American Journal of Epidemiology
Copyright © 2003 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved
Vol. 158, No. 12
Printed in U.S.A.
DOI: 10.1093/aje/kwg274
Sources of Variability in Blood Pressure Measurement using the Dinamap PRO
100 Automated Oscillometric Device
Joseph J. Chang1, Daniel Rabinowitz2, and Steven Shea1,3
1
Department of Medicine, College of Physicians and Surgeons, Columbia University, NY.
Department of Statistics, Columbia University, New York, NY.
3 Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY.
2
Received for publication December 12, 2002; accepted for publication June 18, 2003.
The Dinamap automated oscillometric device (GE Medical Systems Information Technologies, Inc.,
Milwaukee, Wisconsin) for blood pressure measurement is widely employed in epidemiologic studies because it
is easy to use and because it eliminates observer variability. In this study, the authors assessed the variability in
observed blood pressures associated with use of the Dinamap monitor and estimated the contributions of various
factors to that variability. In 60 volunteers (30 aged 23–35 years and 30 aged 54–82 years) from New York, New
York, the authors obtained 30 simultaneous paired blood pressure measurements in both arms at 1-minute
intervals, using three separate Dinamap PRO 100 devices allocated to arm and subject according to a balanced
incomplete block design. Variability, defined as the between-arm difference in blood pressure measurements,
was analyzed using a mixed-effects linear regression model. A total of 1,800 paired blood pressure
measurements were obtained between September 2001 and June 2002. The mean ages of the two groups were
28.3 years (standard deviation, 4.0) and 71.7 years (standard deviation, 8.0). A diagnosis of hypertension was
present in 53% of the older subjects and none of the younger subjects. Fifty percent of paired simultaneous blood
pressure measurements obtained were in agreement within 4 mmHg for systolic blood pressure or within 3 mmHg
for diastolic blood pressure. Residual variability, attributable to the intrinsic inaccuracy of the device, accounted
for 64–82% of the total systolic and diastolic blood pressure variability. The majority of variability in blood pressure
measurement was due to the device as used under the study conditions.
blood pressure; epidemiologic methods; equipment and supplies; observer variation; oscillometry; research
design
Abbreviation: MESA, Multi-Ethnic Study of Atherosclerosis.
The Dinamap automated oscillometric device (GE
Medical Systems Information Technologies, Inc.,
Milwaukee, Wisconsin) is a device for measuring resting
blood pressure. It is easy to use, requires less training than
manual mercury sphygmomanometers, and intuitively
appears to eliminate or reduce interrater variability as a
source of error in blood pressure measurement. These
features have led to the use of Dinamap devices in several
epidemiologic studies, including the Multi-Ethnic Study of
Atherosclerosis (MESA) (1), the Atherosclerosis Risk in
Communities Study (2), the National Heart, Lung, and
Blood Institute Family Heart Study (3), and the Hypertension Genetic Epidemiology Network study (4). These uses
have included measurement of resting blood pressure and
blood pressure response to stressors. In several published
papers (5–20), investigators have assessed the accuracy of
Dinamap devices by comparing Dinamap blood pressure
measurements with those obtained by intraarterial recording
or auscultation and manual sphygmomanometry. No
published data of which we are aware have directly assessed
the variability in measured blood pressures obtained with an
automated oscillometric blood pressure device.
There are at least two ways in which inaccuracy might
contribute to measurement variability. First, a deflation rate
of 2 mmHg per second in a subject with a heart rate of 60
beats per minute means that one cardiac cycle is sampled per
Reprint requests to Dr. Steven Shea, Division of General Medicine, PH 9 East—Room 105, Columbia University College of Physicians and
Surgeons, 622 West 168th Street, New York, NY 10032 (e-mail: [email protected]).
1218
Am J Epidemiol 2003;158:1218–1226
Blood Pressure Measurement Variability 1219
2 mmHg, which sets this as the limit of resolution of the
device under these conditions. Second, the device works by
measuring the arterial pressure oscillation during the cardiac
cycle, identifying the pressure level at which the rate of rise
is maximal (designated mean arterial pressure), and using a
proprietary algorithm to estimate systolic and diastolic blood
pressure (e.g., systolic blood pressure = 1.5 × mean arterial
pressure). Therefore, the imprecision with which the
maximal rate of rise is identified is a second intrinsic source
of error. In addition, the algorithm used to estimate systolic
and diastolic blood pressure from mean arterial pressure is
empirically derived and approximate. However, if all else
were completely accurate, this would not create variability in
observed blood pressure measurements.
