Long-term Mammography Adherence among Uninsured Women Enrolled in the Breast Screening and Patient Navigation (BSPAN) Program

Screening mammography adherence, defined as regular breast cancer screening in the form of mammography every 1 to 2 years (1), has proven to be an effective secondary prevention method with an increased probability of diagnosing breast cancer at an earlier stage, and a consequent improvement in survival by 21% posttreatment (2–8). However, rates of screening adherence are universally lower than rates of women up-to-date for screening (mammogram within the previous 2 years; refs. 9–11). Lower adherence can be attributed to barriers, such as fear of positive mammogram, pain, embarrassment, transportation issues, and lack of physician recommendation (12, 13). However, race/ethnicity and usual source of care are the strongest predictors of mammography adherence; rates for Latinas and Blacks are lower than Whites (14). Being uninsured and lacking a regular source of care are independent predictors of lower adherence (11). Additionally, race and insurance status interact significantly, resulting in lower adherence among uninsured minority women (12).

In 1990, the Centers for Disease Control and Prevention (CDC) created the National Breast and Cervical Cancer Early Detection Program (NBCCEDP); the program applies federal and state funding to subsidize breast and cervical cancer screening to under- and uninsured women across the country. NBCCEDP seeks to decrease cancer health disparities between insured and uninsured women by increasing access to screening services and providing presumptive Medicaid coverage for those found to have a breast cancer diagnosis. As of 2020, NBCCEDP has funded 70 programs, and provided breast and cervical cancer screening services to approximately 6 million women in all 50 U.S. states, District of Columbia, 6 U.S. territories, and 13 tribal organizations (15). Prior evaluation of some of these NBCCEDP programs have reported mammography adherence rates ranging from 24% to 81.5%, depending on screening interval and location of the program (10, 16–18). However, these evaluations have primarily assessed baseline adherence (one on-schedule repeat screen after initial mammogram) rather than longitudinal adherence (long-term on-schedule repeat mammograms).

In 2009, Moncrief Cancer Institute, a community-based nonprofit affiliate of the University of Texas Southwestern Harold C. Simmons Comprehensive Cancer Center, created the Breast Screening and Patient Navigation (BSPAN) Program (19) to increase access among under- and uninsured women in North Texas. BSPAN, which provides access to no-cost patient navigation, breast cancer screening, and diagnostic services, is financed through a combination of federal, state, and philanthropic funds. The purpose of this study is to generate contemporary evidence of baseline and longitudinal adherence among uninsured women enrolled in the NBCCEDP-funded program (BSPAN) of North Texas. We analyzed data from the unique longitudinal BSPAN data to determine both prevalence and correlates of (i) baseline adherence, and (ii) longitudinal adherence to screening mammograms.

In the BSPAN hub-and-spoke model, Moncrief (the “hub”) contracts with Texas Breast and Cervical Cancer Services (BCCS)–the state affiliate of the NBCCEDP program–and then provides central reimbursement for mammography services for a regional network of participating community providers (the “spokes”). These “spoke” provider groups lack capacity to administer a BCCS contract themselves (20). The administrative activities are funded by the Cancer Prevention and Research Institute of Texas (CPRIT), while the screening and diagnostic services are funded by BCCS. Funding partner priorities emphasize targeting never-screened and screen-not-up-to-date (have not received a mammogram at least within the past 5 years) individuals.

Acting as the “hub”, Moncrief also provides community outreach to alert women in rural and underserved communities of the opportunity to receive no-cost mammography services from local partners or through BSPAN mobile mammography events (20, 21). In brief, women contact the BSPAN program via a toll-free telephone number in response to advertisements, pamphlets, word-of-mouth outreach, or community events. BSPAN program staff answer the calls, assess eligibility, and schedule eligible women for screening mammograms at their nearest “spoke” location. Telephone-based patient navigation facilitates result reporting and diagnostic appointments for follow-up, as necessary. All participants receive annual reminder letters for repeat mammography, mailed from the “spoke” location. The “hub” makes telephone calls for annual reminders, subject to available funds (when BCCS reappropriations unexpectedly award additional midyear funds), clinical capacity, and schedule availability at “spoke” locations. We extracted and analyzed electronic health record data for these BSPAN participants during 2012 to 2019. Ethical approval for this study was granted by the Institutional Review Board at University of Texas Southwestern Medical Center in accordance with the U.S. Common Rule.

