OBJECTIVE: To develop and recalibrate an Australian 5-year cardiovascular disease (CVD) mortality risk score to produce contemporary predictions of risk. METHODS: Data were pooled from six Australian cohort studies (n = 54,829), with baseline data collected between 1989 and 2003. Participants included were aged 40-74 years and free of CVD at baseline. Variables were harmonised across studies and missing data were imputed using multiple imputation. Cox proportional hazards models were used to estimate the risk of CVD mortality associated with factors mutually independently predictive (p < 0.05) and a 5-year risk prediction algorithm was constructed. This algorithm was recalibrated to reflect contemporary national levels of CVD mortality and risk factors using national statistics. RESULTS: Over a mean 16.6 years follow-up, 1375 participants in the six studies died from CVD. The prediction model included age, sex, smoking, diabetes, systolic blood pressure, total and high-density lipoprotein cholesterol (HDLC), a social deprivation score, estimated glomerular filtration rate and its square and interactions of sex with diabetes, HDLC and deprivation score, and of age with systolic blood pressure and smoking. This model discriminated well when applied to a Scottish study population (c-statistic (95% confidence interval): 0.751 (0.709, 0.793)). Recalibration generally increased estimated risks, but well below those predicted by the European SCORE models. CONCLUSIONS: The resulting risk score, which includes markers of both chronic kidney disease and socioeconomic deprivation, is the first CVD mortality risk prediction tool for Australia to be derived using Australian data. The primary model, and the method of recalibration, is applicable elsewhere.
AD - Baker IDI Heart and Diabetes Institute, Melbourne, Australia. kathryn.backholer@deakin.edu.au.OBJECTIVE: To develop and recalibrate an Australian 5-year cardiovascular disease (CVD) mortality risk score to produce contemporary predictions of risk. METHODS: Data were pooled from six Australian cohort studies (n = 54,829), with baseline data collected between 1989 and 2003. Participants included were aged 40-74 years and free of CVD at baseline. Variables were harmonised across studies and missing data were imputed using multiple imputation. Cox proportional hazards models were used to estimate the risk of CVD mortality associated with factors mutually independently predictive (p < 0.05) and a 5-year risk prediction algorithm was constructed. This algorithm was recalibrated to reflect contemporary national levels of CVD mortality and risk factors using national statistics. RESULTS: Over a mean 16.6 years follow-up, 1375 participants in the six studies died from CVD. The prediction model included age, sex, smoking, diabetes, systolic blood pressure, total and high-density lipoprotein cholesterol (HDLC), a social deprivation score, estimated glomerular filtration rate and its square and interactions of sex with diabetes, HDLC and deprivation score, and of age with systolic blood pressure and smoking. This model discriminated well when applied to a Scottish study population (c-statistic (95% confidence interval): 0.751 (0.709, 0.793)). Recalibration generally increased estimated risks, but well below those predicted by the European SCORE models. CONCLUSIONS: The resulting risk score, which includes markers of both chronic kidney disease and socioeconomic deprivation, is the first CVD mortality risk prediction tool for Australia to be derived using Australian data. The primary model, and the method of recalibration, is applicable elsewhere.
PY - 2017 SN - 1471-2261 (Electronic)