Racial and Ethnic Disparities in Cervical Cancer Screening From Three U.S. Healthcare Settings

The study investigates racial and ethnic disparities in cervical cancer screening uptake and timely follow-up after abnormal results across multiple U.S. healthcare settings. Motivated by persistent disparities—Black women are about 30% more likely to develop and 60% more likely to die from cervical cancer than non-Hispanic White women, and Hispanic women have a 51% higher incidence and ~20% higher mortality—this work addresses limitations of prior research that relied on self-reports (which overestimate screening and can differentially misclassify by race) and single-site studies that may obscure system-level drivers. The research question is whether and to what extent screening and follow-up differ by race/ethnicity across three distinct health systems, and how clinical, socioeconomic, and structural factors (insurance, site of care) account for observed differences. The purpose is to inform equitable cancer prevention strategies by identifying modifiable drivers of disparities in real-world care settings.

Prior literature demonstrates mixed findings on racial/ethnic differences in self-reported cervical cancer screening, with concerns about recall and social desirability bias and evidence that self-reports overestimate screening relative to EHR data and may underestimate disparities by race. Single-system EHR studies, while internally valid, may miss cross-system variation and structural influences (e.g., site differences, insurance). National patterns indicate Black and Hispanic patients disproportionately receive care in safety-net and federally qualified health centers, which may have fewer resources and different screening practices. The PROSPR consortium provides a framework for multilevel evaluation of screening processes. Guidelines during the study period recommended cytology every 3 years (ages 21–30) and cytology every 3 years or cotesting every 5 years (ages 30–65). Emerging evidence and guidelines suggest a role for primary HPV testing and possible self-sampling, with concerns that unequal implementation may exacerbate disparities. Limited studies have tracked diagnostic follow-up after abnormal screening, reporting follow-up completion rates around 65%–78%, highlighting a system-wide gap.

Design and setting: Retrospective cohort study within the METRICS Research Center (part of the PROSPR consortium), including three sites: Site A (integrated safety-net health system in the southwestern U.S., reported by academic partner), Site B (northwestern mixed-model healthcare system providing insurance and care), and Site C (northeastern integrated healthcare system with multiple affiliated primary care networks). Data timeframe: June 1, 2016–May 31, 2019; analyses conducted in 2022. Data sources: Harmonized electronic health record (EHR) and administrative databases. Population: Female patients aged 18–89 years in METRICS. For this analysis, included cohort members aged 22–64 years on June 1, 2016 (age-eligible per USPSTF), who were average-risk (no prior cervical abnormality or cancer, not living with HIV, intact cervix). To ensure due status, excluded those with a documented cotest in the 2 years before the study window (June 1, 2014–May 31, 2016). Sites A and C required at least one visit to primary care or women’s health clinics (2010–2019); Site B required health plan enrollment and attribution to a Site B primary care provider. Outcomes: (1) Any cervical cancer screening (Pap, HPV, or cotest) within the 3-year window. Among screened patients aged 30–64, modality was categorized as cytology alone versus cotesting. (2) Timely follow-up after high-grade or guideline-defined abnormal screening result (e.g., NILM/HPV16/18+, ASC-US/HPV+, or worse per 2012 ASCCP guidelines): receipt of colposcopy, biopsy, or excisional procedure within 6 months; follow-up procedures ascertained through December 31, 2019. Exposure: Race/ethnicity from EHR, categorized as non-Hispanic White, non-Hispanic Black, Hispanic (Hispanic ethnicity overrides race), Asian/Pacific Islander, multiracial, other (including Native American), or unknown. Covariates (as of June 1, 2016): Age; pregnancy during study period; BMI; smoking status (never/former/current/unknown); comorbidity via adapted Charlson index (0, 1–3, ≥4) using ICD-9/10 codes; health insurance (Medicare only, Medicaid only, commercial only, uninsured/medical assistance, multiple, other government/insurance); number of primary care visits in the two years before the study period; site of care. Statistical analysis: Described population overall and by race/ethnicity. Compared proportions receiving screening, receiving cotest (among screened ages 30–64), having an abnormal result, and completing 6-month follow-up using chi-square tests overall and by site. Estimated risk ratios (RRs) and 95% CIs via modified Poisson regression with log link for: (a) any screening, (b) abnormal result among screened, and (c) timely follow-up among those with abnormalities. Conceptualized race as a social construct; sequential models estimated: Model 1 (unadjusted), Model 2 (clinically adjusted: age, pregnancy, BMI, Charlson score, smoking), Model 3 (fully adjusted: adds insurance type, primary care visits, and site of care) to assess attenuation or amplification of disparities.