Other factors may also contribute to variability in
observed blood pressure values and/or modify the contribution to variability of the intrinsic inaccuracy of the device.
These factors include blood pressure level, systematic differences in blood pressure between the left arm and the right
arm, subject-specific variation in these differences, differences in measurement by different devices, differences in the
order in which the devices are assigned to a particular arm,
and sequence effects (i.e., earlier blood pressure measurements versus later measurements). It has also long been
recognized that the physical conditions under which blood
pressure is measured may affect the observed blood pressure
level. These include cuff size relative to arm size, cuff position relative to the heart, posture, physical activity, the state
of relaxation of the subject, and ambient temperature (21–
25).
The purpose of this study was to quantify the variability in
observed blood pressure associated with use of the Dinamap
PRO 100 and to estimate the contributions of various factors
to that variability under standardized conditions comparable
to those achievable in epidemiologic research.
MATERIALS AND METHODS
Subjects
We conducted the study among 60 volunteers from New
York, New York, between September 2001 and June 2002.
The study was performed in two parts: first in a volunteer
group of 30 healthy younger persons (ages 23–35 years) who
were not using any medication for a chronic health condition, other than treatment for allergies or birth control, and
then in a group of 30 older subjects (ages 54–82 years)
enrolled in MESA at the Columbia University MESA Field
Center. The presence of chronic medical conditions, use of
medications, and the presence of hypertension were ascertained using a brief structured interview. The presence of
hypertension was based on self-report of a physician’s diagnosis or current use of antihypertensive medication. None of
the younger participants had measured blood pressure levels
above 140/90 mmHg. Atrial fibrillation, as determined at
baseline interview or by electrocardiogram, was an exclusion criterion for MESA (1). None of the subjects had atrial
fibrillation, which can interfere with oscillometric measurement of blood pressure. The Institutional Review Board of
Am J Epidemiol 2003;158:1218–1226
the Columbia-Presbyterian Medical Center approved the
study, and all subjects gave written informed consent.
Procedure
Thirty paired simultaneous blood pressure measurements
were obtained at 1-minute intervals in each participant, in
three sequences of 10 each. We used three different Dinamap
PRO 100 devices. Participants were allocated to one of 12
configurations of assignment of the three devices to the left
or right arm using a balanced incomplete block design:
1,2
2,3
3,1
2,3
1,2
3,1
3,1
2,3
1,2
1,2
3,1
2,3
2,3
3,1
1,2
3,1
1,2
2,3
2,1
3,2
1,3
3,2
2,1
1,3
1,3
3,2
2,1
2,1
1,3
3,2
3,2
1,3
2,1
1,3
2,1
3,1
The numbers in this table refer to the three devices. A participant assigned to one of the 12 configurations had the three
devices used on the left and right arms, as shown, in three
pairings such that 10 pairs of blood pressure measurements
were made in each pairing.
Blood pressure measurements were obtained with the
subject in the seated position with 5 minutes’ rest before the
first sequence and between sequences, under standardized
conditions corresponding to consensus recommendations for
blood pressure measurement (23–25). The measurements
were taken in a quiet room with an ambient temperature
between 70°F and 76°F. Subjects were seated in a chair with
back support, with both feet resting comfortably on the floor
and both forearms supported on a level surface. Subjects
were asked not to talk during the procedure. Cuff size was
selected on the basis of a table of cuff sizes and arm circumferences to avoid artifacts arising from selection of an inappropriate-size cuff. The cuff was placed on the arm at the
level of the heart. All blood pressure measurements were
obtained by a single investigator (J. J. C.).
