Statistical analysis of blood pressure measurement errors by oscillometry during surgical operations

OBJECTIVES: Although a validated oscillometry sphygmomanometer satisfies the accuracy criteria of Advancement of Medical Instrumentation (AAMI), its long-term blood pressure (BP) measurement error during operations remains to be determined. We aim to (a) compare the error range throughout surgical operations with the accuracy criteria of AAMI, and (b) investigate the probabilities of occurrence of abnormal, large errors and clinically meaningful errors. METHODS: BP level were measured from 270 participants using oscillometry and arterial cannulation (invasive method) in the same BP monitor throughout surgeries. Mean deviation and SD (oscillometry vs. invasive method) were calculated from 6640 sets of data and presented in the Bland-Altman Plots. Also, the average, the largest, and the smallest measurement errors (errormean, errormax, and errormin) per patient were obtained. The probability distributions of the three types of errors were shown using histograms (percentage vs. SD). In addition, the clinically meaningful large errors (≥ 10 mmHg) of the adult patients when their systolic blood pressure (SBP) values were around 90 mmHg were investigated. RESULTS: The mean deviation (1.98 mmHg for SBP and 4.31 mmHg for diastolic blood pressure (DBP) satisfies the AAMI criterion (≤ 5 mmHg), but the SD (14.87 mmHg for SBP and 11.21 mmHg for DBP) exceeds the AAMI criterion (≤ 8 mmHg). The probability of errormax more than 40 mmHg is 14% for SBP and 6% for DBP. The probability of errormean more than 24 mmHg (4.07% for SBP and 1.48% for DBP), and that of errormin more than 24 mmHg (0.37% for SBP and 0.37% for DBP) are all greater than the criterion of 0.26%. The clinically meaningful errors are found in 28.78% of the adult patients. CONCLUSION: The SD of long-term BP measurement by our oscillometric method during operations exceeds AAMI accuracy criteria. And it is important to be aware of the abnormal large errors and clinically meaningful errors as their probabilities are rather significant. We analyze the possible cause of the major errors and deduce that similar errors may also happen with other oscillometric devices.