Cervical cancer is the third most common cancer in women worldwide, but mortality from cervical cancer has decreased dramatically with widespread screening and treatment.1 Despite the importance of screening, studies demonstrate harm caused by overscreening including complications from procedures, treatment of false-positive results and lesions that will likely regress, anxiety, and resource utilization.2–5 Guidelines from all major organizations therefore universally recommend lengthening intervals between cervical cancer screening. In 2009, the American College of Obstetricians and Gynecologists recommended that the interval between screening increase from every 1 year to every 2 years in those younger than 30 years old and from every 2–3 years to every 3 years in those 30 years old or older.6 In 2012, the American College of Obstetricians and Gynecologists recommended that the interval increase to every 3 years in those younger than 30 years old, stay at 3 years in those 30 years old or older with negative cytology alone, and increase to 5 years in those 30 years old or older with negative human papillomavirus (HPV) cotesting.7 These screening recommendations are also reflected in the most recent update from October 2016.8
Literature on adherence to cervical cancer screening guidelines in the United States is scant and contradictory; self-report data suggest that underscreening is more common,9 whereas case-based and clinic-based studies indicate overscreening is more prevalent.10–15 Because low-income women have a fourfold greater risk for cervical cancer,16,17 understanding guideline adherence is important for state Medicaid programs, which insure 16% of women in the United States ages 19–64 years.18 National self-report data suggest that Medicaid enrollees receive fewer preventive tests than those privately insured,19 but cervical cancer screening guideline adherence has not been fully assessed in a Medicaid population.
Our objectives were to assess changes in cervical cancer screening guideline adherence after the 2009 guideline change among women enrolled in Pennsylvania Medicaid and to determine risk factors associated with underscreening and overscreening.
MATERIALS AND METHODS
We obtained administrative claims and enrollment data from the Pennsylvania Department of Human Services on all fee-for-service and Medicaid managed care enrollees from January 2007 to December 2013. Pennsylvania has the country's sixth largest Medicaid program by enrollment and the fourth largest by expenditure.20 The data included all Medicaid enrollment and claims data for outpatient, inpatient, and professional services and prescription drugs. We defined covariates and outcomes based on procedure and diagnosis codes from claims. We performed an observational cohort analysis of women before and after the guideline changes. The University of Pittsburgh institutional review board approved of the study as exempt because it involved existing data and participants could not be identified.
We included women ages 18–64 years who were continuously enrolled in Pennsylvania Medicaid. To our knowledge, Pennsylvania Medicaid imposes no restrictions such as copayments or prior authorization for Pap tests. To assess adherence to different guidelines during the study period and allow for a sufficiently long follow-up period, we evaluated only those women who had Pap tests in prespecified time periods (the preguideline period was January 1, 2007, to June 30, 2007, the first 6 months of our data set, and the postguideline period was November 1, 2009, to April 30, 2010, the first 6 months after the guideline change), and defined adherence to screening guidelines based on the time interval between Pap tests. Because guidelines differ by age (younger than 30 years old compared with 30 years old or older), we analyzed data stratified by age categories. We required women to be continuously enrolled for a minimum of 2 years for those younger than 30 years old and 4 years for those 30 years old or older. We chose to not require continuous enrollment for the entire 7-year study period because even small lapses in coverage would dramatically limit our study population and introduce selection bias (ie, disproportionately including those with long-term disabilities), leading to less generalizable results.
We imposed some additional exclusion criteria. First, we excluded women who did not have at least one office visit because we wanted to focus on women with some contact with the outpatient medical system. Second, because we were unable to observe services covered by Medicare, we excluded women who were dually enrolled in Medicare and Medicaid. Finally, we excluded women with pre-existing conditions requiring alternative frequencies of Pap screening, including women with diethylstilbestrol exposure, cervical or endometrial cancer history, cervical intraepithelial neoplasia or abnormal cervical cytology, total hysterectomy, human immunodeficiency virus infection, leukemia, neutropenia, organ transplantation, and use of immunosuppressants (see Fig. 1 for sample sizes by exclusion criteria and Table 1 for coding definitions).
To gain a sense of how our study population compared with a population who had no Pap screening, we generated descriptive statistics for women who met all of our inclusion and exclusion criteria but with zero Pap tests.
