Abstract
Oscillometric cuff devices are believed to use fixed ratio algorithms to compute blood pressure (BP) from the oscillogram (cuff pressure oscillation height versus applied cuff pressure function). However, variations in the cuff transducer that occur each time the cuff is wrapped on a given arm and variations in pulse pressure can alter the oscillogram shape and the ratios for computing BP. A popular validated home BP monitor was studied to determine whether and how it considers these effects. The monitor, with its Universal cuff placed on four different mandrels, was analyzed using a non-invasive BP simulator at various settings and an external pressure sensor to construct the oscillograms. Several algorithms were tested to estimate the home monitor BP values. A fixed ratio algorithm estimated the home monitor systolic and diastolic BP with errors of 5.8 and 1.5 mmHg. A variable ratio algorithm in which the ratios go inversely with the maximum oscillogram amplitude estimated the home monitor values with errors of just 1.5 and 0.8 mmHg. Using the maximum oscillogram amplitude may be particularly helpful for mitigating the variable pulse pressure effect. This study shows how oscillometric cuff devices likely work, which is important for understanding and improving their accuracy.
Supplementary Information
The online version contains supplementary material available at 10.1038/s41598-025-18850-w.
Keywords: Automatic cuff device, Blood pressure computation, Fixed ratio algorithm, Maximum oscillogram amplitude, Oscillometry, Variable ratio algorithm
Subject terms: Risk factors, Biomedical engineering
Introduction
Oscillometric cuff devices provide automatic measurement of brachial artery blood pressure (BP). These devices are widely used in office, home, and ambulatory settings for hypertension management and research as well as in the hospital for vital sign monitoring. Although the devices are of proven value in the field of hypertension, it is known that they are not very accurate, especially in unusual individuals such as those with arrhythmias, large arm circumference, and wide pulse pressure (PP)1,2. There has been substantial, very recent interest on identifying the error sources of oscillometric cuff devices including measurement artifact, hydrostatic pressure changes, and cuff size with insights into mitigating these types of errors3–5.
Figure 1 illustrates the conventional oscillometric principle. An inflatable cuff is wrapped around the upper arm. The cuff is rapidly inflated to occlude the brachial artery and then slowly deflated. The air pressure inside the cuff is measured to indicate the pressure applied to the artery (via its slow component) and the arterial blood volume oscillations underneath the cuff during deflation (via its fast component). Since the relationship from transmural pressure (internal BP – external applied cuff pressure) to blood volume in arteries is sigmoidal, the cuff pressure oscillations rise to a maximum and then fall with the declining applied cuff pressure. BP is computed from the cuff pressure measurement typically via a proprietary algorithm that is likely different amongst manufacturers6–8. It is often believed that a fixed ratio algorithm is employed6–8. This algorithm computes systolic BP (SBP) and diastolic BP (DBP) as the cuff pressure at which the oscillation height (ΔO) versus applied cuff pressure (Pc) function (“oscillogram”) is some fixed ratio of its maximum value (e.g., Rs = ΔOs/ΔOm = 0.6 and Rd = ΔOd/ΔOm = 0.75 in Fig. 1).
Fig. 1.
The conventional oscillometric cuff blood pressure (BP) measurement principle. The cuff is wrapped around the upper arm and rapidly inflated and then slowly deflated; the cuff pressure is measured; the oscillogram relating cuff pressure oscillation height (ΔO) to the applied cuff pressure (Pc) is constructed; and a proprietary algorithm is applied to the oscillogram to compute systolic BP (SBP) and diastolic BP (DBP). The presumptive algorithm is the fixed ratio algorithm in which SBP and DBP are each computed as the cuff pressure (PRs or PRd) at which the oscillogram amplitude is a ratio of its maximum value (Rs or Rd) that is fixed to a specific value. The main algorithm presented herein for a popular validated home BP monitor is a variable ratio algorithm in which each ratio is inversely related to the maximum oscillogram amplitude (ΔOm).
However, the cuff pressure-air volume relationship, which approximately transduces the arterial blood volume oscillations to the observed cuff pressure oscillations9, is nonlinear and can change with the cuff, the way the cuff is wrapped, and the characteristics of the arm (which, in turn, may be affected by position and temperature). This variable cuff transducer effect can alter the shape of the oscillogram and consequently the ratios for computing BP. Arterial stiffening can amplify the effect due to the wide PP. Figure 2A shows how the oscillogram shape and ratios can change with variations in the cuff pressure-air volume relationship despite fixed arterial properties and BP levels via simulations with a mathematical model of the cuff-arm-artery system (see Supplementary Methods). Furthermore, wide PP by itself can also alter the shape of the oscillogram and ratios for computing BP8,10,11. Figure 2B shows how the oscillogram shape and ratios can change with variations in PP but fixed arterial and cuff properties via simulations with the mathematical model (see Supplementary Methods).
Fig. 2.
Simulations with a mathematical model of the cuff-arm-artery system illustrating that the ratios for computing BP are not fixed. A) Variations in the cuff pressure-air volume relationship, which can occur each time the cuff is wrapped on a given arm, can alter the shape of the oscillogram and consequently the ratios for computing BP without any change in arterial properties and BP. The oscillograms in the middle panel are representative of an NIBP simulator, while the oscillograms in the lower panel are physiological. B) Variations in pulse pressure (PP) alone can also alter the oscillogram shape and ratios. See Supplementary Methods for details.
