Quality Assessment of Data on Blood Pressure in the WHO MONICA Project

Appendix 2. Estimation of bias due to cuff size

(See Section 4.3.)

The bias (in mmHg) of systolic and diastolic readings for adult blood pressure cuffs of bladder widths 12, 15 and 18 cm at arm circumferences from 26 cm to 50 cm were tabulated in reference (6). (More precisely, the reference tabulates the correction required to remove the bias, but the bias is the opposite number.) In order to facilitate interpolation and extrapolation for cuff widths and arm circumferences not tabulated and to smoothen the biases for the rounding to full mmHg, which was used in the reference, we fitted a regression surface on the biases tabulated. The simplest regression model with a good fit was:

bias = b₀ + b₁ arm + b₂ cuff + b₃ cuff² + b₄ arm × cuff + b₅ arm × cuff² + e,

where

arm is the individual's arm circumference,
cuff is the width of the bladder of the cuff,
e is the model error term, and
the coefficients for systolic and diastolic blood pressure are:

The bias of the population mean value was estimated as the mean of the estimates of the individual biases obtained using the regression model.

The main reservations on these bias estimates are that we are not aware of the details of the way  the bias estimates of reference (6) were obtained, and our estimates for cuffs narrower than 12 cm are based on extrapolation. Nevertheless, we believe that our estimates give a good idea of the magnitude of biases related to the use of different cuff sizes.