The Pap test is a widely available screening test for detecting precursor lesions to cervical cancer, and it has been credited with greatly reducing the cervical cancer death rate.1 Despite this effective test, about 4,000 women die each year in the United States as a result of cervical cancer.2 There are three main factors in the relationship between Pap testing and cervical cancer mortality: lack of Pap testing, failure of the Pap test to detect an abnormality, and lack of adequate follow-up after an abnormal Pap test.3 To help ensure appropriate follow-up after an abnormal Pap test, the American Society for Colposcopy and Cervical Pathology (ASCCP) published evidence-based consensus guidelines for the management of abnormal cytologic findings.4 According to ASCCP 2001 guidelines, an immediate colposcopy is the appropriate recommended follow-up for Pap test results of low-grade squamous intraepithelial lesions (LSIL); atypical squamous cells, cannot exclude high-grade squamous intraepithelial lesions (ASC-H); and high-grade squamous intraepithelial lesions (HSIL).
Despite these guidelines, a recent systematic review regarding follow-up care after abnormal screening tests for cervical, breast, and colon cancer showed that fewer than 75% of women received timely and appropriate follow-up care.5 The rate of nonadherence after abnormal Pap tests varies dramatically across studies, ranging from 7–73%.5–7 Most of the studies we reviewed defined adherence as a dichotomous variable within a certain time; however, the time varied across studies, creating challenges for comparing findings among studies.
In this study, we measured the actual length of time from the abnormal index Pap test until a woman received follow-up care—specifically, colposcopy or related procedures—using a commercially insured population. We examined the factors associated with differences in the timing of receiving an ASCCP-recommended follow-up procedure. Identifying factors that delay appropriate follow-up can help clinicians and policy makers target or design interventions to improve adherence to appropriate follow-up care, thus helping to reduce cancer incidence and death rates.
MATERIALS AND METHODS
We examined administrative and claims data for women aged 18 to 64 years who were enrolled in a large U.S. health care organization8; had a Pap test (ie, index Pap test) between January 1, 2004, and December 31, 2004; and were continuously enrolled in the health care plan 9 months before and 12 months after the index Pap test. We identified a total of 4,824 women who had an abnormal index Pap test of LSIL, ASC-H, and HSIL using International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes 795.03, 795.02, and 795.04, respectively. We restricted the study to data from 2004 because the specific Pap test results of LSIL and HSIL, which were specified in the ICD-9-CM codes, were introduced on October 1, 2004. We excluded women who had had a cervical cancer–related procedure or diagnosis in the 9 months before the index Pap test, because this suggested that the index Pap test was a follow-up to a previous abnormality. Because the study was an analysis of existing deidentified data, this research was exempt from institutional review board review.
Follow-up diagnostic procedures, including colposcopy, biopsy, or conization, were identified by Physicians' Current Procedural Terminology Coding System, 4th edition codes 57420, 57452, 57454, 57455, 57456, 57460, 57461, 57500, 57505, 57510, 57511, 57513, 57520, 57521, 57522, 57530, 57531, 57540, 57545, 57550, 57555, and 57556; or ICD-9-CM procedure codes 672, 674, 673.2, 673.3, 671.1, 671.2, and 671.9. Because the focus of this study was to estimate time to any diagnostic procedure included in the ASCCP 2001 guidelines, we use the term “colposcopy” to refer to all the previously mentioned diagnostic procedures after the index Pap test, recognizing that it will encompass excisions.
We used survival analysis, or time-to-event analysis, to estimate the length of time from the abnormal Pap test to a colposcopy procedure. The event in this type of analysis was defined as the first occurrence of a colposcopy. The “survival” time was defined as the number of days between the index Pap test and the first occurrence of a colposcopy. Women who did not receive a colposcopy were treated as censored at the end of the follow-up period (365 days).
We used the Kaplan-Meier product-limit method,9 plotting the cumulative probability of receiving a colposcopy to estimate survival time of receiving a diagnostic procedure after an abnormal index Pap test. In addition, we used a survival regression model to estimate the relationship between the survival experience and explanatory variables. Although the Cox proportional hazards model is the commonly used approach for time-to-event data, it assumes that the hazards (ie, risk of occurrence of the event at a point in time) among different groups are proportional to each other.10 This assumption was not supported by our data. Among all the parametric models, we found that the lognormal model is the best fit for the data in this study. A lognormal model assumes that the logarithm of the survival time is distributed normally. If there is no censoring of data (eg, loss to follow-up), a lognormal model is equivalent to a multiple linear regression model, with the logarithm of survival time as the dependent variable.
