Shorter-term and longer-term mortality prediction in the Australian Diabetes, Obesity and Lifestyle (AusDiab) study

Background: While identification of key risk factors for mortality has contributed to advances in healthcare, the effect of these risk factors in predicting mortality over different time horizons remains unclear. Aims: We sought to determine how risk factors predicted shorter-term and longer-term mortality across the age spectrum in adults. 

Methods: We used data from 11 247 adults of the Australian Diabetes, Obesity and Lifestyle (AusDiab) study. Cox multivariable regression models were used to estimate hazard ratios (HRs) of shorter-term (0–10 years) and longer-term (10–20 years) all-cause and cardiovascular disease (CVD)-related mortality associated with risk factors. Models with interaction between baseline age and each risk factor were also fitted. 

Results: During a 20-year follow-up, 2185 deaths occurred. Smoking, diabetes, male sex and albuminuria all independently predicted shorter- and longer-term all-cause and CVD mortality. Most associations were stronger in the shorter term compared to the longer term. A notable exception was the association between smoking and CVD mortality, which was stronger in the longer term (HR 3.55, 95% confidence interval (CI) 2.57–4.90) compared to the shorter term (HR 2.06, 95% CI 1.33–3.20). The magnitude of association between most risk factors and mortality attenuated with age. Conclusions: Classical risk factors for total and CVD mortality remain important up to 20 years after their measurement. In unselected adult cohorts, longer-term follow-up (e.g. beyond 10 years) may not provide additional information on associations of risk factors with mortality beyond that obtained in shorter-term follow-up.