Development and external validation of a mortality prediction model for community-dwelling adults aged ≥50 years with frailty or pre-frailty.

BACKGROUND: Frailty and pre-frailty raise mortality risk in older adults, yet clinicians lack tailored prognostic tools. We developed and externally validated a model predicting all-cause mortality in community-dwelling adults aged ≥50 years with frailty or pre-frailty.

METHODS: Using SHARE 2006-07 data (n = 15,195), we fitted a Cox model with backward selection and multiple imputation. Predictors spanned demographics, lifestyle, ADL/IADL limitations, mobility, pain and key comorbidities. We assessed discrimination by integrated and time-specific AUCs and calibration by comparing predicted versus observed mortality. External validation occurred in ELSA (n = 4464) and HRS (n = 7414).

RESULTS: Within the SHARE cohort (61.0% female; median follow-up 3.8 years), 18.5% (n = 2811) of participants died. The final model incorporated age, sex, BMI, smoking status, IADL limitations, mobility impairments, physical activity, pain, and major comorbidities. Internal validation demonstrated robust discrimination (optimism-corrected iAUC 0.81, 95% CI 0.79-0.82), with sustained accuracy at 1-, 5-, and 7-year intervals (AUCs 0.80-0.83). External validation in ELSA (iAUC 0.83) and HRS (iAUC 0.82) confirmed strong generalizability, alongside excellent calibration across all time points.

CONCLUSIONS: We present a rigorously validated prognostic model for mortality in frail and pre-frail older adults, demonstrating high accuracy and clinical utility. By providing personalized risk estimates, this tool can enhance shared decision-making, optimize care planning, and improve outcomes for a growing at-risk population.