Abstract
BACKGROUND The average lifetime risk of breast cancer for an American woman is 12.5%, but individual risks vary significantly. Risk modeling is a standard of care for breast cancer screening and prevention with recommended tools to stratify individual risks based on age, family history, breast density, and a host of other known risk factors. Because of a lack of resources rurally, we have not consistently met this standard of care within all of North Carolina.
METHODS We implemented a quality improvement project to assess the risk for breast cancer by gathering data on community risks. We implemented an evidence-based tool (Tyrer-Cuzick) for quantifying risk within a mostly rural population of Eastern North Carolina and developed customized services for women meeting elevated-risk definition. These services included additional imaging for elevated-risk women and a risk-reduction program. We also assessed genetic risks for hereditary breast and ovarian cancer in our at-risk population using National Comprehensive Cancer Network (NCCN) guidelines based on family history and added local genetics extenders to help test more women. We analyzed data regularly using Plan-Do-Study-Act methods to improve outcomes over 1 year.
RESULTS We screened a population of 4500 women at a community hospital over a 1-year period for their individual lifetime cancer risk and genetic risk. Breast cancer risk was quantitated at the time of mammography, and women were stratified into 3 groups for risk management. Within our screening population, 6.3% of women were at high risk (defined by a lifetime breast cancer risk greater than or equal to 20%) and another 8.1% were above-average risk (defined by a lifetime breast cancer risk of 15%–20%). These women (14.4%) could potentially benefit from additional risk-management strategies. Additionally, 20% of all unaffected women within a typical screening population of Eastern North Carolina met NCCN guidelines for hereditary breast cancer and ovarian cancer testing independent of their cancer risk score. Using a model of targeted intervention within a population with elevated risks can be helpful in improving outcomes.
LIMITATIONS This population within Eastern North Carolina is mostly rural and represents a potentially biased population, as it involves older women undergoing annual mammography. It may not be broadly applicable to the entire population based on age, geography, and other risks.
CONCLUSIONS This model for improving cancer risk assessment and testing at a small community hospital in Eastern North Carolina was successful and addressed a community need. We discovered a high rate of increased-risk women who can benefit from individualized risk management, and a higher percentage of women who potentially benefit from genetic testing. These higher cumulative risks may in part explain some of the disparities seen for breast-cancer-specific outcomes in some parts of the state.
Outer Banks Hospital is a small community hospital within a part of North Carolina with historically worse cancer outcomes from risks we hoped to identify and mitigate. Overall cancer mortality rates are historically 16% higher in the 29-county regiona of Eastern North Carolina than the rest of the state, and breast cancer mortality rates are 19% higher in Eastern North Carolina than the rest of the state’s 71 counties [1]. North Carolina is also historically above the US average for breast cancer mortality rates, and so this represents an opportunity for improvement [2].
We are data-driven cancer providers and speculate that some of the disparity may be due to a clustering of risks geographically that we can quantify and change. Some risks are modifiable (e.g., obesity, physical activity, diet), while many are not (e.g., family history, gender, age). Many of these risks can be modeled using population-validated tools, and we sought to examine these risks within Eastern North Carolina where we postulated these risks accumulate more frequently. Using risk modeling that quantitates individual cancer risks to stratify patient risk within our population means we can offer targeted interventions to detect cancer at earlier stages, lower the overall risk, and prevent cancer(s). This may also help address some of the disparity in cancer outcomes noted geographically.
We could not access population data that quantitatively examines cumulative risks for breast cancer within Eastern North Carolina, so we began our quality initiative by first conducting a review of patients regionally for individual risks. A 4-year retrospective review was done to establish a baseline on 165 local breast cancer patients within Eastern North Carolina [3]. We analyzed patients for established risks, which can then help us choose a model to predict a person’s individual risk for cancer. That retrospective review highlighted several significantly elevated risk factors locally that served as a baseline for this quality improvement initiative. Pareto data analysis revealed the following in rank order as the most common risk factors: female gender, body mass index (BMI) elevation, advanced age, positive family history, increased breast density, and previous breast biopsies [3].
