# Association between working alliance and treatment outcomes in a mobile health intervention with a conversational agent (CanRelax)

## TL;DR

This study found that a stronger working alliance with a conversational agent in a mobile health app is linked to better treatment outcomes for cancer patients.

## Contribution

The study expands the understanding of working alliance in fully automated digital health interventions using a conversational agent.

## Key findings

- Stronger working alliance, especially in goal and task components, is associated with lower distress levels.
- Baseline treatment expectations are positively related to working alliance formation.
- The bond subscale of working alliance showed no significant association with distress.

## Abstract

This study explored the association between the digital working alliance and treatment outcomes in users of a fully automated mobile health app with a conversational agent (CanRelax app 2.0), which was effective in reducing distress in adults with cancer. This expands research about the working alliance association to an innovative setting. Drawing on secondary data from a randomized controlled trial, we examined whether the strength of the working alliance—and its three subcomponents: bond, task, goal—assessed at week 4 using the Working Alliance Inventory for Internet Interventions, was related to distress levels at week 10. The analysis population consisted of adults diagnosed with cancer in the past five years who received the CanRelax app 2.0 (N = 117). Participants accessed a conversational agent–guided intervention featuring mindfulness, relaxation, and personalized weekly coaching modules with behavior change techniques. Mean subscale scores (bond: 4.04, SD = 0.95; goal: 3.46, SD = 0.74; task: 2.79, SD = 0.75; 1–5 scale) indicated good working alliance. Partial correlations adjusting for baseline distress, sex, age, and treatment expectation indicated a small negative association between working alliance and distress (r = −0.18), particularly for the task and goal components. No significant association was found for the bond subscale in the overall sample. Associations were stronger in females, and baseline treatment expectations were positively related to working alliance. These findings suggest a stronger working alliance may contribute to better treatment outcomes in fully automated digital health interventions, and optimizing goal and task agreement could further enhance intervention effectiveness.

German Clinical Trials Register (DRKS00027546) on 23.02.2022.

•Users can form a working alliance with conversational agent-guided interventions.•Stronger alliance is linked to better treatment outcomes.•Goal and task agreement components show the strongest association with outcomes.•Optimizing goal and task agreement may enhance digital intervention effectiveness.•User expectations may influence working alliance development.

Users can form a working alliance with conversational agent-guided interventions.

Stronger alliance is linked to better treatment outcomes.

Goal and task agreement components show the strongest association with outcomes.

Optimizing goal and task agreement may enhance digital intervention effectiveness.

User expectations may influence working alliance development.