Stability of blood pressure - Nature

Journal of Human Hypertension (1999) 13, 637–642  1999 Stockton Press. All rights reserved 0950-9240/99 $15.00 http://www.stockton-press.co.uk/jhh

ORIGINAL ARTICLE

Stability of blood pressure: is a sequential blood pressure reading protocol efficient for a large-scale community screening programme Y-C Huang and DE Morisky Department of Community Health Sciences, UCLA School of Public Health, Los Angeles, CA 90095–1772, USA

The purpose of this study is to assess the relative stability of the systolic and diastolic blood pressure (BP) measures obtained in a state-wide screening programme. A state-wide hypertension survey was conducted in California in 1983 in which a total of 6381 adults were interviewed in their homes. Three BP measurements were taken, with a 5-min interval between each measure. The BP screening protocol used by The Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure recommended a total of three measures, in which the second and the third measures are averaged. Results of the household survey indicated significant differences between the three subsequent BP measurements (P ⬍ 0.001), with much smaller differences between the second and third measure. This result

implies that BP of an individual is approaching stability after the second measure and, consequently, two subsequent BP measurements may be sufficient to identify subjects with elevated BP. We compare the three-measure screening protocol with a two-measure screening protocol in which only the first two BP readings are taken and the second reading is used to indicate BP levels of subjects. The percentage of agreement between the threemeasure and the two-measure screening protocols is 97.57%. Using the three-measure screening protocol as a standard, the sensitivity and false-negative rate are 98.73% and 0.43%, respectively. The two-measurement screening protocol is recommended as an equally sensitive and a more efficient procedure for a large-scale community-screening programme.

Keywords: blood pressure measurement; hypertension; sensitivity; specificity

Introduction The variability of blood pressure (BP) within an individual is a widely recognised phenomenon that has been examined repeatedly for many decades.1–9 However, the stability of BP is still a critical issue not only for clinical treatment of hypertension but also for hypertension screening programme in communities in which individuals are referred for verification and confirmation of elevated BP. To reduce the influence of the fluctuation of BP on the assessment of an elevated pressure, repeated measurements of BP, which frequently consists of three consecutive readings, are a common procedure in determining the BP of a subject. The BP measuring protocol10 used in the 1980s recommended a total of three measures, with the average of the second and third readings being used to assess BP status. This protocol could improve the quality of measurement by averaging out the fluctuation of BP over a short time period. However, multiple readings could increase the possibility of measurement

errors due to variation in measurement instrument or observers’ errors. Furthermore, extra measures increase the duration of the measurement process and, consequently, the unit cost for determining the BP of a subject. This increase could become significant in a large-scale, community-screening programme involving BP measurements of thousands of individuals. How to obtain a balance between the quality and cost of BP measurement is an important concern for public health professionals. This study uses the 1983 California Hypertension Survey Data11 to assess the relative stability of systolic and diastolic BP over a short time period, based on three BP measures which were taken at 5-min intervals. Furthermore, we evaluate the feasibility in using the first two consecutive measures instead of three measures to screen participants with elevated BP.

Materials and methods Sampling design

Correspondence: Donald E Morisky, UCLA School of Public Health, Box 951772 Los Angeles, CA 90095–1772, USA Received 31 March 1999; revised 23 April 1999; accepted 21 May 1999

In 1983, a state-wide hypertension survey was conducted in California to measure the prevalence of elevated BP and hypertension among residents and