We included the following variables in the analysis: age (18–20, 21–30, 31–40, 41–50, and 51–64 years), census region of women's residence (ie, Northeast, South, Midwest, and West), health care settings or facilities (ie, office-based and outpatient clinics, inpatient and laboratories, or others settings), type of physician performing the index Pap test (ie, gynecologist, family practitioner, internist, or other type of physician), and type of insurance (ie, Health Maintenance Organization [HMO], Preferred Provider Organization [PPO], Point of Service, or other types). We also included contextual variables obtained from the U.S. Census Bureau11 describing the proportion of the other races except whites (we will use the term “all-other-races” throughout this article), population, and median household income based on the ZIP code tabulation areas (the smallest geographic area) of the neighborhoods in which the women in the study sample resided. Both the linear and quadratic forms for the proportion of the all-other-races population in each ZIP code tabulation area were included in the model. To reduce correlation between the linear and quadratic forms, we centered the proportion of the all-other-races population by subtracting the data from the sample mean.12 From the final survival model, we obtained the predicted median length of time to receive a colposcopy after an abnormal Pap test for groups of women defined by age and Pap test results, keeping all other factors constant. All the analyses were performed using SAS 9 (SAS Institute Inc., Cary, NC).
Among the total sample of 4,824 women, we identified Pap test results of LSIL, ASC-H, or HSIL, and the majority of the Pap test results in the study were ASC-H (60%), with 33% LSIL and 8% HSIL (Table 1). The age distribution varied slightly across the type of Pap test results, with relatively more women younger than 30 years with an LSIL result as compared with the other test results. Overall, the majority of the patients in the study sample were aged 31 years or older. Although the magnitude of all other sample characteristic proportions varied slightly, the patterns were similar across the three test results. More than 75% of the patients in the study sample had insurance coverage through either a PPO or Point of Service plan. More than half of the women received their index Pap test from a gynecologist; the majority of the women received it in a physician's office or an outpatient clinic. Nearly half of the women resided in the South and, overall, 70% of the women lived in large urban areas. The majority of women lived in areas where the population was mostly white non-Hispanic; the median percentage of the all-other-races population was under 20%. The median household income was about $50,000 across all three test results.
The median length of time from the abnormal Pap test to a colposcopy was 99 days. The proportion of women receiving a follow-up procedure decreased as age increased: about 70% for women younger than 30 years, 58% for women aged 31–40 years, 49% for women aged 41–50 years, and 43% for women aged 51–64 years. The percentage of women receiving the recommended follow-up procedure also differed by type of abnormality: 85% for HSIL, 79% for LSIL, and 45% for ASC-H. Figures 1 and 2 show the plot of cumulative probability of receiving colposcopy by age group for time to follow-up for the different age groups (Fig. 1) and Pap test results (Fig. 2). The Wilcoxon test and log-rank test both show statistically significant differences (P<.001) for age groups and Pap test results in the length of time from the abnormal index Pap test to the first occurrence of a diagnostic procedure.
A final model included all the variables listed in Table 1 and interactions of age by test results. A log likelihood ratio test indicated a good fit of the final model (χ2=0.886, P<.001). Because the time from diagnosis to colposcopy was similar for the two younger age groups (18–20 [adolescent] and 21–30 years old), we combined these two age groups in the survival regression model. Table 2 presents the coefficient and ratio of expected (ie, model predicted) time of each covariate from the lognormal regression model. The regression coefficients show the change of logged time with the change of each covariate while controlling for other factors. For example, the estimate coefficient of −0.5011 means that women aged 18–30 years with an abnormal result of ASC-H had a 0.5011 decrease in logged time compared with women aged 31–40 years with ASC-H test results. Although these coefficients present a precise change of logged time, the meaning of these estimates is not immediately informative. Therefore, we also present the ratio of expected time to show the comparison of length of time to follow-up between each covariate and its corresponding reference group (by exponentiating the regression coefficient). For example, the ratio of 0.606 indicates that the women aged 18–30 years with ASC-H abnormal results have a 39.4% decrease in the length of time as compared with women aged 31–40 years with ASC-H abnormal results while controlling for other factors.