Statistical analysis
Bland-Altman plots of differences between each systolic
and diastolic pair of measurements were used to examine
whether the variability of the difference in paired measurements was related to the level of blood pressure. Mean blood
pressure levels by age group were calculated by dividing the
sum of the subject-specific means by the number of subjects
in the group. Mixed-effects linear models were used to
examine the data on blood pressure measurement differences. The analyses proceeded from exploratory model
fitting and model checking to the development of a final
model. Among the factors considered in the exploratory
models were population and subject-specific differences
between average blood pressure measurements in the right
and left arms, differences between average blood pressure
measurements, differences in blood pressure measurement
variability in the three different machines, and the covariance structure of the serial blood pressure measurements.
1220 Chang et al.
TABLE 1. Mean blood pressure levels and differences, by age group, in a study of variability in blood pressure
measurements, New York, New York, 2001–2002*
Blood pressure (mmHg)
Younger group (n = 30)
Variable
Older group (n = 30)
Left arm
Right arm
Difference
Systolic pressure
115.0 (12.3)†
116.7 (12.8)
1.7 (7.1)
141.3 (22.2)
Left arm
141.7 (22.6)
Right arm
Difference
0.4 (7.6)
Diastolic pressure
68.9 (7.4)
69.0 (8.0)
0.1 (5.5)
78.1 (10.8)
79.1 (10.7)
1.0 (5.5)
* Mean values were computed from the total of 3,600 blood pressure observations (30 in each arm in each subject).
† Numbers in parentheses, standard deviation.
The preponderance of right-handed subjects precluded
simultaneous examination of differences between the right
and left arms and differences between the dominant and
nondominant arms.
The final model took the form
Dijk = µ + αi + βij + γr(i,j) + γl(i,j) + εijk,
where Dijk denotes the kth difference between the right- and
left-hand measurements taken during the jth block of
measurements from the ith subject, µ is the population
difference in average measurements from the right and left
arms, αi is a random subject-specific variance component
representing the deviation of the ith subject’s right-arm
versus left-arm difference from the population right-arm
versus left-arm difference, βij denotes the block-specific
variance components representing the expected deviations of
within-block difference from the expected subject-specific
difference, r(i,j) and l(i,j) denote the machines used on the
right and left arms, respectively, in the jth block of measurements from the ith subject, the γ’s are fixed-effects terms
representing the differences in the expected measurements
between machines, and εijk is a random effect representing
residual variability in the kth measurement in the jth block
from the ith subject. The random effects are defined to be
independent.
The estimable fixed-effect parameters in this model are µ
and the differences, γ1 – γ3 and γ2 – γ3. The estimable randomeffects parameters are the variance of the subject-specific
and block-specific components, σα and σβ, and the variance
of the residual terms, σε. The estimates presented in the table
are based on untransformed data; however, log- and squareroot-transformed data revealed qualitatively similar results.
Tests of significance were made for the presence of a handedness effect, an overall machine effect (with 2 df), the
subject-specific and block-specific variance components,
and the residual variability. Model checking was carried out
through a combination of examining the significance and
magnitude of effect of parameter estimates and examining
graphical displays of residuals and covariance structure
parameter estimates. All analyses were conducted separately
for systolic and diastolic blood pressure measurements. Estimation and significance level checking were carried out
using SAS PROC MIXED (SAS Institute, Inc., Cary, North
Carolina) (26).
Quantiles were calculated for the observed distributions of
the absolute values of the between-arm differences in blood
pressure, both unadjusted and adjusted for blood pressure
level, arm, subject, block, device, and sequence. For assessment of sequential measurement variability rather than
simultaneous variability, the corresponding quantiles were
calculated for unadjusted paired blood pressure measurements, where the pairs were defined as the ith measurement
and the (i + 1)th measurement in the same arm for each
subject using a single Dinamap device (a single block of 10
measurements).
RESULTS
The mean age of the younger group was 28.3 years (standard deviation, 4.0), while that of the older group was 71.7
years (standard deviation, 8.0). Of the 30 younger subjects,
22 (73.3 percent) were female, as compared with 12 (40
percent) of the 30 older subjects. None of the younger
subjects had hypertension, as compared with 16 (53.3
percent) of the older subjects. In both groups, 28 of the 30
subjects were right-handed.