We analyzed a popular validated home monitor to determine whether and how it considers the variable cuff transducer and PP effects in the computation of SBP and DBP. We found that this home monitor uses the maximum oscillogram amplitude (see ΔOm in Fig. 1) to likely alter the ratios for computing SBP and DBP, which could help in mitigating both effects. Our study provides more information on how oscillometric cuff devices compute BP and is therefore important for proper interpretation and scientific scrutiny7 as well as for understanding and improving the accuracy of these devices.
Methods
We rigorously studied a validated home BP monitor (BP7100 3 Series, Omron Healthcare) with a Universal cuff (22–42 cm)12, as shown in Fig. 3A. We used the two-finger approach to properly wrap the cuff13 on four different mandrels, as shown in Fig. 3B. The mandrels were congruent with the cuff in terms of their size and included small, medium, and large rigid mandrels (ProSim NIBP Mandrel Set, Fluke Biomedical) and the small mandrel wrapped with Polyethylene foam padding (small-foam) to simulate arm tissue compression. We did not disturb the cuff wrapping on each mandrel throughout the study.
Fig. 3.
Investigation of a popular validated home BP monitor to determine whether and how it considers the variable cuff transducer and PP effects. A) The home BP monitor for study was the BP7100 3 Series (Omron Healthcare) with Universal cuff (22–42 cm). Photograph by authors, 2025. B) The cuff was properly wrapped around four different mandrels. C) The steady-state cuff pressure-air volume relationship of each mandrel was measured using a syringe system to inject known air volume and an NIBP simulator to measure the cuff pressure. D) The home monitor with each mandrel was analyzed using the NIBP simulator at various simulator settings and an external sensor to measure the cuff pressure and then construct the oscillogram. E) The measured steady-state cuff pressure-air volume relationships differed over the four mandrels. F) An example of the cuff pressure and cuff pressure oscillations via the validated home monitor for one mandrel and one simulator setting is shown. Medium mandrel and 170/80 mmHg/0.6 ml/75 bpm simulator setting.
We first measured the static cuff pressure-air volume relationship of the cuff on each mandrel, as shown in Fig. 3C. We increased the air volume in the cuff in steps of 10 ml using a syringe interfaced to the cuff via tubing with a one-way valve for air flow only in the direction toward the cuff. We recorded the steady-state cuff pressure for each step using the manometer of a non-invasive BP (NIBP) simulator (ProSim 8, Fluke Biomedical) until the cuff pressure reached 200 mmHg.
We then used tubing to interface the home BP monitor with the cuff on each mandrel to the NIBP simulator and a pressure sensor (DPM2Plus Universal Pressure Meter Tester, Fluke Biomedical) connected to a data acquisition unit (NI-6002, National Instruments) to record the cuff pressure, as shown in Fig. 3D. We set the simulator as follows: seven BP levels (90/60, 120/60, 110/70, 140/80, 130/90, 170/80, and 150/100 mmHg, which provided a reasonable sampling of the adult BP distribution in a general outpatient population14); three maximum volume oscillations (0.6, 0.9, 1.25 cc); and one heart rate (75 bpm). We recorded the SBP and DBP measurements of the home monitor and the cuff pressure for each simulator setting and each mandrel three times for a total of 252 (= 7 BP levels × 3 volumes × 4 mandrels × 3 repetitions) measurement sets.
We constructed the oscillogram for each of the 252 cuff pressure measurements. We first lowpass filtered the cuff pressure measurement with cutoff frequency of 0.5 Hz to obtain the applied cuff pressure and bandpass filtered the cuff pressure measurement with cutoff frequencies of 0.5 and 5 Hz to obtain the cuff pressure oscillations. We performed beat detection using a sliding window defined using the simulator heart rate. Within each window, we detected the local maxima and local minima subjected to a minimum amplitude threshold of 0.02 mmHg. We computed the oscillation height as the difference between the detected peaks and valleys and plotted it against the average of the applied cuff pressure over the oscillation. We next uniformly resampled the cuff pressure range into 100 samples and applied linear interpolation to determine the oscillation height for each of the samples. We then applied a five-sample unweighted moving average filter to obtain a smoothed oscillogram. We finally averaged the three repeated oscillograms as well as the three corresponding SBP and DBP values reported by the home monitor.
We started the analysis of the 84 averaged measurement pairs by examining a strategic subset of the data, which basically represented our preconceived notions, to ascertain candidate forms of the home monitor BP computation algorithm. We then accordingly applied various linear regression models to estimate the home monitor SBP and DBP values from all measurement pairs and thereby arrive at precise algorithms. This analysis included testing of the commonly presumed fixed ratio algorithm to estimate the home monitor SBP and DBP values.
We thereafter made more measurements to assess the generality of the findings. First, we employed two mandrels (small and small-foam) at three heart rates (50, 75, and 150 bpm) and two BP levels (120/60 and 170/80 mmHg) with the maximum volume oscillation fixed (0.6 ml). We thus obtained 12 additional averaged measurement pairs from the NIBP simulator. Second, we studied six healthy volunteers under approval from, and in accordance with the relevant guidelines and regulations of, the University of Pittsburgh IRB (STUDY20060267, 2020 –) and with written, informed consent from each volunteer. We obtained measurements from the home monitor and external pressure sensor with the arm cuff wrapped properly, tightly, and loosely on the volunteers for a total of 18 measurement pairs. We applied the previously determined algorithms to predict the home monitor BP values from these further measurements.
