Factors associated with the quality of urban geriatric healthcare services empowered by large language models and improvement pathways: a theoretical framework and cross-sectional study

With the rapid advancement of digital technologies, large language models (LLMs) are increasingly being deployed in the healthcare sector. In specific application scenarios, however, urban older people face a practical paradox between “device intelligence” and “low digital literacy.” Drawing on service quality theory, this study examined the factors associated with the quality of LLM-empowered healthcare services as perceived by urban older people and proposed corresponding improvement pathways. A cross-sectional quantitative study was conducted among urban older adults in Beijing, China, who had used hospital-provided LLM-based medical devices. A perception-based, adapted SERVQUAL questionnaire was developed and pilot-tested. A total of 298 valid responses were analysed (valid response rate, 91.9%). After reliability and validity testing, structural equation modelling (SEM) was used to examine the associations between the five SERVQUAL dimensions (tangibility, reliability, responsiveness, assurance and empathy) and perceived LLM-empowered geriatric healthcare service quality. All five dimensions—tangibility, reliability, responsiveness, assurance and empathy—were significantly and positively associated with older people’s perceived quality of LLM-empowered healthcare services. Assurance and empathy showed the largest standardised path coefficients; however, constrained comparison tests detected no statistically significant differences among the coefficients. The structural model explained 57.0% of the variance in perceived service quality. The five SERVQUAL dimensions were positively associated with perceived quality of LLM-empowered healthcare services among the surveyed urban older adults, supporting the applicability of a perception-based SERVQUAL approach to AI-mediated geriatric care. The findings provide preliminary, perception-based evidence to inform the age-friendly technical optimisation of LLM-based services, hybrid online–offline service models and AI governance policies.