Introduction

Breast oncologists and surgeons have long recognized that breast cancer care should be refined by individual patient needs, preferences, and values, as patients may respond to treatment differently based on a variety of factors. Over the last three decades, personalized care has gained traction with the emergence of genomic medicine [1], ‘big data’ [2], digital health [3, 4], and advanced treatment for breast cancer [5, 6]. In this context, several web-based, interactive decision-making tools have been introduced to clinical practice to support personalized breast cancer care [7,8,9,10,11]. These breast cancer-specific tools were designed to provide tailored outcomes and care recommendations considering individual demographic (e.g., age) [12], genomic (e.g., 21-gene recurrence score) [13], clinical (e.g., tumor size) [14], behavioral (e.g., smoking) [15], and contextual (e.g., insurance status) [16] characteristics together with patient needs, preferences, and values [17]. For example, the ‘BreastCHOICE’ tool is a personalized decision-making tool used to estimate the risk of surgical complications in early-stage breast cancer patients considering breast reconstruction based on their individual height, weight, past medical history, smoking status, and personal preferences/values [15].

Overall, studies have shown that personalized decision-making tools could increase knowledge, reduce negative emotions, such as anxiety and fear, associated with treatment, and improve overall quality of life among breast cancer patients and survivors [7, 18,19,20]. Furthermore, breast cancer decision-making tools that include contextual factors, such as treatment costs, insurance status, and access to treatment facilities, could potentially help address root causes of disparities in clinical settings [21,22,23,24]. For example, decision-making tools for medical situations, including chest pain, diabetes, Graves’ disease, depression, osteoporosis, and cardiovascular risk prevention, have shown that tools that raise cost as an issue could increase the occurrence of conversations related to the costs of drugs, insurance, and health care between patients and their physicians [25].

Recently, the U.S. Food and Drug Administration (FDA) issued a guidance to regulate decision-making tools as medical devices, increasing the focus on using high-quality tools to support clinical care in the U.S. [26]. However, there are several barriers to integrating high-quality personalized decision-making tools into current clinical care [19]. For instance, physicians and patients have reported a lack of understanding of existing tools, limited knowledge on how these tools can be used to support clinical care, and as a result, low motivation to use decision-making tools to guide clinical care [27,28,29]. Studies have also found that both patients and physicians have limited knowledge on the validity, usability, and quality of existing tools to assess their performance in real-world practice settings [30,31,32,33,34].

While breast cancer decision-making tools exist, there is limited information about their quality, validity, usability, feasibility, and acceptability. We aimed to fill this knowledge gap by critically reviewing the characteristics of existing English-language, interactive, web-based personalized decision-making tools available to support breast cancer care. The overarching goal of our review was to present evidence on the existing decision-making tools for breast cancer treatment and survivorship to support the integration of these tools into clinical practice.

Methods

This scoping review followed the methodological framework initially proposed by Arksey and O’Malley, Levac and colleagues, and the Joanna Briggs Institute [35,36,37]. This framework includes six stages to guide scoping review processes: (1) specifying the research question, (2) identifying relevant literature, (3) selecting studies, (4) data mapping, (5) summarizing, synthesizing, and reporting the results, and (6) including expert consultation. Our review was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) Checklist (Supplemental Table 1) [38]. The study was registered in Open Science Framework [39]. Since the study included a review of published articles and study-level results, institutional review board approval or exemption was not required.

Data sources and search strategy

We conducted a search of published literature to identify articles that discussed personalized, interactive, dynamic, web-based decision-making tools designed to support breast cancer treatment and survivorship decisions for physicians and individuals diagnosed with breast cancer. The comprehensive search strategy included a combination of keywords, synonyms, Medical Subject Headings (MeSH), and Emtree terms relating to concepts of clinical decision-making tools, survivorship, treatment, web-based, personalized, and breast cancer (Supplemental Table 2). A trained librarian (GB) at the National Institutes of Health pilot tested 50 articles and refined our search strategy based on the initial search results. We searched PubMed, PsycInfo, Embase, Scopus, Web of Science, and Cochrane Database of Systematic Reviews for relevant articles. After screening all the articles from the database searches, we reviewed the reference lists of the articles to identify any additional tools that may have been missed, and these additional relevant articles were screened based on inclusion/exclusion criteria. The date of our most recent search was May 12, 2023.