We found significant interaction effects between age and Pap test results. For the reference group of abnormal Pap test results, ASC-H, younger women received a colposcopy sooner than older women. Women with Pap test results of LSIL and HSIL received a colposcopy sooner than those with ASC-H results. However, the magnitude of a reduced length of time to receive a colposcopy was greater in older women (with negative signs in coefficients) than in younger women (positive signs in coefficients). To help understand these complicated interaction effects, we computed the predicted median survival time for women in different age groups and with different Pap test results while keeping all other factors constant. Figure 3 shows that the predicted time to receive a colposcopy was similar for women with LSIL and HSIL results. The predicted time to receive a colposcopy increased as age increased for women with ASC-H results.
Table 2 also shows that the length of time to receive a colposcopy increased about 50% for women whose medical coverage was a PPO, point of service, or other than that for women whose coverage was an HMO. Similarly, women who received an index Pap test from family practice physicians or other types of physicians had a longer length of time to colposcopy (36% and 65% longer for family practice physicians and other physicians, respectively) than women who received their index Pap test from a gynecologist.
We found that the effect of the area-level, all-other-races population contextual variable on length of time to receive colposcopy was complex. Specifically, the relationship between the percentage of the all-other-races population at the ZIP code tabulation area level and time to colposcopy for women living in the areas resembles a U shape. As the proportion of the all-other-races population increases, the length of time to colposcopy decreases to a point and then begins to rise. The underlying mechanism for this is reflected in the negative sign for the linear form (−0.2596) and the positive sign for the quadratic form (1.1432) in Table 2. Controlling for all other factors, we estimated that women who lived in an area that is about 38% all-other-races have the shortest expected time to receive a diagnostic procedure (a figure arrived at by summing the lowest value and the mean value).
We used health insurance claims data to investigate the length of time from an abnormal Pap test result of LSIL, ASC-H, or HSIL to a colposcopy. Regardless of age, women with Pap test results of LSIL and HSIL received the recommended follow-up within 3–4 months from an initial index Pap test (median time). Thus, it appears the real-world practice for this population of women adheres to the 2001 ASCCP guidelines.4 Although the model predicted a slightly longer (1 month longer) median time for women aged older than 50 years with LSIL Pap test results than for the younger women with the same abnormality, the difference of time was not statistically significant between age groups (data not shown). This finding may be attributable to the fact that we censored a higher proportion of the older women with LSIL results (ie, they had not received a colposcopy within the study time frame); moreover, some women might have received follow-up care other than a colposcopy. According to the 2001 consensus guidelines, follow-up without colposcopy for postmenopausal women is acceptable if the women receive repeat Pap testing at 6 and 12 months, receive a human papillomavirus DNA test at 12 months, or use vaginal creams and receive repeat Pap tests later.4
Our study found that age is an important factor related to the timing of a follow-up colposcopy for women with a diagnosis of ASC-H. Younger women received follow-up procedures sooner than older women. Several reasons may explain this finding. In our study sample, older women were less likely than younger women to have their index Pap test performed by a gynecologist, and some providers may not appreciate that ASC-H is not ASC of undetermined significance. A survey of a nationally representative sample of clinicians providing Pap screening from 2003–2004 reported that a higher proportion of clinicians with clinical specialties other than gynecology tended not to follow recommended guidelines.13 Furthermore, a higher proportion of clinicians with clinical specialties other than gynecology also reported a lack of available cervical colposcopy capabilities on site.13 Because our study sample included a higher proportion of older women who received care from clinicians other than gynecologists, a large proportion of these clinicians may not have suggested a follow-up colposcopy or may not have had the colposcopy capabilities on site. In fact, among women with ASC-H results who did not receive a colposcopy, a higher proportion of older women did receive follow-up care (including a repeat Pap test or human papillomavirus DNA test; 33% for those aged 51–64, 28% for those aged 41–50, 27% for those aged 31–40, and 21% for those younger than 30 years). The findings that older women with ASC-H results received a colposcopy later or did not receive a colposcopy highlight an area of concern for adherence to ASCCP guidelines. Our model predicted that women older than 40 years would not receive a colposcopy within 1 year (Fig. 3). This delay may put these women at a higher risk of developing cervical cancer. Future research to improve adherence to follow-up procedures for women older than 40 years should focus on factors that affect how well providers of different specialties follow current guidelines and how to improve the availability of diagnostic capabilities.
