pandemic was expected, coinciding with stay-at-home orders and disruptions in health care services, particularly for nonurgent health care needs. The temporary closure of screening sites, concerns about contracting SARS-CoV-2 infection in health care facilities, and recommendations from professional societies to postpone cancer screening visits likely contributed to this decline in screening.13,18,27 However, a persistent decline in screening can lead to delayed diagnoses and identification of late-stage diseases, resulting in worse outcomes for women.28 According to a modeling study by Burger et al, a 6-month disruption in CC screening could lead to an additional 5 to 7 cases per 1 million screened women, and a 24-month disruption could lead to an additional 38 to 45 cases by 2027.29 To date, only one study has found that CC rates have almost rebounded to prepandemic levels in the USA after the lifting of stay-at-home orders.30 In this study, the authors suggested that tracking and reminder systems for follow-up on missed screenings played a role in the normalization of screening rates.
During the pandemic, the need for alternative methods to maintain screening has become evident. CC screening still requires an in-person visit, and, unlike colon cancer, no Food and Drug Administration-approved self-sampling kits are currently available.30 If useful self-sampling kits could be developed and adopted, individuals could receive support through telehealth to collect self-samples, ensuring that screening could continue in extraordinary situations, such as pandemics, so that early diagnosis is not interrupted.5
Our findings differ from those of prior studies that found that race, ethnicity, geographic region, and socioeconomic status were influencing factors in CC screening.5 While the majority of studies suggested that racial/ethnic minorities in the USA experienced lower cancer screening rates, our analysis suggested that Hispanic/Latino and Black/African American patients were more likely to undergo CC screening, at rates compared with those of their white counterparts.13 Our results may be explained by a lack of minority patients in our sample; nearly 75% of the eligible patients in our sample were White. Alternatively, our results may be attributed to the fact that individuals in our sample had initial access to primary care, as emphasized by Kim et al.13 Similarly, in contrast to the findings of previous studies,4,5,23 the lack of differences in terms of rurality and SDI in our study may have been related to already established access to care.
Individuals aged 21 to 29 years had higher odds of screening than did those aged 30 to 64 years. These results may be driven by the fact that older age is associated with a higher risk of COVID-19-related complications, making older individuals more hesitant to visit health care facilities. In addition, younger women, particularly those who are more sexually active, may perceive themselves at greater risk for cervical cancer due to HPV exposure, further driving their likelihood to prioritize screenings.31 In contrast, older women, particularly those in long-term relationships, may perceive themselves at lower risk and thus may not prioritize screening as much.
Clinicians are particularly pivotal in increasing CC screening rates. Effective follow-up, guidance, and promotion of screening by clinicians significantly enhanced the participation of eligible individuals. Although most clinicians in our study were male, approximately 75% of the screenings were conducted by female clinicians. This finding, which is consistent with those of prepandemic studies, indicated that the sex of the clinician is closely related to patient CC screening preferences.6,7 Female clinicians may be more likely to suggest CC screening and address patient concerns. Understanding these underlying mechanisms may highlight approaches for increasing CC screening in future.
Our study reveals a critical gap in cancer screenings during the pandemic. The strengths of our study include the use of a diverse national sample derived from clinical registry data. We conducted a comprehensive analysis, examining 2 full years of follow-up data and comparing the results with 3 prepandemic years to offer a more in-depth understanding of CC screening trends. In addition, our study included detailed patient characteristics, enabling the identification of vulnerable subgroups, along with an exploration of clinician characteristics.
Our study had some limitations. First, the current screening status of the participants could not be evaluated when estimating the CC screening rates. Although the second year showed some return to prepandemic activities, some outbreaks like Delta Wave and ongoing restrictions likely kept some people from seeking health care services. Since our study did not extend beyond the second year, it may not capture the full recovery of screening rates. Other study limitations were related to our data source. Involvement of primary care practices in the PRIME registry is voluntary and thus the registry is not necessarily representative of all primary care practices across the country. In addition, PRIME captures encounters with participating practices rather than consistently following a set of patients over time. Thus, our study was based on repeated cross-sectional data, which lacked the continuity of a longitudinal cohort following the same individuals over time. Finally, the PRIME data exclusively capture the care provided by primary care practices; therefore, we were unable to account for CC screening performed by other clinicians such as gynecologists. Therefore, the screening rates in our study may underestimate the actual rates in this population.
Conclusion
We observed a decline in CC screening rates during the COVID-19 pandemic in the USA. Although the rates quickly rebounded after the initial months of the pandemic, our extended follow-up showed that screening rates remained below prepandemic levels even by the end of the second year. Our findings highlighted the importance of future studies examining the long-term effects of missed or delayed CC screening on cancer diagnosis and mortality. To avoid adverse outcomes, new home-based screening approaches should be developed. We hope that our findings will help policy makers and health care professionals prepare for extraordinary situations, such as the COVID-19 pandemic.
Appendix
Percentage of clinicians conducting at least one screening during the study periods, quarterly.
STROBE Statement—Checklist of Items Included in the Study
Codes Used to Determine Cervical Cancer Screenings
Adjusted Logistic Regression Model Estimating the Odds Ratio for Screening by Patient and Clinician Characteristics, Stratified by Race/Ethnicity
Notes
This article was externally peer reviewed.
Funding: This study was supported by The Scientific and Technological Research Council of Turkiye (TUBITAK) 2219-International Postdoctoral Research Fellowship Program for Turkish Citizens.
Conflict of interest: The authors have no conflicts of interest to declare.
To see this article online, please go to: http://jabfm.org/content/38/2/209.full.
- Received for publication June 21, 2024.
- Revision received September 23, 2024.
- Accepted for publication October 7, 2024.