Inclusion and exclusion criteria

For all articles, the inclusion criteria included: (1) female or male adults (≥ 18 years) diagnosed with breast cancer, (2) breast cancer treatment or survivorship, (3) online, web-based risk prediction models and interactive, personalized, or individualized tools developed from 2013 to 2023, (4) primary empirical research studies, and (5) articles written in English. We limited our search to include tools from 2013 to 2023, as these tools are more likely to consider the most up-to-date information on breast cancer treatment and survivorship care. Additional information is provided in Supplemental Table 3.

Data screening, extraction, and assessment of articles and tools

All titles and abstracts from articles retrieved from the databases were initially screened for eligibility by four authors (KW, DK, JZ, LS) based on the inclusion and exclusion criteria. A second round of screening using the same criteria was conducted via a full text review of the remaining articles. Screening was done using Covidence, an online application that helps streamline the review process [40]. Disagreements between authors were resolved through discussions.

We visited each tool’s publicly available website and tested each tool with pseudo patient characteristics to identify patient inputs used for personalization and breast cancer outcomes included in the tool. For tools that did not have publicly available websites, we reviewed screenshots and examined the tool development section in the methods of each corresponding article to retrieve information. We contacted the corresponding author for missing information. We used the articles, websites, and relevant screenshots to extract information about each tool, including the name and purpose, target population for tool development, interventions, data source and methods, input factors (e.g., individual, clinical, genomic, behavioral, contextual) used for personalization, breast cancer outcome/s, target user/s, and date of last update.

We also reviewed articles that provided information on tool validation, usability, feasibility, and acceptability testing. Personalized, web-based decision-making tools typically use statistical and/or simulation models to estimate outcomes associated with various input factors. After model development, these models are validated in independent, external samples to evaluate model performance and generalizability [41]. Usability testing is designed to capture the user experience and understanding of the tool, while feasibility testing helps infer the likelihood that the decision-making tool will be used to enhance the patient-physician interaction [32,33,34]. Acceptability testing is conducted to evaluate user satisfaction with the tool [32,33,34]. We extracted information on the distribution of race and ethnicity, education, income, marital status, and insurance in the sample of individuals included in validation, usability, feasibility, and acceptability testing of the tools. Data were extracted using Covidence and Excel [40].

Quality assessment

We used the International Patient Decision Aid Standard (IPDAS) instrument to assess the quality of each tool included in our study [42]. The IPDAS collaboration considers a decision aid to be any tool that helps people make decisions about health care [43]. The IPDAS instrument was selected for the quality assessment since it was established to provide a standardized framework and a set of criteria to evaluate the content, development, and implementation of decision tools used to support health care decisions [43]. These criteria may be useful to a wide range of individuals who may use decision tools such as patients, healthcare providers, tool developers, researchers, and policymakers [42, 43].

Accordingly, the IPDAS instrument checklist evaluates tools based on the presentation of information, ability to clarify patient values, tool development process, story usage, the impact of the tool on decision processes, and decision quality [42, 43]. The full IPDAS instrument checklist is accessible in Supplemental Table 4. In our study, the tools were scored from a range of 0 to 63, with increasing scores representing the increasing number of items from the IPDAS instrument checklist represented in each tool. Finally, we summarized the overall strengths and weaknesses of each tool considering the IPDAS instrument checklist [42].

Results

Search results

A total of 5,237 records were identified through PubMed, PsycInfo, Embase, Scopus, Web of Science, and Cochrane Database of Systematic Reviews. After removing duplicates, irrelevant, and ineligible articles, a total of 46 relevant articles were included in this study (Fig. 1). These articles described 54 tools, including 11 tools that provided personalized breast cancer treatment outcomes based on individual factors (e.g., age, tumor characteristics). The remaining 43 tools provided breast cancer outcomes associated with individual factors but did not include treatment-specific personalized breast cancer outcomes.