In our study sample, the proportion of women with a Pap test result of ASC-H was higher than women with LSIL (59.7% compared with 32.7%), which is contradictory to the literature.14 The main reason for this pattern is that the sample we identified as having an LSIL Pap test result was based on the ICD-9-CM code 795.03, which was officially available only in late 2004,15 whereas the code to identify ASC-H has been available since 2002. To test whether our findings were biased because of the underrepresentation of LSIL, we excluded women who received a Pap result of ASC-H before October 1, 2004. The survival model from this restricted sample yielded findings similar to the whole sample.
We identified several factors that provide important information for health care providers and policy makers to improve the adherence to recommended guidelines, including those from the ASCCP. First, physician specialty made a difference: gynecologists tended to deliver the diagnosis procedure sooner than did physicians with other specialties. Second, health care setting or facilities made a difference: women receiving care from office-based or outpatient clinic settings had a colposcopy sooner than women who received care from other settings or facilities. Future studies that examine the characteristics of the health care providers (eg, their knowledge, training, and attitude toward screening abnormalities) and of the health care settings (eg, size, location, and staffing) will help clarify the influence of these factors on timing of delivering diagnostic procedures.
Third, our finding that women who had HMO insurance received more timely follow-up procedures than women who had other types of insurance is consistent with the literature that HMOs are associated with delivering better preventive care.16,17 This suggests that HMOs play an important role as coordinators of care, such as using recall protocols to inform enrollees to receive timely follow-up care; such procedures may be missing in most individual physicians' offices. Future efforts to improve adherence to follow-up may pull from strategies used by HMOs in care coordination.
In addition, our findings suggested that the population density of racial and ethnic groups, reflected in the percentages of the all-other-races population (shown at the ZIP code tabulation area level), may play a critical role for women in these areas to receive a timely colposcopy after an abnormal Pap test result. The U-shaped relationship between the timing and the proportion of all-other-races population in an area indicates that either resources in these areas or social support among the racial or ethnic groups may play an important role in this health care service. Studies that have found similar patterns suggest two lines of explanation for the duality of findings. First, several studies show that economic or residential segregation is an important determinant of health and health care.18–20 Areas with a high proportion of all-other-races population often have less availability of and access to health care resources, and the residents are less likely to receive preventive care; such an effect was not found in the white population.20–22 Our model controlled for area-level median household income and still demonstrated an impeded effect of a high proportion of all-other-races population on follow-up, suggesting that factors related to health care resources in racially diverse areas, independent of poverty, may be critical to preventive care for people living in these areas. Second, recent research has provided evidence that racial and ethnic minorities were mentally and physically healthier if they lived in areas that had a higher concentration of people from the same racial and ethnic group.23 The real mechanism of group density effect is still unclear, but many researchers hypothesize that higher racial and ethnic density may produce increased social support from the same racial and ethnic group or decreased experience of status inconsistency from different racial and ethnic groups.19 These two lines of research provide insight for understanding the U-shaped finding in our study.
This study takes advantage of health insurance claims data for its detailed records of service procedures that a patient receives. However, claims data have limitations. The major limitation of the claims databases is that records are submitted for billing purposes instead of research purposes. Thus, information on a diagnosis or procedure may be incomplete if the claim was not submitted for reimbursement. In addition, claims databases are based on a large convenience sample. The sample is not random; therefore, the results derived from claims data may not generalize well to other populations.
Despite evidence-based consensus guidelines implemented since 2001 for the management of women with cytologic abnormalities, prior research evaluating adherence to these guidelines has not reached a consensus regarding the time frame of follow-up care. Studies using a direct measure of time intervals are scant and limited by small sample sizes, with a focus on underserved populations and narrow geographic areas.24–26 This study directly measured time intervals and used a large sample of the privately insured population across the nation to estimate provider and patient factors related to timing of receiving follow-up care after an abnormal Pap test. These data are needed for conducting policy assessments to evaluate follow-up care and timing to determine whether providers are in fact changing practices as guidelines are updated. From our findings, the timing for follow-up with colposcopy is not consistent for all Pap test results. Interventions that emphasize the importance of timely follow-up of abnormal Pap tests that can lead to cervical cancer are warranted.
1.Gustafsson L, Pontén J, Zack M, Adami HO. International incidence rates of invasive cervical cancer after introduction of cytological screening. Cancer Causes Control 1997;8:755–63.
