The Development, Validation and Application of Policy-Focused Cardiovascular Risk Equations: The Vascular Risk in Adult New Zealanders (VARIANZ) Studies

Abstract

Cardiovascular disease (CVD) risk prediction equations have traditionally been clinically- focused and not ideally suited to inform population-wide health policy. The aim of this thesis was to determine whether policy-focused CVD risk equations could be developed, validated and applied in Aotearoa New Zealand. ‘Policy-focused’ equations in this thesis were defined as equations that can inform health planning for populations at a national, regional or organisational level, to reduce the burden of CVD. The thesis comprised five data linkage studies: i) construction of a novel entire-population cohort; ii) investigation of the importance of CVD preventive pharmacotherapy changes for risk modelling; iii) development and validation of new policy-focused CVD risk equations; and iv) two applications of the new equations. Anonymised individual-level linkage of eight national routinely collected health databases identified a cohort of 2,543,577 New Zealanders (85%) aged ≥20 years in 2006 with five years follow-up of CVD hospitalisations, deaths and pharmacotherapy dispensing. Among around 1.75 million 30-74 year-olds without CVD from this 2006 cohort, one in five were dispensed baseline CVD preventive pharmacotherapy with only modest medication changes subsequently observed over five years. Sex-specific equations estimating 5-year risk of CVD hospitalisation or death were developed with Cox regression from this primary prevention sub-cohort using eight pre-specified predictors routinely collected for all New Zealanders. Accurate calibration and risk discrimination were observed in national, regional, ethnic and age-specific populations, in deprivation quintiles and according to baseline medication dispensing. The first application of the equations examined their performance among people aged 75-89 years without CVD in 2006. Progressively worsening calibration and discrimination of the new equations were demonstrated with increasing age. The second application involved the development of a new 2013 national primary prevention cohort of 2.25 million 30-74 year-olds from the Statistics New Zealand Integrated Data Infrastructure, and investigation of medication management according to CVD risk. CVD pharmacotherapy was strongly associated with predicted CVD risk and, once commenced, was generally continued. However, only half of high-risk individuals received recommended dual therapy. This doctoral research demonstrated that relatively rich individual-level CVD-related data can be quickly and cheaply identified from administrative health databases and used to develop robust policy-focused CVD risk equations. The application of these equations to almost every New Zealander in specified subpopulations indicates that CVD risk equations derived from primarily middle-aged or early-retirement cohorts – as is standard global practice – do not perform well among people over 75 years. Furthermore, although CVD preventive pharmacotherapy strongly tracks CVD risk, there is still a large treatment gap among those at highest risk. Similar policy-focused CVD risk equations could be developed in the increasing number of countries with linkable national databases and used to identify quality improvement opportunities for entire populations.