Introduction

Breast Cancer Susceptibility Gene 1 (BRCA1) and BRCA2 mutations lead to an increased risk of breast and ovarian cancer. By age 70 years, a woman carrying a BRCA1/2 mutation has a lifetime breast cancer risk of 45–65% [1]. In comparison, women without BRCA1/2 mutations have a 12% lifetime risk. In the general population of the USA, around 1 in 400 carry a BRCA1/2 mutation. In the Ashkenazi Jewish population, the rate is 1 in 40 [2].

Genetic testing can identify women with BRCA1/2 mutations, allowing for early intervention to reduce the risk of breast or ovarian cancer [3]. In 2005, the United States Preventive Services Task Force (USPSTF) recommended women with a family history associated with increased risk for BRCA1/2 mutations should be referred for genetic counseling [4]. This statement was modified in 2013 and reaffirmed in 2019 to further recommend that primary care physicians conduct risk assessment with their female patients for personal, family, or ancestral (e.g., Ashkenazi Jewish or French Canadian) history of breast, ovarian, tubal, or peritoneal cancer for referral to genetic counseling for BRCA1/2 testing when appropriate [5, 6].

Few studies have evaluated BRCA1/2 testing referral rates. While it is estimated that between 5 and 25% of primary care patients meet criteria for referral [7,8,9], studies have shown that 80–90% of those women had no discussion with their primary care provider about genetic counseling and were not referred [7, 9, 10]. There was an increase in referral rates throughout the 2010 decade, with a notable jump in 2013 following Angelina Jolie’s editorial in the New York Times and the USPSTF updated recommendations [11, 12]. Research regarding variation in referral by patient factors is limited. In one study of women who were interviewed about genetic counseling utilization, those with graduate degrees and those who discussed their family history with their physician were most likely to report being referred [13]. Yet survey studies are subject to selection bias, potentially obscuring important differences in referral practices in the real world. Moreover, very little is known about individual physician referral rates, particularly among patients without a prior cancer diagnosis. To improve appropriate BRCA1/2 testing referral rates in accordance with the USPSTF recommendation, understanding how referral practices operate within primary care is critical.

The objective of our study was to describe referrals for BRCA1/2 testing in a large integrated health system and to assess patient and physician factors associated with referral provision.

Methods

This is a retrospective cohort study using data from the electronic health record (EHR) of the Cleveland Clinic Health System (CCHS). This study includes female patients between 18 and 50 years with a primary care visit to any CCHS Northeast Ohio Internal Medicine or Family Medicine physician between 2010 and 2019. Most of the assessment tools consider breast cancer at age less than 50 years to be more suspect of a genetic mutation, and therefore we included this age range for our cohort. We excluded patients with a breast cancer diagnosis prior to their genetic counseling referral.

This study was approved by Cleveland Clinic’s Institutional Review Board.

BRCA1/2 testing referrals

We identified whether study-eligible patients ever received a BRCA1/2 genetic testing referral by assessing the “consult to genetic counseling” order at each primary care visit and noting the ordering provider and the department. We defined referral for BRCA1/2 genetic testing as those that were referred to a certified cancer genetic counselor. Patients who attended a genetic counseling appointment for reasons unrelated to BRCA1/2 testing were considered not referred.

Patient and physician characteristics

We collected patient characteristics (age, race, marital status, and insurance type), family history of cancer (yes/no), genetic counseling referrals and appointments, and associated dates from the EHR. Because patients differed in the number of primary care appointments they had over the course of the study, we used each patient’s age and insurance type associated with their initial primary care appointment during the study period for analysis. For insurance type, we separated Cleveland Clinic from other commercial insurance because Cleveland Clinic has more generous coverage for genetic testing than many commercial plans. Public insurance includes Medicaid and Medicare.

Because of their higher-than-average rates of BRCA1/2 mutations, the USPSTF recommends that Ashkenazi Jewish heritage be taken into consideration when assessing risk. We collected religion from the EHR, which is self-reported under a patient’s social history, as an indicator of Ashkenazi Jewish ancestry. Approximately 50% of individuals had a documented religion. We assumed Jewish patients were Ashkenazi because most of Greater Cleveland’s Jewish population is Ashkenazi [14]. We categorized ancestry as Jewish or non-Jewish. Those that did not have a documented religion were categorized as non-Jewish.