We included BSPAN program participants ages 40 to 64 years (corresponding to BCCS eligibility criteria) in our analyses if they had at least one mammogram (screening or diagnostic) with a corresponding BIRADS (Breast Imaging-Reporting and Data System) score during the years 2012 to 2019. Women with the first mammogram (index mammogram) after April 1, 2019 were excluded, due to insufficient follow-up time. Women with a personal history of breast cancer prior to, and those diagnosed with breast cancer following their first encounter in BSPAN were also excluded, because guidelines recommend they receive surveillance rather than screening mammograms.

The outcomes–baseline and longitudinal mammography adherence–were defined as per Hubbard and colleagues (1). Baseline adherence was defined as a binary outcome with women categorized as adherent if they received their second mammogram 9 to 30 months after the index mammogram. This is consistent with other studies wherein repeat screening within 24 months is considered guideline-concordant and an additional 6 months provides a buffer to adjust for scheduling queues and wait periods not under the patient's control (1). Longitudinal adherence, assessed only among women categorized as baseline adherent, was defined as being adherent 39 months from the index mammogram. Baseline adherent women received at least one repeat mammogram 9 months from the index mammogram, and are considered guideline adherent up to 30 months from that period. The earliest a woman could become non-adherent would be 39 months from the index mammogram. Thus, longitudinal adherence helps to assess the time that women remain adherent to screening (instead of the number of screening mammograms), because it quantifies person-time coverage of screening in the population (1).

We assessed potential correlates including demographic factors such as age, race/ethnicity, marital status, urbanization [based on 2013 rural-urban continuum codes, defined as urban (codes 1–2), near metro (codes 3–6), and rural (codes 7–9)] (22, 23), language preference, reading ability, and years lived in the United States, as well as clinical factors such as hysterectomy status, personal history of non-breast cancer, menopausal status, hormone replacement therapy use, prior mammograms (screening and diagnostic), prior Papanicolaou (PAP) test for cervical cancer, and prior breast symptoms.

We used |\chi $|² and independent samples t test to tabulate and depict patient characteristics and baseline adherence among women enrolled in BSPAN, and univariate logistic regression to assess associations of sociodemographic and clinical characteristics with baseline adherence. Predictors with a P value of <0.2 were included in the multivariable logistic regression. Two variables in BSPAN (reading ability and years lived in the United States) were added only in the year 2015. There is a possibility that the missing data in previous years for these variables might affect the results of logistic regression. So, we conducted sensitivity analysis excluding those two covariates.

Because longitudinal adherence is based on person-time coverage of screening, predictors affecting longitudinal adherence were assessed using a Cox proportional hazards model. Predictors used in the Cox model were the same as that of logistic regression, and those with a P value < 0.2 in the univariate model were entered into the multivariable Cox model. Failure was defined as diagnosis of breast cancer, loss to follow-up, or death. The proportional hazards assumptions were tested for the model. All statistical tests were performed using SAS Software version 9.4 and all P values < 0.05 were considered statistically significant.

Deidentified data can be made available upon request.

We identified 21,897 women aged 40 to 64 years who received mammograms through BSPAN. Of these, 2,605 women were excluded based on eligibility criteria–846 had prior diagnosis of breast cancer, 1,597 had insufficient follow-up time, and 162 were diagnosed with breast cancer after an index mammogram in BSPAN. Thus, 19,292 women were included for analyses (Fig. 1).