We categorized guideline adherence by comparing the time interval between Pap tests and a guideline-based interval that includes a window of ±6 months for primary analyses. We defined adherence to guidelines as underscreening, appropriate screening, and overscreening if the interval between Pap tests was longer, within, or shorter than the guideline-based window period, respectively (Table 2). Some studies define adherence to guidelines by the number of Pap tests in defined time periods, but this method does not capture time intervals between Pap tests, which is a core element of the screening guidelines. To reflect the realities of health care delivery, which do not follow strict time intervals,20 we chose to include a 6-month window period in the guideline-based interval to provide some flexibility for patients to receive Pap tests slightly before or after the guideline-based interval as a result of scheduling and appointment availability. For example, women 30 years old or older who had a Pap test after November 1, 2009 should have had screening every 36 months based on guidelines; we defined 30–42 months as our appropriate guideline-based interval for this population. We considered a Pap interval as underscreening if only one Pap test was done during an enrollment period and overscreening if a woman younger than 21 years old had a Pap test after the 2009 guideline change.
Patient demographics included patient race, ethnicity, age (continuous), residence in urban compared with nonurban areas,22 and whether the patient was enrolled in fee-for-service or Medicaid managed care programs. In addition, we included an indicator for residence in institutional facilities and an indicator of whether the reason for Medicaid enrollment was receipt of Supplemental Security Income (a potential proxy for disability) compared with other Medicaid eligibility groups (Temporary Assistance for Needy Families program, General Assistance, programs for individuals just above income eligibility thresholds, and pregnancy). Patient health status covariates included annual outpatient visits, mental health and substance abuse disorder history (defined in Table 3), and a comorbidity measure using the Elixhauser Comorbidity Index.23 We excluded mental health and substance abuse comorbidities from the Elixhauser Comorbidity Index (Table 4) because we assessed them separately. Finally, we included measures likely to be highly related to cervical cancer screening including sexually transmitted infection (STI) history, STI testing frequency, and pregnancy status (defined in Table 3).
Pap-level covariates included timing of Pap test (before or after the 2009 guideline change) as well as family planning services delivered on the same day as the Pap test.
We used patient-level data to calculate descriptive statistics of the cohort and interval-based data to provide more information on time intervals between Pap tests and to define our categorical outcomes. To determine patient demographic and health status factors associated with underscreening or overscreening, we performed multivariable multinomial logistic regression. We used appropriate screening as our base outcome for multinomial models and built models using clinically meaningful covariates. The primary independent variable was timing of a Pap test (performed before or after the 2009 guideline change). Within-patient correlation was accounted for using robust standard error estimation. We then derived average predicted probabilities from multinomial models. As discussed previously, we included a 6-month window period in the guideline intervals. We performed sensitivity analyses using no window period and 3-month window periods to vary definitions for guideline-based intervals. Results from these analyses were similar in magnitude and direction to our primary analysis; we present the 6-month window period as our main results. Statistical significance was determined using a P value of <.05. We performed analyses using Stata 13.
Our sample included 53,160 Pap tests performed in 29,650 women who met our continuous enrollment and inclusion criteria. Table 4 presents clinical characteristics of the cohort stratified by age group. Mean age was 22 years among 17,360 younger than 30 years old and 39 among 12,290 30 years old or older. The 168,909 continuously enrolled women who had zero Pap tests in our study period were more likely to be white (63%), less likely to be black (25%), and were similarly likely to be Hispanic (11%) compared with our cohorts.
Table 5 presents the mean and median days between Pap tests among women who had at least two Pap tests in their respective study periods relative to the guideline-based interval. Although the guideline-based interval increased from 1 to 2 years among those younger than 30 years old, the median days between Pap tests increased only from 371 to 405. Similarly, although the guideline-based interval increased from every 2–3 years to every 3 years among those 30 years old or older, the median days between Pap tests decreased from 432 to 430.
Figure 2 reflects average predicted probabilities for our three categorical outcomes (underscreening, appropriate screening, and overscreening) using multinomial multivariable models as defined by the interval between Pap tests (±6-month window period). In those younger than 30 years old, we observed that underscreening decreased from 49% before the guideline change to 21% after. In contrast, underscreening increased postguideline change in those 30 years old or older from 20% to 29%. Among women who were underscreened, 40% had one Pap test over the course of 4 years. Appropriate screening decreased in both those younger than 30 years old (45 to 11%) and those 30 years old or older (27 to 6%). Consistent with the minimal increase in the time interval between Pap tests for those younger than 30 years old (Table 5), overscreening markedly increased in that age group from 6% to 67%. Similarly, overscreening increased in those 30 years old or older (from 54 to 65%). Among those younger than 30 years old who were overscreened after 2009, 46% were younger than 21 years. All differences were statistically significant (P<.001).