Fig. 1
figure 1

Article identification process using research framework. From: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. https://doi.org/10.1136/bmj.n71

Full size image

Personalized tools for treatment outcomes (N = 11)

These tools varied by target population, inputs, outcomes, and treatment considerations (Table 1) [13,14,15, 17, 44,45,46,47,48,49]. The tools were developed for adult women (≥ 18 years) with early-stage breast cancer [13,14,15, 17, 44, 45, 47,48,49] or ductal carcinoma in situ (DCIS) [46] considering different types of treatment. The target users for two tools were only patients [15, 17], while four tools were developed for physicians only [47,48,49], and nine tools were developed for both physicians and patients [13, 14, 44,45,46]. Two tools were developed specifically for older women [45, 49]. Four tools predicted treatment outcomes for local–regional or distant recurrence risk [13, 14, 44, 45], and five tools predicted breast cancer mortality [44,45,46, 49], while the other tools predicted other treatment outcomes. Several tools (N = 4) included multiple outcomes [14, 44,45,46]. For example, ‘BTxChoice’ provided estimates for the 10-year risk of distant recurrence and life-years gained with and without chemotherapy treatment [14].

Table 1 Web-based decision-making tools for personalized treatment outcomes (N = 11)
Full size table

The tools varied by inputs used to estimate breast cancer treatment outcomes. All tools included individual and clinical characteristics, such as age and tumor size. Two tools considered genomic features measured by the 21-gene recurrence score [13, 14], and two tools considered health behaviors [15, 45]. No tools considered the impact of contextual factors, such as insurance status or access to a treatment facility. One tool helped elicit patient preferences and values by providing a brief survey outlining patients’ thoughts and feelings about treatment options [15]. We found one tool considering the variation of breast cancer outcomes based on race and ethnicity [46].

Validation, usability, feasibility, and acceptability testing

Six tools were externally validated [13,14,15, 46,47,48], three tools were internally validated [13, 14, 44], and three tools did not undergo any validation [17, 45, 49]. Five tools provided results from usability, feasibility, and/or acceptability testing [14, 15, 17, 45, 46]. ‘BreastCHOICE’ had a high mean usability score of 6.3, which was measured using the Computer System Usability Questionnaire, providing a score ranging from 1.0 (lowest) to 7.0 (highest) [15, 50]. ‘Which treatment for DCIS is right for you?’ had a mean usability score of 3.7 out of 5.0 measured using the System Usability Scale and the Preparation for Decision-Making Scale [46, 51, 52]. ‘BTxChoice’ and ‘Radiotherapy for Older Women’ did not report results from usability testing, but the authors stated that the tools were in the process of undergoing testing [14, 45]. ‘BRECONDA’ underwent acceptability and feasibility testing; it was assessed for usefulness and relevancy on a Likert-scale from 1 (lowest) to 5 (highest), with the tool receiving mean scores of 4.8 and 4.4, respectively [17]. Follow-up studies confirmed acceptability of the tool [53, 54].

Supplemental Table 5 provides the distribution of race and ethnicity, income, education, marital status, and insurance status of the individuals included in the validation and usability testing of the tools. Most patients included in validation and usability testing were White (68.2–83.9%) and married (71.1–86.0%).

Personalized tools for other outcomes (N = 43)

A total of 43 tools included models to estimate breast cancer outcomes associated with individual, tumor, and contextual characteristics, but did not include treatment-specific personalized breast cancer outcomes (Table 2) [16, 55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87]. These tools were created for adult (≥ 18 years) female and/or male breast cancer patients who had undergone treatment for DCIS or invasive breast cancer. Ten tools were developed for patients with bone or lung metastases after a breast cancer diagnosis [56, 57, 69, 73, 75]. Four tools were created for young breast cancer patients (18–40 years) [16, 68, 78], and another three were created for elderly patients (≥ 65 years) [64, 65, 74]. Three tools were developed specifically for male breast cancer patients with bone metastases [73]. The target user for four tools was patients [55, 78, 80, 81], while 31 were developed only for physicians [16, 56,57,58,59, 62,63,64,65, 67,68,69,70,71, 74,75,76,77, 79, 82,83,84, 86, 87], and eight were developed for both physicians and patients [60, 61, 66, 72, 73, 85]. The most common outcomes estimated in these tools included overall survival (N = 20) [16, 57, 62,63,64, 66,67,68,69,70,71,72,73,74,75,76,77, 79, 85, 86], breast-cancer specific survival (N = 7) [16, 57, 65, 70, 73, 75, 79], and risk of bone metastasis (N = 3) [56, 69, 73]. The ‘After Cancer Education and Support Operations’ tool was the only tool developed to support breast cancer survivors by providing health alerts and follow-up care recommendations after treatment [55].