2.U.S. Cancer Statistics Working Group. United States Cancer Statistics: 1999–2005 Incidence and Mortality Web-based Report. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute; 2009.
3.Leyden WA, Manos MM, Geiger AM, Weinmann S, Mouchawar J, Bischoff K, et al. Cervical cancer in women with comprehensive health care access: attributable factors in the screening process. J Natl Cancer Inst 2005;97:675–83.
4.Wright TC Jr, Cox JT, Massad LS, Twiggs LB, Wilkinson EJ; ASCCP-sponsored Consensus Conference. 2001 Consensus Guidelines for the management of women with cervical cytological abnormalities. JAMA 2002;287:2120–9.
5.Yabroff KR, Washington KS, Leader A, Neilson E, Mandelblatt J. Is the promise of cancer-screening programs being compromised? Quality of follow-up care after abnormal screening results. Med Care Res Rev 2003;60:294–331.
6.Khanna N, Phillips MD. Adherence to care plan in women with abnormal Papanicolaou smears: a review of barriers and interventions. J Am Board Fam Pract 2001;14:123–30.
7.Eggleston KS, Coker AL, Das IP, Cordray ST, Luchok KJ. Understanding barriers for adherence to follow-up care for abnormal pap tests. J Womens Health (Larchmt) 2007;16:311–30.
8.Shatin D, Rawson N, Stergachis A. United Health Group. In: Strom B, editor. Pharmacoepidemiology. Sussex: John Wiley; 2005. p. 271–80.
9.Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457–81.
10.Cox DR. Regression models and life tables. J R Stat Soc B 1972;34:187–220.
12.Kleibaum D, Kupper L, Muller K, Nizam A. Applied regression analysis and other multivariable methods. 3rd ed. Pacific Grover (CA): Duxberry Press; 1998.
13.Irwin K, Montaρo D, Kasprzyk D, Carlin L, Freeman C, Barnes R, et al. Cervical cancer screening, abnormal cytology management, and counseling practices in the United States. Obstet Gynecol 2006;108:397–409.
14.Davey DD, Neal MH, Wilbur DC, Colgan TJ, Styer PE, Mody DR. Bethesda 2001 implementation and reporting rates: 2003 practices of participants in the College of American Pathologists Interlaboratory Comparison Program in Cervicovaginal Cytology. Arch Pathol Lab Med 2004;128:1224–9.
16.Phillips KA, Haas JS, Liang SY, Baker LC, Tye S, Kerlikowske K, et al. Are gatekeeper requirements associated with cancer screening utilization? Health Serv Res 2004;39:153–78.
17.Landon BE, Zaslavsky AM, Bernard SL, Cioffi MJ, Cleary PD. Comparison of performance of traditional medicare vs medicare managed care. JAMA 2004;291:1744–52.
18.Waitzman NJ, Smith KR. Separate but lethal: the effects of economic segregation on mortality in metropolitan America. Milbank Q 1998;76:341–73, 304.
19.Lobmayer P, Wilkinson RG. Inequality, residential segregation by income, and mortality in US cities. J Epidemiol Community Health 2002;56:165–6.
20.Polednak AP. Trends in US urban black infant mortality, by degree of residential segregation. Am J Public Health 1996;86:723–6.
21.Pickett KE, Ahern JE, Selvin S, Abrams B. Neighborhood socioeconomic status, maternal race and preterm delivery: a case-control study. Ann Epidemiol 2002;12:410–8.
22.Prentice JC. Neighborhood effects on primary care access in Los Angeles. Soc Sci Med 2006;62:1291–303.
23.Pickett KE, Wilkinson RG. People like us: ethnic group density effects on health. Ethn Health 2008;13:321–34.
24.Coker AL, Eggleston KS, Meyer TE, Luchok K, Das IP. What predicts adherence to follow-up recommendations for abnormal Pap tests among older women? Gynecol Oncol 2007;105:74–80.
25.Eggleston KS, Coker AL, Luchok KJ, Meyer TE. Adherence to recommendations for follow-up to abnormal Pap tests. Obstet Gynecol 2007;109:1332–41.
26.Howell LP, Gurusinghe S, Tabnak F. Impact of ASC-H terminology on the detection of HSILs in medically underserved California women. Diagn Cytopathol 2009;37:103–10.
© 2010 The American College of Obstetricians and Gynecologists
This article has been cited
Obstetrics & GynecologyCorrectionsObstetrics & Gynecology