A total of 1,800 paired blood pressure measurements were
obtained. Table 1 shows the mean blood pressure levels for
the right and left arms in each group. As expected, the
younger group had lower mean systolic and diastolic blood
pressures than the older group. Figures 1, 2, 3, and 4 represent the Bland-Altman plots of the differences in each
systolic and diastolic pair for the younger and older groups.
No apparent relation between variability in the differences in
paired blood pressure measurements and level of blood pressure was noted. Linear regressions of the squared differences
on the average systolic and diastolic blood pressure values
indicated no significant relation, which was consistent with
the visual impression. The multivariate models for variability did not include terms for systolic or diastolic blood
pressure level, because these terms did not contribute significantly to variability in these models.
Table 2 shows the parameter estimates from the mixedeffects models for each group. There was ambiguous
evidence of a systematic difference in the average blood
pressure measurement between the right arm and the left
arm. There was no statistically significant difference among
the three machines in average blood pressure measurements.
There was evidence of significant subject-specific variability
(variance of α). This variance estimate represents the variAm J Epidemiol 2003;158:1218–1226
Blood Pressure Measurement Variability 1221
FIGURE 1. Bland-Altman plot of between-arm differences in systolic blood pressure (S_DIFF) (mmHg) and mean systolic blood pressure
(S_AVG) (mmHg) in the younger age group, New York, New York, 2001–2002.
ance of the deviation of the ith subject’s right-arm versus
left-arm difference from the population average right-arm
versus left-arm difference. There was also evidence of
block-specific variability (variance of β). This variance estimate represents the variance of the deviation of a block of 10
blood pressure measurements’ right- versus left-arm difference from the population average right- versus left-arm
difference. This contribution may represent cuff repositioning or changes in posture. The variance of the random
error (ε) is the residual variability. The interpretation of the
residual variability is variability due to the intrinsic inaccuracy of the blood pressure device as used under the conditions of the study. We cannot be sure that, in addition, a
component of variability due to rapidly fluctuating differ-
FIGURE 2. Bland-Altman plot of between-arm differences in diastolic blood pressure (D_DIFF) (mmHg) and mean diastolic blood pressure
(D_AVG) (mmHg) in the younger age group, New York, New York, 2001–2002.
Am J Epidemiol 2003;158:1218–1226
1222 Chang et al.
FIGURE 3. Bland-Altman plot of between-arm differences in systolic blood pressure (S_DIFF) (mmHg) and mean systolic blood pressure
(S_AVG) (mmHg) in the older age group, New York, New York, 2001–2002.
ences between arms in the true underlying blood pressure is
not also contributing to the estimated residual variability, but
to our knowledge this has not been described.
In additional analyses, we found no significant evidence of
serial correlation in the measurement differences beyond
that which could be accounted for by the between-subject
and between-block variance components. There was also no
strong evidence for changes in the variability of the blood
pressure measurements with sequential measurements or for
differences in the variability of measurements taken from
different machines. However, these aspects of the analysis
had low statistical power.
As is shown in table 2, residual variability was the largest
component of the three variance components that contributed to the variability of the paired differences in blood pressure measurements. After removal of the fixed effects, which
were small, the residual variability accounted for 64–82
percent of the total variability in systolic and diastolic blood
pressure differences between arms in the younger and older
groups, respectively. It is interesting to compare the
FIGURE 4. Bland-Altman plot of between-arm differences in diastolic blood pressure (D_DIFF) (mmHg) and mean diastolic blood pressure
(D_AVG) (mmHg) in the older age group, New York, New York, 2001–2002.
