What Factors Are Associated with Diagnostic Follow-Up after Abnormal Mammograms? Findings from a U.S. National Survey

Regular mammography use is associated with earlier stage of breast cancer at diagnosis and improved breast cancer survival in clinical trials (1, 2). Although the magnitude of the mortality benefit associated with screening has been debated (3), most evidence-based guidelines continue to recommend regular screening (4, 5). Mammography use has increased dramatically in the past two decades in the United States (6, 7), yet in some populations, particularly racial minorities, and the poor or medically underserved, the percentage of women with advanced disease at diagnosis remains high (8–11). A potential explanation is that follow-up after an abnormal mammogram is delayed or incomplete, and thus, the benefits of screening are not being realized.

Prior studies have reported that between 32% and 98% of women with abnormal mammograms receive at least some follow-up (12–26). This wide range may be due, in part, to study differences in the severity of the underlying abnormal mammogram, the time period for follow-up assessment, and the definition of follow-up services (27). Patient characteristics associated with incomplete follow-up of abnormal screening tests include being poor or uneducated (13, 20), a racial or ethnic minority (13, 22, 28), and under- or uninsured (13). Women with delayed or incomplete follow-up of abnormal screening tests have also been reported to have high levels of cancer-specific anxiety (12) or fatalistic beliefs and misconceptions about cancer (29). Such attitudes and beliefs are more common among minority, low income, or uninsured women (29, 30). However, most studies of follow-up after abnormal mammograms have been conducted in populations within health insurance systems (24), health care facilities (15, 17), programs with income-eligibility requirements (22, 25), or racial groups (12), and as a result, little work has been done to disentangle the effects of race, socioeconomic status, health beliefs and behaviors, and health care access factors on follow-up of abnormal mammograms.

We used a national sample of women in the U.S. who had ever had an abnormal mammogram to explore the roles of race, socioeconomic status, health behaviors, health care access factors, and cancer risk perceptions on completion of diagnostic follow-up.

The sample for this study was selected from the year 2000 National Health Interview Survey (NHIS), an annual multistage survey conducted in the civilian population of the U.S. The sampling strategy is designed to yield nationally representative estimates of health and health behavior. The NHIS is an in-person, household survey and all data are self-report. Both Black and Hispanic populations are oversampled. The Cancer Control Module contains questions about cancer screening and follow-up, and had an overall response rate of 72.1%. More information on survey design and content is available from http://www.cdc.gov/nchs/nhis.htm.

Women aged 30 and above were asked if they had ever had a mammogram, and those that had were asked the age at the first mammogram, number of mammograms within the past 6 years, and a series of questions related to the most recent mammogram. Women were then asked “Have you ever had a mammogram where the results were NOT normal?” The study sample included all women aged 30 and above who reported ever having had an abnormal mammogram and who responded to a follow-up question about the type of diagnostic follow-up (N = 1901; weighted N = 11.5 million women).

We incorporated elements of several theoretical models of health care utilization in this analysis. First, we used a modified version of the Andersen and Aday model of realized access to health care (31–33) to guide selection of variables potentially associated with follow-up of abnormal mammograms. This model depicts the interaction of patients and providers within the health care system and their social environment. An important aspect of the model is the outcome of realized access to care or the use of needed medical services. Patient level components of the model that may either facilitate or impede the receipt of follow-up care include sociodemographic characteristics, general health and health behaviors, and health care utilization characteristics. Additionally, because behavioral models of health care utilization, such as the Health Belief Model (34), include perceived susceptibility or severity of disease as components of health behaviors, we also included variables related to cancer risk and risk perceptions.

The outcome measure in this study was receipt of at least some diagnostic follow-up after an abnormal mammogram. Respondents were asked to list all additional tests or procedures they received, which were coded into the following categories: additional imaging; clinical examination or surgical consult; biopsy or fine-needle aspiration; breast or lump removed; or no follow-up. We categorized follow-up as a binary variable—completed at least some follow-up or did not complete any follow-up (yes, no).