Of the 19,292 women, 13,910 women (72.1%) were baseline non-adherent, meaning they did not receive a second screen within 30 months from the index mammogram. Only 5,382 (27.9%) were baseline adherent (See Fig. 1). Demographic and clinical characteristics of women by baseline adherence are shown in Table 1. Both adherent and non-adherent women were mostly young, predominantly Hispanic, single, lived in urban counties, and had good reading ability. Most women were also asymptomatic, had no prior history of cancer, were menopausal and had some screening encounter in the past (PAP test or screening mammogram).

Table 2 shows results of multivariable logistic regression assessing correlates of baseline adherence. Being baseline adherent was more likely among women who were partnered, lived in urban counties, preferred speaking Spanish, had poor reading ability, lived longer in the United States, were asymptomatic, were menopausal, had prior PAP testing, and had prior screening or diagnostic mammograms, compared with women who were non-partnered, lived in rural counties, preferred speaking English, had good reading ability, lived in the United States for shorter duration, were symptomatic, nonmenopausal, had no prior PAP testing, and had no prior screening or diagnostic mammograms, respectively. Sensitivity analyses revealed similar results (Supplementary Table S1).

Of the 5,382 women who were baseline adherent, 4,364 (81.1%) women demonstrated longitudinal adherence (See Fig. 1).

HRs computed from multivariable Cox proportional hazards model appear in Table 3. All assumptions of proportional hazards model were met. Correlates of longitudinal adherence are similar to those from the multivariable logistic regression model for baseline adherence. Women who preferred Spanish, were asymptomatic, menopausal, or had prior screening mammograms had higher odds of maintaining longitudinal adherence compared with women who preferred English, were symptomatic, nonmenopausal, and had no prior screening mammograms, respectively. The only covariate that changed direction was relationship status, where women who were partnered showed lower longitudinal adherence, compared with non-partnered women.

To the best of our knowledge, our study is the first to assess longitudinal adherence to screening mammography among under- and uninsured women served through an opportunistic community-based outreach program designed to increase access to publicly subsidized clinical services. Prior to BSPAN, BCCS reimbursement was only available in two urban counties with integrated public health systems and a few urban federally qualified health centers. Individual community providers and commercial breast imaging centers in North Texas did not have access to BCCS reimbursement prior to BSPAN, whose hub-and-spoke model addressed barriers both at the provider level (limited number across vast geographic regions and lack of infrastructure to compete for BCCS contracts with the Texas Department of State Health Services), and at the patient level (lack of knowledge about no-cost screening services, distance from providers, and language barriers; refs. 19, 24). Results from this unique real-world dataset provide realistic mammography adherence rates and demonstrate behavioral patterns among women independent of any regular patient-centered medical home. Results may be generalizable to new areas introducing similar no-cost screening through CDC funding or other programs.

Our analyses of data from the BSPAN program showed mixed results. On the one hand, the program was able to provide screening mammograms for more than 19,000 under- or uninsured women in North Texas. However, only about 28% received a second mammogram 9 to 30 months after the index mammogram (baseline adherence). This is consistent with results from other NBCCEDP-funded programs. An initial study conducted in New York State used administrative data from 9,485 women screened between 1988 and 1991 and reported a rescreening rate of 27% within 36 to 60 months (16). Another study using administrative data conducted in Washington State reported a rescreening rate of 25.7% within 15 months and 40.3% within 27 months among 2,888 women who received their index mammogram between 1994 and 1995 (17). One study that used self-reporting coupled with administrative data from four CDC programs, 1991 to 1998, reported a rescreening rate of 72.4% within 18 months and 81.5% within 30 months; this single study showed high rescreening rates and may be partially attributed to inclusion of women receiving repeat screening outside the NBCCEDP program (10). Another study compared rescreening rates among women in the NBCCEDP program alone versus those enrolled for the combined NBCCEDP and WISEWOMAN program (Well-Integrated Screening and Evaluation for Women Across the Nation program that provides cardiovascular screening along with cancer screening services) across 10 locations. Outcomes from those analyses, using data from 2000 to 2004, reported rescreening rates from 24% to 80%, depending on the location and the type of program (NBCCEDP alone or NBCCEDP and WISEWOMAN combined; ref. 18).