Am J Epidemiol 2003;158:1218–1226
Blood Pressure Measurement Variability 1223
TABLE 2. Parameter estimates from a mixed-effects model predicting variability in between-arm differences in blood
pressure among 60 subjects, New York, New York, 2001–2002
Systolic blood pressure
Parameter
Younger group
(n = 30)
Diastolic blood pressure
Older group
(n = 30)
Younger group
(n = 30)
Older group
(n = 30)
Fixed effects (mmHg)
Right arm vs. left arm (µ)
1.93 (0.71)**,†
0.35 (0.61)
0.02 (0.60)
1.16 (0.44)*
Differences between machine 1 and
machine 3 (γ1 – γ3)
0.31 (0.49)
0.12 (0.40)
–0.29 (0.31)
0.41 (0.35)
Differences between machine 2 and
machine 3 (γ2 – γ3)
–0.40 (0.49)
–0.77 (0.40)
–0.30 (0.31)
0.15 (0.35)
Subject-specific variability (variance of α)
10.48 (3.75)**
7.98 (2.76)*
8.48 (2.63)**
3.71 (1.48)*
Block-specific variability (variance of β)
7.49 (1.99)***
2.40 (1.35)**
2.40 (0.82)**
3.03 (1.00)**
Variance of the random error (variance of ε)
31.89 (1.59)***
47.68 (2.37)***
19.58 (0.97)***
23.31 (1.16)***
Random effects (mmHg2)
* p < 0.05; ** p < 0.01; *** p < 0.001.
† Numbers in parentheses, standard error.
proportion of variability due to the residual terms and the
subject- and block-specific components of variability. The
ratios for systolic and diastolic pressure and for the two
groups ranged from approximately 2:5 to 1:2. The estimated
contribution of systematic differences in the machines and
between the right and left arms was on the order of only one
tenth of the contribution of the random-effects components.
While some aspects of the parameter estimates differed
between the older and younger subjects, these differences,
when considered in light of the large number of possible
comparisons, did not appear to be statistically significant.
Table 3 shows the absolute values for between-arm differences, unadjusted and adjusted for factors contributing to the
variability in blood pressure measurements, categorized by
age group. In both groups, 50 percent of paired unadjusted
measurements agreed within 4 mmHg for systolic blood
pressure and within 3 mmHg for diastolic blood pressure.
Adjustment for factors considered in the mixed-effects
models reduced the paired differences by approximately 25
percent at the 50th percentile values, consistent with the estimate of residual variability of 64–82 percent.
Table 4 shows the absolute values for differences between
the ith and (i + 1)th blood pressure measurements taken 1
minute apart within blocks. The unadjusted values in tables
3 and 4 are whole numbers, because they are quantiles of raw
differences, and all measurements were recorded to the
nearest whole number. As is shown in table 4, in both age
groups, 50 percent of sequential measurements agreed
within 4–6 mmHg for systolic blood pressure and within 3–4
mmHg for diastolic blood pressure. The interpretation of the
unadjusted variability in this analysis is variability due to the
intrinsic inaccuracy of the blood pressure device as used
under the conditions of the study, plus variability due to
minute-to-minute variability in blood pressure. The very
close agreement between the values in tables 3 and 4
suggests that, on average, the minute-to-minute variability in
blood pressure is quite small among subjects sitting quietly
under controlled conditions.
Am J Epidemiol 2003;158:1218–1226
DISCUSSION
Our analysis of 1,800 simultaneous paired blood pressure
measurements from 60 adults, obtained using the Dinamap
PRO 100 oscillometric device, showed that the largest
component of variability of between-arm differences in
blood pressure was residual variability. Residual variability
may be viewed as corresponding to a combination of random
TABLE 3. Absolute values for between-arm differences in
blood pressure among 60 subjects, New York, New York, 2001–
2002
Difference in blood pressure (mmHg)
Age group and
percentile
Systolic pressure
Unadjusted
Adjusted*
Diastolic pressure
Unadjusted
Adjusted*
Younger group
5th
0
0.3
0
0.2
10th
1.0
0.5
0
0.4
25th
2.0
1.3
1.0
1.0
50th
4.0
2.7
3.0
2.3
75th
7.0
5.0
6.0
4.2
90th
12.0
8.8
9.0
6.7
95th
15.0
12.2
11.0
8.9
5th
0
0.3
0
0.2
10th
1.0
0.6
0
0.4
25th
2.0
1.4
1.0
1.0
50th
4.0
3.0
3.0
2.3
75th
7.0
5.9
6.0
4.4
90th
12.0
10.6
9.0
7.4
95th
17.0
15.0
11.5
9.7
Older group
* Adjusted for blood pressure level, arm, subject, block, device,
and sequence.
