Skip main navigation

Quantifying Blood Pressure Visit-to-Visit Variability in the Real-World Setting: A Retrospective Cohort Study

Originally publishedhttps://doi.org/10.1161/CIRCOUTCOMES.122.009258Circulation: Cardiovascular Quality and Outcomes. 2023;16

Abstract

Background:

Visit-to-visit variability (VVV) in blood pressure values has been reported in clinical studies. However, little is known about VVV in clinical practice and whether it is associated with patient characteristics in real-world setting.

Methods:

We conducted a retrospective cohort study to quantify VVV in systolic blood pressure (SBP) values in a real-world setting. We included adults (age ≥18 years) with at least 2 outpatient visits between January 1, 2014 and October 31, 2018 from Yale New Haven Health System. Patient-level measures of VVV included SD and coefficient of variation of a given patient’s SBP across visits. We calculated patient-level VVV overall and by patient subgroups. We further developed a multilevel regression model to assess the extent to which VVV in SBP was explained by patient characteristics.

Results:

The study population included 537 218 adults, with a total of 7 721 864 SBP measurements. The mean age was 53.4 (SD 19.0) years, 60.4% were women, 69.4% were non-Hispanic White, and 18.1% were on antihypertensive medications. Patients had a mean body mass index of 28.4 (5.9) kg/m2 and 22.6%, 8.0%, 9.7%, and 5.6% had a history of hypertension, diabetes, hyperlipidemia, and coronary artery disease, respectively. The mean number of visits per patient was 13.3, over an average period of 2.4 years. The mean (SD) intraindividual SD and coefficient of variation of SBP across visits were 10.6 (5.1) mm Hg and 0.08 (0.04). These measures of blood pressure variation were consistent across patient subgroups defined by demographic characteristics and medical history. In the multivariable linear regression model, only 4% of the variance in absolute standardized difference was attributable to patient characteristics.

Conclusions:

The VVV in real-world practice poses challenges for management of patients with hypertension based on blood pressure readings in outpatient settings and suggest the need to go beyond episodic clinic evaluation.

What is Known

  • Based on the data from 537 218 adults and 7 721 864 office-based blood pressure (BP) measurements from a large health system in the United States, the mean absolute change in systolic BP between 2 consecutive visits was about 12 mm Hg.

  • Given the observed visit-to-visit variability (VVV), if an antihypertensive medication truly reduced a patient’s systolic BP by 10 mm Hg, clinicians would expect to observe a reduction of systolic BP <5 mm Hg at the next visit 37% of the time.

  • Less than 5% of VVV in BP values was attributable to patient characteristics.

What the Study Adds

  • The large VVV in office-based BP values poses a great challenge for clinicians to determine hypertension status, therapy indications, and intervention responses in real-world practice.

  • There is a critical need to go beyond episodic clinic evaluation and use home BP monitoring and ambulatory BP monitoring as out-of-office alternatives to establish diagnosis of hypertension and BP control.

In current practice, clinicians generally rely on blood pressure (BP) readings obtained during clinical encounters to detect elevated BP levels and the response to treatment strategies. The challenge is that office-based BP measurements vary in ways that can obscure a hypertension diagnosis or overwhelm a signal of treatment response.1,2 This variation can derive from biological fluctuations, variation in timing, context, and method of measurement.3,4 This variation has led to the recommendation that BP changes should be evaluated with home BP monitoring (HBPM) and ambulatory BP monitoring (ABPM) as out-of-office alternatives,3,5 but that is not common practice due to the cost and limited access to testing.6

The prognostic value of individual BP measurement variation has been well documented. Several studies, primarily using data collected in prospective studies, have documented visit-to-visit variability (VVV) in BP values and its association with the risk of adverse outcomes.7–14 For example, in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) the middle quintile of the systolic BP (SBP) SD was ~10 mm Hg and a higher VVV of SBP was associated with an increased risk of mortality.11 However, information is lacking on the individual BP measurement variation in the real-world settings, where BP measurement, its timing, and context are less standardized. The information is important because, beyond its prognostic value, the real-world BP variation can represent noise that impairs clinical management. Understanding the VVV in office-based BP values in real-world practice is a critical first step toward developing the necessary strategies to support decision-making amid an amplitude of change that may frequently cross BP categories and be larger than the expected effect of many interventions.

Accordingly, we sought to describe real-world VVV using data from 537 218 outpatients with 7 million office-based BP measurements during a 5-years period from a large health system with a diverse patient population. With these data, we employed a novel approach of creating dyads, which reflects what clinicians see when they assess the current visit and compare it to the prior visit. We quantified the VVV using different measures and investigated patient factors associated with VVV in real-world practice. We also estimated the proportion of times that clinicians would expect to observe a clinically meaningful change in BP at the next visit given the observed VVV and the number of visits to be 80% certain that the true BP had decreased by 10 mm Hg.