We lastly studied an unvalidated home BP monitor (Series 100, CVS), a validated ambulatory BP monitor (Oscar 2, Suntech Medical)12, and another validated home monitor (H-BP100SBP, Welch-Allyn)12. Based on our earlier findings, we strategically analyzed these BP monitors using the same experimental setup (see Fig. 3D) to efficiently glean some information on how they work.
Results
Figure 3E illustrates the cuff pressure-air volume relationships for the four mandrels (see Fig. 3B), which may be thought of as representing different arms. Each relationship was increasing concave up over the cuff pressure range and became nearly linear at higher cuff pressures. As expected, the four nonlinear relationships substantially differed. The small mandrel required the least air volume to reach a fixed cuff pressure, because the cuff volume was restricted by its smaller surface area. It therefore produced the largest slope or “cuff transducer gain” over its linear range. The small-foam mandrel required the most air volume and yielded the smallest gain and most nonlinear relationship (i.e., its local slope changed most with increasing cuff pressure) due to the compressible foam.
Figure 3F illustrates an example of the cuff pressure and cuff pressure oscillations for one mandrel at one simulator setting. (The cuff pressure inflation and deflation patterns are imposed by the validated home monitor (see Fig. 3A), and the NIBP simulator perturbs the cuff air volume in a prescribed manner to mimic arterial blood volume pulsations, which are transduced to cuff pressure oscillations.) The home monitor detects the cuff pressure oscillations in real time to minimize the measurement time. It uses a previous SBP measurement, if available, to determine the maximum cuff pressure for inflation and then checks for oscillations to determine if further inflation is required to achieve arterial occlusion. The monitor then linearly deflates the cuff at a rate of 4 mmHg/sec. The monitor rapidly deflates the cuff one beat after it determines DBP to end the measurement. Also, per the user’s manual, the heart rate range of the home monitor is 40–180 bpm.
Figure 4 presents the results of the home monitor from a strategic subset of the mandrels and NIBP simulator settings that we employed. The results from 4A to 4G progressively indicated candidate forms of its BP computation algorithm.
Fig. 4.
Measurements from a strategic subset of mandrels and simulator settings to gain insight into how the home monitor computes BP. A) Oscillograms, B) normalized oscillograms, and C) estimated home monitor SBP and DBP values via a fixed ratio algorithm (see Fig. 1) versus the respective measured home monitor values (along with the identity line) for one mandrel at simulator settings of seven BP levels but with the maximum volume oscillation and heart rate fixed. Large mandrel and 0.6 ml/75 bpm simulator setting. D) Normalized oscillograms, E) unnormalized oscillograms, and F) cuff pressure during inflation only versus time for the four mandrels at simulator settings of two BP levels but with the maximum volume oscillation and heart rate fixed. 0.6 ml/75 bpm simulator setting. G) Unnormalized oscillograms, normalized oscillograms, and cuff pressure during inflation for one mandrel at simulator settings of three maximum volume oscillations with the BP level and heart rate fixed. Small-foam mandrel and 110/70 mmHg/75 bpm simulator setting. H) Normalized and unnormalized (raw) cuff pressure oscillations (from which the oscillograms in G) were constructed) versus the applied cuff pressure.
Figure 4A illustrates the oscillograms for one mandrel at simulator settings of seven BP levels but with the maximum volume oscillation and heart rate fixed. The oscillograms differed in maximum amplitude positions (consistent with the differences in mean BP) and widths to the left and right of the maximum amplitude (consistent with the differences in DBP and SBP from the mean BP). Furthermore, these oscillograms increased and decreased in an approximately linear manner, which was consistent with our expectation that the NIBP simulator varies the volume oscillations linearly with cuff pressure. Therefore, these results suggested that the home monitor may be using ratios (rather than slopes) to compute BP. Figure 4B illustrates the same oscillograms but normalized to unity amplitude along with the home monitor SBP and DBP values. These data specifically suggested fixed ratios of 0.73 for DBP and 0.61 for SBP. Figure 4C shows the SBP and DBP values computed from the oscillograms via these fixed ratios versus the home monitor SBP and DBP values along with the line of identity. This plot shows more clearly than Fig. 4B that fixed ratios may be the algorithm used by the home monitor to compute SBP and DBP for this subset of measurements.
Figure 4D illustrates the normalized oscillograms for the four mandrels at simulator settings of two BP levels with the maximum volume oscillation and heart rate fixed. The normalized oscillograms were not identical over the mandrels due to the differing cuff pressure-air volume relationships (see Fig. 3E). For the small-foam mandrel, the normalized oscillograms were widest to the right of the maximum amplitude, yet the home monitor SBP values were lowest. Further, for the three rigid mandrels at the 130/90 mmHg setting, the normalized oscillograms were superimposable, yet the home monitor SBP values ranged from 123 to 130 mmHg. These results suggested that the home monitor is not simply applying fixed ratios but is rather using one or more other features to compute BP and that the feature(s) could not have come from the normalized oscillogram. The other feature(s) likewise could not have come from the oscillation shape (e.g., area or width) changes that also occur with cuff pressure 15, because the oscillation shape is maintained by the simulator (see Fig. 3F).
