Cardiovascular medication changes over 5 years in a national data linkage study: implications for risk prediction models

Suneela Mehta,1 Rod Jackson,1 Sue Wells,1 Jeff Harrison,2 Daniel J Exeter,1 Andrew J Kerr1,3 1Section of Epidemiology and Biostatistics, 2School of Pharmacy, University of Auckland, 3Cardiology Services, Middlemore Hospital, Auckland, New Zealand Background: Despite widespread use of cardiovascular disease (CVD) preventive medications in cohorts used to develop CVD risk prediction models, only some incorporate baseline CVD pharmacotherapy and none account for treatment changes during study follow-up. Therefore, current risk prediction scores may underestimate the true CVD event risk. We examined changes in CVD pharmacotherapy over 5 years in preparation for developing new 5-year risk prediction models.Methods: Anonymized individual-level linkage of eight national administrative health datasets enabled identification of all New Zealanders aged 30–74 years, without prior hospitalization for CVD or heart failure, who utilized publicly funded health services during 2006. We determined proportions of participants dispensed blood pressure lowering, lipid lowering, and antiplatelet/anticoagulant pharmacotherapy at baseline in 2006, and the proportion of person years of follow-up (2007–2011) where dispensing occurred.Results: The study population comprised of 1,766,584 individuals, representing ~85% of all New Zealanders aged 30–74 years without prior CVD or heart failure in 2006, with mean follow-up of 4.9 years (standard deviation 0.6 years; 8,589,931 total person years). CVD medications were dispensed to 21% of people at baseline, with most single or combination pharmacotherapies continuing for ≥80% of follow-up. Complete discontinuation of baseline treatment accounted for 2% of follow-up time while CVD pharmacotherapy that commenced after baseline accounted for 7% of total follow-up time. Conclusion: In a national primary prevention cohort of 30–74 year olds, one in five received baseline CVD primary preventive pharmacotherapy and medication changes over the subsequent 5 years were modest. Baseline medication use is an important consideration when estimating CVD risk from modern cohorts. It is currently unclear how to incorporate available methods to account for treatment changes during follow-up into risk prediction scores, but this study demonstrates that baseline therapy captures most of the effect of treatment in 5-year risk models. However, the impact of treatment changes on the more common 10-year risk models requires further investigation. Keywords: cardiovascular diseases, primary prevention, drug therapy, routine data, record linkage&nbsp