A noteworthy finding from our study is that, among those who were baseline adherent, an overwhelming majority (>80%) demonstrated longitudinal adherence 39 months after the index mammogram. Longitudinal adherence has not been previously assessed in a NBCCEDP-funded program, but studies reporting longitudinal adherence rates among insured women have found rates ranging from 42% to 85%, depending on duration of the follow-up and the screening intervals (1, 9, 25, 26). The fact that longitudinal adherence rates in the unique BSPAN program are comparable with those achieved among insured women in previous studies recommends the merit of this unique “hub-and-spoke” outreach strategy providing access to the CDC funding for screening. The findings also emphasize the importance of the second mammogram in the progression toward continued adherence.

Our assessment of correlates of both baseline and longitudinal adherence identified findings consistent with other studies. Social support has been associated with higher rates of repeat screening, especially among women at lower income and education levels (27), and could explain the higher odds of baseline adherence among women who were partnered compared with those who were non-partnered. Consistent with our findings, previous research has also found women living in urban areas have better access to healthcare and higher utilization of preventive services, including repeat screening, compared with rural women (28). Although BSPAN sought to mitigate the difference between women in rural and urban areas, outcomes show that the outreach program was not able to do so for repeat adherence.

We found language, literacy, and time lived in the United States were associated with repeat adherence. Consistent with previous studies (29), women who preferred speaking Spanish had higher rates of adherence. We also identified a previously unreported association of self-reported poor reading ability with adherence. This is contrary to findings from prior research conducted in non-NBCCEDP samples, where low English fluency and inability to read and write English was associated with poor adherence in some studies (30–32), whereas one study showed no association between English fluency and screening adherence (33). We also found baseline adherence was more likely among women who had lived longer in the United States. Brown and colleagues, who previously also found a positive association between time in the United States and adherence, suggested greater adherence could be attributed to acculturation (34). Our unique finding, where Spanish-speaking women with poor reading ability and longer period of residence in the United States have higher rates of adherence, could be due to two reasons. One possible reason could be that these women reside in Spanish-speaking enclaves. Women who live in enclaves have better social support (35) and are more likely to practice protective health if they have access to health care (36). Our BSPAN program provided health care access using the hub-and-spoke model along with no-cost screening services and could possibly explain the high rates of screening adherence. A second reason could be that the low reading ability in our group could be associated with lack of awareness of potential risks of screening, thereby leading to high rates of adherence. Kadivar and colleagues have suggested that more acculturated Hispanic women and those with higher literacy may be more aware of potential risks of screening and thereby have lower repeat screening rates (37). Further research is required to understand this phenomenon and learn how to facilitate adherence in other areas with large Spanish-speaking enclaves.

Historically, prior mammography usage has been a strong predictor of repeat screening, also evident in our results (38, 39). Similarly, it has been shown that women who practice other screening behaviors such as PAP test and colorectal cancer screening are also likely to undergo both initial and repeat breast cancer screening (18, 40). Although race has been consistently shown to be associated with repeat screening rates, we did not see such an association in our study. This could be due to our unique dataset with a large number of Hispanic women and proportionately lower number of Black women.

Covariates were similar for baseline and longitudinal adherence, except one. Having a partner was associated with higher baseline adherence, but lower longitudinal adherence. One possible explanation for this could be that the effect of marital status in the long-term is based on marital transitions and life course and not just social support, as discussed by prior studies (41, 42). Further research is needed to understand the effect of marital transitions on cancer screening behaviors in the long-term.

