Background

The predominant oncologist-led model in many countries is unsustainable to meet the needs of a growing cohort of breast cancer survivors (BCS). Despite available alternative models, adoption rates have been poor. To help BCS navigate survivorship care, we aimed to systematically develop a decision aid (DA) to guide their choice of follow-up care model and evaluate its acceptability and usability among BCS and health care providers (HCPs).

Methods

We recruited BCS aged ≥ 21 years who have completed primary treatment and understand English. BCS receiving palliative care or with cognitive impairment were excluded. HCPs who routinely discussed post-treatment care with BCS were purposively sampled based on disciplines. Each participant reviewed the DA during a semi-structured interview using the 'think aloud' approach and completed an acceptability questionnaire. Descriptive statistics and directed content analysis were used.

Results

We conducted three rounds of alpha testing with 15 BCS and 8 HCPs. All BCS found the final DA prototype easy to navigate with sufficient interactivity. The information imbalance favouring the shared care option perceived by 60% of BCS in early rounds was rectified. The length of DA was optimized to be 'just right'. Key revisions made included (1) presenting care options side-by-side to improve perceived information balance, (2) creating dedicated sections explaining HCPs' care roles to address gaps in health system contextual knowledge, and (3) employing a multicriteria decision analysis method for preference clarification exercise to reflect the user's openness towards shared care. Most BCS (73%) found the DA useful for decision-making, and 93% were willing to discuss the DA with their HCPs. Most HCPs (88%) agreed that the DA was a reliable tool and would be easily integrated into routine care.

Conclusions

Our experience highlighted the need to provide contextual information on the health care system for decisions related to care delivery. Developers should address potential variability within the care model and clarify inherent biases, such as low confidence levels in primary care. Future work could expand on the developed DA's informational structure to apply to other care models and leverage artificial intelligence to optimize information delivery.