Impact of Online Interactive Decision Tools on Women's Decision-Making Regarding Breast Cancer Screening: Systematic Review and Meta-Analysis

Abstract

Background: The online nature of decision aids (DAs) and related e-tools supporting women's decision-making regarding breast cancer screening (BCS) through mammography may facilitate broader access, making them a valuable addition to BCS programs.

Objective: This systematic review and meta-analysis aims to evaluate the scientific evidence on the impacts of these e-tools and to provide a comprehensive assessment of the factors associated with their increased utility and efficacy.

Methods: We followed the 2020 PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and conducted a search of MEDLINE, PsycINFO, Embase, CINAHL, and Web of Science databases from August 2010 to April 2023. We included studies reporting on populations at average risk of breast cancer, which utilized DAs or related e-tools, and assessed women's participation in BCS by mammography or other key cognitive determinants of decision-making as primary or secondary outcomes. We conducted meta-analyses on the identified randomized controlled trials, which were assessed using the revised Cochrane Risk of Bias 2 (RoB 2) tool. We further explored intermediate and high heterogeneity between studies to enhance the validity of our results.

Results: In total, 22 different e-tools were identified across 31 papers. The degree of tailoring in the e-tools, specifically whether the tool was fully tailored or featured with tailoring, was the most influential factor in women's decision-making regarding BCS. Compared with control groups, tailored e-tools significantly increased women's long-term participation in BCS (risk ratio 1.14, 95% CI 1.07-1.23, P<.001, I²=0%). Tailored-to-breast-cancer-risk e-tools increased women's level of worry (mean difference 0.31, 95% CI 0.13-0.48, P<.001, I²=0%). E-tools also improved women's adequate knowledge of BCS, with features-with-tailoring e-tools designed and tested with the general population being more effective than tailored e-tools designed for or tested with non-BCS participants (χ²₁=5.1, P=.02). Features-with-tailoring e-tools increased both the rate of women who intended not to undergo BCS (risk ratio 1.88, 95% CI 1.43-2.48, P<.001, I²=0%) and the rate of women who had made an informed choice regarding their intention to undergo BCS (risk ratio 1.60, 95% CI 1.09-2.33, P=.02, I²=91%). Additionally, these tools decreased the proportion of women with decision conflict (risk ratio 0.77, 95% CI 0.65-0.91, P=.002, I²=0%). Shared decision-making was not formally evaluated. This review is limited by small sample sizes, including only a few studies in the meta-analysis, some with a high risk of bias, and high heterogeneity between the studies and e-tools.

Conclusions: Features-with-tailoring e-tools could potentially negatively impact BCS programs by fostering negative intentions and attitudes toward BCS participation. Conversely, tailored e-tools may increase women's participation in BCS but, when tailored to risk, they may elevate their levels of worry. To maximize the effectiveness of e-tools while minimizing potential negative impacts, we advocate for an "on-demand" layered approach to their design.

Keywords: average risk; breast cancer screening; cognitive determinants; decision aid; decision-making; online interactive; screening participation; shared decision-making; women.

Figure 1

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 flow diagram. The diagram was generated using the Covidence software [44].

Figure 2

Forest plots of studies comparing the effect of e-tools with that of control on women’s (A) participation in breast cancer screening (BCS) by mammography at long term [61,63-66], (B) intention to undergo BCS [54,55,68-71], (C) intention not to undergo BCS [69,70], and (D) adequate knowledge about BCS [54,63,69,70]. We reported in (B) the data corresponding to Lin et al.’s “tailored message intervention” (TMI) [68] and Seitz et al.’s “extended information with untailored exemplars” e–tools [71]. Data for the other Lin et al.’s [68] and Seitz et al’s [71] e-tools are presented in in Appendix S4 in Multimedia Appendix 1. In (D) subgrouping was based on tailoring complexity of the e–tools (ie, tailored e-tools compared with features–with–tailoring e–tools), which end up being identical to subgrouping based on previous breast cancer screening status (ie, nonparticipants in BCS [tailored e–tools] compared with general population [features–with–tailoring e–tools]). Control was usual care, except in the studies by Lin and Wang [68], Roberto et al [55], and Seitz et al [71] (website), and Walsh et al [64] (video). The risk of bias was assessed using the Risk-of-Bias 2 (RoB 2) tool [52]. The risk level for each domain of RoB 2 (A: randomization process, B: deviations from the intended protocol, C: missing data, D: outcome measurement, and E: reporting results) and the (F) overall risk were evaluated as low (green), moderate (yellow), or high (red). M-H: Mantel-Haenszel.

Figure 3

Forest plots of studies comparing the effect of e-tools with that of control on women's (A) level of worry [63,71], (B) accuracy of perception of individual breast cancer risk [63,71], (C) decisional conflict [55,70], and (D) informed choice [55,69,70]. We reported in (A) and (B) the data corresponding to Seitz et al’s [71] “extended information with untailored exemplars” e-tool, which was the most comparable to Schapira et al’s [63] e-tool. Data for the other Seitz et al’s [71] e-tools are presented in Figures S6 and S7 in Multimedia Appendix 1. The control group was usual care, except for Seitz et al’s [71] and Roberto et al’s [55] e-tools, which used a website as the control. The risk of bias was assessed using the Risk-of-Bias 2 (RoB 2) tool [52]. The risk level for each domain of RoB 2 (A: randomization process, B: deviations from the intended protocol, C: missing data, D: outcome measurement, and E: reporting results) and the (F) overall risk were evaluated as low (green), moderate (yellow), or high (red). M-H: Mantel-Haenszel.