Using Digital Humans to Deliver Stress Management to Women With Breast Cancer: Feasibility and Acceptability

Abstract

Breast cancer patients experience a myriad of social, emotional, and physical stressors following diagnosis and treatment. Cognitive Behavioural Stress Management (CBSM) is a 10-week in-person group programme that improves stress management skills and decreases the risk of adverse outcomes of stress in breast cancer patients. However, CBSM is limited in its availability. Digital humans offer a novel method to deliver CBSM to patients remotely and autonomously to improve accessibility, but feasibility and acceptability must first be evaluated. The aim of this thesis was to examine the acceptability and feasibility of a digital human in delivering one module of CBSM to breast cancer patients post-treatment. Fifteen female patients who had completed active treatment for breast cancer and fifteen female healthcare clinicians working within cancer-related spaces were recruited from New Zealand. Patients and clinicians completed one 60-minute module of CBSM delivered by a digital human, containing psychoeducation about stress and deep breathing exercises. Patients and clinicians completed a questionnaire post- session on acceptability. Feasibility was measured through the practicality of recruitment and the trial methods. Pre- and post- intervention changes in stress, relaxation, and anxiety were assessed using visual analogue scales. Perspectives about Large Language Models (LLMs) were assessed pre- and post- intervention to inform potential incorporation into future iterations. Conventional content analysis was conducted on responses to open-ended questions collected post-intervention. Findings indicated moderate acceptability and feasibility of the digital human intervention. Patients were more accepting of the intervention than clinicians. Stress and anxiety decreased from baseline to post-intervention, and relaxation increased. Content analysis found two dominant themes regarding (1) the intervention, and (2) the digital human. Participants liked the interactive session and the digital human’s warm appearance but disliked the long length, information density, and impersonalised nature of the session. Participants expressed cautious curiosity about LLMs and wanted them to be further researched. This was the first study to assess the feasibility and acceptability of CBSM delivered by a digital human in breast cancer patient populations. Findings supported feasibility and acceptability, and suggested developments to be made, including shortening the length of the session and personalising it to the user. Future research should investigate the integration of LLMs and develop further modules of digital human delivered CBSM.