A simulation-based evaluation of nine oscillometric blood pressure monitors for self-measurement

OBJECTIVE: To evaluate the performance of nine self-measurement oscillometric blood pressure monitors using a simulator. METHODS: For each monitor, simulation data from 48 sets of simulated waveforms with four simulations for each set were used for analysis. The waveforms represent a wide range of blood pressure. The monitor-simulator blood pressure differences were analyzed according to the 1992 Association for the Advancement of Medical Instrumentation (AAMI) and 1993 British Hypertension Society (BHS) protocols, except that corrections were made to take into account simulation variability. One-way analysis of variance was used to compare the differences for various combinations of monitors. The monitors' heart rate readings were compared with the rated accuracy. RESULTS: First, the mean blood pressure differences in general vary from monitor to monitor, the absolute mean differences ranging from 1.2 to 18.2mmHg. This can be partly explained by the likely use of different blood pressure determination criteria. Second, the Omron HEM-711 and HEM-712C gave about the same mean difference for systolic pressure and for diastolic pressure, suggesting that the two monitors may be using the same or approximately the same set of determination criteria. Third, the quantitative assessments for some of the monitors, for systolic or diastolic pressure or both, satisfy the clinical-use accuracy criteria of the AAMI or BHS protocol or both. These assessments alone cannot, however, be used to conclude whether or not any of the monitors fulfils the accuracy requirements of either protocol. Fourth, the corrected standard deviations of the differences range from 1.5 to 16.6mmHg, most of them being substantially less than the 8mmHg limit stipulated in the AAMI protocol. The different standard deviations suggest a varying robustness of the signal-processing methods used by the monitors. Fifth, for 7 out of the 9 monitors, more than 94% of the heart rate readings fall within +/-5% of their reference readings. CONCLUSIONS: The nine monitors in general performed differently with respect to the simulator. The results cannot be used fully to reflect the monitors' performance on human subjects because they were not based on clinical evaluation. Studies that lead to a more realistic simulation of oscillometric blood pressure are needed. Protocols integrating both clinical and simulation-based evaluations need to be developed.