Decision aid development
BSD was created using the Ottawa Decision Support Framework [9] and evidence-based recommendations for communication of health-related risk information [10]. BSD content was developed with input from breast oncologists and radiologists, internists and gynecologists. Prior to using BSD in this study, usability testing was conducted with clinicians and volunteers in the target age group, and the site was modified per their feedback. BSD meets fully 26 of 28 International Patient Decision Aid Standards (IPDAS) Content Criteria and 17 of 20 relevant IPDAS Development Process criteria (see Additional file 1: Table S1) [11]. Five of six criteria for internet-based decision aids were fully met and one was partially met.
The landing page of BSD frames the two decisions the tool is designed to address: when to start (age 40 or age 50) and how often to have (every year or every other year) routine screening mammograms. BSD users next complete a breast cancer risk assessment based on the Gail model [12, 13]. In addition to eliciting the Gail model risk factor information, BSD identifies women at increased risk due to: 1) a first-degree relative diagnosed with breast cancer before age 50 or ovarian cancer at any age; 2) a prior diagnosis of ductal carcinoma in situ or lobular carcinoma in situ; 3) a prior diagnosis of atypical ductal hyperplasia; 4) a known BRCA1/2 mutation; or 5) a history of chest radiation. Women who report any high-risk criteria and women with a predicted five-year breast cancer risk ≥1.66% – consistent with accepted definitions of high risk [14] – may not proceed with the decision aid. Instead, they are told that based on the information they provided, they may be at greater-than-average risk of developing breast cancer, and that for women like them, routine mammography and possibly other screening tests may be recommended. They are encouraged to speak with their doctors about their breast cancer risk and screening options.
For women who report no high-risk criteria, BSD displays their personal predicted risk of developing breast cancer in the next 5 years. Risk information is presented using female icon arrays, percentages and numeric frequencies with explanatory text (“Of 1,000 women like you, X will develop breast cancer in the next 5 years”).
In subsequent pages, BSD provides information about the accuracy of screening mammograms and the benefits and harms of routine screening. Mammogram accuracy is expressed with icon arrays and frequencies to describe test results and true cancer status in 1,000 women in their 40s who have a screening mammogram, based on published estimates from the Breast Cancer Surveillance Consortium [1, 15].
Quantitatively, benefit is described as reduction in the number of deaths from breast cancer associated with the four different screening schedules reflecting different starting ages (40 vs. 50) and screening intervals (every year vs. every other year). This information, based on published estimates from the Cancer Intervention and Surveillance Modeling Network [16], is displayed using female icon arrays and frequencies with explanatory text. Differences in breast cancer death rates are estimated as a function of the relative risk reduction associated with each screening schedule and the underlying breast cancer mortality risk for women ages 40–49 [1].
The potential harms of routine screening described in BSD include diagnostic work-ups associated with false-positive results; overdiagnosis and unnecessary treatment; delayed cancer diagnosis and false reassurance associated with false-negative results; and cumulative radiation exposure associated with repeat screening, which is described as “extremely low.” False-positive and false-negative rates are given quantitatively. Other harms are described qualitatively and in plain language, with links to additional information about these outcomes.
Following the information about benefits and harms of screening, BSD users can explore their attitudes toward breast cancer, screening and involvement in health care decisions in a values-clarification exercise. Users are asked to consider a series of 10 statements and indicate their level of agreement or disagreement with each statement. Finally, BSD provides a one-page summary of the session, with options to save and print the summary document, which users are encouraged to share with their clinician.
A public-use version of BSD is available at www.breastscreeningdecisions.com.
Decision aid evaluation
Subjects and recruitment
BSD was evaluated in a prospective, single-arm trial at a large, urban, academic medical center between March 2013 and April 2014. The study was approved by the Institutional Review Board at Weill Cornell Medical College.
We identified women ages 40–49 with a scheduled appointment for routine preventive care at one of three participating primary care and gynecology practices. Electronic health records (EHRs) were screened to exclude women with a personal history of breast cancer and women at increased risk of breast cancer due to one of the five aforementioned high-risk criteria. All other women were invited to participate and to use BSD at their convenience before their scheduled visit. Invitations, mailed and emailed 4–6 weeks before the scheduled visit, included a personalized letter signed by the woman’s physician with instructions for accessing the BSD website and a unique username and password. On-site access to BSD at each participating clinic was offered to women who did not have internet access elsewhere.
Informed consent was obtained immediately after logging into the BSD website. Women could log in more than once and resume or repeat a session any time prior to their scheduled visit. Women who logged in, gave consent, met eligibility criteria and used the decision aid (BSD users) received follow-up as described below. There was no active follow-up with women who were invited to use BSD but did not log in and give consent prior to their scheduled visit (non-users), although information in the EHR about their age and mammography history was collected for eligibility screening and comparison with BSD users.
Assessments
At the conclusion of the BSD session, users completed an online survey about their impressions of the site, with seven statements about ease of navigation, clarity and importance of information, usefulness of graphs and figures, time burden, and whether they would recommend the site to other women. Responses to each statement were given on a 5-point scale from “strongly agree” to “strongly disagree.”
One month after each user’s scheduled preventive care visit, she was contacted by telephone and asked to complete a follow-up survey by phone or online. In order to identify potential unintended impacts of the decision aid on behavior, actual or intended use of screening mammography was ascertained by asking women whether they had a screening mammogram since their visit, scheduled an appointment for a screening mammogram, or if they were planning to schedule an appointment within the next six months. Women who responded “no” to each of these successive questions were considered to have no plan for a screening mammogram within the next 6 months, unless they explicitly stated that they were unsure about screening mammography in response to the final question. Decisional conflict was assessed using O'Connor’s Decisional Conflict Scale (DCS), a 16-item instrument developed to evaluate health care decision support interventions [17]. The DCS has high reliability, and DCS scores below 25 have been associated with implementing decisions. Breast cancer worry was assessed using items adapted from McCaul’s Breast Cancer Worry Scale (BCWS) [18]. Subjects were asked how often they worry about breast cancer, how often worry about breast cancer affects their mood, and how often worry about breast cancer affects their performance of daily activities, with responses given on a 5-point scale from “not at all” to “almost all the time”. Knowledge and beliefs about breast cancer and screening mammography were assessed using individual items developed for this study or selected from existing instruments [19]. Information about each subject’s age, race, marital status, number of children, type of health insurance and prior use of mammography – for screening or diagnostic purposes – was obtained from the EHR.
Clinicians were asked to complete a brief paper survey at the conclusion of each woman’s scheduled preventive care visit. They were asked if screening mammography was discussed during the visit, the duration of the discussion in minutes, whether the patient seemed informed about screening mammography and whether the patient seemed anxious about mammography or breast cancer. Responses to these questions were given on a 5-point scale. Clinicians were blinded to information about each woman’s use of the decision aid, unless a woman volunteered this information during her scheduled visit. Thus, clinician assessments were conducted for both BSD users and non-users.
Analysis
All survey responses were analyzed using descriptive statistics, including frequencies, means and medians. DCS scores were estimated using that instrument’s scoring algorithm. Associations between predicted breast cancer risk and responses to selected items in the follow-up survey were assessed using one-way analysis of variance (ANOVA). Analysis of associations between between other outcomes were assessed using Fisher’s exact test. All analyses were performed in SAS (version 9.4, Cary, NC).