1224 Chang et al.
TABLE 4. Absolute values for within-pair* differences in blood
pressure among 60 subjects, New York, New York, 2001–2002
Age group
and percentile
Difference (mmHg)
SBP†
DBP†
Younger group
5th
0
0
10th
1.0
0
25th
2.0
1.0
50th
4.0
3.0
75th
7.0
6.0
90th
12.0
9.0
95th
15.0
11.0
5th
0
0
10th
1.0
0
25th
2.0
1.0
50th
4.0
3.0
75th
7.0
6.0
90th
12.0
9.0
95th
17.0
11.5
Older group
* A pair was defined as the i th and (i + 1)th blood pressure
observations in the same arm taken 1 minute apart within a block of
observations.
† SBP, systolic blood pressure; DBP, diastolic blood pressure.
measurement error inherent in the action of the machines,
random rapidly fluctuating differences between blood pressure values in the right and left arms, and possibly also transient effects of posture. The subject-specific and blockspecific components of variability were estimated to be of
smaller magnitude than the residual variability. The subjectspecific components of variability may be viewed as corresponding to a combination of true subject-specific differences between average blood pressures in the right and left
arms or systematic differences in posture or placement of the
cuffs on the right and left arms. The block-specific components of variability may be viewed as a combination of fairly
slowly changing differences in actual blood pressure
between the right and left arms and changes between blocks
of measurement in posture or position of the cuffs on the
right and left arms.
Subject-specific variability, which made the largest contribution to variability for systolic and diastolic pressure in
both groups after residual variability, represents differences
among subjects in cuff placement, posture, state of relaxation, and other aspects of the subject’s state at the time of
measurement or of the interaction between the subject and
the device. The persistence of subject-specific variability,
despite careful standardization of the conditions and procedures for measuring the blood pressure, suggests that even
more meticulous standardization may have the potential to
reduce variability somewhat.
We did not find the variability of simultaneous blood pressure measurements and sequential blood pressure measure-
ments obtained in the same arm 1 minute apart to be very
different. This suggests that short-term biologic variation in
blood pressure is small compared with other sources of variability.
Our findings on between-arm differences in blood pressure measurements using the Dinamap device are comparable to those from other studies of simultaneous
measurements using different devices. In 1960, using
aneroid sphygmomanometers, Harrison et al. (27) performed
three simultaneous blood pressure measurements on both
arms at 1-minute intervals in 447 patients. Fifty percent of
these subjects had between-arm differences in systolic blood
pressure greater than or equal to 5 mmHg (versus 45 percent
in our study), and 44 percent had differences in diastolic
blood pressure of at least 5 mmHg (versus 34 percent in our
study). Harrison et al. also obtained simultaneous intraarterial pressure measurements of the brachial artery in both
arms in a subsample of 53 subjects. In this subgroup, 29
percent of the subjects were noted to have between-arm
differences in systolic blood pressure of at least 5 mmHg.
Similarly, in 1985, Gould et al. (28) published results from a
study of eight simultaneous paired blood pressure measurements made on both arms using a random-zero mercury
sphygmomanometer in 91 hypertensive subjects. They
reported that only 8 percent of the between-arm differences
in systolic pressure and 3 percent of differences in diastolic
pressure exceeded 10 mmHg, as compared with 15 percent
and 8 percent for systolic and diastolic pressure differences
in our study. Kaufman et al. (29) analyzed paired arm blood
pressure measurements obtained with three different oscillometric devices, one of which was an earlier model of the
Dinamap, in 25 anesthetized patients. On the basis of evaluation of Bland-Altman data plots, these investigators
concluded that there was no important relation between level
of blood pressure and variability. In addition, they concluded
that different devices have different degrees of variability
and that therefore measurements obtained by different
devices are not interchangeable (29).
Several limitations of our study may be considered in
interpreting these findings. Biologic and nonbiologic factors
other than those considered in the analyses may contribute to
blood pressure variability. However, it seems less likely that
such factors would affect between-arm differences in simultaneous blood pressure measurements. Our data also do not
address the reliability of data obtained with Dinamap models
other than the PRO 100 or with automated oscillometric
devices sold by other manufacturers. An earlier model of the
Dinamap, the 1846-SX, was used in the Atherosclerosis Risk
in Communities Study (2) and the Hypertension Genetic
Epidemiology Network study (4), while the PRO 100 model
is being used in MESA (1) and was used in this study.