Stability of blood pressure Y-C Huang and DE Morisky

638

to obtain additional information about the participants’ health-related behaviours. A three-stage probability sample was drawn to represent the target population of all persons 18 years of age or older residing in ordinary housing units in California during the period of survey.12 Using the 1980 census figures, primary sampling areas were divided into 100 strata. Two primary sampling units were selected from each stratum. Four blocks were selected from each primary sampling unit. Approximately six housing units were selected from each block. In each selected housing unit, one adult was selected at random and asked to complete the full version of the 30-min questionnaire, including measurements of BP, by the interviewer. All other adults in the unit were interviewed with a short version of the questionnaire, also including measurements of BP. Both questionnaires contained basic questions addressing knowledge, awareness, treatment and control of hypertension. The full questionnaire included many additional items regarding health history and health practices. A total of 6381 adults completed the short version of questionnaire. Of these participants, 3346 adults completed the full version of the questionnaire as well. In this study we used the BP readings of 6381 participants to evaluate the stability of BP. Measurement of blood pressure Blood pressure measures were taken approximately 15 min into the questionnaire survey. Participants were asked if they had ever been told by a physician that they had a high BP. Blood pressure was then measured using a standardised, calibrated mercury sphygmomanometer, taking the reading on the right arm with the person seated. Two additional measures were taken at 5 min intervals. During the interval, participants were asked additional questions regarding their knowledge, awareness and practices related to high BP. A total of three measures were recorded for each participant. The average of the second and third measures is used to establish the mean systolic and diastolic BP and to identify the individuals with an elevated BP. Classification of elevated blood pressure Because of the fluctuation of BP, absolutely satisfactory criteria for classification of elevated BP are impossible. Nevertheless, a reasonable and acceptable criterion is still necessary for measuring the prevalence of elevated BP pressure and investigating the fluctuation of BP. In this study, the classification criteria of hypertension recommended from The American Heart Association13 and The Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure10,14,15 are used to identify the individuals with an elevated BP. An individual will be classified as having an elevated BP if his/her systolic BP (SBP) is greater than or equal to 140 mm Hg or his/her diastolic BP (DBP) is greater than or equal to 90 mm Hg or the individual is currently taking antihypertensive medicine. On

the other hand, the individual is normotensive if his/her systolic and diastolic BP is less than 140 mm Hg and 90 mm Hg, respectively. Analytical methods Our analyses sought to describe the fluctuation of BP over a short time period and evaluate the feasibility in using the first two BP measures instead of three measures to assess the BP status of participants. Descriptive analysis: The differences of systolic and diastolic BP among three measures are used to evaluate the variation of BP. The SP12 and SP23 indicate the differences of SBP, (reading 1 − reading 2) and (reading 2 − reading 3), respectively. The DP12 and DP23 indicate the differences of DBP, (reading 1 − reading 2) and (reading 2 − reading 3), respectively. Univariate statistics, including mean and standard deviation of SP12, SP23, DP12 and DP23, are used to describe the variation of BP among the three readings. Furthermore, the fluctuation of BP within each demographic subgroup is discussed. Agreement analysis: The three-BP-measuring protocol recommends a total of three measures, with the average of the second and third readings being used to assess BP status of participants. This protocol is called protocol A. In this study, we propose an alternative protocol that only records two consecutive BP measures and uses the second measure to indicate the BP status of participants. The first reading is discarded due to its instability. This setting is designed not only to reduce the number of measurements but also to avoid the influence of the fluctuation of BP. This alternative protocol is called protocol B. To evaluate the performance of protocol B, we compare the agreement of screening results from protocol B with those from protocol A. We assume protocol A can assess the ‘true’ BP status. The BP reading from protocol A precisely indicates participant’s BP level. Thereafter, we compare the agreement of screening results between the two protocols. The percentages of agreement of all subjects, including sensitivity,16 specificity,16 false positive16 and false negative16 and Kappa statistic16 are calculated to evaluate the consistency of the two protocols. A high sensitivity but a low false negative are particularly useful for our assessment since the primary goal of a screening programme is to identify the higher-risk subjects with undetected elevated BP levels from the community and refer them to verification, screening and confirmation. And if diagnosed with high BP, provide them with early treatment or lifestyle interventions.