Many women receive primary care from their OB/GYN physician [15]. To account for this, we assessed the number of visits patients had with Internal/Family Medicine relative to OB/GYN over the study period. If the number of Internal/Family Medicine appointments outnumbered the number of OB/GYN appointments by 10, we considered their primary provider to be from Internal/Family Medicine. If the patient had more than 10 OB/GYN appointments compared to Internal/Family Medicine appointments, we considered them managed by OB/GYN. If the number of OB/GYN and Internal/Family Medicine appointments were within 10 of each other, we considered them as co-managed.

High-risk status

The USPSTF recommendation for testing applies only to patients identified as high risk, primarily defined by family history of breast cancer. We defined high-risk patients as those with a documented family history of breast cancer in their medical record. Those without a documented family history of breast cancer were considered average risk. Yet because documentation of family history is inconsistent [16, 17], we also conducted the analysis on the average-risk cohort.

Statistical analyses

We described referrals for genetic counseling using chi-squared statistics or t-tests to assess differences across groups. We first used multivariable logistic regression to estimate differences in the odds of referral receipt on the population overall. We then used multivariable logistic regression to estimate differences in the odds of a patient being referred for genetic counseling by patient characteristics and physician specialty, stratified by patient risk status. We described the unadjusted rate of genetic counseling referrals by year among the high-risk cohort. For patients who were referred for BRCA1/2 testing, we described the referring physician’s specialty, as well as variation in individual physician referral rates. Analyses were conducted in R.

Results

Our study included 279,568 patients, 5% of whom met criteria for being high risk based on family history. Patient characteristics appear in Table 1. For the overall sample, the mean age was 34 years (standard deviation (SD): 9.6) and majority were white (77%). In terms of primary provider, 46% of patients had an Internal/Family Medicine physician, 16% used their OB/GYN, and 38% of patients were co-managed.

Table 1 Patient characteristics, overall and by risk level
Full size table

Among the high-risk patients (i.e., those who had a documented family history of breast cancer), 22% were referred for genetic counseling. High-risk patients differed from average-risk patients in that they were older (mean age 37 versus 34 years old), a greater proportion were white (80% high risk versus 77% average risk), and Jewish (2% versus 1%) (p < 0.001 for all comparisons). Eight-percent of patients who had an OB/GYN as their primary provider and 7% of patients who were co-managed were high risk, while only 3% of those managed solely by an Internal/Family Medicine physician were high risk (p < 0.001).

High-risk patients

Table 2 presents bivariate comparisons between patient characteristics and referral for genetic counseling by risk status. Twenty-two percent of high-risk patients were referred compared to 1% of average-risk patients. Among the high-risk patient cohort, 40% of Jewish patients were referred compared to 22% of non-Jewish patients. Twenty-four percent of high-risk patients managed by an OB/GYN, 22% of those managed by an Internal/Family Medicine physician, and 21% of those co-managed were referred (p = 0.03). Among high-risk patients, the referral rate steadily increased from 1.5% in 2010 to 5.9% in 2018, decreasing in 2019 to 3.6%. The largest increase occurred between 2012 and 2013 (1.7 to 3.2%).

Table 2 Association between patient factors and referral to genetic counseling, by risk level
Full size table

In the adjusted logistic regression models on the high-risk sample (Table 3), Jewish patients were more likely to be referred than non-Jewish patients (aOR 2.1, 95% CI 1.7, 2.7). Compared to white patients, black patients were significantly less likely to be referred (aOR 0.87; 95% CI 0.77, 0.98), as were not married/partnered patients compared to married/partnered patients (aOR 0.87; 95% CI 0.79, 0.94). Patients who used an OB/GYN as their primary provider were significantly more likely to be referred (aOR 1.5; 95% CI 1.3, 1.6) compared to those patients primarily managed in Internal/Family Medicine.

Table 3 Adjusted odds of being referred for genetic counseling based on patient characteristics
Full size table

Overall and average-risk patients

In the adjusted logistic regression model in the overall sample, family history was significantly associated with being referred for genetic testing (aOR 30.0; 95% CI 28.3, 31.8).

The adjusted logistic regression model for average-risk patients was comparable to that of the high-risk cohort (Table 3). Yet the association between being managed by an OB/GYN versus Internal/Family Medicine and screening receipt was stronger in this group (aOR 9.31; 95% CI 8.62, 10.1) compared to the high-risk patients.