The other feature(s) could have come from the unnormalized oscillograms or the applied cuff pressure during inflation. Figure 4E illustrates the same oscillograms in Fig. 4D but without normalization. The maximum oscillogram amplitudes did differ substantially over the mandrels in line with the cuff transducer gains (see Fig. 3E). Figure 4F shows the corresponding cuff pressure during inflation. The cuff pressure inflation rate did also differ over the mandrels and was lower when more air volume was needed to reach a given cuff pressure as per the cuff pressure-air volume relationships (see Fig. 3E). In particular, the diastolic and systolic ratios progressively decreased with increasing maximum oscillogram amplitude and cuff pressure inflation rate. These results suggested that the home monitor is also using the maximum oscillogram amplitude and/or a feature of the cuff pressure during inflation such as the initial or average slope to compute SBP and DBP.
Figure 4G illustrates the unnormalized oscillograms, normalized oscillograms, and cuff pressure during inflation for one mandrel at simulator settings of three maximum volume oscillations but with the BP level and heart rate fixed. The home monitor BP values and the maximum oscillogram amplitudes differed substantially, whereas the cuff pressure inflation pattern as well as the normalized oscillograms were virtually superimposable. Figure 4H illustrates the normalized and raw cuff pressure oscillations (from which the oscillograms were constructed) versus the applied cuff pressure. The cuff pressure oscillations only changed in amplitude. These results indicated that the home monitor must be using the maximum oscillogram amplitude to compute SBP and DBP.
In sum, the data from the strategic subset of mandrels and simulator settings suggested that the home monitor uses ratios, the maximum oscillogram amplitude, and possibly the initial or average slope of the cuff pressure during inflation to compute SBP and DBP. Further, this subset of data collectively suggested that the systolic and diastolic ratios are inversely related to the maximum oscillogram amplitude and the cuff pressure inflation slopes and may therefore be computed from one or more of these features for a “variable ratio” algorithm. So, we developed variable as well as fixed ratio algorithms based on linear regression models to estimate the home monitor SBP and DBP values over all mandrels at all simulator settings. We first determined the values of the systolic and diastolic ratios (see Rs and Rd in Fig. 1) that corresponded to the SBP and DBP values reported by home monitor for each oscillogram (see, e.g., squares and circles in Fig. 4). We then averaged these home monitor ratios for a pure intercept regression model and estimated the home monitor SBP and DBP values at these fixed ratios. We next developed slope-intercept regression models to estimate the home monitor ratios from various combinations of the maximum oscillogram amplitude and cuff pressure inflation slopes. We finally computed the SBP and DBP values at these variable ratios to determine if they could improve the estimation of the corresponding home monitor values. We also similarly investigated linear regression models to estimate the home monitor SBP and DBP values directly from the cuff pressure at a fixed ratio of the oscillogram (see PRs or PRd in Fig. 1), the maximum oscillogram amplitude, and the cuff pressure inflation slopes for a “direct regression” algorithm.
Figure 5A shows the home monitor SBP and DBP values estimated from the measured oscillograms using fixed ratios (Rs = 0.66 and Rd = 0.55) versus the corresponding home monitor values. The root-mean-squared-errors of the estimations were 5.8 mmHg for SBP and 1.5 mmHg for DBP. Figure 5B shows the home monitor SBP and DBP values estimated from the measured oscillograms using variable ratios versus the corresponding home monitor values. The variable ratios here were computed from a linear regression model with the reciprocal of the maximum oscillogram amplitude as the independent variable or input (see legends in Fig. 5B). Adding just this one feature to the model substantially reduced the estimation errors to 1.5 mmHg for SBP and 0.8 mmHg for DBP. However, adding the initial or average slope of the cuff pressure during inflation as a second input to the model hardly reduced the estimation errors (best case of 1.4 mmHg for SBP and 0.8 mmHg for DBP). Further, computing variable ratios via a linear regression model with a cuff pressure inflation feature alone as input could at best only reduce the estimation error to 4.1 mmHg for SBP and 1.3 mmHg for DBP. Figure 5C shows the home monitor SBP and DBP values estimated by the best direct regression algorithm from the measured oscillograms versus the corresponding home monitor values. The algorithm inputs here were the cuff pressure at a fixed ratio of the oscillogram and the maximum oscillogram amplitude (see legends in Fig. 5C). This algorithm used more model parameters than the variable ratio algorithm to yield similar errors of 1.3 mmHg for SBP and 0.7 mmHg for DBP.
Fig. 5.
Estimated versus measured home monitor SBP and DBP values for all four mandrels at all simulator settings but with heart rate fixed. A) Fixed ratio algorithm results. B) Results of a variable ratio algorithm in which the ratio is computed via a slope-intercept model with the reciprocal of the maximum oscillogram amplitude (ΔOm) as input. See Fig. 1. C) Results of a direct regression algorithm in which SBP and DBP are computed via slope-intercept models with the cuff pressure at a fixed ratio (PRs or PRd) and the cuff pressure at a fixed ratio divided by the maximum oscillogram amplitude (PRs/ΔOm or PRd/ΔOm) as inputs. RMSE is root-mean-squared-error of the estimates against the measurements. 90/60, 120/60, 110/70, 140/80, 130/90, 170/80, 150/100 mmHg/0.6, 0.9, 1.25 ml/75 bpm simulator settings.
