Context: The prevalence of hypertension in developing countries is coming closer to values found in developed countries. However, surveys usually rely on readings taken at a single visit, the option to implement the diagnosis on readings taken at multiple visits, being limited by costs. Objective: To estimate more accurately the magnitude and extent of the resource that should be allocated to the prevention of hypertension. Design: Population-based cross-sectional survey with triplicate blood pressure (BP) readings taken on two separate home-visits. Setting: Rural and urban locations in three areas of Yemen (capital, inland and coast). Participants: A nationally representative sample of the Yemen population aged 15-69 years (5063 men and 5179 women), with an overall response rate of 92% in urban and 94% in rural locations. Main outcome measure: Hypertension diagnosed as systolic BP >= 140 mm Hg and/or diastolic BP >= 90 mm Hg and/or self-reported use of antihypertensive drugs. Results: Hypertension prevalence (age-standardised to the WHO world population 2001) based on fulfilling the same criteria on both visits (11.3%; 95% Cl 10.7% to 11.9%), was 35% lower than estimation based on the first visit (17.3%; 16.5% to 18.0%). Advanced age, blood glucose >= 7 mmol/l or proteinuria >= 1+ at dipstick test at visit one were significant predictors of confirmation at visit 2. The 959 participants found to be hypertensive at visit 1 or at visit 2 only and thus excluded from the final diagnosis had a rate of proteinuria (5.0%; 3.8% to 6.5%) comparable to rates of the general population (6.1%; 5.6% to 6.6%), and of subjects normotensive at both visits (5.6%; 5.1% to 6.2%). Only 1.9% of Yemen population classified at high or very high cardiovascular (CV) risk at visit 1 moved to average, low or moderate CV risk categories after two visits. Conclusions: Hypertension prevalence based on readings obtained after two visits is 35% lower than estimation based on the first visit, subjects were excluded from final diagnosis belonging to low CV risk classes.