Abstract
<title>Abstract</title> <p>Background: Digital transformation, population aging, and the expansion of long-term and technology-supported care are changing the competencies, skill mix, and deployment requirements of the health workforce. Smart health services require workers who can combine health knowledge, digital literacy, rehabilitation and elderly-care skills, service coordination, and data-informed practice. Objective: This study examined whether perceived support from AI-enabled multidimensional knowledge graphs (AI-MKGs) was associated with curriculum-workforce alignment and smart health workforce development quality in China, and whether supplementary HRH indicators and a proof-of-concept AI-MKG evaluation strengthened the workforce-development interpretation of the model. Methods: A cross-sectional survey was conducted among 250 smart health workforce stakeholders recruited from education, service-delivery, management, research, and digital health contexts. Supplementary HRH indicators were added to the SEM framework, including graduate role readiness, job-role fit, skill-mix coverage, deployment flexibility, retention intention, service-delivery readiness, and employer/practitioner practical-readiness evaluation. A proof-of-concept AI-MKG was also evaluated through expert validation and a pilot curriculum-planning task. Results: The sample included educators (31.2%), students or trainees (21.6%), educational/workforce administrators (15.2%), health or elder-care practitioners (18.8%), and researchers or policy experts (13.2%) from six institutional settings and six geographic regions. Perceived AI-MKG support was positively associated with curriculum-workforce alignment (beta = 0.533, p < 0.001) and workforce development quality (beta = 0.423, p < 0.001). Curriculum-workforce alignment was associated with graduate role readiness (beta = 0.41, p < 0.001), job-role fit (beta = 0.46, p < 0.001), skill-mix coverage (beta = 0.38, p < 0.001), deployment flexibility (beta = 0.35, p < 0.001), service-delivery readiness (beta = 0.43, p < 0.001), and retention intention (beta = 0.27, p < 0.001). The AI-MKG contained 2,083 nodes and 5,734 edges; expert validation showed triple precision of 87.9%, relation recall of 75.8%, and F1 of 81.4%. In the pilot task, average curriculum-gap identification time decreased from 41.2 to 26.5 minutes, a 35.7% reduction. Conclusions: AI-MKGs may be most relevant to human resources for health when they function as workforce-development infrastructure rather than as isolated educational technology. By linking curriculum content, competency standards, job tasks, assessment evidence, and service-delivery needs, AI-MKGs could support curriculum renewal, skill-mix planning, deployment preparation, and governance coordination across education and health-service systems. The cross-sectional survey supports associative rather than causal interpretation, and future studies should use longitudinal, administrative, employer, and service-performance data.</p>