Analytical approach:
The analysis was based on a Markov model, with a lifetime horizon and 10 cohorts of men or women aged 40, 50, 60, 70, or 75 years. The authors stated that it was carried out from the perspective of the UK NHS and personal social services.
Effectiveness data:
The clinical data were from a selection of published sources. The specificity of blood pressure monitoring for each option was a key input for the model and the data were from a recent meta-analysis. The treatment effect for antihypertensive drugs was from a meta-analysis of clinical trials. The long-term risk of cardiovascular events due to hypertension was derived using Framingham risk equations and data from the Health Survey for England (HSE) 2006 and other published sources. Some assumptions were needed; for example, the accuracy of continuous blood pressure monitoring was assumed to be 100%.
Monetary benefit and utility valuations:
The utility values for cardiovascular conditions were from a published study. Those for the general population without a cardiovascular event were age and gender specific and were from the HSE, which collected them using the European Quality of life (EQ-5D) instrument.
Measure of benefit:
Quality-adjusted life-years (QALYs) were the summary benefit measure and were discounted at an annual rate of 3.5%.
Cost data:
The economic analysis included the costs of diagnosis (equipment, consumables, maintenance and staff time), antihypertensive treatment (drugs and annual clinical review), and management of cardiovascular disease (CVD). The costs of CVD were from published reports and national estimates. Most of the other costs were from official NHS sources. The drug costs were based on the most commonly used generic drugs, and the proportion of people on one, two, or three drugs, from the HSE. All costs were in UK £ and were at 2009 to 2010 prices. A 3.5% annual discount rate was applied.
Analysis of uncertainty:
The overall uncertainty was investigated, in a probabilistic sensitivity analysis, using Monte Carlo simulation. Each model input was assigned a distribution based on its point estimate and standard error. Deterministic sensitivity analyses were performed to identify the most influential inputs.