Observed variability may be different for measurements
obtained under conditions that differ from those in our study.
Our study did not address the accuracy of blood pressure
measurements taken with the Dinamap PRO 100 device.
Other investigators have examined the accuracy of oscillometric devices compared with intraarterial blood pressure
recordings or mercury sphygmomanometry (5–20). A recent
review summarized findings on the accuracy of a number of
blood pressure devices (13). The Dinamap 8100 device was
Am J Epidemiol 2003;158:1218–1226
Blood Pressure Measurement Variability 1225
included in the review, which summarized data previously
reported (5–8). The Dinamap 8100 did not satisfactorily
meet the criteria of either the British Hypertension Society or
the US-based Association for the Advancement of Medical
Instrumentation, both of which are based on comparison of
values observed using the test device with values obtained
by manual mercury sphygmomanometry (30, 31). We are
not aware of published data on the accuracy of the Dinamap
PRO 100 device, the model used in our study. None of the
subjects in our study had atrial fibrillation or other significant cardiac arrhythmias. Oscillometric devices are inaccurate in subjects with arrhythmias.
Epidemiologists conducting studies in which blood pressure data are collected must choose between manual
methods involving sphygmomanometry and automated
oscillometric methods. Inaccuracy may arise from random
error, systematic error, or both. Systematic error, which may
be of great importance in clinical situations involving monitoring of acutely ill patients or treatment decisions for
patients with hypertension, may be less critical in epidemiologic studies in that regression coefficients may not be much
affected if heteroscedasticity is not present. Random error
and observer bias may matter more in epidemiologic studies.
In summary, 50 percent of paired simultaneous blood pressure measurements obtained with an automated oscillometric device, the Dinamap PRO 100, agreed within 4
mmHg for systolic blood pressure or 3 mmHg for diastolic
blood pressure. The majority of this measurement variability
was due to the device as used in the experimental procedure.
Other factors explained 36 percent or less of the total variability. Overall, in terms of the observed level of measurement variability, we believe that the Dinamap PRO 100
device is probably adequate for epidemiologic studies of
associations between blood pressure and other variables.
Measurement variability will be reduced when two or more
measurements are averaged, as is done in MESA (1) and
other epidemiologic studies.
ACKNOWLEDGMENTS
This research was sponsored by the National Heart, Lung,
and Blood Institute (contract N0-HC-95161) and the Health
Resources and Services Administration (grant D28-PE50011).
REFERENCES
1. Bild DE, Bluemke DA, Burke GL, et al. Multi-Ethnic Study of
Atherosclerosis (MESA): objectives and design. Am J Epidemiol 2002;156:871–81.
2. The ARIC Investigators. ARIC manuals of operation. No. 11.
Sitting blood pressure and postural changes in blood pressure
and heart rate. Chapel Hill, NC: ARIC Coordinating Center,
Department of Biostatistics, University of North Carolina at
Chapel Hill, 1987.
3. The FHS Investigators. NHLBI Family Heart Study: manuals
of procedures. St. Louis, MO: Division of Biostatistics, Washington University, 1993.
Am J Epidemiol 2003;158:1218–1226
4. The HyperGEN Investigators. Hypertension Genetic Epidemiology Network manuals of operation: clinical examination.
Version 2.0. St. Louis, MO: Division of Biostatistics, Washington University, 1998.
5. O’Brien E, Mee F, Atkins N, et al. Short report: accuracy of the
Dinamap portable monitor, model 8100 determined by the British Hypertension Society protocol. J Hypertens 1993;11:761–3.
6. O’Brien E, Atkins N. Accuracy of the Dinamap portable monitor, model 8100: a review of the evidence for accuracy. Blood
Press Monit 1997;2:31–3.
7. O’Brien E, Atkins N. Inaccuracy of the Dinamap 8100 portable
monitor. (Letter). Lancet 1997;349:1026.
8. Friedman B. Accuracy of Dinamap monitors. (Letter). Lancet
1997;350:217–18.