Table 6 shows risk factors associated with underscreening and overscreening compared with appropriate screening through multinomial multivariable regression for our age groups. Compared with appropriate screening, those younger than 30 years old who had Pap tests after the guideline change were more likely to experience both underscreening and overscreening (relative risk ratio [RR] 1.72, 95% confidence interval [CI] 1.58–1.88 and relative RR 48.7, 95% CI 42.8–55.4, respectively). Women 30 years of age or older who had Pap tests after the guideline change were also more likely to experience underscreening and overscreening relative to appropriate screening (relative RR 6.30, 95% CI 5.55–7.14 and relative RR 5.28, 95% CI 4.70–5.94, respectively). Notably, white women had a significantly decreased relative risk for underscreening (relative to appropriate) compared with black women (relative RR 0.81, 95% CI 0.74–0.90) among those younger than 30 years old. On the other hand, among those 30 years old or older, white women had a significantly higher relative risk for overscreening compared with black women (relative RR 1.61, 95% CI 1.42–1.82). Finally, higher STI testing frequency was associated with increased overscreening in both age groups (relative RR 2.16, 95% CI 1.92–2.43 and relative RR 1.89, 95% CI 1.61–2.22, respectively).
Our findings demonstrate a decrease in guideline-based cervical cancer screening after the 2009 guideline change. The interval between Pap tests did not meaningfully change, but the proportion of appropriately screened women decreased sharply. Our results reflect a dichotomy: overscreening increased dramatically, accounting for 60% of screening, whereas underscreening decreased in those younger than 30 years old but increased in those 30 years old or older. Women who are pregnant, black women younger than 30 years, and women younger than 30 years enrolled in managed care were more likely to experience underscreening, whereas white women older than 30 years and those with more office visits and more STI testing were more likely to be overscreened.
Our results reflect lower rates of appropriate screening and higher rates of overscreening than survey-based studies with self-reported measures of screening.10–15 Our population-based study of Medicaid claims, which is better suited than survey data to measure intervals between screening,24 uniquely identifies dramatic overscreening. One laboratory database study reported median years between Pap tests increased from 1.5 in 2008 to 1.87 in 2011, consistent with our results.24 Among our study's participants younger than 30 years old, premature screening substantially contributed to overscreening post-2009, which is consistent with studies showing high rates of screening in younger women.25
Potential reasons for overscreening include limited health care provider knowledge of guidelines, patient expectations, health care provider concerns that lengthening the screening interval causes loss of health system engagement, and continued reimbursement.4,5,26–30 It is possible that increased overscreening after guideline changes was the result of an expected lag in guideline implementation. However, we ensured that all postguideline women had an index Pap test in the first 6 months after the guideline change and defined adherence based on the time to subsequent Pap tests, thus giving health care providers 2–3 years to adapt.
Worsening underscreening among those 30 years old or older was concerning given that low-income women are at fourfold increased risk for developing cervical cancer. We hypothesize that for a segment of our population, health care providers were already following the U.S. Preventive Services Task Force and American Cancer Society 2012 guidelines that recommended screening every 5 years among women with negative cytology and HPV cotesting. Although HPV results were not widely available in 2009–2010 in clinical settings, health care providers with access to cotesting results potentially followed the newest guidelines.
Strengths in our design include large sample size, assessment of multiple guidelines, and interval-based adherence measurement. It is worth noting some limitations to our analysis. First, we limited our analyses to women with Pap tests in “index” periods with one or more outpatient visits. Although this approach may underestimate underscreening because we did not capture women without health system engagement, our interval-based method allows for accurate measurement of underscreening and overscreening in a defined population. It is very unlikely that these inclusion criteria biased our estimates of the effect of the guideline change on adherence because these criteria were applied equally to women enrolled in Medicaid before and after the change. Second, because the newest 2012 guidelines lengthened the screening interval for some to 5 years, not enough time had elapsed to permit assessment of adherence to these guidelines. If anything, rates of appropriate screening may be even lower, and overscreening may be even higher under the newest guidelines. Third, as a result of limitations with claims data, we were unable to assess how health care provider-related factors and HPV vaccination affected adherence outcomes, although it is notable that screening guidelines in our study period were consistent across specialties and did not differ based on vaccination status. Finally, our study design did not allow inference of causal relationships between patient characteristics and cervical cancer screening.
