| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
1 Division of Cancer Prevention and Control, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, GA; 2 Battelle Centers for Public Health Research and Evaluation, Atlanta, GA; and 3 Battelle Centers for Public Health Research and Evaluation, Baltimore, MD
Requests for reprints: Janet Kay Bobo, Battelle Centers for Public Health Research and Evaluation, 4500 Sand Point Way Northeast, Suite 100, Seattle, WA 98105-3949. Phone: (206) 528-3141; Fax: (206) 528-3550. E-mail: boboj{at}battelle.org
 |
Abstract |
|---|
 Top Abstract BackgroundMaterials and MethodsResultsDiscussionReferences |
|---|
|
Background |
|---|
|
TopAbstractBackgroundMaterials and MethodsResultsDiscussionReferences |
|---|
The U.S. Preventive Services Task Force (10), the National Cancer Institute (11), and the American Cancer Society (12) have all recommended mammography rescreening for age-eligible women every 1–2 years. The NBCCEDP requires that all participating programs implement plans to facilitate annual or biennial mammography rescreening consistent with U.S. Preventive Services Task Force guidelines. These plans must include sending appropriate rescreening reminders.
Since 1997, CDC has routinely analyzed electronic records submitted semiannually from each participating state, tribe, and territory (program) to monitor the percentage of women rescreened within 18 and 30 months of their previous annual or biennial examination. The 6-month grace period accommodates appointment scheduling difficulties and other minor concerns that may have delayed rescreening beyond the recommended 1–2-year interval. The available data have suggested satisfactory levels of cervical cancer rescreening but potentially inadequate levels of mammography rescreening.
Further evaluation of the findings related to mammography rescreening has been problematic because the amount of information that can be derived from the routinely submitted records is limited in several crucial respects. First, CDC does not collect data on subsequent mammograms that NBCCEDP enrollees receive through nonprogram providers such as Medicare. Similarly, CDC does not monitor subsequent mammography use among women who lose their NBCCEDP eligibility due to enrollment in an employer-based health plan. Program data also cannot be used to identify risk factors for delayed rescreening because the submitted records contain only the minimum number of variables needed to monitor cancer detection rates and ensure compliance with NBCCEDP policies on age-related eligibility, minority access, and provision of appropriate diagnostic and follow-up care.
To collect the data required for an accurate estimation of the percentage of enrollees rescreened on schedule, whether through the NBCCEDP or another program, and to identify factors that encourage timely rescreening, CDC supported a 4-year study in Maryland, New York, Ohio, and Texas known as the Survey on Mammography Rescreening project. Using NBCCEDP records submitted from these states and the Breast Imaging Reporting and Data Systems lexicon (13) developed by the American College of Radiology, cohorts of women who had a study-eligible index mammogram in 1997 were identified and interviewed at least 30 months after that examination. Here, we report findings for the cohort of 2024 women with a negative or benign index mammogram.
|
Materials and Methods |
|---|
|
TopAbstractBackgroundMaterials and MethodsResultsDiscussionReferences |
|---|
To select the four programs, CDC reviewed data from all NBCCEDP states, tribes, and territories to identify those that had provided sufficient numbers of mammograms in 1997 coded negative/benign and probably benign. (Data from the probably benign cohort were collected concurrently with the information reported here but will not be discussed further in this report.) Most programs had submitted thousands of records reporting negative or benign examinations, but only a small subset had the minimum number of probably benign mammograms needed for the second cohort. The Maryland, New York, Ohio, and Texas programs were chosen from this subset because the combined racial and Hispanic ethnicity distribution of their enrollees best approximated the corresponding distribution observed across all NBCCEDP-funded mammograms completed between July 1991 and March 1998. In the latter group, about 54% of the mammograms were provided to White women, 16% to Black women, 19% to Hispanic women, 6% to American Indians/Alaskan Natives, 3% to Asians/Pacific Islanders, and 2% to women who did not indicate their race or ethnicity (3).