Our study has some limitations. Because it included low-income, under- and uninsured predominantly Hispanic women served by NBCCEDP in North Texas, the results cannot be generalized to other dissimilar populations. The data for this study were extracted from electronic health records of women enrolled in an outreach program, and not a research study. This limited our ability to include important qualitative variables, such as self-efficacy, self-reported health status, and interaction with the physician, which could provide a comprehensive understanding of the longitudinal adherence among women in BSPAN. Another limitation is the lack of investigation into women with delayed repeat screening until after 30 months from the index mammogram (Fig. 1). The low prevalence of this group (205 out of 19,292 women or 1.1%) in our sample precluded the use of meaningful statistical methods to assess and understand this group. Additionally, we could not capture information about women who received screening services outside BSPAN, which could partially explain the lower rates of baseline adherence in our sample (i.e., women could have accessed additional screens outside the BSPAN program). However, this likelihood is lower in BSPAN than for insured samples, because the majority of women using BSPAN were very low-income and, in a state that did not expand Medicaid under the Affordable Care Act, lacked coverage for routine preventive services such as mammography. The Community Preventive Services Task Force recommends the use of client reminders to increase repeat screening rates, and emphasizes that telephone reminders are more effective than reminder letters to increase rescreening rates (43). Although our program mailed annual reminder letters for repeat screening to all BSPAN participants, telephone reminders were not universally implemented and could have attributed to our low baseline adherence rates. However, inconsistent implementation of navigation and telephone reminders are customary in real-world outreach programs that primarily serve under- and uninsured minority populations compared with well-controlled research programs (44).

Despite the limitations, this study is important as it quantifies mammography rates, both baseline and longitudinal adherence, in under- and uninsured women in North Texas, and provides insight into the appropriate use of scarce resources to improve longitudinal adherence. Although BSPAN was able to overcome administrative difficulties and provide breast cancer screening and diagnostic services to more than 19,000 women, rates of baseline adherence were lower, especially compared with other NBCCEDP-partner programs across the United States. However, an encouraging finding was that a large proportion of baseline adherent women continued screening to achieve longitudinal adherence. This highlights the importance of concentrating resources and implementing cost-effective interventions, especially during the second screen. For instance, at the individual level, we can improve adherence by implementing tailored print and telephone reminders universally and increasing physician-patient communication, while at the program level, we can reduce barriers such as transportation by increasing access to mobile mammography clinics. Further research is required to understand optimal implementation strategies of these interventions using scarce NBCCEDP resources to increase baseline adherence in BSPAN. These studies need to be replicated in other populations as well, to improve overall longitudinal adherence rates, which could help to achieve mortality reduction through early detection of breast cancer.

R.G. Nair reports grants from CPRIT, NCI, and National Center for Advancing Translational Sciences (NCATS) during the conduct of the study. S.J.C. Lee reports grants from CPRIT, NCI, and NCATS during the conduct of the study. K.E. Argenbright reports grants from CPRIT during the conduct of the study. J.A. Tiro reports grants from CPRIT, NCI, and NIH during the conduct of the study. C.S. Skinner reports grants from CPRIT, NCI, and NCATS during the conduct of the study. No disclosures were reported by the other authors.

R.G. Nair: Conceptualization, formal analysis, supervision, methodology, writing–original draft, project administration, writing–review and editing. S.J.C. Lee: Conceptualization, resources, supervision, funding acquisition, methodology, writing–review and editing. E. Berry: Project administration, writing–review and editing. K.E. Argenbright: Funding acquisition, writing–review and editing. J.A. Tiro: Funding acquisition, writing–review and editing. C.S. Skinner: Conceptualization, resources, supervision, funding acquisition, visualization, methodology, writing–original draft, writing–review and editing.

The authors thank Hannah Fullington, Lorrie Burkhalter, and Caroline Mejias for their help with data extraction, as well as Melanie Carithers and Early Detection Program colleagues at Moncrief Cancer Institute and the many county partner organizations for their ongoing collaboration and dedication to patients. S.J.C. Lee and K.E. Argenbright received Cancer Prevention Research Institute of Texas (PP120097; PP150053; PP180018), C.S. Skinner and S.J.C. Lee received NCI (5P30CA142543-10), and S.J.C. Lee received National Center for Advancing Translational Sciences (CTSA UL1TR001105).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.