Trust divergence in elderly care robot acceptance in middle-aged and older adults: A latent profile and network analysis.

BACKGROUND: Acceptance of elderly care robots among middle-aged and older adults is heterogeneous, yet subgroup-specific patterns are often overlooked in variable-centered research. In particular, less is known about how trust, competence, and perception-related factors are organized within different acceptance subgroups. OBJECTIVE: To identify acceptance patterns toward elderly care robots among middle-aged and older adults and to examine subgroup-specific association structures underlying acceptance. METHODS: A cross-sectional survey was conducted among 577 middle-aged and older adults in Shanghai, China. Latent profile analysis was used to identify acceptance patterns using 4 care-robot acceptance items as profile indicators. Mixed-data profile-specific networks were then estimated using a semiparametric Gaussian-copula/nonparanormal approach. Bootstrap-based edge checks and follow-up regression analyses were used to examine subgroup-specific trust patterns more directly. RESULTS: Three acceptance patterns were identified: Proactive-Advocate (341/577, 59.1%), Cautious-Observer (164/577, 28.4%), and Negative-Resister (72/577, 12.5%). Across profiles, perceived competence and competence-/benevolence-related trust were more central to acceptance than sociodemographic factors. The Proactive-Advocate profile showed a relatively dense and positively integrated trust-perception-acceptance structure, whereas the Cautious-Observer profile showed a weaker and more conditional structure. The Negative-Resister profile showed the sparsest network and an exploratory subgroup-specific divergent trust pattern, which was further examined using follow-up regression analyses. CONCLUSIONS: Acceptance of elderly care robots among middle-aged and older adults is shaped by subgroup-specific configurations of competence perceptions, trust, and relational evaluation rather than by a single uniform pathway. By combining latent profile analysis with network analysis, this study provides a more differentiated account of subgroup heterogeneity in care-robot acceptance and supports subgroup-sensitive, person-centered, and ethically grounded implementation strategies. REGISTRATION: Not applicable.