After receiving Institutional Review Board approvals and clearance from the Office of Management and Budget, separate sampling frames were derived for each of the four states. Each frame was restricted to mammograms with a Breast Imaging Reporting and Data Systems result of negative or benign provided in 1997 to women aged 50–69 years at the time of their examination. For most study participants, the index mammogram was not the initial NBCCEDP mammogram they had received at program enrollment. If two or more mammograms had been provided in 1997, only the earliest one was retained in the sampling frame. Mammograms that could be linked with either a clinical breast examination coded "abnormal, suspicious for cancer" within 3 months of the mammogram date or a mammogram coded "probably benign, short-term follow-up recommended" in the preceding 9 months were also excluded because they may not have been true screening mammograms. The number of study-eligible mammograms per state ranged from a low of 4351 to a high of 12,639. After all eligible mammograms were sorted by date to limit bias associated with month of the index screen, 500 index mammograms in each state were identified by selecting every nth report (14). Oversampling ensured that index mammography recipients who could not participate in the study due to mental incompetence, serious illness, or death prior to the end of their 30-month interval could be replaced with similarly eligible women.
Detailed protocols guided all aspects of subject enrollment. English and Spanish versions of an invitational letter were mailed to each woman shortly after the end of her 30-month interval. A $25.00 incentive was offered, and bilingual speakers were available during all telephone contacts. Repeated attempts were made to reach nonrespondents, and extensive field tracing was done to locate women who had moved since their index mammogram.
All telephone interviews were conducted by the Battelle Survey Operations Center in Baltimore. Women who were not fluent in Spanish or English were interviewed with an interpreter in their preferred language. Proxy interviews were not attempted for women who died after their 30 months of follow-up. The 30-minute interview was based on questions derived from the literature, input from program staff in the study states, and six focus groups conducted among NBCCEDP enrollees in Texas (15). The variables used in the analyses reported here include mammography use (dates and results), demographic factors (age, race, ethnicity, marital status, income, education, urbanization, residential mobility, interview language, and country of birth), medical history [obesity, hysterectomy status, personal and family history of cancer, personal history of health problems or disabilities, self-reported health status, hormone replacement therapy (HRT) use, and number of mammograms previously received], health care access (insurance status at time of interview, usual source of care, and knowledge of NBCCEDP eligibility status), and social support (received screening reminders and encouragements from medical providers and other sources).
Women who reported the locations and approximate dates of their index examination and any subsequent mammograms were asked to sign consent forms authorizing our review of their mammography reports. Corresponding mammography records were operationally defined as those occurring within 6 months of the screening dates reported during the interview. Reports for index examinations were obtained to confirm dates reported in the electronic records. Procedure dates, views taken, results, and recommendations were abstracted, but only the mammography dates were used in the analyses reported here.
To be consistent with NBCCEDP rescreening policies, we classified all participants according to their self-report data as "rescreened on schedule," "not rescreened on schedule," or "status unknown" at 18 and 30 months after their index mammography. To illustrate, a woman who was rescreened 20 months after her index mammography was coded as "not rescreened on schedule" at the 18-month end point and "rescreened on schedule" at the 30-month end point. Some women who could not be classified at 18 months could be classified at 30 months. For example, a woman with an index mammogram in February 1997 who reported that she had a subsequent mammogram in 1998 but did not remember the month or day could not be classified at 18 months as either rescreened or not rescreened but was classified as rescreened at 30 months. If the mammography date from the medical record was available, it was used to classify such women more precisely. Women with ambiguous rescreening status were removed from the 18-month and/or the 30-month analyses as appropriate.
We evaluated the accuracy of our self-report data in the subset of women with medical record confirmation of both the index examination and at least one subsequent mammography. Participants were cross-classified according to their rescreening status at 18 and 30 months (on-schedule or not) based on self-report data and the dates from their mammography record. Many women had to be excluded from these analyses because we were able to obtain only one of their two required mammography records. In other studies of cancer screening behavior, self-report data have been found to be reasonably accurate (16–20), although some evidence of overreporting (18) and telescoping has been noted (16).
Rescreening proportions and 95% confidence intervals (95% CI) were determined for all interviewed women with sufficient data and for subsets defined by demographic, medical history, health care access, and social support variables. The 18- and 30-month findings were similar, but both sets of results are reported here to assist screening facilities that recommend annual or biennial rescreening. SUDAAN was used to derive point estimates and standard errors, taking the stratified survey design into account (21). To identify factors significantly associated with on-schedule rescreening, we used the two-tailed 
2 test.