Figure 6A shows the home monitor SBP and DBP values predicted by the variable ratio algorithm from the measured oscillograms versus the corresponding home monitor values for two mandrels at simulator settings of three heart rate and two BP levels with the maximum volume oscillation fixed. These prediction results suggested that the home monitor may not be using heart rate to compute BP. Figure 6B shows the home monitor SBP and DBP values predicted by the fixed ratio, variable ratio, and direct regression algorithms from the measured oscillograms versus the corresponding home monitor values from human volunteers. The variable ratio algorithm predicted the home monitor BP values reasonably well with errors of 4.6 mmHg for SBP, which was better than the fixed ratio algorithm, and 3.5 mmHg for DBP, while the direct regression algorithm predicted the home monitor DBP values with an error of 2.2 mmHg. After further smoothing the noisier oscillograms from the volunteers, the SBP error of the variable ratio algorithm reduced to 3.7 mmHg (not shown). These prediction results suggested that the home monitor may not be using oscillation shape variations to compute BP.
Fig. 6.
Predicted versus measured home monitor SBP and DBP values for different algorithms and data. A) Variable ratio algorithm (see Fig. 5B) results for two mandrels at simulator settings of three heart rate (HR) and two BP levels with maximum volume oscillation fixed. Small and small-foam mandrels and 120/60, 170/80 mmHg/0.6 ml simulator setting. B) Results of all three algorithms (see Fig. 5) for six volunteers with the cuff placed properly, tightly, and loosely around the arm.
Figure 7 shows the BP values of an unvalidated home monitor, which operates via linear deflation, and the unnormalized and normalized oscillograms for the small and small-foam mandrels at fixed simulator settings and for the small mandrel at two maximum volume oscillations but with the BP level and heart rate fixed. These results suggested that this home monitor may be using a fixed ratio algorithm to compute BP. Table 1 shows similar BP values from a validated ambulatory monitor, which operates via step deflation, for a given mandrel at two maximum volume oscillations with the BP level and heart rate fixed. These results suggested that this ambulatory monitor may not be using the maximum oscillogram amplitude to compute BP. Table 2 shows differing BP values from another validated home monitor, which applied linear inflation for the small mandrel and step deflation for the small-foam mandrel, for a given mandrel at two maximum volume oscillations with the BP level and heart rate fixed. These results suggested that this home monitor may be using the maximum oscillogram amplitude to compute BP. The oscillograms are not shown for the latter two BP monitors, because oscillogram construction was uncertain for the step deflation measurements.
Fig. 7.
Measurements from a strategic subset of mandrels and simulator settings to gain insight into how an unvalidated home monitor (Series 100, CVS) computes BP. Unnormalized and normalized oscillograms and BP values via the home monitor, which operates via linear deflation, for two mandrels at simulator settings of two maximum volume oscillations and two BP levels with heart rate fixed. 75 bpm simulator setting.
Table 1.
Measurements from a strategic subset of mandrels and simulator settings to gain insight into how a validated ambulatory monitor (Oscar 2, Suntech Medical) computes BP.
| Oscillometric cuff device | Mandrel | Maximum volume oscillation (ml) | Monitor SBP (mmHg) | Monitor DBP (mmHg) |
|---|---|---|---|---|
|
Ambulatory BP monitor (Suntech medical) |
Small | 0.4 | 116 | 77 |
| 1.25 | 117 | 77 | ||
| Small-Foam | 0.4 | 117 | 80 | |
| 1.25 | 119 | 78 |
This monitor employs step deflation. BP is blood pressure; SBP and DBP are systolic and diastolic BP. Values at 110/70 mmHg/75 bpm simulator setting.
Table 2.
Measurements from a strategic subset of mandrels and simulator settings to gain insight into how another validated home monitor (H-BP100SBP, Welch-Allyn) computes BP.
| Oscillometric cuff device | Mandrel | Maximum volume oscillation (ml) | Monitor SBP (mmHg) | Monitor DBP (mmHg) |
|---|---|---|---|---|
| Home BP monitor (Welch-Allyn) | Small | 0.4 | 114 | 77 |
| 1.25 | 118 | 70 | ||
| Small-Foam | 0.4 | 109 | 75 | |
| 1.25 | 118 | 77 |
This monitor employed linear inflation for the small mandrel and step deflation for the small-foam mandrel. Values at 110/70 mmHg/75 bpm simulator setting.
Discussion
The field of hypertension has been built to a significant extent on oscillometric cuff devices. Yet, it is unknown how these devices compute BP from cuff pressure measurements.
It is believed that oscillometric cuff devices apply a fixed ratio algorithm to compute SBP and DBP from the oscillogram6–8. However, variations in the nonlinear cuff pressure-air volume relationship that occur each time the cuff is wrapped on a different arm or each time it is rewrapped on the same arm and variations in pulse pressure can alter the oscillogram shape and the ratios for computing BP (see Fig. 2). We therefore wondered whether and how current devices consider these effects in the BP computation? We rigorously studied a popular validated home BP monitor (Omron Healthcare). We initially broke apart the monitor to ascertain that it only includes one sensor that measures cuff pressure. We placed its Universal cuff on four mandrels (small, medium, large, and small-foam) to mimic different arms and studied an intact monitor using an NIBP simulator at various settings and an external pressure sensor to construct the oscillogram.