Results Demographic characteristics and the prevalence of hypertension A total of 6381 adults, which comprise 2993 males (46.9%) and 3388 females (53.1%), are included into


this study. Ages of participants range from 18 to 96 years, with an average age of 42 years. About 32% of the participants are over the age of 50 years. The major ethnic subgroup are White subjects (65.5%). The demographic characteristics of the participants and the prevalence of elevated BP by each demographic characteristic are summarised in Table 1. Of the 6381 participants, 1570 adults were found to have elevated BP. The prevalence increases with ages of participants. Males are more likely to have elevated BP than females. Elevated BP is also more common among Black subjects than that among the other ethnic subgroups.17–19 The differences among three blood pressure readings The means (standard deviations) of SP12, SP23, DP12 and DP23 given 6381 participants are shown in Table 2. The means are all significantly unequal to zero by paired t-test (P ⬍ 0.001). This implies that the variation of BP really exists among the three readings. We note that the variations between the three BP readings are inclined to decrease and the standard deviations of SP12 and SP23 are greater than those of DP12 and DP23. The BP is less stable between the first two BP readings and the fluctuation in SBP is greater than that in DBP. We also investigate the absolute values of SP12, SP23, DP12 and DP23, which indicate the magnitude of changes between three readings. For 84.6% of participants, the absolute value of SP23 is 5 mm Hg or less; however, only 77.7% of participants show an absolute value of SP12 of 5 mm Hg or less. Similarly, the percentages of participants with the absolute values of DP12 and DP23 being 5 mm Hg or less are 81.7% and 87.2%, respectively. The magnitude of the fluctuation of the BP over a short time is 5 mm Hg or less for most participants. Furthermore, we compare the means and standard Table 1 The demographic characteristics of participants and the percentages of participants who have elevated BP by each demographic factors No. of participants (%)

Total number of participants

6381

No. of participants with elevated BP

Prevalence rate (%)

1570

24.6

Gender Male Females

2993 (46.9) 3388 (53.1)

856 714

28.6 21.1

Age* 18–34 years old 35– 49 years old 50–64 years old 65+ years old

2727 1572 1236 837

(42.8) (24.7) (19.4) (13.1)

222 324 533 486

8.1 20.6 43.1 58.1

Ethnicity Blacks Whites Others

795 (12.5) 4181 (65.5) 1405 (22.0)

253 1065 252

31.8 25.5 17.9

*Nine participants have missing values on age.

deviations of SP12, SP23, DP12 and DP23 by elevated BP status and demographic factors. Table 2 shows that the standard deviations of SP12, SP23, DP12 and DP23 are very different by elevated BP status and demographic factors. The standard deviations among subjects without an elevated BP are less than those among subjects with an elevated BP. The BP of a subject with an elevated BP is less stable over a short time period. Furthermore, the standard deviations are also much higher among males and Black subjects. From Table 2, we note that the fluctuation of SBP is quite different between age groups since the standard deviations of SP12 and SP23 are much higher among individuals 50 years old or older. As age increases, the stability of the SBP decreases. In summary, elevated BP status is a major indicator on the variability of BP. The variability of BP is also altered given different demographic factors. Blood pressure is usually less stable among the demographic subgroups with higher prevalence of elevated BP. For evaluating the interaction effect of elevated BP status and demographic factors, these factors were simultaneously analysed in the analysis of variance model (ANOVA). Table 2 indicates that elevated BP status and ethnicity have the most significant effects on the change of systolic and diastolic BP. Screening of subjects with elevated blood pressure using protocol A and protocol B Table 3 shows the agreement of screening results using protocols A and B. By protocol A, 1570 participants are identified as having elevated BP. The prevalence is 24.6%. By protocol B, 1684 participants are identified as having elevated BP, with prevalence being 26.39%. The prevalence from protocol B could be overestimated by 1.79%, if the BP by protocol A represents the ‘true’ status. Of the 6381 participants, the percentage of agreement on screening subjects with elevated BP is 97.59% and Kappa statistic is 0.9361. Only 154 (2.41%) participants have inconsistent results by the two protocols. This implies that protocol B is nearly as effective as protocol A in screening subjects with elevated BP. Using the results from protocol A as a standard reference, the sensitivity and the false negative are 98.73% and 0.43%, respectively. Using protocol B, 1.37% of subjects who actually have an elevated BP may be incorrectly identified as not having an elevated BP and 0.43% of subjects would have been incorrectly identified as not having an elevated BP. Protocol B is quite effective in determining the ‘true’ subjects with elevated BP due to its relatively high level of sensitivity and low false negative rate. Of 154 participants who have inconsistent results, 58.4% and 61.0% have systolic and diastolic BPs from 135 to 145 mm Hg and from 85 to 95 mm Hg, respectively. Results from this analysis imply that for most adults, two BP readings are sufficient to precisely indicate their BP status. The third BP may only be needed to improve the identification of elevated BP if participant’s second BP reading is

639


640

Table 2 Means and (standard deviations) of differences between BP readings 1, 2 and 3 by elevated BP status, age, gender and ethnic groups Variables

No.