Table 2 Web-based decision-making tools for other outcomes (N = 43)
Full size table

All tools considered individual and clinical factors such as age and tumor stage [16, 55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87]. Four tools considered health behaviors, such as smoking status [59, 60, 78, 82]. Twenty-one tools incorporated contextual factors, including marital status (N = 20) [16, 57, 62,63,64,65, 69, 70, 72,73,74,75,76, 79], insurance status (N = 4) [16, 69], education (N = 1) [78], employment status (N = 1) [78], and financial status (N = 1) [78]. Only two tools included components to incorporate patient preferences or values into decision-making [78, 80]. We found 17 tools considering Black, White, and other race categories to estimate breast cancer outcomes [16, 56, 57, 59, 62,63,64,65, 69, 70, 75, 76]. One tool considered Hispanic and non-Hispanic ethnicities [59].

Validation, usability, feasibility, and acceptability testing

We found that 15 tools were externally validated [55, 56, 58, 59, 61, 65, 68, 69, 71, 76, 79, 84], and 35 tools were internally validated [16, 56, 57, 59, 60, 62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77, 79, 83, 84, 86, 87]. Five tools did not undergo any validation testing [78, 80,81,82, 85]. Only four tools underwent usability, feasibility, and/or acceptability testing [55, 78, 80, 81]. ‘After Cancer Education and Support Operations’ assessed usability and acceptability using ‘Consistency’, ‘Stressfulness’, and ‘Simplicity’ with scores ranging from 1 (most positive) to 7 (most negative) [55, 88]. Consistency refers to the ability to use the tool in the same way over time, while stressfulness refers to the amount of worry or tension caused by the use of the tool, and simplicity refers to the ease of use of the tool [55, 89]. Users reported a mean consistency score of 1.2, a mean stressfulness score of 1.2, and a mean simplicity score of 1.4 for this tool [55]. The tool developed by Graetz et al. was tested for feasibility based on reports from physicians and nurses who used it; providers stated that the tool was easy to implement and did not significantly impact workflow [81]. The ‘Resources and Education for Adherence to Cancer Hormonal Therapy’ tool was assessed for feasibility and acceptability [80], where the study aimed to have 80% of eligible individuals enroll with 80% completing at least one online session. Both goals were exceeded for this tool, with 85.4% of eligible individuals enrolling and 83.7% of individuals completing at least one session [80]. Acceptability was measured using the ‘Client Satisfaction Questionnaire’ and the ‘Intervention Feedback Questionnaire’ [90]. The ‘Resources and Education for Adherence to Cancer Hormonal Therapy’ tool had a mean acceptability score of 3.0 (range 1–4) and 3.4 (range 1–5) on both questionnaires, respectively [80].

Supplemental Table 5 provides the distribution of race and ethnicity, income, education, marital status, and insurance status of the individuals included in validation, usability, feasibility, and acceptability testing of these tools. Most patients were White (0–93.0%), married (41.4–94.0%), and had insurance (93.7–94.9%).

Quality assessment

The sum of the scores for each tool in each dimension on the IPDAS instrument checklist is reported in Supplemental Table 6. The tools could receive scores ranging from 0 (lowest quality) to 63 (highest quality). Most tools provided information about options (N = 48) and outcome probabilities (N = 48), were written in plain language (N = 49), and were easy to navigate online (N = 51). However, only six tools provided disclosure information about funding or conflicts of interest, and only two tools used stories. In our sample, the average quality assessment score for the tools was 16 (range: 6–46; potential maximum: 63). The tool with the highest IPDAS instrument score was ‘BreastCHOICE’, with 46 points. ‘BreastCHOICE’ provided information on different options and the development process while also sufficiently incorporating patient values and preferences into the decision-making tool by asking patients what matters most to them, what their concerns were, and how they feel about different treatments [15].