Physician characteristics associated with referral practices

Of the 5317 patients who were referred for BRCA1/2 genetic counseling, 55 were self-referrals. Among the remaining referrals, 41% were made by an OB/GYN, 13% were made by an Internal Medicine physician, 13% by a Breast Center/Breast Surgery physician, 12% by a Family Medicine physician, and 20% from other departments. Out of a total 1789 Internal/Family Medicine physicians, only 422 (24%) ever referred a patient for BRCA1/2 testing. Of those physicians who referred, the mean number of referrals was 3.1 (SD: 5.2) and the median was 2. The number of referrals ranged from 1 to 54.

Discussion

In our study of BRCA1/2 genetic testing referrals among patients seen in primary care, less than a quarter of women at high risk of breast cancer were referred for BRCA1/2 testing in accordance with the USPSTF recommendation. Referral practices varied by patient demographic and clinical characteristics. Family history of breast cancer was the strongest predictor of BRCA1/2 referral. Compared to white patients, black patients were significantly less likely to receive a referral for BRCA1/2 testing. Jewish patients were more likely to be referred for testing compared to non-Jewish patients. Where women received most of their care also mattered—those who received most of their care via OB/GYN were significantly more likely to receive a referral compared to those managed primarily in Internal/Family Medicine. Referral rates for individual physicians were overall very low. Three-quarters of primary care physicians never referred a single patient for testing.

The low rate of BRCA1/2 testing referral among the average-risk population and the high-risk population in our study is consistent with the findings of others. One study using the National Health Interview Survey found while 24% met criteria for high risk of breast cancer, only 9% discussed genetic counseling with their primary provider and were referred [10]. An interview study found nearly 90% of high-risk women discussed their family history with their primary care provider. Yet only 2% of average-risk patients and 20% of high-risk patients were referred for genetic counseling [13]. This is similar to our findings of 1% of average-risk patients and 22% of high-risk patients. While the proportion of high-risk patients who were referred rose slightly over the 10 year period, in any given year this constituted a small minority. Family history of breast cancer was the strongest predictor of referral in our study. Other studies of BRCA1/2 testing referrals similarly found that family history was a strong predictor of referral receipt [13, 18]. Our study extends these findings by examining other patient characteristics that impact referral rates, including race, Ashkenazi Jewish ethnicity, and marital status.

Studies have pointed to physician knowledge deficits and patient unawareness of testing availability as obstacles to patients being referred and tested [18,19,20]. In one study, primary care physicians reported lack of education/training and associated discomfort as the main barriers to referring patients [21]. It is important that primary care physicians are educated about the USPSTF guidelines and the importance of referring high-risk patients. Multiple screening tools have been developed to aid physicians with this task [5, 22]. Furthermore, EHR-based calculators can also help physicians estimate a patient’s risk. The Cleveland Clinic EHR does not include a calculator for risk assessment. Physician engagement with EHR-based decision support tools is suboptimal in general, and little is known about use of breast cancer risk screening tools in routine practice. Understanding reasons for low rate of physician referrals for BRCA1/2 testing is necessary for designing interventions to improve appropriate referrals.

Identifying high-risk patients may not be the only barrier. Physicians may be unaware of the current landscape surrounding BRCA1/2 testing. Previous impediments to testing such as cost, concerns about discrimination, and failure of the test to change management, have changed drastically in the past twenty years. Under the Affordable Care Act, BRCA1/2 testing is considered a covered preventive service for those who meet eligibly per the USPSTF recommendations [23], and the Genetic Information Nondiscrimination Act (GINA) prohibits discrimination by health insurance plans based on genetic information [24]. Importantly, patients frequently act based on testing results. If positive, women can elect for bilateral prophylactic mastectomy or increased surveillance, both of which greatly decrease mortality from breast cancer [3]. Further research is warranted on both physician- and patient-level barriers that may contribute to the low rate of referrals.

To our knowledge, ours is one of the first studies to account for Ashkenazi Jewish ancestry using data from the EHR to assess variation in BRCA1/2 testing referral rates. The USPSTF recommends that Ashkenazi Jewish ancestry be considered when assessing a woman’s risk of a BRCA1/2 mutation. One study found 40% of Ashkenazi Jewish BRCA1/2 carriers had no family history of breast or ovarian cancer, while another found 56% had no family history [25, 26]. Even in the absence of family history, Ashkenazi Jewish heritage alone likely warrants a genetic counseling referral [27]. We found Jewish women were significantly more likely to be referred than non-Jewish women. While this may indicate that physicians use ancestry as a factor in referral decisions, it is also possible that because Ashkenazi Jewish women are aware of their higher risk, they request referrals for screening. While they were referred at higher rates than non-Jewish women, overall, only 8% of Ashkenazi Jewish women received referrals. Perhaps primary care providers are unfamiliar with the elevated risk of BRCA1/2 in this group or that being Jewish is an ethnicity in addition to a religion. It also may be that physicians do not know how to broach the subject of religion or ethnicity with patients [28, 29]. It is likely that physicians need more support in having these conversations. More comprehensive documentation of ancestry in the EHR may be helpful.

