Terminal Digit Bias Is Not An Issue For Properly Trained Healthcare Personnel Using Manual or Semi-Automated Devices

摘要

The objective of this study was to evaluate terminal digit preference in blood pressure (BP) measurements taken from a sample of clinics at a large academic health sciences center. We hypothesized that terminal digit preference would occur more frequently in BP measurements taken with manual mercury sphygmomanometry compared to those obtained with semi-automated instruments. A total of 1,393 BP measures were obtained in 16 ambulatory and inpatient sites by personnel using both mercury (n= 1,286) and semi-automated (n=107) devices For the semi-automated devices, a trained observer repeated the patient's BP following American Heart Association recommendations using a similar device with a known calibration history. At least two recorded systolic and diastolic blood pressures (average of two or more readings for each) were obtained for all manual mercury readings. Data were evaluated using descriptive statistics and Chi square as appropriate (SPSS software, 17.0). Overall, zero and other terminal digit preference was observed more frequently in systolic (X~2 = 883.21, df = 9, p < 0.001) and diastolic readings (X~2 = 1076.77, df = 9, p < 0.001) from manual instruments, while all end digits obtained by clinic staff using semi-automated devices were more evenly distributed (X~2 = 8.23, df = 9, p = 0.511 for systolic and X~2 = 10.48, df = 9, p = 0.313 for diastolic). In addition to zero digit bias in mercury readings, even numbers were reported with significantly higher frequency than odd numbers. There was no detectable digit preference observed when examining semi-automated measurements by clinic staff or device type for either systolic or diastolic BP measures. These findings demonstrate that terminal digit preference was more likely to occur with manual mercury sphygmomanometry. This phenomenon was most likely the result of mercury column graduation in 2 mm Hg increments producing a higher than expected frequency of even digits.