Sociodemographic characteristics were measured by age at the time of the survey (<50, 50–64, 65+), race (White, non-Hispanic, Black, non-Hispanic, and Hispanic/Other), immigrant status (born in the U.S. versus other), marital status (married or living as married versus other), educational attainment (less than high school graduate, high school graduate, at least some college), and income level in relation to the poverty level (0–149%, 150–299%, 300–449%, and 450%+). Residence was also characterized by geographic region (Northeast, Midwest, South, or West), and community size [Metropolitan Statistical Area (MSA), non-MSA].

General health and health behaviors were measured by health status (excellent/very good, good, fair/poor/missing), problems with activities of daily living (yes, no), problems with instrumental activities of daily living (yes, no), body mass index (<24.9, 25–29.9, 30+), chronic anxiety or depression (yes, no), and cigarette smoking status (current, former, or never smoker). Compliance with regular mammography was calculated for the subset of women aged 55–79, and was defined as having the most recent mammogram within the past 2 years and at least three mammograms in the past 6 years (yes, no) (35). This age range was chosen because it includes a sufficient number of years for a woman to have received three mammograms on a biannual schedule starting at age 50, the age where most guidelines have consistently recommended mammography screening initiation (4, 5).

Breast cancer risk was measured by the number of first-degree relatives with breast cancer under the age of 50 (0, 1+) and the Gail breast cancer risk score (36, 37). This score is based on current age, age at first live birth, age at menarche, number of first-degree relatives with breast cancer, and number of breast biopsies, and was used to calculate 5-year risk of breast cancer (<1.67%, 1.67%+). Respondents reported their perceived risk of getting cancer in the future as low, medium, or high and the amount of cancer they perceived in their family as low, medium, or high. Both variables were coded in these response categories (low, medium, high).

Health care utilization characteristics were measured by type of health insurance coverage [private/military, public only (i.e., Medicaid, Medicare without additional Medigap policies), no health insurance], whether or not the health insurance required approval for specialty care (yes, no), and whether the woman had a usual source of sick care (yes, no), or preventive care (yes, no). Primary care (yes, no) or specialist visits (yes, no) in the past year were also used to measure heath care utilization. Barriers to care included delayed or incomplete medical care due to cost or worry about cost (yes, no) or delayed or incomplete medical care for any other reason (yes, no).

Descriptive statistics were calculated for the study sample of women who ever had an abnormal mammogram. Bivariate and multivariate logistic regression analyses were used to explore whether follow-up after an abnormal mammogram was associated with sociodemographic characteristics, general health and health behaviors, health care access factors, and cancer risk and risk perceptions.

Multivariate models were fitted in a stage-wise fashion within each component of the theoretical framework. Covariates with a dose-response association with the outcome variable or an association with the outcome variable in bivariate analyses at P < 0.20 were retained for separate component-specific intermediate multivariate models (i.e., sociodemographic characteristics, general health and health behaviors, health care utilization characteristics, and cancer risk perceptions). Covariates with associations with the outcome variable of P < 0.20 in intermediate models were retained for the final multivariate model. Potential interactions were assessed with stratified analyses and interaction terms for race/ethnicity and immigrant status, geographic region and MSA, and type of health insurance and requirement for specialty referral. The statistical software SUDAAN (38) was used to incorporate the complex NHIS sample design and weights in descriptive statistics and logistic regressions.

Compliance with regular mammography use and recent specialist visit were not included in bivariate or multivariate analyses because additional mammographic imaging or consultation with a breast surgeon are also follow-up procedures for an abnormal mammogram. Similarly, the Gail risk score was reported descriptively only because biopsy is a component of the Gail risk score and is a procedure used in the follow-up of abnormal mammograms.

Approximately 40% of women in the study sample were between the ages of 50 and 64, with about 34% aged less than 50 and the remaining 26% aged 65 or above (Table 1). The vast majority of women who reported having an abnormal mammogram were White, non-Hispanic, born in the U.S., with at least a high school education. Most women in the sample rated their health as excellent or very good, had few health limitations as measured by activities of daily living or instrumental activities of daily living, and had never smoked. Among the subset of women aged 55–79 (N = 763), 85.2% were compliant with regular, on-going mammography screening.

Approximately 14% of the sample had one or more first-degree relatives with breast cancer, and less than one-quarter had elevated 5-year breast cancer risk as measured by the Gail model, estimated their risk of getting cancer in the future as high, or perceived the amount of cancer in their family as high. More than 90% of the sample had access to care as measured by health insurance coverage, a usual source of preventive care, sick care, or having seen a primary care physician within the past 12 months. Less than 15% delayed care in the past 12 months because of cost or other reasons.