As expected, properly wrapping the same cuff on the different mandrels altered the nonlinear cuff pressure-air volume relationships (see Fig. 3E), and this variable cuff transducer effect, in turn, altered the shape of the oscillograms at fixed simulator settings (see Fig. 4D). Most notably, the normalized oscillograms for the small-foam mandrel were widest to right of the maximum amplitude, as the local slope of the cuff pressure-air volume relationship for this mandrel increased more with cuff pressure than the rigid mandrels to soften the decline in the blood volume oscillation amplitude. Our a priori expectations were that the oscillograms would differ more to the left of the maximum amplitude (see bottom panel of Fig. 2A), because the cuff pressure-air volume relationships are most nonlinear in the lower cuff pressure range. However, the simulator produced volume pulsations and consequently an oscillogram that is steeper to the left than the right of the maximum amplitude, which can reduce the effect of cuff nonlinearity in the lower cuff pressure range. Our simulations with a mathematical model indeed showed that the nonlinear cuff transducer can impact the left side of the oscillogram less in this case (see middle panel of Fig. 2A). It should also be mentioned that steady-state cuff pressure-air volume relationships may only approximately represent the cuff transducer during a dynamic oscillogram measurement.
We found that the home monitor does not employ a fixed ratio algorithm but is rather using the maximum oscillogram amplitude to compute BP. We arrived at this conclusion by examining essentially raw measurements from one mandrel while varying only the maximum volume oscillation with all other simulator settings fixed. The oscillogram changed only in amplitude and not in any other way (see Fig. 4G). The raw cuff pressure oscillations likewise changed in amplitude alone (see Fig. 4H). The cuff pressure during inflation also did not change (see Fig. 4G). Yet, the home monitor SBP and DBP values did change (see Fig. 4G). Since the home monitor just measures the cuff pressure, the only explanation can be that it uses the maximum oscillogram amplitude to compute BP.
By examining a few more of the simulator measurements (see Figs. 4A-F), we deduced that the home monitor may be using a variable ratio algorithm to compute SBP and DBP in which the variable ratios are determined using the maximum oscillogram amplitude. By then analyzing the measurements from all mandrels at various BP level and maximum volume oscillation settings, we found that a fixed ratio algorithm estimated the home monitor SBP values with an appreciable error of 5.8 mmHg and the home monitor DBP values with an error of 1.5 mmHg (see Fig. 5A). In contrast, a variable ratio algorithm in which the ratio is computed via a slope-intercept regression model with the reciprocal of the maximum oscillogram amplitude as input (see Fig. 1) reduced the errors in the home monitor SBP and DBP values all the way down to 1.5 and 0.8 mmHg, respectively (see Fig. 5B). It is noteworthy that only one extra model parameter was needed to significantly reduce the errors. From an error reduction perspective, the maximum oscillogram amplitude may seem less important in estimating the home monitor DBP values than SBP values. However, this result may be explained by the simulator producing steeper oscillograms on the left side as mentioned above.
The maximum oscillogram amplitude decreases with increasing nonlinearity of the cuff pressure-air volume relationship due to the concomitant decrease in the cuff transducer gain (see Figs. 3E and 4E). Increasing cuff nonlinearity can widen the oscillogram to the right of the maximum amplitude, as mentioned above, as well as narrow the oscillogram to the left of the maximum amplitude (see Figs. 2A and 4D) such that a fixed ratio algorithm would overestimate SBP and DBP. The home monitor increases the systolic and diastolic ratios based on the higher reciprocal of the maximum oscillogram amplitude to produce a lower SBP but an even higher DBP. The maximum oscillogram amplitude also increases with the maximum volume oscillation (see Fig. 4G), which increases with increasing PP. (Note, however, that PP does not impact the magnitude of volume oscillations in the simulator (see Fig. 4A), because it does not include an arterial model.) Increasing PP decreases both the systolic and diastolic ratios such that a fixed ratio algorithm would underestimate SBP and overestimate DBP (see Fig. 2B). The home monitor uses lower systolic and diastolic ratios based on the lower reciprocal of the maximum oscillogram amplitude to yield a higher SBP and a lower DBP. In sum, by using the maximum oscillogram amplitude, the home monitor may be able to compensate for both the variable cuff transducer and PP effects in computing SBP but only the variable PP effect in computing DBP. Importantly, the maximum volume oscillation can also increase following an increase in arterial compliance via, for example, smooth muscle relaxation-induced increase in the amplitude of the sigmoidal arterial blood volume-transmural pressure relationship. However, increasing the sigmoidal relationship amplitude would only scale the oscillogram such that using the maximum oscillogram amplitude to adjust the ratios would be detrimental to the BP computation (see Fig. 4G). That the home monitor uses the maximum oscillogram amplitude to compute BP suggested that it was empirically found to yield a net benefit to BP measurement accuracy.