We found black women were significantly less likely to be referred to genetic counseling than white women, even accounting for family history of breast cancer. There is mixed evidence regarding the association between patient race and BRCA1/2 testing referral practices [10, 13, 18, 19]. Studies have found lower referral and counseling rates among black women due to lack of awareness, cost-related factors, distrust, and family concerns [18, 19]. Physician-level or institutional-level factors may also play a role, such as implicit bias or structural racism. Several studies found that family history is less complete among black patients, either because of less awareness of family history or because physicians are less likely to ask about it [30,31,32]. As we were unable to account for the content of the primary care encounters, we do not know if a referral for genetic counseling was offered to black patients at a lower rate than white patients, or if black patients were more likely to decline referrals. However, a population-based study of white and black women in Florida showed that black women were 16 times less likely to discuss genetic testing with their provider [33]. There is limited data on rates of BRCA1/2 mutations among black women. However, it is known that black women do not have a lower risk than other races [34,35,36], further underscoring the need for expanded use of appropriate testing in this population.

Patients who used an OB/GYN as their primary provider were more likely to be high risk, as well as referred for BRCA1/2 testing than those who were managed solely in Internal/Family Medicine. This likely reflects a difference in the salience of breast cancer risk across specialties. OB/GYNs are perhaps more likely to ask about family history of breast cancer than Internal/Family Medicine physicians given their expertise in this area. Furthermore, a vignette-based study found OB/GYNs were more likely to report adherence to guidelines regarding genetic testing for BRCA1/2 than Family Medicine physicians [37]. While we found both average-risk patients and high-risk patients were significantly more likely to be referred if they were primarily managed by an OB/GYN, this difference was pronounced for average-risk women. This finding suggests that either some of the average-risk patients are actually high risk or that OB/GYNs are more aggressive than Internal/Family Medicine physicians in referring to genetic counseling, even for women who do not fit traditional criteria for screening. Concurrent use of OB/GYN and Internal/Family Medicine for primary care among women of childbearing age is common. However, there in an inherent ambiguity among the providers as to who is responsible for different preventive care services. Understanding how OB/GYN and Internal/Family Medicine physicians either consciously or subconsciously allocate responsibility for breast care is needed to ensure high-risk women who are not regularly seeing OB/GYN physicians receive appropriate BRCA1/2 screening referrals.

Our study has several limitations. We gleaned personal and family history of breast cancer from the EHR. Consistent with other family history measures in the EHR [16, 17], this measure likely undercounted the true number of women with a family history. Furthermore, family history documentation is inconsistent, with some providers giving more details than others (e.g., relationship to patient, age at onset). We included all cases in this study, even though perhaps some are in distant relatives or were diagnosed at an old age. Similarly, religion is variably documented in the EHR, and we used religion as a marker for ethnicity; not all secular Jews would report “Jewish” as their religion. Therefore, the true number of Jewish women has been undercounted, making the true referral rate in this group even lower than we documented in this study. Additionally, our definition of primary management was unable to account for the fact some women may have had numerous OB/GYN visits due to pre- or post-natal care. Yet OB/GYNs can and do refer patients to genetic counseling during such visits. These data come from a single health system in Northeast Ohio and may not be generalizable to other settings. That said, according to family history alone, we found approximately 5% of patients were high risk and therefore eligible for referral. This rate is consistent with two other studies [8, 9], suggesting our study may have appropriate EHR documentation and be representative of the wider population of primary care patients. Finally, in our bivariate analyses, we found statistically significant differences in referral rates by patient factors which are likely the result of our very large sample size, and are not clinically significant. Our results should be interpreted with that in mind.

Conclusion

Given the high likelihood of a breast or ovarian cancer diagnosis among women with BRCA1/2 mutations, genetic testing is among high-risk women is critical. The USPSTF has identified primary care physicians as the entry point for testing referrals when appropriate. Yet in our study in a major health system, referral rates from primary care physicians were extremely low, even among patients with a documented family history of breast cancer, and most physicians never referred a single patient. Physician-focused interventions are needed to ensure all women likely to benefit from testing are appropriately identified and referred.