In response to the type of follow-up completed, 8.6% [95% confidence interval: 7.2%, 10.0%] of the sample reported they received no follow-up. The weighted estimate is approximately 1 million women nationally who ever had an abnormal mammogram and received no follow-up. Most women who completed at least some follow-up had more than one type of test including additional mammographic or ultrasound imaging (55%), removal of a lump (32%), biopsy or fine needle aspiration (30%), or a clinical examination or surgical consultation (13%). These percentages add to more than 100% because the process of diagnostic follow-up may include a series of tests or procedures. Of the group of women who completed at least some follow-up, 15% reported that the additional results indicated breast cancer (data not shown).

Sociodemographic Characteristics. Black, non-Hispanic, and Hispanic women were less likely to report follow-up than White, non-Hispanic women in bivariate analyses (Table 2), but when the other demographic characteristics were included in the intermediate multivariate model, this association was no longer statistically significant. Low socioeconomic status, measured by either educational attainment or poverty level income threshold was associated with reported follow-up in bivariate analyses with a clear dose response pattern. Educational attainment was included in the final, multivariate model (Table 3). Women with less than a high school education were less likely to report follow-up compared with women with at least a college degree, controlling for all other factors. Women living in the South and Northeast were less likely to report follow-up compared with women living in the Midwest, although this association was of borderline significance for women in the Northeast. Finally, younger women (<50) were less likely to complete at least some follow-up than women aged 65 and above in multivariate analyses.

General Health and Health Behaviors. Women who reported their health as being good or fair/poor compared with very good/excellent were less likely to complete at least some diagnostic follow-up in bivariate and multivariate analyses.

Breast Cancer Risk Factors and Cancer Risk Perceptions. In general, as risk or perceived risk of cancer increased, so did likelihood of completing at least some diagnostic follow-up. Women who indicated the amount of cancer in their family was high were more likely to complete at least some follow-up compared with women with a low amount of cancer in the family.

Health Care Utilization Characteristics. Women with no health insurance or only public health insurance were less likely to complete at least some follow-up in bivariate analyses, but this estimate was no longer significant after controlling for other variables in the final model. Women who did not get needed medical care or who delayed care because of worry about cost were less likely to complete at least some follow-up in bivariate analyses, although the estimate was not significant in multivariate analyses.

None of the interactions tested was significantly associated with completion of at least some follow-up in intermediate multivariate models and were therefore not included in the final model.

This is one of the first studies in the U.S. to use nationally representative data to explore the association between receipt of follow-up care after abnormal mammograms and sociodemographic factors, general health and health behaviors, cancer risk and risk perceptions, and health care utilization. Approximately 9% of our sample, almost 1 million women, did not complete any diagnostic follow-up after abnormal mammograms. Even though lack of follow-up for a specific abnormal mammogram may not be associated with a diagnosis of advanced breast cancer, women without cancer who do not complete any diagnostic follow-up may also fail to return to breast cancer screening, may delay seeking care should they have breast cancer symptoms in the future, or may fail to initiate other types of cancer screening. Women who have breast cancer at the time of an abnormal mammogram, but do not comply with diagnostic follow-up may have more advanced disease at diagnosis. Studies have shown that diagnostic delays as short as 3 months following symptom identification are associated with poorer breast cancer survival (39). Thus, delayed or incomplete follow-up after an abnormal screening test is an important component of the breast cancer screening process and its role in future screening behaviors and stage of disease at diagnosis is an important area for additional research. Mammography registries with linkages to tumor registries, such as the Breast Cancer Surveillance Consortium (40), or other state-based or health plan-based programs (41–43) might be useful data sources for this research because they contain women-level longitudinal data on mammography screening, abnormal mammograms, subsequent screening, and breast cancer diagnoses.

Another strength of the study was the breadth of detail in the NHIS that allowed us to use a theoretical framework to explore factors associated with follow-up after abnormal mammograms. Our findings support the usefulness of this framework in that variables representing major components of the model, including sociodemographic factors, general health and health behaviors, and cancer risk and risk perceptions were associated with follow-up after abnormal mammograms. More specifically, women who were younger, less educated, in poorer health, and who perceived low levels of cancer in their families were less likely to complete any follow-up after an abnormal mammogram. We also identified geographic variation in completion of at least some follow-up. Our findings help to identify additional variables that might more fully elaborate model components in future research studies of factors associated with receipt of follow-up care after abnormal screening tests.