SP12

SP23

DP12

DP23

All participants

6381

0.81 (5.09)

0.57 (4.07)

0.37 (4.52)

0.21 (3.67)

Elevated BP status Participants without elevated BP

4811

Participants with elevated BP

1570

* 0.42 (6.45) 0.95 (4.55)

0.63 (4.98) 0.55 (3.73)

*** −0.01 (5.18) 0.50 (4.27)

0.36 (4.15) 0.16 (3.50)

0.80 (4.53) 0.88 (6.08)

0.53 (3.71) 0.66 (4.72)

0.37 (4.51) 0.39 (4.54)

0.17 (3.70) 0.29 (3.61)

0.87 (5.04) 0.77 (5.12)

0.50 (4.16) 0.63 (4.00)

0.30 (4.68) 0.43 (4.37)

0.18 (3.84) 0.23 (3.52)

0.58 (5.12) 0.91 (5.27) 0.66 (4.47)

** 0.08 (4.21) 0.67 (4.15) 0.53 (3.73)

*** −0.08 (5.00) 0.45 (4.45) 0.41 (4.44)

**** 0.32 (4.19) 0.27 (3.59) −0.02 (3.59)

Age ⬍50 years old

4299

⭓50 years old

2073

Gender Males Females Ethnic Groups Blacks

2993 3388

795

Whites

4181

Others

1405

*In ANOVA analysis, elevated BP status has statistically significant effect on SP12 (P ⬍ 0.05). **In ANOVA analysis, ethnicity has statistically significant effect on SP23 (P ⬍ 0.05). ***In ANOVA analysis, elevated BP status and ethnicity have statistically significant effects on DP12 (P ⬍ 0.05). ****In ANOVA analysis, ethnicity has statistically significant effect on DP23 (P ⬍ 0.05). Table 3 The percentage of agreement on screening subjects with elevated BP by two protocols Protocol B Subjects with elevated BP

Totals

Subjects without elevated BP

Subjects with elevated BP

1550

20

1570

Subjects without elevated BP

134

4677

4811

Totals

1684

4697

6381

Protocol A

The percentage of agreement between two protocols: (1550 + 4677)/6381 = 97.57%. Sensitivity, 1550/1570 = 98.73%; specificity, 4677/4811 = 97.21%; false-negative, 20/4697 = 0.43%; false-positive, 134/1684 = 7.96%. Kappa statistic = 0.9361.

elevated. This implication is particularly important for a large-scale community screening programme because using protocol B can significantly reduce the aggregated cost.

Discussion The result of this study has reconfirmed previous findings that BP varies not only from day-to-day but also throughout the day.20 Though the fluctuation of the BP over a short time period exists, the magnitude of variation is only 2 mm Hg or less for most of the subjects in our analysis. This phenomenon is particularly manifest for the variation between readings two and three, in which approximately 70% of participants had a variation of 2 mm Hg or less. This

implies that the BP of most of adults tends to be more stable after the second reading. The phenomenon of a higher BP reading on the first measurement rather than that on the second or subsequent measurement, have been reported by many investigators.6,20 It has been suggested that this phenomenon is attributed to the temporary stress from the unfamiliar surroundings. Consequently, most BP measuring protocols recommend repeated measurements and discard the first reading. However, the decreasing variation among the repeated measurements could be due to the explanation of regression to the mean or the bias during repeated measures. Therefore, the efficacy of repeated measurements is still debated.21,22 To circumvent this dilemma, this study advocates the two-measure


641

Table 4 The cost estimations for screening 6381 participants using protocol A and protocol B

