Development and validation of the co-designed social frailty 10-Item screening tool (SF-10)

Social frailty is a multidimensional determinant of adverse outcomes in later life, yet existing measures often lack clear conceptual foundations, cultural relevance, and psychometric rigour.

To develop and evaluate the psychometric properties of a co-designed social frailty assessment tool and its short forms for community-dwelling older adults, and to establish the SF-10 as a brief screening instrument.

Cross-sectional psychometric study.

Using a structured three-phase scale development framework, items were generated deductively from the social needs framework and social vulnerability concepts, informed by a scoping review and co-design workshops with older adults, caregivers, and professionals in Hong Kong. Content validity was refined through stakeholder workshops and cognitive interviewing. The initial 15-item Social Frailty Index (SF-15) was administered to 234 adults aged ≥60 years. Confirmatory factor analysis within a Bayesian bifact or structural equation model evaluated fit, latent structure, and reliability. A 10-item short form (SF-10) was then developed and tested for internal consistency, structural validity, and domain coverage.

The SF-15 aligned with the proposed dimensions but showed poor fit (PPp = 0.001). The SF-10 demonstrated acceptable fit (PPp = 0.216), excellent composite reliability (ω = 0.913), strong internal consistency (α = 0.824; ordinal α = 0.869), moderate reliable variance attributable to the general factor (ωH = 0.625), and high factor score quality (ρη^,η = 0.94). Retained items covered resources, participation, connections, relationships, and self-management. Subscale scores may support domain-specific profiling, though interpretation requires further validation.

The SF-10 is a theoretically grounded, culturally sensitive brief screening tool with promising psychometric properties among Hong Kong Chinese older adults. Its total score can identify overall social vulnerability, while domain-level data may guide tailored assessment, referral, and intervention. Independent, longitudinal, and cross-cultural validation is needed before clinical use.