Most breast cancers were in women (99%) as expected, and that was not an area for intervention. Elevated body mass index (BMI greater than or equal to 25) was present in 80% of patients, with obesity present in 40% of breast cancer patients [3]. Obesity (BMI greater than or equal to 30) is causally linked to cancer and adverse outcomes in both pre-menopausal women and post-menopausal women, and this may represent an area for strategic intervention regionally [4]. Familial cancer histories are also common within our affected patient population and likewise increase cancer risk. Literature reports indicate less than 25% of women with breast cancer have breast cancer in other family members [5], and we found more than 52% of women affected with breast cancer in our region had familial histories of breast cancer. Having an affected first-degree relative statistically increases the risk of breast cancer 2-fold [5], and we noted 33% of patients with breast cancer reported affected first-degree relatives regionally [3], compared to a rate of 13%–16% nationally [5]. Based on this baseline data, we were awarded a partnership grant with Pfizer and the Association of Community Cancer Centers to explore genetic testing in our community. Age is another risk, and the median age in Eastern North Carolina is 10 years older than the state median [6]. Breast density in our affected population was similar to other reports where 40% of women have the highest 2 categories of density on mammograms [7]. Other examined factors were not statistically different in our review other than previous biopsies of the breast reported in 30% of screened women [3].
Anticipating that similar risks accumulate within the unaffected population, we sought ways to quantitate these overall risks and offer evidence-based interventions in our at-risk population before a diagnosis of cancer [8–10]. By identifying individual factors that can elevate risk, we can intervene with models to improve outcomes by adding risk-adapted screening, testing for genetic risks to cancer susceptibility, and introducing risk-mitigation strategies, including prevention. Examples of risk-adapted screening include breast magnetic resonance imaging for increased-risk women and whole-breast ultrasound imaging in addition to conventional mammography [9]. Examples of risk-mitigation include chemoprevention with Tamoxifen (and Raloxifene), preventive surgery (i.e., mastectomy and oophorectomy in pathogenic mutation carriers), and lifestyle management (optimizing BMI, improved diet, increased activity, reduced alcohol and tobacco) [10]. While this concept is not novel, it is often lagging in small communities where resources are lacking and where it is needed to help improve some of the observed regional disparity in cancer outcomes.
Our specific aim was to review existing evidence-based strategies and to implement a quality improvement project to assess and address small community risks where roughly 1 in 5 (20%) Americans live [11], and where access to quality care is limited. More specifically, our aim was to pilot a risk assessment tool within Eastern North Carolina, where it could have a more meaningful impact based on suspected higher cumulative risks. Furthermore, we hoped to be able to manage this identified risk with targeted additions to breast care pathways regionally.
Methods
We recruited a breast care team to examine an identified gap in care with the specific aim of integrating risk modeling for breast cancer and finding opportunities for intervention. This team included surgeons, radiology personnel, medical oncology, genetics, radiation oncology, a nurse, and an integrative medicine/lifestyle medicine physician. We strategically reviewed baseline data and our current process for individual risk assessment using Strengths, Weaknesses, Opportunities, and Threats (SWOT) Analysis (Figure 1) [12].
SWOT (Strengths, Weaknesses, Opportunities, Threats) Analysis of Current Screening Process
Note. No current process for individualized risk assessment was the greatest weakness, and simultaneously represented a great opportunity.
There was no current risk assessment model within our hospital or our regional network using recommended guidelines for quantitating breast cancer risk or customized risk management [9, 10]. Rather, we discovered our providers mostly follow general age-based guidelines for screenings that are neither risk-adapted nor modified based on true individualized risks. Furthermore, genetic testing, which can further help identify at-risk women, was sparsely offered regionally when we began this program with affected patients and non-existent for unaffected patients. We therefore created an improvement plan to implement breast cancer risk assessment at a small community hospital and created the workflow to address the identified risks following risk stratification. We formulated a quality improvement process after consulting with a certified breast cancer surgeon already doing this in Virginia and used repeating Plan-Do-Study-Act (PDSA) cycles to improve upon the process over 1 year.
To assess the risk in the largest possible at-risk group, we administered a questionnaire in radiology at the time of mammography, which we previously demonstrated represents an ideal navigation point for coordinating breast care [13]. Based on our population of overweight, older women with dense breasts and high familial cancer rates, we chose a risk assessment tool that modeled these relevant factors in a population. We chose the Tyrer-Cuzick risk assessment tool because it satisfied this requirement [14]. This evidence-based risk model is 1 of several currently recommended by NCCN [9]. We then designed a questionnaire that includes the relevant questions to calculate the individual’s risk: age, weight, height, age at menarche, history of pregnancy and age, menopause status, use of hormone replacement therapy, known BRCA gene mutations, history of ovarian cancer, previous breast biopsy, family history (including first-degree and second-degree relatives) and family heritage (Ashkenazi), as well as breast-density corrections. Our questionnaire asked these data metrics of individual patients and was given to all mammography patients at intake. From our baseline data, we know many of these risks occurred within our affected population and we needed an inclusive model of these local risks. Tyrer-Cuzick risk scores were calculated on the day of imaging using an online calculator that requires the above metrics [14], including estimated absolute lifetime risk of breast cancer (as a percentage) and relative risks that are based on an age-adjusted population risk.
