Background: We have previously developed a score for predicting cardiovascular events in the intermediate term in an elderly hypertensive population. In this study, we aimed to extend this work to predict 10-year cardiovascular and all-cause mortality in the hypertensive aged population.

Methods: Ten-year follow-up data of 5,378 hypertensive participants in the Second Australian National Blood Pressure study who were aged 65-84 years at baseline (1995-2001) and without prior cardiovascular events were analyzed. By using bootstrap resampling variable selection methods and comparing the Akaike and Bayesian information criterion and C-indices of the potential models, optimal and parsimonious multivariable Cox proportional hazards models were developed to predict 10-year cardiovascular and all-cause mortality. The models were validated using bootstrap validation method internally and using the Dubbo Study dataset externally.

Results: The final model for cardiovascular mortality included detrimental (age, smoking, diabetes, waist-hip ratio, and disadvantaged socioeconomic status) and protective factors (female sex, alcohol consumption, and physical activity). The final model for all-cause mortality also included detrimental (age, smoking, random blood glucose, and disadvantaged socioeconomic status) and protective factors (female sex, alcohol consumption, body mass index, and statin use). Blood pressure did not appear in either model in this patient group. The C-statistics for internal validation were 0.707 (cardiovascular mortality) and 0.678 (all-cause mortality), and for external validation were 0.729 (cardiovascular mortality) and 0.772 (all-cause mortality).

Conclusions: These algorithms allow reliable estimation of 10-year risk of cardiovascular and all-cause mortality for hypertensive aged individuals.