Sources of Error in Office Blood Pressure Measurement

Roy N. Morcos, MD, FAAFP; Kimbroe J. Carter, MD; Frank Castro, MS; Sumira Koirala, MD; Deepti Sharma, MD; Haroon Syed, MD


J Am Board Fam Med. 2019;32(5):732-738. 

In This Article


Terminal Digit Bias (TDB)

Observational Study. BPs recorded by 3 nurses using manual (aneroid) and automated devices were analyzed for TDB. BP measurements with manual and automated devices were obtained from patient charts and categorized by observer and type of device used. The manual device used in the study was the Welch Allyn CE0297 aneroid sphygmomanometer. The automated device was the Omron Digital BP Monitor, Model HEM-907 XL, which has been certified and used in several major hypertension studies.[12] Overall, 3000 BP terminal digit observations were evaluated. There were 250 systolic and 250 diastolic observations obtained by each of the 3 nurses using manual and automated devices. The frequency of terminal digits was calculated for both sets of data.

Statistical Analysis. Manual BP observations having terminal digits 0, 2, 4, 6, and 8 were analyzed using the χ 2 test for independence with 4 degrees of freedom. Automated BP observations having terminal digits 0 through 9 were analyzed using the χ 2 test for independence with 9 degrees of freedom. χ 2 tests were performed in Microsoft Excel for each nurse with P < .01 considered statistically significant.

Effects of Patient Positioning on BP Measurements

Randomized controlled trial. A randomized controlled trial ( Identifier: NCT03460249) was initially conducted to evaluate the effect of sequence of patient positioning on BP measurements, that is, the sequence of table followed by chair BP measurements versus chair followed by table BP measurements. Thirty patients were randomized, and the results were analyzed with χ 2 tests, finding no difference in the sequence. As a result, a table to chair sequence was adopted for the study.

Standard Deviation Estimate. To estimate the standard deviation of the automated device and observer, the BP of a healthy, nonhypertensive individual was repeatedly measured in the standard seated position. One hundred twenty BP measurements were obtained by the same observer over 2 days to minimize subject and operator fatigue. For the chair and table positions, the systolic BP standard deviations were 4.20 mm Hg and 4.33 mm Hg, respectively, and the diastolic BP standard deviations were 3.62 mm Hg and 4.26 mm Hg, respectively.

Patients. The study population consisted of patients who consecutively presented to a teaching family medicine center for a scheduled appointment. Adult patients, aged 18 years and older, were informed about the study and invited to participate. Exclusion criteria were patients who declined participation for any reason, those in significant pain or distress who may have been unable to complete the protocol, and those with limited mobility who may have had difficulty getting up to the examination table. The study was approved by a regional institutional review board for human subjects, and informed consent was obtained from the patients. There were a total of 1176 BP measurements from 294 patients. Of the 294 participants, 188 (63.9%) were female, 58 (19.7%) were diabetic, 141 (48.0%) were hypertensive, 106 (36.1%) were hyperlipidemic, 36 (12.2%) had cardiovascular disease, and 158 (53.7%) were current or formerly smokers. The median age was 50.5 years with the youngest being 18 years and the oldest, 90 years.

Data Collection. Four BP readings were obtained for each individual using an automated device, 2 in the table position followed by 2 in the chair position. BPs were measured with an automated device for most patients. Overall, 294 individuals agreed to participate and were tested. A manual device was used for patients with significant arrhythmia, such as atrial fibrillation and tachycardia, or if a measurement could not be obtained with the automated device.[13] The guidelines of the American Heart Association (AHA) for patient positioning and BP measurement technique were followed.[14] Before beginning the study, detailed instructions regarding proper BP measurement technique were provided to the nurses who obtained the data. A medical chart review was performed for all patients, and clinical data were obtained including age, sex, smoking status and presence of diabetes, cardiovascular disease, hypertension, and hyperlipidemia. Other than age, all collected data were binary, categorized as present or absent.

Classification of Hypertension. In accordance with the commonly used Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC-7) guideline definitions, a normal BP is below 120/80 mm Hg, prehypertension is a BP between 120 to 139 and 80 to 89 mm Hg, and hypertension is a BP equal to or above 140/90 mm Hg.[4] The average systolic and diastolic BP values in the 2 positions were calculated for each patient's readings, and the patient was classified as having normal BP, prehypertension, or hypertension. Classification was repeated using another guideline published by the American College of Cardiology (ACC)/AHA in 2017, in which normal BP is defined as below 120/80 mm Hg, elevated BP as 120 to 129 and below 80 mm Hg, and hypertension as 130/80 mm Hg or above.[12] Patients whose BPs showed significant lowering between average table and chair readings resulting in a change of classification from prehypertension (elevated BP) to normal or from hypertension to either prehypertension or normal were considered misclassified. For example, a patient with a table average reading of 128/87 mm Hg and a chair average reading of 118/78 mm Hg would be misclassified as having prehypertension according to the JNC-7 guideline.

Statistical Analysis. The difference between the average BP values from the table and chair positions was calculated for each patient's systolic and diastolic readings, and an unpaired t-test performed to analyze the data. The standard error of the mean of these differences was determined and a 2-sided 95% confidence upper bound for the standard error of the difference (SEd ) was based on the normal distribution's upper bound of 1.96 × SEd . Using the normal distribution assumption, this results in upper bounds of 8.36 mm Hg for systolic BP and 7.74 mm Hg for diastolic BP. Patient observed differences were compared against the confidence upper bound to identify significant systolic and diastolic BP changes due to patient positioning. For example, consider a patient with 2 systolic automated BP readings in the table position averaging 129 mm Hg and 2 systolic automated BP readings in the chair position averaging 119 mm Hg. Assuming an SEd of 4.26 mm Hg and an upper bound of 8.36 mm Hg, the observed 10 mm Hg difference in average systolic BP is considered significant since the degree of random variation from the device and observer is not expected to exceed 8.36 mm Hg for systolic BP.

Data from the observational study were processed using the Binary Logistic and Probit Regression function with significant BP lowering as the dependent variable. Clinical factors identified in the data set were used as independent variables. Logistic regression was performed using the Excel Add-in Real Statistics Resource Pack software (Release 4.3, to determine if any of these clinical factors could predict significant lowering of BP due to change in patient positioning. We also evaluated misclassification of hypertensive disease according to clinical data subgroups.