Methods

The authors declare that all supporting data are available within the article and its Supplemental Material.

Study Population and Data Source

This is a population-based, retrospective cohort study. The study population consisted of adult patients (age ≥18 years) with at least 2 outpatient visits in the Yale-New Haven Health System (YNHHS), from January 1, 2014 through December 31, 2018. YNHHS is the Connecticut’s largest health care system, which provides care for ~3 million residents across Connecticut, New York, and Rhode Island. All YNHHS practice sites use Epic Corporations (Madison, WI) electronic health record system, so that every visit is integrated into the data repository queried for this study.

Assessment of VVV of BP

The primary outcome was VVV, defined as variability in BP between visits. BP levels were measured in the clinics primarily using automated sphygmomanometers (OMRON Healthcare, Inc), although the fidelity to this protocol across different sites was not assessed. We focused on VVV of SBP based on patient-level and dyad-level measures of variation. Consistent with previous studies of VVV,7–11 patient-level measures of variation included SD and coefficient of variation (CV) of a given patient’s SBP across visits. SD was defined as the SD of all SBP measurements for each patient, without restriction on the time duration between visits; CV was defined as SD divided by the mean of all SBP measurements for each patient. In a secondary analysis, we also assessed the average real variability and variability independent of the mean (VIM) across visits. Average real variability was defined as the average of absolute changes between consecutive BP readings.15 VIM was a transformation of SD that was defined to be uncorrelated with mean levels.16 VIM was calculated as the SD of BP readings divided by the mean BP raised to the power of x, where x is obtained from fitting a nonlinear regression model SD SBP against mean SBP among all individuals in the cohort.

We further evaluated the VVV in consecutive visit dyads. A dyad was defined as a pair of any 2 consecutive visits for the same patient occurring within a 90-day window. For example, if a patient had 3 visits during a 90-day window, then the first and second visits were considered as one dyad and the second and third visits were considered as another dyad. We considered that BP values in a dyad with >90 days would not be reliable. The rationale was that clinicians commonly compare the current BP with the BP in the previous visit for a given patient when evaluating an intervention. By using dyad-level measures, we replicated the situation that clinicians see a patient and make assessment about whether there has been interval improvement in real-world practice.

Dyad-level measures of variation included difference, absolute difference, standardized difference, and absolute standardized difference.17 The difference was defined as the difference between SBP measured at the first and second visits of a dyad; absolute difference was defined as the absolute value of difference. Standardized difference was defined as the difference between BP measured at the first and second visits of a dyad, divided by the BP at the first visit; absolute standardized difference was defined as the absolute value of standardized difference. Detailed definitions of patient-level and dyad-level measures of variation are described in Table S1.

Covariates

Among the overall population, we assessed the variations of VVV across subgroups formed by age category (18–44 years, 45–59 years, and ≥60 years), sex (male versus female), race/ethnicity (Asian patients, Hispanic patients, non-Hispanic Black patients, non-Hispanic White patients, and other races), time between visits (<6 weeks versus ≥6 weeks), diagnosis of hypertension (yes versus no), hypertensive stage (normal BP, elevated BP, stage 1 hypertension, stage 2 hypertension, and hypertension crisis), and comorbidities including hyperlipidemia, diabetes, coronary artery disease, and chronic kidney disease. Among patients with hypertension, we assessed the variations of VVV by number of medications changed (increased number of medications, decreased number of medications, same number of medications, no medication) between 2 consecutive visits.

Specifically, hypertension stages included normal BP (SBP <120 mm Hg), elevated BP (SBP 120–129 mm Hg), stage 1 hypertension (SBP 130–139 mm Hg), stage 2 hypertension (SBP ≥140 mm Hg), and hypertension crisis (SBP ≥180 mm Hg). We determined the number of antihypertensive medications changed between 2 visits. A dyad was labeled as increased number of medications, decreased number of medications, same number of medications, or no medications depending on whether between the 2 visits, the number of antihypertensive medications increased, decreased, were both the same non-zero value, and were both zero, respectively. Comorbidities were identified using International Classification of Diseases, Ninth Revision or International Classification of Diseases, Tenth Revision codes listed in Table S2. Comorbidities were defined as any diagnosis during the study period.