Most risk factor questions were anchored by date [e.g., Please think back to (INDEX MAMM DATE PLUS 1 YEAR), that is about 1 year after you had the mammogram on (INDEX MAMM DATE). As far as you know, were you eligible on (INDEX MAMM DATE PLUS 1 YEAR) to have another mammogram that was fully or partially paid for by (PROGRAM NAME) or a similar program?]. For several variables (HRT use, hysterectomy, breast problems, and cancer history), we asked about the month(s) and year(s) when the factor was experienced. These data were used to classify women who reported experiencing the factor either before or both before and after the index date as "yes, before." Women who reported experiencing the factor only after their index mammogram were coded as "yes, but only after." Because some women could recall the relevant year but not the month, some women in the "yes, only after" group may have experienced the factor after their first rescreening mammography.
Like other mammography utilization studies (22–24), we used logistic regression methods to evaluate multiple risk factors concurrently (14). Demographic, medical history, health care access, and social support variables were tested in hierarchical models that controlled for the NBCCEDP enrollment state. Variables were excluded from the final models if they did not improve model fit and were not significantly associated with rescreening or if they were highly correlated with other terms in the model (e.g., language preference at interview and race/ethnicity). All models compared the odds of on-schedule rescreening among women with the specified characteristic to the odds of on-schedule rescreening among women without that characteristic. Odds ratios (OR) above 1.0 with 95% CIs that exclude the null value indicate factors that may be associated with an increased likelihood of completing timely rescreening. An OR below 1.0 suggests a factor that may reduce this likelihood.
|
Results |
|---|
|
TopAbstractBackgroundMaterials and MethodsResultsDiscussionReferences |
|---|
|
Almost all (96.9%) interviewed women gave verbal permission for the release of medical record data and 86.4% also provided written consent. The signed consent forms authorized collection of 3662 records, including the index screen and one or more rescreening mammograms completed during the 30-month follow-up. Some mammography reports could not be obtained because the facilities had closed or could not be located, the dates provided by the women could not be matched to clinic files, or the facility refused to provide the information. Ultimately, we obtained 78.4% of the records (2870 mammography reports).
Although finding the index mammography report more than 3 years after the examination was often difficult, we obtained mammography records for both the index screen and at least one subsequent mammogram for 757 women (45% of the participants). In this subset with complete data, 98% of those who reported having been rescreened within 18 months had dates in their medical records that confirmed this status. All 757 women who reported a subsequent mammogram within 30 months had confirming medical record dates.
Rescreening Status
Using the self-report information and any available medical record data in ambiguous situations, we determined rescreening status for 1630 of the interviewed women (96.7%) at 18 months and for 1680 (99.7%) women at 30 months. Within these subsets, we found that 72.4% were rescreened within 18 months and 81.5% within 30 months (Table 2). For completeness, we also reviewed the NBCCEDP records submitted semiannually to CDC by the participating states to identify any additional mammograms during the follow-up interval that were not mentioned in the telephone interviews. Although a few additional mammograms were identified, the overall rescreening percentages increased by less than 2% (data not shown).
|
Several medical history variables were also significantly associated with on-schedule rescreening at both end points (Table 3). Women who described their health as good to excellent were more likely than those who described it as poor to fair to have been rescreened. The number of prior mammograms was strongly and positively associated with our observed rescreening percentages. Women who had used HRT or had a hysterectomy since their index mammogram were more likely to have been rescreened than those who had not. As we expected, a history of breast cancer significantly increased the likelihood of rescreening. For the variables indicating a history of other cancers and a family history of breast cancer, a similar pattern was observed but the 2 values were not statistically significant at the 0.05 value.
|
Finally, we found that a supportive environment improves rescreening (Table 4). Supportive factors include receiving a reminder such as a postcard or letter and being strongly encouraged by their physician, a nurse, a friend, or a relative to have another mammogram.
|
|
With one exception, these findings were not materially altered when the logistic models excluded women with HRT, hysterectomy, or cancer diagnoses dates that potentially could have occurred after the first rescreening date (data not shown). In the restricted 18-month model, the OR for first use of HRT after the index date dropped to 0.97 (95% CI = 0.37–2.55).
|
Discussion |
|---|
|
TopAbstractBackgroundMaterials and MethodsResultsDiscussionReferences |
|---|
Despite these strengths, the findings reported here should be generalized with caution. The NBCCEDP serves only 15–20% of the eligible national population and our participants were selected from only four programs. Although many minority women were included in our sample, the data set imperfectly mirrored the racial and ethnic balance of the NBCCEDP because American Indians/Alaskan Natives and Asians/Pacific Islanders were underrepresented. Furthermore, we were unable to obtain the records needed to confirm all index and rescreening mammography dates and thus cannot exclude the possibility that the subgroup with complete information differs in some way from the remainder of the sample.