- Cohort: Of 1,027,128 METRICS members, 188,415 (18.3%) met eligibility. Racial/ethnic distribution: 44.7% non-Hispanic White, 12.4% non-Hispanic Black, 32.5% Hispanic, 6.7% Asian/Pacific Islander, 2.1% multiracial/other (including Native American/Alaska Native), 1.6% unknown. - Insurance and site differences: Black and Hispanic patients were more often uninsured (26.0% and 37.8%) or on Medicaid (26.2% and 19.7%) than non-Hispanic White patients (2.4% uninsured; 9.2% Medicaid) and were concentrated at Site A (safety-net), which had the lowest screening prevalence (58.6%). Site C had the highest screening prevalence (71.1%). - Screening uptake over 3 years (overall 62.8%): non-Hispanic White 63.5% (reference), non-Hispanic Black 53.2% (lower), Hispanic 65.4% (higher), Asian/Pacific Islander ~66.0%–66.5% (higher), unknown 49.6% (lowest). Within each site, Black patients had screening rates comparable to or higher than Whites, but overall lower due to concentration at the lowest-performing site. Hispanic patients had higher screening than Whites across all sites, with the largest difference at Site A (64.6% vs 40.0%, p<0.001). - Adjusted screening RRs: Hispanic patients’ association reversed across models—slightly lower after clinical adjustment (Model 2 RR≈0.98 [0.97–0.99]) but substantially higher after adding insurance, visits, and site (Model 3 RR≈1.18 [1.16–1.20]; abstract also reports RR≈1.14 [1.12–1.16]). Black patients remained less likely than Whites after clinical adjustment (Model 2 RR≈0.86 [0.84–0.88]), but this disparity was fully attenuated after adding insurance and site (Model 3 RR≈1.00 [0.98–1.03]). The higher screening for Asian/Pacific Islander patients was no longer significant after clinical adjustment (RR≈1.02 [1.00–1.04]). - Abnormal results among those screened (~3.5% overall): higher among Black (4.1%) and Hispanic (3.7%) than White (3.3%) and Asian (2.7%) patients (p≤0.01). In fully adjusted models, Black RR≈1.26 [1.12–1.41] and Hispanic RR≈1.18 [1.06–1.31] vs White; Asian patients were slightly less likely than Whites to have abnormalities. - Timely follow-up within 6 months among those with abnormalities: overall 72.5%; by site, Site B 65.6% vs Site A 77.2% and Site C 77.0%. By race/ethnicity, Hispanic patients had higher unadjusted follow-up than Whites (78.8% vs 69.8%, p<0.001); this difference was not significant after adjustment for insurance and site. Discussion text notes Black patients had the lowest follow-up (~65%) and Hispanic the highest (~79%). - Screening modality (ages 30–64): Cotesting prevalence varied widely by site (Site A 12.7%, Site B 68.7%, Site C 84.8%). Within sites, differences by race/ethnicity were modest; however, across sites, Black (44.0%) and Hispanic (22.2%) patients were far less likely to receive cotesting than White patients (73.6%) (p<0.001). - Overall, screening and follow-up fell below the 80% coverage target across all groups and settings.

Findings demonstrate that disparities in cervical cancer screening are strongly shaped by structural factors—particularly site of care and insurance coverage. Although Black patients’ within-site screening was comparable to or exceeded White patients’ screening, overall screening was lower among Black patients because they were more likely to receive care at the safety-net site with the lowest screening rates. Adjustment for insurance and site fully attenuated Black–White differences in screening, underscoring systemic inequities in access and resource allocation across healthcare systems. Hispanic patients had higher screening uptake than non-Hispanic White patients, and this advantage increased after accounting for clinical characteristics and structural factors, suggesting protective factors or effective engagement strategies within this population that merit further study. Both Black and Hispanic patients had persistently higher risks of screen-detected abnormalities after full adjustment, highlighting underlying differences in risk exposure, screening histories, or other unmeasured social determinants. Timely diagnostic follow-up was suboptimal overall (≈73%) and particularly low for Black patients and at Site B, indicating critical process failures post-screening that can compromise the benefits of early detection. Marked cross-site variation in cotesting and the very low cotesting at the safety-net site contributed to racial/ethnic differences in modality exposure; with evolving guidelines and the potential for primary HPV testing and self-sampling, inequitable implementation could exacerbate or alleviate disparities. Collectively, results support multilevel interventions targeting system-level drivers (insurance coverage, care delivery resources, workflows) alongside patient-level supports to improve both screening uptake and diagnostic follow-up equitably.

Across three diverse U.S. healthcare systems, cervical cancer screening and follow-up were below target levels (<80%), with overall lower screening among Black patients driven by concentration in lower-performing, under-resourced settings, and higher screening among Hispanic patients that persisted after adjustment. Black and Hispanic patients had higher adjusted risks of abnormal findings, and follow-up within 6 months was inadequate for all groups, particularly Black patients. These findings highlight systemic racism and inequities in insurance and site resources as root drivers of disparities and suggest that interventions must address where and how care is delivered, not only who receives it. Future research should: (1) identify and scale successful practices contributing to higher screening among Hispanic patients; (2) evaluate equitable implementation of primary HPV testing and potential self-sampling in safety-net and other settings; (3) develop multilevel strategies to improve timely follow-up after abnormal results; and (4) examine local context and structural racism measures to disentangle site and community effects.

- Possible misclassification of screening if patients received services outside attributed systems. - Requiring 3 years of data improves internal validity but may introduce selection bias by including only continuously enrolled/engaged patients. - Unmeasured mediators, including structural and interpersonal racism, were not directly captured. - Site effects may conflate healthcare system characteristics with local/state policy and community context, limiting causal attribution. - Race/ethnicity from EHR can be misclassified, with ~1.6% unknown; small numbers precluded separate reporting for American Indian/Alaska Native patients despite high burden. - Generalizability is limited to participating systems; results are not national estimates. - Potential residual confounding from socioeconomic or access factors not fully measured.