Our findings suggest opportunities to improve adherence in Medicaid through an examination of risk factors for underscreening and overscreening identified in administrative data. For example, pregnancy was associated with underscreening, which is consistent with national quality data suggesting that women access health care less postpartum.31 Interventions to address underscreening can target postpartum mothers who should continue accessing the health care system after delivery. Given that Medicaid programs contract with managed care in most states32 and managed care enrollment was significantly associated with underscreening, states may incentivize plans to increase guideline adherence among their enrollees. Black women (younger than 30 years) were more likely to be underscreened, and white women (30 years or older) were more likely to be overscreened. Importantly, black women are more likely to develop and to die from cervical cancer.17 System-based efforts are therefore essential to eliminating this disparity.
Interventions to reduce overscreening could ensure that women with greater health care utilization and those with frequent STI testing not be overscreened. Specific interventions to address overscreening include electronic medical record alerts that flag overscreening (like existing alerts for underscreening) at a clinical level and disincentivizing overscreening in quality-based reimbursements at a health-system level. A reduction in overscreening could lead to a more than 50% reduction in Pap tests performed annually, which will lead to reductions in expenditures, unnecessary procedures with potential complications, and inconvenience.
1. Fact sheet-cervical cancer. Bethesda (MD): National Institutes of Health; 2010. Available at: https://report.nih.gov/nihfactsheets/Pdfs/CervicalCancer(NCI).pdf
. Retrieved May 20, 2016.
2. Kulasingam SL, Havrilesky LJ, Ghebre R, Myers ER. Screening for cervical cancer: a modeling study for the US Preventive Services Task Force. J Low Genit Tract Dis 2013;17:193–202.
3. Vesco KK, Whitlock EP, Eder M, Lin J, Burda BU, Senger CA, et al. Screening for cervical cancer: a systematic evidence review for the US Preventive Services Task Force. Evidence report no. 86. AHRQ publication no. 11-05156-EF-1. 2011. Rockville (MD): Agency for Healthcare Research and Quality; 2011.
4. Goldhaber-Fiebert JD, Stout NK, Salomon JA, Kuntz KM, Goldie SJ. Cost-effectiveness of cervical cancer screening with human papillomavirus DNA testing and HPV-16,18 vaccination. J Natl Cancer Inst 2008;100:308–20.
5. Sawaya GF, Kulasingam S, Denberg TD, Qaseem A; Clinical Guidelines Committee of American College of Physicians. Cervical cancer screening in average-risk women: best practice advice from the Clinical Guidelines Committee of the American College of Physicians. Ann Intern Med 2015;162:851–9.
6. Cervical cytology screening. ACOG Practice Bulletin No. 109. American College of Obstetricians and Gynecologists. Obstet Gynecol 2009;114:1409–20.
7. Screening for cervical cancer. Practice Bulletin No. 131. American College of Obstetricians and Gynecologists. Obstet Gynecol 2012;120:1222–38.
8. Cervical cancer screening and prevention. Practice Bulletin No. 168. American College of Obstetricians and Gynecologists. Obstet Gynecol 2016;128:e111–30.
9. Sabatino SA, White MC, Thompson TD, Klabunde CN; Centers for Disease Control and Prevention (CDC). Cancer screening test use—United States, 2013. MMWR Morb Mortal Wkly Rep 2015;64:464–8.
10. Almeida CM, Rodriguez MA, Skootsky S, Pregler J, Steers N, Wenger NS. Cervical cancer screening overuse and underuse: patient and physician factors. Am J Manag Care 2013;19:482–9.
11. Saint M, Gildengorin G, Sawaya GF. Current cervical neoplasia screening practices of obstetrician/gynecologists in the US. Am J Obstet Gynecol 2005;192:414–21.
12. Roland KB, Soman A, Benard VB, Saraiya M. Human papillomavirus and Papanicolaou tests screening interval recommendations in the United States. Am J Obstet Gynecol 2011;205:447.e1–8.
13. Berkowitz Z, Saraiya M, Sawaya GF. Cervical cancer screening intervals, 2006 to 2009: moving beyond annual testing. JAMA Intern Med 2013;173:922–4.