Statistical Analysis

We first described the sociodemographic and clinical characteristics of the study population. We plotted the mean SBP and diastolic BP distributions across each patient’s visits, time between visits, length of follow-up, and number of visits. We calculated patient-level measures of variation including SD and CV of a given patient’s SBP across visits, and dyad-level measures of variation including difference, absolute difference, standardized difference, and absolute standardized difference of SBP between 2 consecutive visits. We then assessed how these measures varied across patient subgroup. Among the overall population, we assessed the variations of VVV across patient subgroup formed by age, sex, race/ethnicity, time between visits, diagnosis of hypertension, and comorbidities. Among patients with hypertension, we assessed the variations of VVV by the number of medications changed between 2 consecutive visits. Since patients might have different number of visits and dyads during the study period, we randomly selected a dyad for each patient when calculating the summary statistics of dyad-level measures of variation. We conducted 2 additional sensitivity analyses to test robustness of our results—the first used a different random dyad for each patient and the second used all dyads for each patient in calculating the dyad-level measures of variation.

To assess the extent to which VVV in SBP was explained by patient characteristics, we further developed a multilevel regression model, partitioning the variance by the individual, department, and hospital levels to account for the hierarchical structure of the data.18 Specifically, we fitted a random intercept model to incorporate the random effect of the clustering variables (departments and hospitals) and considered the absolute standardized difference in SBP between 2 consecutive visits as the dependent variable. The individual-level variables included patients’ sex, age, race/ethnicity, treatment status, time from last visit, and comorbidities. The conditional R2 was then reported as a metric of the variance in VVV explained by these covariates and their corresponding structure.

Previous systematic review of clinical trials reported that different antihypertensive mediations had an average effect of lowering SBP by ~10 mm Hg.19 Assuming an antihypertensive medication truly reduced a patient’s SBP by 10 mm Hg, we estimated the proportion of times that clinicians would observe a SBP reduction of <5 mm Hg (a change indicating no difference in BP had occurred) at the next visit (Supplemental Material). We also estimated the proportion of times that clinicians would expect to observe no reduction in patients’ SBP at the next visit. Lastly, we estimated the number of visits it would take to be 80% certain that the true SBP had decreased by 10 mm Hg, assuming SBP measurements in these visits were independent.

All analyses were conducted using R 4.0. All statistical testing was 2-sided at a significance level of 0.05. Institutional review board approval for this study was obtained through the Yale University Human Investigation Committee. The study followed the guidelines for cohort studies, described in the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.20

Results

Population Characteristics

A total of 859 716 adult patients seen in outpatient clinics at YNHHS from January 1, 2014 through December 31, 2018, of which we excluded 21 653 patients with all SBP measurements missing, <40 mm Hg, or >300 mm Hg, and 300 818 patients with all visits separated by >90 days (Figure S1). The final study sample included 537 218 adults, with a total of 7 721 864 SBP measurements over the study period. The mean age was 53.4 (SD 19.0) years, 60.4% were women, 69.4% were non-Hispanic White, and 18.1% were on antihypertensive medications. Patients had a mean body mass index of 28.4 (5.9) kg/m2 and 22.6%, 8.0%, 9.7%, and 5.6% had a history of hypertension, diabetes, hyperlipidemia, and coronary artery disease, respectively (Table).

Table. Characteristics of Study Population at the Patient-Level

CharacteristicsOverall (N=537 218)
Age, y; mean (SD)53.4 (19.0)
Body mass index, mean (SD)28.4 (5.9)
Systolic BP across visits, mean (SD)125.2 (12.8)
Diastolic BP across visits, mean (SD)75.1 (7.5)
Number of visits per patient, mean (SD)13.3 (15.6)
Sex, N (%)
 Female324 543 (60.4%)
 Male212 675 (39.6%)
Race/ethnicity, N (%)
 Non-Hispanic Asian14 880 (2.8%)
 Non-Hispanic Black61 968 (11.5%)
 Non-Hispanic White373 081 (69.4%)
 Hispanic59 796 (11.1%)
 Other/unknown27 493 (5.1%)
Insurance, N (%)
 Private294 714 (54.9%)
 Public216 286 (40.3%)
 Self-pay11 763 (2.2%)
 Other/unknown14 455 (2.7%)
Comorbidities, N (%)
 Hypertension121 485 (22.6%)
 Hyperlipidemia52 148 (9.7%)
 Diabetes43 063 (8.0%)
 Chronic kidney disease9141 (1.7%)
 Coronary artery disease30 082 (5.6%)

BP indicates blood pressure.

Distribution of SBP Measurements

The distributions of SBP, time between visits, length of follow-up, and number of visits are shown in Figure 1. Overall, the mean (SD) number of visits per patient was 13.3 (15.6), over an average period of 2.4 (1.6) years. The average of the mean SBP and diastolic BP measurements across each patient’s visits was 125 (12.8) mm Hg and 75.1 (7.5) mm Hg. The 20 most common primary diagnoses are listed in Table S3, among which essential hypertension, hyperlipidemia, diabetes, cough, acute upper respiratory infection were the top 5 common diagnoses.