(1) Time for BP measurements (2) Labour fee for health personnel and facilities (3) Labour fee for screened subjects The direct cost due to personnel and facilities The indirect cost due to subject’s work time loss

The screening test using three measurements (protocol A)

The screening test using two measurements (protocol B)

16 minutes 20 dollars/per h 15 dollars/per h

9 minutes 20 dollar/per h 15 dollars/per h

16 min * 20 dollars = 5.33 dollars

9 min * 20 dollars = 3 dollars

16 min * 15 dollars = 4 dollars

9 min * 15 dollars = 2.25 dollars

The unit cost of the screening on each subject Total cost for screening 6381 subjects

9.33 dollars

5.25 dollars

59534.73 dollars

33500.25 dollars

The difference of unit cost between protocol A and protocol B is 9.33 − 5.25 = 4.08 dollars. The aggregated difference of cost for screening 6381 subjects is 26034.48 dollars.

protocol to replace the single or triple measurement approach. This protocol intends to avoid the influence of the instability of the first BP reading and to minimise the possible biases that occurred during the process of repeated measurements. Our analysis shows that the two-measure protocol is effective to screen for subjects with elevated BP, given that the percentage of consistency between the two-measure and three-measure protocols is 97%. The implication of our findings is particularly important from a cost-effectiveness perspective. The American Heart Association has identified the several steps an individual must take to obtain an accurate measurement.23 This includes preparation and rest period, proper cuff size determination and placement, maximum inflation level, stethoscope placement, inflation/deflation and measurement recording. The approximate time required for a third BP measurement is 7 min. Extrapolating this time to the 6381 participants who were given a third measure amounts to 744.45 hours (7 min × 6381 participants/60 min). Inclusion of other costs associated with screening (labour fee for health personnel and screened subjects), the direct cost for screening each subject could be reduced from 5.33 dollars to 3 dollars. If the indirect cost is considered as well, using protocol B will result in a savings of greater than 4 dollars. While the savings seem to be minimal for each subject, the aggregated savings for a large number of subjects will become significant. Not included in this cost analysis is the additional cost incurred due to misclassification of BP level. However, since the false-negative and false-positive rates are relatively low, the aggregate cost due to misclassification is minimal compared to the cost savings obtained from protocol B (Table 4). Data to conduct a detailed assessment of the actual cost is not available and further research is recommended in order to quantify these costs. From our analysis, most of the inconsistencies between protocols A and B occur when participant’s systolic or diastolic BP is around 135–145 mm Hg or 85–95 mm Hg. This suggests that protocol B is less effective in accurately classifying subjects with BP near the borderline of elevated BP. Though extra

BP readings may improve the accuracy of screening, the same shortcoming could still exist in a protocol utilising three or more measurements. However, this problem generally will not become an obstacle in a community screening programme because the major purpose of a screening programme is to identify the high-risk undetected/unaware population, which includes healthy, asymptomatic individuals with elevated BP. Based on our analysis, protocol B is effective in screening and identifying all possible high-risk participants, although the protocol may slightly overestimate the prevalence of elevated BP. Our two-measurement protocol (protocol B) is also supported by the revised protocol (protocol C) in The Sixth Report of The Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure, 1997.15 Protocol C recommends the average of two or more readings separated by a 2min interval. If the first two readings differ by more than 5 mm Hg, additional readings should be obtained. We re-examine all participants using protocol C and find that 6184 of 6381 participants (96.9%) have consistent screening results by protocols B and C. Protocol C slightly differs from protocol B due to the fact that protocol C requires additional measures when an individual has a difference of greater than 5 mm Hg between the first and second readings. Of the 2215 participants who were found to have a difference of more than 5 mm Hg between the first two readings, the inconsistency of the screening results by the two protocols is 4.47%. This implies that extra BP readings only slightly improve the accuracy of screening results. Even though the first two BP readings differ by more than 5 mm Hg, the two measures are still very effective in indicating BP levels for most individuals. In conclusion, an individual’s BP does tend to fluctuate over a short time period. However, the magnitude of variation is minor in most adults and the two-measure protocol is generally an effective screening procedure among adults. Consequently, the two-measure screening protocol is recommended as an equally sensitive and a more efficient procedure for a large-scale community screening programme.


