The effect of alternative criteria for hypertension on estimates of prevalence and control

OBJECTIVES: To assess the impact of various criteria used to define hypertension in community surveys on estimates of prevalence, treatment and control. In particular, this paper examines the effect of using mean versus minimum values; one, two or three examinations; and three different blood pressure levels. DESIGN: A cross-sectional community survey. METHODS: A multistage area sample of households in Hamilton, Canada was selected, yielding 2770 potential interviewees. Data were obtained from 2140 people (77.2%). About 25% of the sample satisfied the criteria for either one or two follow-up visits (95% completion rate). Estimates of the prevalence of hypertension and its control were computed using 28 different criteria to define hypertension. Estimation methods employed analyses that adjusted for each individual respondent's sampling probability and the effect of area sampling on variance estimates. RESULTS: The selection of mean or minimum readings had little impact on the estimates. Prevalence estimates decreased by up to 20% when follow-up information was included but were similar under all three of the studied blood pressure cut-off points. Inclusion of the follow-up information reduced the proportion of hypertensives estimated to be 'unaware' of their condition by over 60% while raising the proportion 'under control' by around 18%. Application of the Third National Health and Nutrition Examination Survey analysis criteria to the present study demonstrated that alterations in criteria can have profound effects on estimates, the prevalence increasing by about 100% and the proportion 'unaware' by 500%. The proportion 'under control' dropped from 69.0 to 21.5%. CONCLUSIONS: Different criteria to define hypertension can have important effects on the estimates of prevalence and control. Authors need to be explicit concerning the criteria used. Readers should be aware of the risk of overinterpreting results based on criteria that do not reflect their objectives (e.g. using a single visit estimate to determine control of clinically relevant hypertension).