We stratified women into 3 risk groups based on their calculated lifetime breast cancer risk: high risk (≥ 20%), intermediate risk (≥ 15% and < 20%), and low-normal risk (<15%). We notified all patients deemed high risk by sending a letter and attempting 2 phone calls to arrange consultation to review their lifetime risk and their genetic risk (if criteria for testing were also met).
One physician and a nurse with additional training as genetic extenders performed all risk assessments [15], referring appropriate patients to a newly created high-risk clinic for further imaging annually and risk-reduction discussions with a breast surgeon. We additionally referred any interested patients to a risk-reduction clinic designed specifically to discuss lifestyle risk modifiers. This outpatient clinic is part of our Center for Wellness and Healthy Living at the Outer Banks Hospital and is run by a physician who is board certified in lifestyle medicine. Modifiable risks include diet, physical activity, weight loss and BMI improvement, alcohol reduction, tobacco cessation, stress reduction, and other measures to mitigate risks. We have previously reported on a pilot model that resulted in women with breast cancer successfully reducing their modifiable risks [16]. We additionally referred all women with very high risk (defined as greater than or equal to 30% lifetime breast cancer risk) to medical oncology to discuss chemoprevention, and any woman with a greater than 40% lifetime risk was referred for risk-reducing surgery discussions based on national guidelines. Genetic clinic testing involved pre-and post-testing counseling and local testing with salivary samples, and included a panel of genes associated with hereditary breast and ovarian cancer following NCCN guidelines. This project was approved in July 2019 through an institutional review board at East Carolina University (UMCIRB 19-001052).
For the study period of 1 year, we tracked all patients meeting high-risk criteria, patients agreeable to further consultation, and patients potentially impacted by further screening and risk mitigation. Separately, we quantitatively analyzed all women in the screening population for family histories of breast cancer and for the percentage qualifying for genetic testing using evidence-based guidelines. We used current NCCN guidelines to decide on eligibility for genetic testing and contacted these women similarly [8]. Statistical methods used to analyze results included chi-square analysis, where appropriate.
We used Plan-Do-Study-Act (PDSA) methods cyclically throughout the year to continuously assess our data and optimize outcomes. PDSA cycles are a way to test changes once implemented and are useful in quality initiative projects [17]. Using these 4 steps helps break down the plans for improvement and then evaluate the outcome, improving on it and testing it again repeatedly.
Results
We analyzed 4500 women at time of mammography over 1 year, with 99.9% completing questionnaires to quantitate lifetime risk for breast cancer (N = 4497). Family histories and personal histories were also analyzed. Patients with previous breast cancer (n = 303) accounted for 6.7% of population and we excluded them from Tyrer-Cuzick risk analysis (since they already have cancer), and we excluded patients too old for the model to calculate their risk (n = 33, aged 85 or older), as risk screening is not recommended in this population. All remaining unaffected patients (n = 4160) appropriate for screening underwent individual quantitative risk assessment and risk stratification using Tyrer-Cuzick model v8 (Figure 2), which was later updated to v8b during this study period. High-risk patients (n = 262) overall accounted for 6.3% of our screened unaffected population, and these women were offered further targeted screening with magnetic resonance imaging (MRI) of the breast, or whole-breast ultrasound, as well as discussions about risk reduction and genetics testing. Intermediate-risk patients (n = 339) accounted for 8.1% of our unaffected screening population, and some of these women were additionally offered risk reduction (e.g., women who had elevated BMI) and additional imaging (e.g., whole-breast ultrasound annually if very dense breasts noted on mammography). Low-normal risk women accounted for 85.6% of the total, and no changes were made in their annual screening. Collectively, 14.4% of women were above-average risk in their estimated lifetime risk for breast cancer.