The cuff pressure inflation pattern is a more specific indicator of cuff nonlinearity. The initial and average slopes of the cuff pressure during inflation indicate the volume of air pumped into the cuff to achieve a given cuff pressure. These slopes generally decrease with increasing cuff nonlinearity (see Figs. 3E and 4F) and are not impacted by the maximum volume oscillation (see Fig. 4G). The duty cycle of the air pump, which is also available to the home monitor for BP computation but could not be observed by us, is likewise a specific indicator of cuff nonlinearity. However, we could not find a feature of the cuff pressure during inflation that was better than the maximum oscillogram amplitude or even helpful in estimating the home monitor BP values.
The average value of the diastolic ratio for the home monitor here was 0.55 (see Fig. 5A), which is lower than previously reported values of around 0.751,8,9. The home monitor also rapidly deflates the cuff one beat after it determines DP. So, we suspected that it could actually be using a higher ratio to compute DP. We studied a direct regression algorithm to estimate the home monitor SBP and DBP values using the cuff pressure at a fixed ratio of the oscillogram and the maximum oscillogram amplitude as inputs. This algorithm yielded errors similar to the variable ratio algorithm (see Fig. 5C) but used more model parameters. On the other hand, the diastolic ratio was higher at 0.70 (see Fig. 5C). It may therefore be possible that the home monitor is using this type of algorithm instead of a variable ratio algorithm, especially for DBP. Either way, although the home monitor is not simply using a fixed ratio algorithm to compute BP, it may not be as complicated as we imagined that it might be at the outset of the study.
By analyzing further measurements, we found that the variable ratio and direct regression algorithms could predict (not estimate) the home monitor BP values with errors of 0.9–4.6 mmHg from low simulator heart rate, which introduces sampling error in the constructed oscillogram, to high simulator heart rate (see Fig. 6A) and in volunteers (see Fig. 6B). These results may confirm that the home monitor is neither using heart rate nor oscillation shape variations to compute BP.
We did not have access to the oscillograms constructed by the home monitor. Therefore, the main source of algorithm error here was the difference in our construction of the oscillogram versus the home monitor construction of the oscillogram. However, given that the data generated using the simulator had minimal artifacts, we expected that our constructed oscillograms and the home monitor oscillograms would be similar. The variable ratio algorithm, with only two model parameters, was indeed able to estimate the home monitor SBP and DBP values in our main simulator study with errors of only about 1 mmHg (see Fig. 5B). These results may not have been probable, if our oscillograms differed appreciably from those of the home monitor. However, the oscillograms from the human volunteers were contaminated by respiration, heart rate variability, and motion artifact, so our constructed oscillograms may not have been as representative of the home monitor oscillograms. For this reason, the algorithms did not predict the home monitor BP values as well (but still with reasonable errors of around 3.5 mmHg). Note that the objective of our study was not to exactly reverse engineer the entire algorithm of the home monitor but rather to understand how it computes BP. Use of the simulator in which the oscillograms could be measured without any physiological or significant measurement artifacts was pivotal in accomplishing our objective.
The home monitor yielded variable SBP values over the four mandrels at the same simulator settings (see Fig. 4D). The home monitor also outputted different BP values when the maximum volume oscillation simulator setting alone was varied (see Fig. 4G). However, the home monitor was designed for accuracy against dual-observer auscultation in humans. Therefore, it was irrelevant to compare the home monitor BP values against the simulator BP settings. That said, we found that the home monitor uses a simple, empirical BP computation algorithm (see Fig. 5B) and therefore may not consider the variable PP and cuff transducer effects well in humans. For example, over-cuffing and under-cuffing are well known problems 16.
Overall, we found the popular validated home monitor of Omron Healthcare to be robust in that it often produced repeatable and sensible BP measurements. We wondered whether the BP monitors of other manufacturers likewise use the maximum oscillogram amplitude to compute BP. So, we conducted strategic analyses of three other BP monitors using the same mandrels and NIBP simulator. We found that another validated home monitor (Welch-Allyn) may also be using the maximum oscillogram amplitude to compute BP (see Table 2), whereas the other two BP monitors may possibly be using a fixed ratio algorithm (see Fig. 7 and Table 1). Oscillometric cuff devices for special populations such as children could employ different BP computation algorithms and may therefore be worthwhile to study in the future using similar methodologies to those employed herein.
One concern of our study is whether we exercised the home monitor enough to capture its exact BP computation algorithm correctly or not. The simulator only allowed us to study the home monitor with steep oscillograms to the left of the maximum amplitude, but physiologic oscillograms may not be right skewed (see bottom panel of Fig. 2A). It is therefore possible that the algorithm especially for computing DBP could use additional features as input (e.g., a feature of the cuff pressure during inflation). However, the home monitor DBP prediction results in volunteers do help alleviate this concern (e.g., see 2.2 mmHg error in Fig. 6B). Furthermore, whether we have fully captured the algorithm of the home monitor or not, our study still showed with certainty that the maximum oscillogram amplitude is an important input to oscillometric BP computation algorithms.
In conclusion, oscillometric cuff devices have been used in countless people and studies and are the primary workhorse in the field of hypertension, yet it is unknown how these devices compute BP from the raw measurements. It is believed that the devices apply a fixed ratio algorithm to the oscillogram to compute BP. Here, we showed that a popular validated home monitor instead uses the maximum oscillogram amplitude in the BP computation. We explained that this additional oscillogram feature could be particularly helpful when PP changes. Our study provides more information on how oscillometric cuff devices work and is therefore important for proper interpretation and scientific scrutiny7. Just as importantly, we must know how these devices work to be able to try to improve upon the incredible but error-prone technology.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
This work was supported by the NIH Grant HL163691.
