Stroke prevalence surveys are more and more needed for health care and facility planning. Prevalence estimates and costs of the definition procedure may vary depending on different screening strategies. We evaluated the impact of these different strategies on the overall diagnostic procedure and on stroke prevalence estimates in the Italian Longitudinal Study on Aging. A population sample of 5, 632 individuals aged 65-84 years was screened for stroke by a simple question on previous stroke diagnosis, questions on possible stroke symptoms and a simple neurological examination. Those screened positive by any of these procedures were fully examined by a neurologist for conclusive diagnosis. We determined the positive predictive value of each procedure on the final stroke diagnosis and calculated prevalence as if each procedure had been used separately. Using the three procedures combined, the prevalence rate was 6.0% (95% confidence interval, 5.4-6.7%). If each procedure had been used as the unique screening tool, the rates would have been 5.1% (4.5-5. 7%), 4.1% (3.6-4.7%) and 2.3% (1.9-2.7%), and positive predictive values 66.4, 55.2 and 45.1%, respectively. Different screening procedures can affect stroke prevalence estimates. Compared to more complex screening strategies, the use of a simple question about previous diagnosis as the unique screening tool leads to only a slight underestimation of stroke prevalence and avoids a 66% increase in the number of subjects to be examined in a second-level specialist evaluation, potentially reducing the costs of the overall diagnostic procedure.