PDSA (Plan Do Study Act) Flow Chart Summarizes the Program Outcomes with Lifetime Risk Stratification for Breast Cancer
In addition to the overall lifetime breast cancer risk assessment, we also assessed the genetic risk for hereditary breast and ovarian cancer (HBOC) in this same population of women. After the first month of our PDSA cycles, we noted that many women had familial risks of cancers that necessitated genetic risk assessment, and we quickly learned that the Tyrer-Cuzick model does not consider the genetic risk fully. We therefore used current NCCN guidelines for genetic risk assessment separately and independently of the lifetime risk of breast cancer. Of 4500 women screened for genetic risk, 303 women had personal histories of breast cancer (affected), and 4197 women were unaffected. Twenty percent of unaffected women (840 out of 4197) met NCCN criteria for testing based on family history, and 55% (166 out of 303) of affected women met NCCN guidelines for genetic testing (Figure 3). To date, we have offered genetic testing to 322 women in total, which represents 32% of eligible women (322 out of 1006). Our pathogenic mutation rate in the unaffected population is 4% currently, and we follow these women in a high-risk clinic and risk-reduction clinic.
PDSA (Plan Do Study Act) Flow Chart for Separate Genetics Risk Assessment
Of the high-risk population agreeable to additional evaluation and follow-up (n = 51 out of 262), all have received additional targeted imaging as part of their annual risk evaluation. This was most commonly MRI of the breasts (n = 49 out of 51, 96%) or whole-breast ultrasound (n = 2 out of 51, 4%). Women who were very high risk (greater than 30% lifetime risk, n = 17) were seen by medical oncology as well for chemoprevention. Twenty-seven women were seen additionally in the risk-reduction clinic for lifestyle modification. Many women with borderline (20%–30% lifetime) high risks did not respond to our efforts to discuss risk management, and our hope with this pilot year of data was to capture these women on subsequent imaging visits. The various PDSA cycles and lessons learned along the way are included with Table 1.
PDSA Cycles and Lessons Learned Over 1 Year
Discussion
Risk assessment is useful for determining who needs additional screening efforts, who benefits from risk reduction, and who qualifies for genetic testing as part of cancer prevention and treatment models [8–10]. We observed high cumulative risks regionally, which we anticipated could be addressed using evidence-based approaches. We had no preexisting services for quantitating risk assessment rurally, and neither counseling nor additional testing locally prior to this project.
Following the quality intervention, we were able to quantitatively assess the breast cancer risk within our population and offer risk-adapted screening and risk reduction through a high-risk clinic, a risk-reduction clinic, and a genetics clinic. After 12 months, 6.3% of unaffected patients met high-risk criteria (lifetime risk greater than 20%), and 8.1% were intermediate risk (lifetime risk greater than 15%). More surprisingly, 22% of all screened women, and 20% of unaffected women, consistently meet HBOC criteria for genetics assessment. Knowing this information can dramatically alter a woman’s risk and is equally important in the risk assessment process.
The strengths of this project included our ability to implement a program given limited resources, and the ability to gather quantitative data that we believe shows risks are inherently higher in Eastern North Carolina. Being a small cancer program also helped, as it took only a few providers to make happen. Another strength of this program was the use of a genetic nurse extender trained locally through City of Hope to be local counselor, navigator, and tester. This model can easily be duplicated by any hospital with an interest in providing genetics services locally. A final strength was anticipating barriers, particularly in higher-risk women, where we adapted a pre-certification process for additional imaging (MRI), which helped with compliance in women agreeable to participation in the program. Additionally, we found no major concerns with billing for genetic testing, with 85% of women paying no out-of-pocket expenses.
A larger number of women than we expected reported familial cancers, with 22% of all women meeting criteria for genetic testing. Much of this was due to close relatives with breast cancer, as expected, but we were surprised by the numbers of ovarian and pancreatic cancers we had not considered. Having a first- or second-degree relative with either of these is an automatic qualifier for genetic testing. After 1 year, 32% of eligible unaffected patients are now germ-line tested for their hereditary risk, and we currently test 100% of affected breast cancer patients for HBOC. Many at-risk patients who met guidelines for additional screening did not respond to our initial attempts, and we changed our process to capture them in a recent PDSA cycle. Of unaffected patients tested, 4% had identified pathogenic mutations—half BRCA and half non-BRCA—which was similar in magnitude to results in our affected patients locally (9%). However, we have only tested one-third of eligible patients to date, so we cannot conclude whether this number will change significantly given more testing compliance.
