Pilot study of decision support tools on breast cancer chemoprevention for high-risk women and healthcare providers in the primary care setting

Women (N = 19,026) presented for screening mammography at Columbia University Medical Center (CUMC) in New York, NY between 2014 and 2016 (Fig. 1) [22]. Among them, 3743 (19.7%) were approached for enrollment into the Know Your Risk: Assessment at Screening (KYRAS) for breast cancer study, and 3077 (82.2%) consented and completed a baseline survey on demographics and breast cancer risk factors. Of the 3077, 511 (16.6%) women were identified as high-risk for breast cancer according to the Gail model. These high-risk women were informed of their breast cancer risk status and given a brochure to the CUMC breast clinic for a high-risk consultation. A subset of participants (N = 50) from this larger study was recruited into a single-arm pilot intervention study of the chemoprevention decision support tools. Eligibility criteria for the pilot included the following: 1) women, age 35–75 years; 2) 5-year invasive breast cancer risk ≥1.67% based upon the Gail model [3]; 3) having a primary care provider (PCP) at CUMC; 4) English or Spanish-speaking. Those with any history of breast cancer or with current or prior use of a selective estrogen receptor modulator (SERM) or aromatase inhibitor (AI) for breast cancer risk reduction were excluded. Having a PCP at CUMC was an inclusion criteria to ensure that both patient and provider would have access to the decision support tools and we could access the electronic health record at CUMC. Providers from primary care clinics at CUMC were invited to participate if they had an identified patient eligible for breast cancer chemoprevention. This study was approved by the institutional review board at CUMC and registered at clinicaltrials.​gov (NCT02954900).

RealRisks is a web-based patient-centered decision aid (DA) designed to increase the following: 1) accurate breast cancer risk perceptions; 2) breast cancer and chemoprevention knowledge; and 3) self-efficacy to engage in a collaborative dialogue about breast cancer risk and chemoprevention decisions (Fig. 2). We designed RealRisks by involving both patients and providers in multiple design sessions and usability studies to arrive at guiding principles that focused on the following: 1) options for providing information of breast cancer risk and chemoprevention, 2) “interactive games” to communicate breast cancer risk, and 3) patient preference elicitation to weigh the risks and benefits of SERMs and AIs (Fig. 2). An early prototype was evaluated in focus groups among women of various ethnicities from New York City. In this evaluation, accuracy of breast cancer risk perception (perceived minus actual breast cancer risk according to the Gail model [3]) significantly improved after interacting with RealRisks, even in the subgroup of women with low numeracy [23]. After the initial prototype was developed, we conducted usability studies with English- and Spanish-speaking women to determine how they navigated, engaged with and understood the information in RealRisks [21]. Using surveys, think-aloud protocols, and subject recordings, we identified several themes relating to the usability of RealRisks, specifically in the content, ease of use, and navigability of the application. By conducting studies in two languages with a diverse multi-ethnic population, we were able to implement interface changes to make RealRisks accessible to users with varying levels of health literacy and acculturation.

RealRisks includes the option to view the material in a text-heavy, or “information dense,” format or in an “information light” format that primarily uses pictures. The DA also has an audio option and Spanish translations. The educational modules include: 1) Breast cancer risk (breast cancer risk factors, calculation of personal breast cancer risk according to the Gail model, interactive games on risk communication); and 2) Chemoprevention (what is chemoprevention, risks and benefits of SERMs and AIs for chemoprevention, preference elicitation of chemoprevention). Through the RealRisks DA, we collect information on breast cancer risk factors to calculate a patient’s Gail risk score. We also collect factors that influence the decision making about chemoprevention through the preference elicitation game. RealRisks generates an action plan for patients summarizing their personalized breast cancer risk profile and preferences for chemoprevention. A provider view of the action plan, which is much shorter in length and designed to be actionable, is also made available to the provider through BNAV.

The provider-facing BNAV tool (Fig. 3) uses a two-pronged strategy to improve knowledge among healthcare providers about breast cancer risk assessment and chemoprevention. After patients complete RealRisks, their providers receive the tailored action plan via secure health messaging and are offered access the web-based BNAV toolbox. Based on the Theory of Planned Behavior [24], the toolbox contains a collection of information and resources that includes: 1) up-to-date guidelines and interactive educational presentations (attitudes); 2) video testimonials from experts in the field and a social component that enables a provider to compare his or her performance against aggregate, anonymous data of his or her peers (subjective norm); and 3) a repository of their patients’ breast cancer risk, together with the action plans generated by the patients’ interactions with RealRisks (perceived behavioral control). The BNAV chemoprevention module includes resources on breast cancer risk assessment, benefits and risks of chemoprevention, and how to manage the side effects of SERMs and AIs. Over time, other modules have been integrated into RealRisks (genetic testing for hereditary breast and ovarian cancer syndrome (HBOC), screening recommendations and guidelines, and risky health behaviors and lifestyle modification) and BNAV (genetic testing for HBOC, screening recommendations and guidelines, and patient-centered care), which allow providers to self-direct viewing resources outside of the clinical encounter. Each module takes about 10–20 min to view and can be completed during multiple sittings. To evaluate BNAV, individual interviews were conducted with 10 PCPs [25]. We found that few providers routinely used breast cancer risk calculators in their practice and they expressed concerns about the added burden of incorporating these tools into the clinic visit and being unfamiliar with chemoprevention.

