An Estimation Framework for Cuff-based Blood Pressure Measurement: Analysis and Correction of Systematic Biases

Abstract Objective: Modern digital oscillometry and traditional auscultation-based blood pressure (BP) measurement share a similar cuff-based occlusive technique to sense BP: an external arm cuff is inflated above arterial pressure and then deflated while pressure changes due to arterial reopening and pulsation are detected and used to estimate BP. This process introduces inherent biases that can affect clinical assessments. This study models and characterizes these biases and develops an estimation-theoretic framework to mitigate them. Approach: Using the MIMIC database, we examine two major sources of BP measurement error: (1) systematic biases introduced by cuff-based occlusive measurement, and (2) respiratory-driven modulation of BP. Using a model of cuff–artery pressure interaction, we characterize errors in systolic BP (SBP) and diastolic BP (DBP) relative to a-line measurements, and analyze the association between respiration and short-term BP variability. We then evaluate least squares (LS), maximum likelihood (ML), and Bayesian estimators for single and multiple BP measurements using leave-one-subject-out cross-validation. Main Results: Simulated cuff-based measurements underestimated SBP by 9.7±7.8 mmHg and overestimated DBP by 6.6±4.8 mmHg on average. The proposed estimation frameworks substantially improved BP estimation errors in the simulated measurements. LS reduced estimation variability but retained the systematic bias, whereas ML corrected this bias, producing mean errors close to zero. Bayesian estimation further reduced mean absolute error (MAE) for most BP metrics. With three measurements, LS achieved MAEs of 9.7±5.7 mmHg and 6.6±3.6 mmHg for SBP and DBP, respectively, compared with 4.1±4.0 mmHg and 2.8±2.3 mmHg using ML. Bayesian estimation further reduced the MAEs to 3.6±4.4 mmHg and 1.9±2.1 mmHg, respectively. Significance: Cuff-based occlusive BP measurements are subject to inherent systematic biases. Modeling and correcting their systematic measurement errors can improve the accuracy and reliability of cuff-based noninvasive BP estimation, particularly when combined with repeated measurements and prior statistical information.