One of our objectives for this study was to provide insight into the relative importance of race/ethnicity, other sociodemographic characteristics, and health care utilization characteristics on completion of at least some follow-up after abnormal mammograms. Although minority women were less likely to report follow-up after abnormal mammograms in bivariate analyses, once educational attainment, age, and geographic region were included in an intermediate multivariate model of socioeconomic characteristics, this effect diminished. Educational attainment, on the other hand, was associated with reported follow-up, controlling for all other factors. We did not have measures of health literacy or numeracy, communication about abnormal findings, or comprehension of follow-up recommendation, factors likely to be reflected in an association between low educational attainment and incomplete follow-up. Future in-depth studies of follow-up care after abnormal screening tests might include these measures in addition to race/ethnicity and measures of sociodemographic status. Inclusion of measures of health literacy and comprehension of abnormal findings will also be important elements for future enhancements of theoretical models of health care access and utilization.

In this study, we found that women who perceived the level of cancer in their family to be high were more likely to complete at least some diagnostic follow-up than were similar women who perceived low levels of cancer in their family. Perceived level of cancer in the family is a subjective measure of general cancer risk, however, and does not necessarily reflect objective breast cancer risk, such as the Gail model risk score (36). Others have reported that fatalism, cancer-specific anxiety, fear of diagnostic tests, or worry about cancer are associated with delayed or incomplete follow-up of abnormal screening tests (12, 29), although associations with objective measures of cancer risk are less clear. The inter-relationship between subjective and objective measures of breast cancer risk in relation to follow-up behaviors is likely to be complex and is an important area for additional research, particularly within a theoretical framework.

Our sample included only women who understood and reported they had an abnormal mammogram. Several studies have reported that physicians do not always inform patients about abnormal screening results (44) or recommend complete diagnostic evaluation (45–49). Two findings are potentially of interest in relation to physician communication about, and recommendation for follow-up care—younger women and women in poorer health were less likely to report any follow-up care. Physicians may be less inclined to strongly recommend diagnostic follow-up for women younger than 50, in whom breast cancer incidence is low, or women of poorer health status, who may be perceived as having less benefit from treatment. Inclusion of physician factors such as training, screening practice patterns, or beliefs about screening efficacy, and characteristics of the physician-patient interaction and follow-up recommendation are particularly relevant in future, in-depth research.

We also identified geographic variation in completion of at least some follow-up. Compared with women living in the Midwest, women living in the South and Northeast were less likely to complete at least some follow-up after an abnormal mammogram. Geographic variation in receipt of follow-up care may reflect patterns of unmeasured population characteristics; differences in state-based insurance, screening legislation, or health care programs; physician practice patterns or practice styles, medical care infrastructure, or other area-level factors. Such variation might be explored further with multi-level models with woman-level characteristics, physician characteristics, and state-based or area level medical and socioeconomic characteristics.

Our finding that approximately 9% of women did not complete at least some follow-up after an abnormal mammogram is not directly comparable to prior estimates of receipt of follow-up care after abnormal mammograms (12, 20–22, 24–26, 28). This estimate is based on self-report, includes abnormal mammograms of all levels of severity, is not linked to a single abnormal mammogram, and is not based on a specific time period (e.g., 30 days, 60 days). The majority of prior studies identified women through medical records, grouped abnormal mammograms by type of follow-up recommendation, and used a specific time period to assess receipt of follow-up care.