Due to limitations in the precision of our data on dates for use of HRT, hysterectomy, and cancer diagnoses, some participants in our "yes, but only after index mammogram" groups may have completed their rescreening mammography before experiencing the risk factor of interest. Results from these analyses should thus be interpreted with caution. Finally, we note that as with most retrospective studies based on self-report data, recall bias may be present in some findings, especially given the advanced ages of some of our participants.
The first important finding is that in this sample of women who received a free mammogram in 1997, the percentages rescreened on schedule (72.4% within 18 months and 81.5% within 30 months) equal or substantially exceed rescreening values reported for six other population-based samples. Only 5% of 3240 women in four Los Angeles medical facilities were rescreened within 21 months of an index mammogram in 1990 based on analyses of medical record data (26). About 82% of 36,123 women in Albuquerque who had an index examination between 1994 and 1997 and about 82% of 121,889 women in Victoria, Australia screened initially between 1995 and 1996 were rescreened within 27 months (27, 28). In a study using health maintenance organization (HMO) medical record data from Detroit, about 69% of 3888 women with an index mammogram between 1989 and 1996 were rescreened within 24 months (29). Similarly, a HMO study in Seattle based on index mammograms from 1990 to 1992 and medical record data found that 73% of 5059 women were rescreened within 6 months of their recommended 12-, 24-, or 36-month intervals (30). Using only administrative data, two NBCCEDP-funded state programs have independently reported mammography rescreening rates for their enrollees. Investigators in New York followed 9485 women initially screened between 1988 and 1991 and found that 27% had been rescreened within 36–60 months of their index mammogram (23). Similarly, investigators in Washington followed 2888 women who received their index mammogram between July 1994 and December 1995 (31). Survival analyses found that 25.7% were rescreened within 15 months and 40.3% were rescreened within 27 months. However, because none of these studies based their findings on self-report data coupled with both medical record review and program administration data, they may have underestimated true rescreening proportions. The higher rescreening rates observed in our study include mammograms that were not funded by the NBCCEDP and may also reflect secular trends in growing public awareness of the importance of repeated assessments.
The observation that age was not associated with rescreening in the bivariate analyses and did not improve fit in the logistic models provides important reassurance about women who lose NBCCEDP eligibility when they reach age 65. CDC policy does not permit programs to continue screening such women with NBCCEDP funds after they become eligible for Medicare. Our finding suggests that in the four states in this study, low-income women are successfully transitioning from the NBCCEDP to Medicare and are as likely to rescreen on-schedule as their younger counterparts. Age has been identified as a significant factor in some reports (30, 32–34) but not in others (22, 29). A study of 8749 women in a large HMO in Michigan used the same age categories that we did and found no association with rescreening status after 2 years (29).
In our multivariate models, only two demographic variables, "moved since index mammogram" and "race/ethnicity," were significantly associated with rescreening. Women who had moved were less likely to have been rescreened in the 18-month model. After controlling for the available confounders including the NBCCEDP enrollment state, Hispanic women were significantly more likely than White women to have been rescreened within 30 months of their index mammogram. The literature is inconsistent on the relationship between race/ethnicity and mammography use (23, 26, 34, 35), which may reflect regional variations in adoption of breast screening behaviors among subgroups, differences in how potential confounders were addressed during data analysis, and intrinsic limitations in our ability to quantify and analyze racial and ethnic characteristics. Our finding suggests that when other factors are held constant, Hispanic women may be more likely than White women to continue using health promotion opportunities after their initial enrollment.
Several studies have found a positive association between years of education and mammography use (32, 34, 36), but others have seen a drop in rates among more educated women (23, 37). Our finding of a nonstatistically significant OR of 0.53 at 30 months for women with 4 or more years of college compared with those with a high school education is consistent with the latter reports. Older women who are well educated but have low incomes are rarely studied and little is known about their health behaviors.