Author contributions
V. D. analyzed the data, performed the computer simulations, and helped prepare the manuscript. M. J. performed the computer simulations and collected the data. R. K. designed the experimental setup and collected the data. H. D. collected the data. A. M. edited the manuscript. J-O. H. provided technical feedback and edited the manuscript. S. G. S. provided technical feedback and helped prepare the manuscript. R. M. led the study and prepared the manuscript.
Funding
National Heart, Lung, and Blood Institute, HL163691, HL163691, HL163691, HL163691
Data availability
The data and code described in this manuscript may be made available upon reasonable request to R.M. (rmukkamala@pitt.edu).
Declarations
Competing interests
V. D., A. M, J-O. H., S. G. S. and R. M. have pending and/or issued patents on oscillometric BP computation algorithms. A. M., J-O. H., S. G. S., and R. M. have an NIH grant on oscillometric cuff devices. All other authors declare no conflicts of interest.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Mahdi Jazini and Ravinder Kumar are equally contributing second authors.
References
- 1.Babbs, C. F. Oscillometric measurement of systolic and diastolic blood pressures validated in a physiologic mathematical model. Biomed. Eng. Online11, 56. 10.1186/1475-925X-11-56 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Liu, J. et al. Patient-specific oscillometric blood pressure measurement: Validation for accuracy and repeatability. IEEE J. Transl. Eng. Health Med.5, 1900110. 10.1109/JTEHM.2016.2639481 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bothe, T. L. et al. Impact of oscillometric measurement artefacts in ambulatory blood pressure monitoring on estimates of average blood pressure and of its variability: A pilot study. J Hypertens41(1), 140–149. 10.1097/HJH.0000000000003315 (2023). [DOI] [PubMed] [Google Scholar]
- 4.Pilz, N. et al. Blood pressure measurement and nocturnal dipping patterns are heavily affected by body posture through changes in hydrostatic pressure between the arm and heart. Hypertens Res48(3), 1144–1154. 10.1038/s41440-024-02056-0 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Bovien Gørlitz, K. et al. One size does not fit all: Universal cuff overestimates oscillometric blood pressure in persons with large arm circumference. Blood Press33(1), 2338208. 10.1080/08037051.2024.2338208 (2024). [DOI] [PubMed] [Google Scholar]
- 6.Muntner, P. et al. Measurement of blood pressure in humans: A scientific statement from the american heart association. Hypertension73(5), E35–E66. 10.1161/HYP.0000000000000087 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Izzo, J. L. and Mitchell, G. F. Pitfalls of Pulse Wave Analysis and Oscillometric Blood Pressure Derivatives. Lippincott Williams and Wilkins. 10.1161/HYPERTENSIONAHA.121.17228 (2021). [DOI] [PubMed]
- 8.Sharman, J. E. et al. Automated ‘Oscillometric’ Blood Pressure Measuring Devices: How They Work and What They Measure. Springer Nature.10.1038/s41371-022-00693-x (2023). [DOI] [PMC free article] [PubMed]
- 9.Drzewiecki, G., Hood, R. & Applet, H. Theory of the oscillometric maximum and the systolic and diastolic detection ratios. Ann Biomed Eng22, 88–96. 10.1114/1.130 (1994). [DOI] [PubMed] [Google Scholar]
- 10.Geddes, L. A., Voelz, M., Combs, C., Reiner, D. & Babbs, C. F. Characterization of the oscillometric method for measuring indirect blood pressure. Ann Biomed Eng10(6), 271–280. 10.1007/BF02367308 (1982). [DOI] [PubMed] [Google Scholar]
- 11.Chandrasekhar, A. et al. Formulas to explain popular oscillometric blood pressure estimation algorithms. Front. Physiol.10, 1415. 10.3389/fphys.2019.01415 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Stergiou, G. S. et al. STRIDE BP international initiative for accurate blood pressure measurement: Systematic review of published validation studies of blood pressure measuring devices. J. Clin. Hypertens21(11), 1616–1622. 10.1111/jch.13710 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Williams, J. S., Brown, S. M. & Conlin, P. R. Blood-pressure measurement. New England J. Med.360(5), e6. 10.1056/NEJMvcm0800157 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Flint, A. C. et al. Effect of systolic and diastolic blood pressure on cardiovascular outcomes. N. Engl. J. Med.381(3), 243–251. 10.1056/nejmoa1803180 (2019). [DOI] [PubMed] [Google Scholar]
- 15.Dhamotharan, V. et al. Oscillometric blood pressure measurement: Modeling and analysis of area and height osillograms. Front. Physiol.16, 1611096. 10.3389/fphys.2025.1611096 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ishigami, J. et al. Effects of cuff size on the accuracy of blood pressure readings: The cuff(SZ) randomized crossover trial. JAMA Intern. Med.183(10), 1061–1068. 10.1001/jamainternmed.2023.3264 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Dhamotharan, V. et al. Mathematical modeling of oscillometric blood pressure measurement: A complete, reduced oscillogram model. IEEE Trans. Biomed. Eng.70(2), 715–722. 10.1109/TBME.2022.3201433 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data and code described in this manuscript may be made available upon reasonable request to R.M. (rmukkamala@pitt.edu).