Finally, our findings may be broadly applicable to improving understanding of factors affecting diagnostic follow-up of other cancer screening tests. Effective screening tests are also available for early detection of cervical and colorectal cancer (50–53), and receipt of follow-up after abnormal results with these screening tests has also been reported to be sub-optimal (17, 45, 47, 48, 54–81); with the vast majority of studies reporting that fewer than 75% of patients completed follow-up (27). As with studies of receipt of follow-up care after abnormal mammograms, variation in these estimates may be due in part to study differences in the severity of the underlying abnormal finding, the time period for follow-up assessment, and the criteria used to define completeness of follow-up services (27). Similar to studies of receipt of follow-up care after abnormal mammograms, patient factors associated with incomplete follow-up of abnormal cervical cancer screening tests also include low socioeconomic status, lack of health insurance, and belonging to a racial or ethnic minority group (54, 69, 82, 83). Additionally, women who have problems coping with abnormal Pap tests, fear painful invasive procedures, or who worry about cancer are less likely to complete follow-up (62, 67, 82, 84). Less work has been done to assess patient-level factors associated with follow-up care after abnormal colorectal cancer screening tests (27), although it is likely that similar patient factors may also be important for the receipt of follow-up care after abnormal colorectal cancer screening results.

The majority of patient-targeted interventions developed to improve follow-up care after abnormal screening tests are effective; strategies include telephone counseling, educational pamphlets, telephone and mailed reminders, and lay health workers (85). To date, the vast majority of these patient-targeted interventions have been developed to improve adherence to receipt of follow-up care after abnormal Pap tests (27). Adaptation of these successful strategies to improve follow-up care after abnormal mammograms will be an important area for future research.

Despite the strengths of having a large nationally representative sample with detail on patient characteristics, health, and health behaviors to assess factors associated with the receipt of follow-up care, this study has several limitations related to the lack of detailed information about the abnormal mammogram and use of self-report data. No data were available on recommended follow-up for the abnormal mammogram, and as a result, follow-up procedures cannot be linked to their underlying recommendations. Additionally, information was not available about the timing of the abnormal mammogram in relation to the survey or about whether it was a screening mammogram or the result of a specific breast symptom. It is also possible that some women may have had an abnormal finding with a 6-month follow-up recommendation just before participation in the survey and not had the opportunity to complete diagnostic follow-up.

The questions we used to measure different components of the theoretical framework were not specific to the abnormal mammogram and in some cases were based on health care use in the past 12 months, whereas abnormal mammograms could have occurred at any time in the past. Additionally, health care access and utilization characteristics may change over time, potentially limiting the identification of associations between variables measuring these constructs and the receipt of follow-up care. For example, compared with individuals with continuous health insurance coverage, individuals with intermittent coverage have been shown to have lower preventive service utilization, although not as low as individuals who are continuously uninsured (86). Grouping individuals with intermittent insurance coverage with either individuals continuously insured or individuals continuously uninsured would diminish observed differences between these groups. Other variables, such as educational attainment, are less likely to change over time in an adult population, minimizing limitations in measurement.

These data in the NHIS are self-report and women may have difficulties remembering the specific procedure or a complicated series of procedures that constitute diagnostic follow-up (87). We used completion of at least some follow-up as the outcome variable, thus minimizing any problems with recall of specific procedures. Although other studies found that patients overreport screening compared with the medical record (88–91), potentially due, in part, to the desire to appear compliant with recognized health behaviors (87), it is unlikely that patients would overidentify themselves as having an abnormal mammogram for social desirability reasons. Finally, although the NHIS oversampled African Americans and Hispanics, there were too few women with abnormal mammograms in other minority groups, such as Asian Americans, to estimate their receipt of follow-up care separately. Such associations might be explored in greater detail with the 2001 California Health Interview Survey, a state-based survey with a similar cancer control module conducted in the nation's most racially, ethnically, and linguistically diverse population (92). A major goal of federal co-sponsorship of this survey was to enable investigators to better estimate cancer risk factors and cancer screening behaviors in racial and ethnic subgroups, such as Asian Americans and Native Americans.

In conclusion, the results of our study indicate that approximately 1 million women in the U.S. reported having had an abnormal mammogram without any diagnostic follow-up, potentially reducing the effectiveness of mass screening programs. Patient-level sociodemographic factors, general health and health behaviors, and cancer risk perceptions were associated with completion of at least some follow-up in multivariate analysis. Our findings may be broadly applicable to receipt of follow-up care after other abnormal cancer screening tests. Exploration of the association between incomplete follow-up and future screening behaviors, as well as the roles of health literacy, provider communication and recommendations, risk perceptions, and health care utilization factors are important areas for additional in-depth research. Ideally, such research should be designed and conducted within theoretical models of health care utilization that reflect health care delivery in the U.S.