9. Ramsey M. Noninvasive automatic determination of mean arterial pressure. Med Biol Eng Comput 1979;17:11–18.
10. Hutton P, Dye J, Prys-Roberts C. An assessment of the
Dinamap 845. Anesthesia 1984;39:261–7.
11. Rutten AJ, Ilsley AH, Skowronski GA, et al. A comparative
study of the measurement of mean arterial blood pressure using
automated oscillometers, arterial cannulation and auscultation.
Anaesth Intensive Care 1986;14:58–65.
12. Manolio TA, Fishel SC, Beattie C, et al. Evaluation of the
Dinamap continuous blood pressure monitor. Am J Hypertens
1988;1:161S–7S.
13. O’Brien E, Waeber B, Parati G, et al. Blood pressure measuring
devices: recommendations of the European Society of Hypertension. BMJ 2001;322:531–6.
14. Gravlee GP, Brockschmidt JK. Accuracy of four indirect methods of blood pressure measurement, with hemodynamic correlation. J Clin Monit 1990;6:284–98.
15. White WB, Lund-Johansen P, Omvik P. Assessment of four
ambulatory blood pressure monitors and measurements by clinicians versus intraarterial blood pressure during rest and exercise. Am J Cardiol 1990;65:60–6.
16. Ornstein S, Markert G, Litchfield L, et al. Evaluation of the
DINAMAP blood pressure monitor in an ambulatory primary
care setting. J Fam Pract 1988;26:517–21.
17. Appel LJ, Marwaqha S, Whelton PK, et al. The impact of automated blood devices on the efficiency of clinical trials. Control
Clin Trials 1992;13:240–7.
18. Sturrock ND, Pound N, Peck GM, et al. An assessment of blood
pressure measurement in a diabetic clinic using random-zero,
semi-automated, and 24-hour monitoring. Diabet Med 1997;14:
370–5.
19. Gorback MS, Quill TJ, Lavine ML. The relative accuracies of
two automated noninvasive arterial pressure measurement
devices. J Clin Monit 1992;17:13–22.
20. Barker ME, Shiell AW, Law CM. Evaluation of the Dinamap
8100 and Omron M1 blood pressure monitors for use in children. Pediatr Perinat Epidemiol 2000;14:179–86.
21. Hunyor SN. Blood pressure measurement in clinical practice.
In: O’Brien E, O’Malley K, eds. Handbook of hypertension.
New York, NY: Elsevier Science Publishers, 1991:95–111.
22. Perry IJ, Beevers DG. Measurement of blood pressure in epidemiological surveys. In: O’Brien E, O’Malley K, eds. Handbook
of hypertension. New York, NY: Elsevier Science Publishers,
1991:174–83.
23. American Society of Hypertension. Recommendations for routine blood pressure measurement by indirect cuff sphygmomanometry. Am J Hypertens 1992;5:207–9.
24. Perloff D, Grim C, Flack J, et al. Human blood pressure determination by sphygmomanometry. Circulation 1993;88:2462–
70.
25. Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. The sixth report of
1226 Chang et al.
26.
27.
28.
29.
the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. Arch Intern Med
1997;157:2413–46.
SAS Institute, Inc. SAS OnlineDoc. Version 8. Cary, NC: SAS
Institute, Inc, 1999.
Harrison EG, Roth GM, Hines EA. Bilateral indirect and direct
arterial pressures. Circulation 1960;22:419–36.
Gould BA, Hornung RS, Kieso HA, et al. Is the blood pressure
the same in both arms? Clin Cardiol 1985;8:423–6.
Kaufmann MA, Pargger H, Drop LJ. Oscillometric blood pres-
sure measurements by different devices are not interchangeable. Anesth Analg 1996;82:377–81.
30. O’Brien E, Petrie J, Littler WA, et al. The British Hypertension
Society protocol for the evaluation of blood pressure measuring
devices. J Hypertens 1993;11(suppl 2):S43–63.
31. Association for the Advancement of Medical Instrumentation.
The American national standard for electronic or automated
sphygmomanometers. (ANSI/AAMI Standard SP 10-1992).
Arlington, VA: Association for the Advancement of Medical
Instrumentation, 1993.
Am J Epidemiol 2003;158:1218–1226