Summary: strengths and weaknesses

We provided a list of strengths and weaknesses of the web-based decision-making tools included in our study in Table 3. In terms of strengths, we found that most tools were written in plain language (N = 49), were validated (N = 45), and provided information about breast cancer outcomes (N = 48). However, usability, feasibility, and acceptability of the tools were evaluated using different measures. As a result, it was not possible to compare the performance of the tools. There was also limited information on the validity and usability testing of the tools in underserved (e.g., uninsured, low education) and underrepresented (e.g., Alaska Native, Pacific Islanders) populations.

Table 3 Key strengths and weaknesses of treatment and survivorship web-based decision-making tools available from 2013 to 2023
Full size table

Discussion

Breast cancer care decisions are complex and often require the consideration of individual, clinical, genetic, health behavioral, and contextual characteristics, as well as personal preferences and values, to achieve optimal treatment outcomes. In this scoping review, we identified 54 web-based, personalized, interactive decision-making tools that could be used to support breast cancer care in clinical settings.

Comparison with other literature

Previous studies have reviewed up to 21 tools, including risk prediction models, to support breast cancer treatment decisions [7, 91, 92]. In contrast, we identified a broader set of tools that could potentially be useful to support breast cancer treatment and survivorship care decisions in clinical settings. Like previous reviews, we also found that most tools still need to undergo usability, feasibility, and acceptability testing [7, 91, 92]. However, in this study, in addition to an appraisal of tool validity, usability, feasibility, and acceptability, we also evaluated the inclusion of underrepresented and underserved populations in tool development and testing. We found that individuals included in post-testing of the tools were mostly White, insured, married, and had higher levels of education. Moreover, previous reviews have provided limited information on health behaviors and contextual factors that may also influence breast cancer outcomes [7, 91, 92]. To our knowledge, this is the first to provide a detailed and comprehensive evaluation of the web-based decision tools considering health behaviors, contextual factors, and the characteristics of the populations included in validity and usability testing of these tools.

Summary of main findings

Tool validation is a necessary step in decision-making tool development, as it provides critical information on the tools’ ability to accurately estimate outcomes of interest in independent cohorts [93]. A tool’s performance (e.g., sensitivity, specificity) may vary based on the distribution of individual, clinical, and contextual characteristics of a given cohort [94]. Therefore, it is important to test the external validity of the decision-making tools (and related algorithms) in independent cohorts prior to the introduction of these tools into practice settings. Validation could also help identify additional important features that may have been missed in the initial development of the tool, which could help further increase the accuracy of the prediction. The validation samples for the tools in our review included mostly White, married, and insured populations. For example, ‘BTxChoice’ was validated in two populations, both with a White majority (73.0–83.9%) [14]. These findings were consistent with previous studies reporting that only 14% of decision tools were tested with a significant representation of underserved and underrepresented groups [95]. The lack of representation in validation samples could limit the ability to assess the performance of these tools in diverse settings [95]. Importantly, if the tools are unable to generate accurate estimates for certain subgroups of the population, using them to guide clinical decisions could perpetuate disparities in cancer care and outcomes. Therefore, it is necessary to develop and validate tools in diverse cohorts including underserved and underrepresented individuals.

Usability testing is a necessary step in tool development to help identify and fix problems with the website/mobile application, [96] but few tools in our review had undergone usability testing. During usability testing, tool developers should assess the tools’ ease of use and the presentation of information considering health literacy and numeracy [97, 98]. Studies have shown that tools that are difficult to use are often neglected despite their utility [99]. Usability testing that includes individuals with different levels of health literacy and numeracy could potentially enhance the long-term utility of these tools in clinical settings [96,97,98].

Several tools considered health behaviors, such as smoking status and alcohol intake. Health behaviors are important predictors of breast cancer mortality and survivorship [100]. While physical activity was not considered a health behavior in most of the decision-making tools included in our study [78], previous studies have shown that increased physical activity could lower breast cancer recurrence and mortality [101, 102]. Current smoking, dietary intake, sedentary behavior, and poor sleep are also known to be associated with breast cancer mortality [103,104,105]. Inclusion of these factors in breast cancer decision-making tools could potentially help patients identify resources (e.g., smoking cessation interventions for quitting) to improve behavior and help physicians develop survivorship care plans considering these factors.