642

References 1 Clark EG, Schweitzer MD, Glock CY, Vought RL. Studies in hypertension II: variability of daily blood pressure measurements in the same individuals over a three-week period. J Chronic Dis 1956; 4: 469– 476. 2 Clark EG, Schweitzer, MD, Glock CY, Vought RL. Studies in hypertension III: analysis of individual blood pressure changes. J Chronic Dis 1956; 4: 469– 476. 3 Armitage P, Fox W, Rose GA, Tinker CM. The variability of measurement of causal blood pressure, I. A laboratory study. Clin Sci 1966; 30: 325–335. 4 Armitage P, Fox W, Rose GA, Tinker CM. The variability of measurement of causal blood pressure, II. Survey experience. Clin Sci 1966; 30: 337–344. 5 Carey RM, Reid RA, Ayers CR. The Charlottesville blood pressure survey, value of repeated blood-pressure measurements. J Am Med Assoc 1976; 236: 847– 851. 6 Gordon T, Sorlie P, Kannel WB. Problem in assessment of blood pressure: The Framingham Study. Int J Epidemiol 1976; 5: 327–334. 7 Rosner B, Polk BF. The implications of blood pressure variability for clinical and screening purposes. J Chronic Dis 1979; 32: 451– 461. 8 Slaby A, Josifko M. Does sequential automated measurements improve the estimate of resting blood pressure? J Hum Hypertens 1992; 6: 31–34. 9 Celis H et al. For how many days should blood pressure be measured at home in older patients before steady levels are obtained? J Hum Hypertens 1997; 11: 673–677. 10 The Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. The 1988 report of the joint national committee on detection, evaluation and treatment of high blood pressure. Archive Internal Med 1988; 148: 1023–1038. 11 Leonard AR. Final report on the demonstration program on the impact of statewide coordination of high blood pressure control. Volume I: Final Narrative Report, California Department of Health Services, Sacramento, CA, 1985.

12 Piazza T. Methods of sampling and estimating sampling error for 1983 California Hypertension Survey. Technical Report No. 29, 1983. 13 Frohlich ED et al. Report of a special task force appointed by steering committee; American Heart Association: recommendations for human blood pressure determinations by sphygmomanometers. Hypertension 1988; 11: 209A–222A. 14 The Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. The 1993 report of the joint national committee on detection, evaluation and treatment of high blood pressure. Archive Internal Med 1993; 153: 154 –183. 15 The Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. The 1997 report of the joint national committee on detection, evaluation and treatment of high blood pressure. Archive Internal Med 1997; 157: 2413–2446. 16 Fleiss JL. Statistical methods for rates and proportions. 2nd edn. Wiley, New York, 1981. 17 Bohnstedt M, Leonard AR, Trudeau MJ, Bal DG. Hypertension– unexpected research results and program redirection. Am J Prevent Med 1987; 3: 200–205. 18 Cornoni-Huntley J, LaCroix AZ, Havlik RJ. Race and sex differentials in the impact of hypertension in the United States. The National Health and Nutrition Examination Survey I: epidemiologic follow-up study. Archive Internal Med 1989; 149: 780–788. 19 Krieger N, Sidney S. Racial discrimination and blood pressure: The CARDIA study of young black and white adults. Am J Pub Health 1996; 86: 1370–1378. 20 Hypertension Detection and Follow-up Program Cooperative Group. variability of blood pressure and the results of screening in the hypertension detection and follow-up program. J Chron Dis 1978; 31: 651–667. 21 Fagan TC et al. Single versus triplicate measurements of blood pressure and heart rate. Hypertension 1988; 11: 282–284. 22 Shapiro D et al. Single versus triplicate measurements: a commentary on Fagan. Hypertension 1988; 16: 103–105. 23 Hill MN, Grim CM. How to take a precise blood pressure measurement. Am J Nursing 1991; 91: 38– 42.