High-risk women completed self-administered questionnaires at baseline, immediately after completing RealRisks, and 6 months after baseline or after the clinical encounter with their PCP. Health literacy, subjective and objective numeracy, and acculturation were assessed at baseline using brief validated measures of each construct [26, 27]. Breast cancer and chemoprevention knowledge was assessed using a 13-item scale, with adequate knowledge defined as at least 50% correct responses [28]. A participant’s confidence in communicating with her PCP was measured as an average of three items (range: 0–100%) [29]. Breast cancer risk perception was assessed by four items of absolute estimate, comparative risk assessed on a 3-point Likert scale, and numeric 5-year and lifetime risk on a scale of 0 to 100% [30]. Breast cancer worry was assessed using responses to two questions on a 7-point Likert scale [31, 32]. Self-efficacy was measured as an average of 5 items to assess perceived confidence in making a choice (range: 0–100%) [33]. Decision conflict was measured using the low literacy version of the Decision Conflict Scale (DCS) after exposure to RealRisks and at 6 months or after the PCP clinical encounter. The total score was calculated by averaging the 10 individual item scores so that higher scores indicated higher decisional conflict (range: 0–100) [33]. Scores lower than 25 on the scale are associated with implementing decisions whereas scores greater than 37.5 indicate high decision conflict, which is characterized by decision delay and/or uncertainty. Subscores of the scale that focus on factors contributing to uncertainty (e.g., feeling uncertain, uninformed, unclear about values, and unsupported in decision-making) were also reported on a range from 0 to 100 [33]. Behavioral intent for chemoprevention was assessed as previously described [14]. Referral to the high-risk breast clinic and chemoprevention uptake were assessed at 6 months via the electronic health record (EHR).

Our primary endpoint was accuracy of perceived breast cancer risk, defined as perceived lifetime risk within ±10% of actual lifetime risk according to the Gail model, at 6 months compared to baseline. The 10% range on either side of the Gail model risk estimate has been commonly used to provide a reasonable margin within which responses are labeled as accurate [34‐37].McNemar’s test was used to compare accurate breast cancer risk perception and adequate breast cancer and chemoprevention knowledge at baseline and follow-up. Paired t-tests were utilized to assess changes in continuous measures of differences in perceived and actual lifetime breast cancer risk, breast cancer and chemoprevention knowledge, breast cancer worry, self-efficacy, and decision conflict from baseline to post-RealRisks, from baseline to post-clinical encounter or 6 months, and from post-RealRisks to post-clinical encounter or 6 months. All analyses were conducted using SAS version 9.4 (Cary, NC) and a p-value < 0.05 was considered to be statistically significant.

Table 1 summarizes the baseline characteristics of the forty women who met all eligibility criteria and completed the baseline survey. The majority (95%) were postmenopausal women with a median age of 64.5 years (range, 49–72 years). Our sample was racially and ethnically diverse with 42.5% Hispanics, 40.0% non-Hispanic white, 12.5% non-Hispanic black and 5.0% Asian or other. The mean 5-year invasive breast cancer risk, based on the Gail model, was 2.38 ± 0.81%. Thirty-five percent reported having benign breast disease and half of them had a first-degree family history of breast cancer. Compared to non-Hispanic women, Hispanics were less educated and had lower acculturation, health literacy, and numeracy.

Table 2 summarizes the patient-reported outcomes at baseline, post-intervention (RealRisks) and at 6 months or after the clinical encounter with their primary care provider (PCP). At baseline, the majority of patients (75%) overestimated their perceived risk of breast cancer. Accurate breast cancer risk perceptions increased from 38% at baseline to 63% at 6 months (p = 0.03). The mean difference between perceived and actual lifetime breast cancer risk according to the Gail model decreased from 23.7% (SD 24.7) at baseline to 12.1% (SD 18.1) post-RealRisks (p = 0.01). After exposure to RealRisks, there was no significant change in breast cancer knowledge; however, mean chemoprevention knowledge scores significantly improved from baseline to follow-up, but this difference diminished over time (0.71 at baseline, 3.69 post-RealRisks, and 1.94 at 6 months, p < 0.01 for both comparisons). Breast cancer worry and self-efficacy in chemoprevention decision making did not change significantly. However, decision conflict increased from post-intervention to 6 months (17.92 vs. 43.44, p < .01). After exposure to RealRisks, 23 women (64%) experienced some or no decisional conflict with scores below 25 compared with 9 women (28%) at 6 months. Sub-scores for every decisional conflict domain were in the same direction, with significantly higher scores at 6 months compared to post-intervention. In terms of chemoprevention intention, 24% of participants expressed interest in taking chemoprevention, 47% were unsure, and 29% were not interested. Three (7.5%) women were referred for high-risk consultation and none had initiated chemoprevention at 6 months. There were no significant differences in patient-reported outcomes between Hispanic and non-Hispanic women, except that Hispanic women tended to have higher mean breast cancer worry scores (4.72 vs. 2.09 at baseline, 5.02 vs. 2.50 post-RealRisks, and 4.82 vs. 2.11 at 6 months, respectively).