Abstract:
Most cardiovascular disease (CVD) risk prediction equations used today were developed from data collected decades ago in populations at higher CVD risk, but less ethnic and socioeconomic diversity, than patients to whom they are now applied. These equations commonly overestimate risk in contemporary populations and underperform when applied in sub-populations. Between 2002 and 2015, the PREDICT cohort study collected CVD risk profiles on a large, representative New Zealand primary care population, to develop CVD risk prediction equations for general populations and a range of sub-populations. Data were recorded using a web-based clinical decision support system in routine practice and linked to outcomes using an encrypted national health identifier. The aim of the research presented in this thesis was to develop new sex-specific models for predicting CVD risk, in the general population (PREDICT-1o models) and a type 2 diabetes (T2D) sub-population (PREDICT-1o T2D models). The new models were also compared to determine whether the performance of the T2Dspecific models justified their development.
