JMIR Publications

ABSTRACT

Background:

Medical and welfare facilities in the Noto region of Japan were severely affected by the 2024 Noto Peninsula earthquake and the subsequent torrential rains. Staff members working in these facilities have been disaster victims and frontline caregivers and face prolonged restoration work with limited psychological support. Nonverbal social robots have been designed to provide companionship and emotional comfort. However, their effects on health-related quality of life (QoL) and well-being among care staff in disaster-affected settings are unknown.

Objective:

This study aimed to investigate whether introducing a nonverbal artificial intelligence (AI) communication robot can improve QoL and subjective well‑being in care facility staff working under disaster conditions. The secondary objective was to assess the safety, acceptability, and intention to continue using the robot.

Methods:

An ABAB intervention design was implemented between February and June 2025. After a 2‑week baseline, staff in dementia care, general care, and short‑stay units received the robot intervention for 2 weeks (A1), followed by a 2‑week withdrawal (B1), re‑intervention (A2), and final withdrawal (B2). The questionnaires were administered at the end of each phase. Primary outcomes were health‑related QoL (EQ‑5D‑5L), well‑being (WHO‑5 Well‑Being Index), and mental health continuum (MHC‑SF). Secondary outcomes included safety (three Likert‑scale items), acceptability (17 semantic‑differential items), and interaction frequency. Friedman tests were used to compare outcomes across phases, with Wilcoxon signed-rank tests and Bonferroni correction for post-hoc comparisons. Only participants with complete data across all phases were analyzed.

Results:

Of the 58 staff completing baseline assessments, 49 provided complete data (25 dementia care, 12 general care, 12 short‑stay). The participants were predominantly female, with a median age in the fifth decade; 75.7% reported personal disaster damage. The median baseline EQ‑5D‑5L utility, WHO‑5 percentage, and MHC‑SF scores were approximately 0.93, 60%, and 35 points, respectively. Interaction frequency with the robot significantly increased during the intervention phases, but Friedman tests showed no significant differences in EQ‑5D‑5L, WHO‑5, or MHC‑SF scores across the ABAB phases within or across units. Safety outcomes and the intention to continue use did not differ between the intervention and withdrawal phases, and no adverse events were reported. Acceptability improved for items, such as “felt calm,” “liked,” and “felt peaceful” in the dementia care unit and for “competent” and “peaceful” in the pooled analysis. However, these effects were insignificant after Bonferroni correction.

Conclusions:

In this study, the short-term use of a nonverbal AI communication robot did not lead to measurable improvements in QoL or well-being. Nonetheless, the increased interaction and positive acceptability ratings suggest that the robot was well-received and could be safely and feasibly deployed in disaster settings. Long-term studies with larger samples are required to determine whether such robots can provide meaningful mental health support to healthcare workers. Clinical Trial: Not applicable.