JMIR Publications

ABSTRACT

Background:

Mobile health (mHealth) interventions with virtual coaches offer scalable and potentially cost-effective solutions for health behavior change. However, these interventions commonly present challenges such as limited personalization and insufficient grounding in evidence-based strategies. Perfect Fit (PF), a personalized mHealth intervention with a text-based virtual coach, supports adults in quitting smoking and becoming more physically active. By combining innovative techniques including sensor technology, end-user involvement, and evidence-based strategies, PF aims to address common challenges faced by mHealth interventions, including those with virtual coaches.

Objective:

The study primarily investigated the feasibility and acceptability of PF. The secondary aim was to explore associations between socio-demographic, smoking-, and physical activity-related characteristics and the feasibility and acceptability outcomes. The third aim was to evaluate the feasibility of conducting the research study.

Methods:

A single-arm, pre-post, mixed-methods study was conducted in the Netherlands with 100 adults who smoked. The intervention lasted approximately 16 weeks. Data were collected at baseline, during the intervention, and post-intervention (4 months). Quantitative data included usage data and self-report questionnaires on feasibility, acceptability, and baseline characteristics. Qualitative data were gathered through post-intervention semi-structured interviews. Analyses included descriptive and inferential analyses, as well as the Framework Approach for the qualitative data.

Results:

PF usage varied considerably across participants (n = 87). The mean satisfaction rating was 2.79 (SD = 0.73; scale range 1–4), and perceived usability had a median score of 67.50 (range = 12.50–87.50; scoring range 0–100), indicating OK-to-good usability. The mean virtual coach acceptance rating was -0.27 (SD = 1.30; scale range -3–3; n = 77). Higher PF usage was associated with greater satisfaction, usability, and coach acceptance (all p ≤ .004). Frequent connection issues with the smartwatch were a disruptive factor. Qualitative findings (n = 12) provided in-depth insights into PF’s feasibility and acceptability, encompassing both positive and negative experiences. For instance, some participants valued the virtual coach for its anonymity, low-threshold access, and the sense of control it offered, while others preferred a human coach for greater accountability. Suggested improvements included more varied content and enhanced adaptability of the coach to users’ input and personal situations. Exploratory analyses suggested that high PF users were older than moderate (p = .014) and low PF users (p = .050). Importantly, PF was perceived as similarly feasible and acceptable across socioeconomic groups (p > .05), aligning with one of the project’s goals. Finally, research procedures and recruitment strategies proved feasible.

Conclusions:

PF shows potential as an accessible and inclusive strategy for multiple health behavior change, contributing to public health. Findings highlight areas for improvement and can guide the future development of virtual coach interventions. Clinical Trial: ClinicalTrials.gov NCT06095999