Blood pressure measurement error: its effect on cross-sectional and trend analyses.

The measurement of blood pressure in epidemiological studies is difficult to standardize between centres in multi-centre studies and between repeat surveys over time. The use of standard mercury sphygmomanometers is common but especially prone to measurement error in terms of departure from the protocol and variation in measurement technique. Data from Australia's cardiovascular risk factor prevalence surveys on 21 independent populations, distributed geographically and temporally, has been examined to assess the effect of these errors on cross-sectional and trend analyses. The examination showed that last digit preference for zero may inflate estimates of proportions having high blood pressure. A tendency to record identical duplicate measurements could contribute 0.85 mmHg to time trends or geographic differences in mean systolic blood pressure (but not diastolic blood pressure). Epidemiological studies for geographic and trend differentials in systolic blood pressure need to be mindful of these effects in their analysis. There was some evidence of deterioration in data quality during data collection but no evidence that observers were influenced in their recording practice by observable respondents' characteristics. Training procedures for blood pressure measurement are of critical importance and adherence to the measurement protocol should be continuously monitored during data collection to ensure comparability of results.