Results for the medical history variables suggest that health care providers are encouraging their patients to be rescreened. The highest 30-month values were observed among women who reported a factor that would have required contact with a physician since the index mammography such as having had a hysterectomy (92.6%), using HRT (85.8%), and a diagnosis of breast cancer (94.7%). The group that described their health status as fair to poor was less likely to report on-schedule rescreening, but it cannot be assumed that they had seen a physician since their index date. Women with other health problems or disabilities, including obesity, were not significantly less likely to have been rescreened.
Results pertaining to the health care access variables are particularly important. Like others (24, 32, 38), we observed that women who do not report having a usual source for medical care were much less likely to adhere to rescreening recommendations. Similarly, women who did not know whether they were still eligible for a free mammogram 1 and 2 years after their index examination were far less likely to have been rescreened than those who believed they remained eligible. In the 18-month multivariate model, the OR was only 0.16 (95% CI = 0.09–0.26). Unfortunately, true program eligibility status could not be determined from the data available.
The other important observation is that women who recalled receiving a rescreening reminder, usually a postcard or letter, were much more likely to have been rescreened than women who indicated they did not receive a reminder or did not know if they had received one. In both multivariate models, the association remained highly significant even after controlling for differences in demographic, medical history, and health care access factors and the NBCCEDP state that had provided the index mammogram. All four of the participating state programs were routinely mailing out rescreening reminders during the follow-up period so it is likely that some women had simply forgotten about receiving one. Interestingly, intervention studies have provided mixed support for the value of sending rescreening reminders to low-income women. Lantz et al. (39) found that those who received reminders were much more likely to complete mammography screening (OR = 4.0, 95% CI = 2.6–6.2). In contrast, Simon et al. (40) reported that reminder letters did not have a beneficial effect on utilization above and beyond physician medical record reminders.
Taken together, the results of this study suggest that low-income women who enroll in a free mammography screening program are as likely as women in the general population to be rescreened every 1–2 years. The NBCCEDP has succeeded not only in increasing the number of low-income women who have been screened for breast cancer at least once but also in encouraging breast cancer screening behaviors that persist over time. Nevertheless, our data indicate that nearly one of every five enrollees was not adhering to the rescreening schedule considered most beneficial for the early detection of breast cancer. Although much is currently being done to improve rescreening rates, additional efforts may be required to locate and assist the missing 20%. To encourage these women to complete routine rescreening, providers and public health programs may need to communicate more frequently or more effectively with their enrollees. All age-eligible, low-income women need to know that they should have another mammogram every 1–2 years and that they may be eligible for another free mammogram whether through the national program or another service provider.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received 8/20/03; revised 12/16/03; accepted 12/30/03.
|
References |
|---|
|
TopAbstractBackgroundMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. A. Edwards, A. M. Chiarelli, L. Stewart, V. Majpruz, P. Ritvo, and V. Mai Predisposing Factors Associated with Compliance to Biennial Breast Screening among Centers with and without Nurses Cancer Epidemiol. Biomarkers Prev., March 1, 2009; 18(3): 739 - 747. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 U. Montes, L. M. Seijo, A. Campo, A. B. Alcaide, G. Bastarrika, and J. J. Zulueta Factors determining early adherence to a lung cancer screening protocol Eur. Respir. J., September 1, 2007; 30(3): 532 - 537. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. Rakowski, H. Meissner, S. W. Vernon, N. Breen, B. Rimer, and M. A. Clark Correlates of Repeat and Recent Mammography for Women Ages 45 to 75 in the 2002 to 2003 Health Information National Trends Survey (HINTS 2003). Cancer Epidemiol. Biomarkers Prev., November 1, 2006; 15(11): 2093 - 2101. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Calvocoressi, M. Stolar, S. V. Kasl, E. B. Claus, and B. A. Jones Applying Recursive Partitioning to a Prospective Study of Factors Associated with Adherence to Mammography Screening Guidelines Am. J. Epidemiol., December 15, 2005; 162(12): 1215 - 1224. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Rosenberg, L. A. Wise, J. R. Palmer, N. J. Horton, and L. L. Adams-Campbell A Multilevel Study of Socioeconomic Predictors of Regular Mammography Use Among African-American Women Cancer Epidemiol. Biomarkers Prev., November 1, 2005; 14(11): 2628 - 2633. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