Few tools considered patients’ preferences and values by asking patients their thoughts and concerns about different treatments and what matters most to them. Patients may have a wide range of preferences and values when considering the benefits and harms of treatment. Patients who receive their preferred treatment have been shown to be half as likely to stop treatment, and patients who are actively involved in decision making throughout their cancer care by voicing their preferences and values report a higher quality of life [18, 106]. Additionally, tools that incorporate patient values, such as cultural values, spirituality, and community, often improve the communication between patients and physicians, leading to improved shared decision making [107].

The debate over whether to include race and ethnicity in risk prediction models is ongoing, and not many tools included race or ethnicity as input variables. Race-based medicine has been used to deliver healthcare for years based on epigenetics, but it has a deeply problematic history used to reinforce and justify slavery and perpetuate racial discrimination [108]. Furthermore, racial categories change over time, which may mean that older tools that have not been updated may not be as relevant or accurate [108]. Currently, there is a push to consider race as an input factor only when it is directly connected to racism and contextual factors [109]. Studies have shown that contextual factors such as lack of health insurance, income, food insecurity, and access to treatment facilities contribute to the racial and ethnic disparities in breast cancer mortality [110, 111]. Therefore, the consideration of these factors in decision-making tools could potentially provide a means to reduce racial and ethnic disparities in breast cancer outcomes in the U.S. [112].

Less than half of the decision-making tools personalized breast cancer outcomes based on individual contextual factors such as insurance, education, employment, marital status, and financial status/burden. We considered marital status as a contextual factor due to the marriage protection theory [113], which posits that marriage may lead to improved breast cancer survival through the strengthening of interpersonal relationships, providing social and financial support, and reducing risky behaviors [114, 115]. Studies have also shown that living in highly segregated neighborhoods in the U.S. are associated with lower rates of breast cancer survival [116, 117]. The inclusion of these factors in decision-making tools may provide an opportunity for physicians to discuss, advocate, and ensure that patients’ full range of circumstances are accounted for when making informed decisions about breast cancer care.

Strengths and limitations

Our review has several limitations that should be considered when evaluating our findings. We did not consider web tools created prior to 2013 or in any language other than English because we wanted to limit our review to include the most recent, relevant tools. However, this means that our search likely did not encompass the full range of personalized decision-making web tools that are currently available for breast cancer care. Additionally, we only assessed tools that were developed in the U.S., Europe, Australia, Japan, and Korea. Because of this, tools may not be generalizable or applicable to all populations. We were unable to access 23 tools due to payment barriers or because only screenshots with incomplete information were available in the publications. As a result, we were unable to assess the quality of all the components of those tools that were not easily accessible. Also, we were unable to report the characteristics of the samples included in the validation, usability, feasibility, and acceptability testing of 18 tools, as this information was not readily available in the original studies.

Despite these limitations, we conducted a robust search for personalized web-based clinical tools and identified a significant number of tools that assessed breast cancer treatment and survivorship outcomes. To our knowledge, this is the first scoping review providing a detailed assessment and comparison of the web-based decision tools available to support breast cancer care in clinical settings.

Conclusions

There was wide variation in the characteristics, validity, usability, and quality of web-based, interactive decision-making tools available to support breast cancer care. We found that the quality assessment tool (i.e., the IPDAS instrument checklist) did not include components to evaluate contextual factors which may influence patient decisions, the ability to seek health care, and patient outcomes [42]. The inclusion of contextual factors in the IPDAS instrument checklist could motivate tool developers to include these factors in new decision-making tools.

We expect the quality and the use of these tools to increase with the new U.S. FDA regulation [26]. However, it is important to concurrently provide training to patients and physicians to ensure that these tools are used for their intended purposes [27,28,29, 118]. Further, integrating decision tools into electronic medical records systems could improve clinical workflow, the speed and quality of decision making, and communication between physicians and their patients [119].