RESUMO
OBJECTIVE: To evaluate the validity of blood pressure measurement by studying the presence of preference bias in last digits and also of the bias which tends to situate measurements below the established limit. To evaluate the interobserver (doctor-nurse) reliability of the measurements. DESIGN: A cross-sectional [correction of crossover] observation study with descriptive and analytic components. SETTING: A nursing station in the Health Centre. PATIENTS: The patients who attended the nursing station by appointment during the months of September and October 1993. MEASUREMENTS AND MAIN RESULTS: The patients who attended the nursing station by appointment had their blood pressure measured by a doctor or nurse in a random fashion. We detected a preference in last digits, which was very significant for both systolic and diastolic pressure (p < 0.0001). The tendency to reduce diastolic figures to below the preestablished limit of pressure was also very significant (p < 0.05). The nurse found lower figures than the doctor, with a very significant difference for diastolic pressure (p < 0.001). The index of kappa concordance between nurse and doctor on evaluating whether pressure was high or not was 0.4864, with a C.I. of 95%, from 0.3530 to 0.6198. CONCLUSIONS: The existence of two important biases in the measurements of pressure questions their validity. Additionally the significant difference between the pressure averages and the low concordance poses a serious question over the interobserver reliability and should make us restate blood pressure measurement and the treatment of patients suffering Hypertension.