JMIR Publications

ABSTRACT

Background:

Early lifestyle interventions may reduce dementia-related cognitive decline, but individuals with cognitive impairment often face barriers to accessing and maintaining such interventions. Smartphone applications and wearables may offer low-burden, real-time support.

Objective:

This study examined the feasibility and acceptability of a two-week sensor-augmented just-in-time adaptive intervention (JITAI) delivered via smartphone-based conversational agents (CAs) and a wearable fitness tracker in older adults with mild cognitive impairment (MCI).

Methods:

In this single-arm, feasibility study, 24 MCI participants (mean age: 75.9 years) used the intervention with a rule-based CA delivering two daily JITAI interventions: (1) a sensor-informed primary physical-activity component, in which continuously collected step-count data from a wrist-worn fitness tracker (Garmin Vivoactive 6) were used to assess vulnerability against a personalized step goal and deliver tailored walking prompts; and (2) an ecological momentary assessment-based secondary health-promoting component targeting one of six additional health domains (hydration, nutrition, cognitive exercise, sleep, social engagement, movement), assessed via momentary self-report. Both components followed the same JITAI decision logic: vulnerability assessment in the afternoon, receptivity evaluation if vulnerable, tailored prompt delivery if receptive, and adherence assessment at an evening check-in. Post-intervention questionnaires assessed technology acceptance, usability, working alliances and qualitative feedback.

Results:

Across the study period, engagement with the CA was high for both components. Afternoon response rates were 86.6% for the primary physical-activity component and 84.2% for the secondary health-promoting-activities component. Evening check-in response rates were 85.6% and 79.3%. Among all responses classified as vulnerable (38.7% and 32.1% of afternoon interactions on the physical-activity and secondary-activity components, respectively), 85.4% and 75.9% were receptive, of which 76.6% and 70.7% resulted in adherence at the evening check-in. On days satisfying the full JITAI chain for physical activity, participants exceeded their personalized step goal, reaching a mean of 170% of goal (SD ± 95) and a mean evening step count of 7,718 steps (SD ± 3,453). The wearable-based data acquisition pipeline was feasible and largely reliable under study conditions, with the Garmin device supplying step-count data in 80.8% of decision points, the smartphone in 17.4%, and manual self-report required in only 1.8% of cases. Participants rated the CA favorably across all technology acceptance and usability domains, with all mean scores above the neutral midpoint on seven-point Likert scales. Qualitative analysis described an intervention experienced as accessible and structuring.

Conclusions:

This pilot study supports the short-term feasibility and acceptability of integrating wearable-derived step data into a smartphone-based CA-delivered JITAI for older adults with MCI. Passive sensing may support personalized intervention delivery while reducing reliance on self-report. However, as a feasibility study, it was not designed to provide evidence of efficacy. Controlled studies with adaptive decision times, sensor-derived baselines, and more diverse samples are needed to evaluate efficacy and sustained behavioral benefits. Clinical Trial: NA