We did not anticipate the potential impact locally from this project. We expected less than 5% of a reference population would be at increased risk based on known population risk factors, and we were ready for the observed 6.3% incidence for the high-risk group (defined as absolute lifetime risk greater than 20%), which is still higher than reported elsewhere, supporting our hypothesis. We did not anticipate an additional 8.1% who are above-average risk and carry risks that should also be addressed. Collectively, almost 15% of our population had not been receiving appropriate risk-based screening. What we also clearly had not anticipated was the familial factor that resulted in 1:5 women being eligible for genetic testing independent of their lifetime breast cancer risk. We did not have the personnel to call, counsel, and test an additional 840 unaffected women in the first year, and therefore we have included an innovative solution in our new PDSA cycle, piloting a triaging system using software with vendor-mediated counselors to perform much of this future work. The one true weakness in the study was therefore reaching more eligible patients and thus compliance, which was the result of our initial weak efforts to contact patients retrospectively after their mammogram, and our limited resources for the higher-than-expected numbers. Current models exploring patients completing questionnaires using online tablets at the time of screening offer a potential solution to this problem by automating results and referrals for both lifetime breast cancer risk assessment and genetic risk assessment.
Little comparative data exists regarding measures of quantitated breast cancer risk in the at-risk population. A large population study over a 19-year period revealed 1.9% of its population was deemed high risk using Tyrer-Cuzick risk modeling that did not include breast density, and 3.5% were considered high risk when breast density was included [18]. We have almost double that percentage in our community. Other reports note higher scores using the Tyrer-Cuzick (5.6% in another study) model than BRCAPRO (1%) or Claus models (0.9%), which also may account for our higher calculated rates [19]. Our observed high rates may also be attributed to an older population (median age in Eastern North Carolina is 48.6 versus 38.4 years in the United States and in the rest of North Carolina) and other collective risks, including obesity and familial history, again underscoring the need for this type of program within our region [6].
Using a PDSA framework for quality improvement, we were able to greatly impact our breast care practice management. Each organization has its own challenges with implementation of quality initiatives, and using the feedback cycles that are part of this PDSA approach to repeatedly analyze results and outcomes allowed us to achieve our desired outcomes. Prior to this project, we were like many community cancer programs rurally, where screening is simply age-based and not risk-based. With the addition of this process and the continued improvements through the various PDSA cycles, we were able to create a lasting model of risk assessment and risk mitigation for breast care within our rural population. We anticipate this model will help remove disparity in breast care over time.
Limitations of this study include its selected nature, which may introduce bias. Although our community is a typical screening population for breast cancer, it is hard to generalize the results to the entire at-risk screening population. Women who undergo annual screening are typically older, and more apt to consider screening based on existing familial risks. The data would be more representative of a true population risk if we expanded this model in the primary care setting. Additionally, compliance was clearly an issue. We found many women were not interested enough in the risks weeks later when they were contacted by our nurse to merit further discussions formally, and so we have changed the process to include sharing this information with patients in the body of the mammogram report on the day of the interpretation. This information is then available to both the ordering provider and the patient, which should improve future compliance. Lastly, it would be helpful to use more than 1 risk model, as population differences can occur with these models, and the software tools that we mention here include multiple risk models, including Gail, Tyrer-Cuzick, and others.
To be sustainable, this model benefits from a system-wide approach with a team of providers and an automated process whereby intake is prospective so results are available to the patient before leaving radiology. The next steps include integrating this into a care pathway in our electronic medical record to obtain better compliance and potentially assess a younger demographic of women not presenting for mammography through primary care, and trialing this working model at other hospitals in network within Eastern North Carolina. We also plan to analyze long-term results of intervention concerning risk reduction and cancer prevention, which will require more years of study. Currently, this model continues after 2.5 years of piloting, and we hope this improvement in breast care will ultimately affect the disparity in breast mortality rates.
Acknowledgments
This quality improvement project was initially supported by a grant from Pfizer and the Association for Community Cancer Centers based on the baseline data for our population referenced in the introduction. We thank Theresa Smith, RN, for her contributions to the preparation of this manuscript. We thank Caroline Dixon, premedical student, for helping with the 4-year data review on affected patients that served as baseline information locally.
Financial support. This study was supported by an IRB-approved grant from ECU, with Pfizer, Inc., and the Association of Community Cancer Centers jointly supporting this project.
Disclosure of interests: All authors report no relevant disclosures.
Footnotes
↵a Beaufort, Bertie, Camden, Carteret, Chowan, Craven, Currituck, Dare, Duplin, Edgecombe, Gates, Greene, Halifax, Hertford, Hyde, Jones, Lenoir, Martin, Nash, Northampton, Onslow, Pamlico, Pasquotank, Perquimans, Pitt, Tyrrell, Washington, Wayne, and Wilson counties
- ©2022 by the North Carolina Institute of Medicine and The Duke Endowment. All rights reserved.
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.