14. Meissner HI, Tiro JA, Yabroff KR, Haggstrom DA, Coughlin SS. Too much of a good thing? Physician practices and patient willingness for less frequent pap test screening intervals. Med Care 2010;48:249–59.
15. Saraiya M, Berkowitz Z, Yabroff KR, Wideroff L, Kobrin S, Benard V. Cervical cancer screening with both human papillomavirus and Papanicolaou testing vs Papanicolaou testing alone: what screening intervals are physicians recommending? Arch Intern Med 2010;170:977–85.
16. Krieger N, Quesenberry C Jr, Peng T, Horn-Ross P, Stewart S, Brown S, et al. Social class, race/ethnicity, and incidence of breast, cervix, colon, lung, and prostate cancer among Asian, Black, Hispanic, and White residents of the San Francisco Bay Area, 1988-92 (United States). Cancer Causes Control 1999;10:525–37.
17. Simard EP, Fedewa S, Ma J, Siegel R, Jemal A. Widening socioeconomic disparities in cervical cancer mortality among women in 26 states, 1993-2007. Cancer 2012;118:5110–6.
18. Kaiser Family Foundation. Women's health insurance coverage; 2016. Available at: http://kff.org/womens-health-policy/fact-sheet/womens-health-insurance-coverage-fact-sheet/
. Retrieved May 20, 2016.
19. McMorrow S, Long SK, Fogel A. Primary care providers ordered fewer preventive services for women with Medicaid than for women with private coverage. Health Aff (Millwood) 2015;34:1001–9.
20. Kaiser Family Foundation. Total monthly Medicaid and CHIP enrollment; 2016. Available at: KFF.org/health-reform/state-indicator/total-monthly-medicaid-and-chip-enrollment
. Retrieved May 20, 2016.
21. Partin MR, Casey-Paal AL, Slater JS, Korn JE. Measuring mammography compliance: lessons learned from a survival analysis of screening behavior. Cancer Epidemiol Biomarkers Prev 1998;7:681–7.
22. U.S. Department of Agriculture Economic Research Service. Rural-urban commuting area codes; 2014. Available at: http://www.ers.usda.gov/data-products/rural-urban-commuting-area-codes.aspx
. Retrieved May 20, 2016.
23. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care 1998;36:8–27.
24. Cuzick J, Myers O, Hunt WC, Robertson M, Joste NE, Castle PE, et al. A population-based evaluation of cervical screening in the United States: 2008-2011. Cancer Epidemiol Biomarkers Prev 2014;23:765–73.
25. Hirth JM, Tan A, Wilkinson GS, Berenson AB. Compliance with cervical cancer screening and human papillomavirus testing guidelines among insured young women. Am J Obstet Gynecol 2013;209:200.e1–7.
26. Corbelli J, Borrero S, Bonnema R, McNamara M, Kraemer K, Rubio D, et al. Differences among primary care physicians' adherence to 2009 ACOG guidelines for cervical cancer screening. J Womens Health (Larchmt) 2014;23:397–403.
27. Hawkins NA, Benard VB, Greek A, Roland KB, Manninen D, Saraiya M. Patient knowledge and beliefs as barriers to extending cervical cancer screening intervals in federally qualified health centers. Prev Med 2013;57:641–5.
28. Henderson JT, Saraiya M, Martinez G, Harper CC, Sawaya GF. Changes to cervical cancer prevention guidelines: effects on screening among U.S. women ages 15-29. Prev Med 2013;56:25–9.
29. Roland KB, Greek A, Hawkins NA, Lin L, Benard VB. Provider beliefs associated with cervical cancer screening interval recommendations: a pilot study in federally qualified health centers. Prev Med Rep 2015;2:444–7.
30. Perkins RB, Anderson BL, Gorin SS, Schulkin JA. Challenges in cervical cancer prevention: survey of U.S. obstetrician-gynecologists. Am J Prev Med 2013;45:175–81.
31. National Committee for Quality Assurance. Prenatal and postpartum care. Washington, DC: National Committee for Quality Assurance; 1994–2009.
32. Kaiser Family Foundation. Medicaid managed care enrollment; 2014. Available at: http://kff.org/medicaid/state-indicator/total-medicaid-mc-enrollment/
. Retrieved May 20, 2016.