Figure 1.

Figure 1. Distributions of systolic blood pressure (mm Hg) across all visits, time (days) between visits, number of visits per patient, and total follow-up time (days) per patient. The y axis represents the probability density of each measure. BP indicates blood pressure.

VVV in SBP Measurements

The distributions of patient-level and dyad-level measures of variation are shown in Figure 2. Overall, the mean (SD) intraindividual SD and CV of SBP across visits were 10.6 (5.1) mm Hg and 0.08 (0.04). This indicated that about 95% of the time SBP values were on average within 21.2 mm Hg between the 2 visits. The secondary analysis of average real variability (11.7 [6.1] mm Hg) and VIM (10.6 [4.6]) confirmed the large variability in office BP measurements. These patient-level measures of variation were consistent across patient subgroups defined by demographic characteristics (Figure 3 and Table S4). For example, the mean (SD) intraindividual SD of SBP across visits was 10.3 (4.9) mm Hg among women and 10.9 (5.2) mm Hg men, and the intraindividual CV of SBP across visits was 0.08 (0.04) for both women and men. Similarly, the average real variability and VIM were 11.4 (5.9) mm Hg and 10.6 (4.4) among women and were 12.1 (6.3) mm Hg and 10.5 (4.8) among men. Notably, there was no patient subgroup with an exceptionally low VVV.

Figure 2.

Figure 2. Distributions of patient-level and dyad-level measures of variation. The y axis represents the probability density of each measure. Abs Diff indicates absolute difference; Abs Std Diff, absolute standardized difference; ARV, average real variability; CV, coefficient of variation; Diff, difference; Std Diff, standardized difference; and VIM, variability independent of the mean.

Figure 3.

Figure 3. Patient-level measures of variation (mean, 25th percentile and 75th percentile) across patient subgroups. BP indicates blood pressure; CAD, coronary artery disease; CKD, chronic kidney disease; and CV, coefficient of variation.

At the dyad-level, the mean (SD) difference in SBP between 2 consecutive visits was −0.7 (15.6) mm Hg, the absolute difference was 11.6 (10.4) mm Hg, the standardized difference was 0 (0.1) mm Hg, and the absolute standardized difference was 0.09 (0.08) mm Hg, respectively (Figure 2). These dyad-level measures of variation were also consistent across patient subgroups defined by demographic characteristics and treatment status (Figure 4 and Table S5). For example, among patients with diagnosis of hypertension, the mean (SD) difference in SBP between 2 consecutive visits was −1.4 (18.4) mm Hg, the absolute difference was 14.0 (12.1) mm Hg, the standardized difference was 0 (0.1) mm Hg, and the absolute standardized difference was 0.1 (0.09) mm Hg, respectively. These numbers were not significantly different from those among patients without diagnosis of hypertension, where the mean (SD) difference in SBP between 2 consecutive visits was −0.4 (14.7) mm Hg, the absolute difference was 11.0 (9.8) mmHg, the standardized difference was 0 (0.1) mm Hg, and the absolute standardized difference was 0.09 (0.08) mm Hg, respectively.

Figure 4.

Figure 4. Dyad-level measures of variation (mean, 25th percentile and 75th percentile) across patient subgroups. Abs Diff indicates absolute difference; Abs Std Diff, absolute standardized difference; CAD, coronary artery disease; CKD, chronic kidney disease; Diff, difference; and Std Diff, standardized difference.

However, these measures differed by hypertension stage of previous visit. Among patients with normal BP (SBP <119 mm Hg), the mean (SD) difference in SBP between 2 consecutive visits was 6.3 (12.8) mm Hg, the absolute difference was 10.5 (9.6) mm Hg, the standardized difference was 0.06 (0.12) mm Hg, and the absolute standardized difference was 0.10 (0.09) mm Hg, respectively. The VVV increased among those with hypertension crisis (SBP >180 mm Hg), where the mean (SD) difference in SBP between 2 consecutive visits was −32.3 (23.5) mm Hg, the absolute difference was 33.9 (21.1) mm Hg, the standardized difference was −0.17 (0.12) mm Hg, and the absolute standardized difference was 0.18 (0.11) mm Hg, respectively. The sensitivity analyses using a different random dyad or all dyads for each patient in calculating the dyad-level measures of variation showed consistent results as the main analysis (Tables S6 and S7).

Variance in VVV Explained by Covariates

The conditional R2 for the multilevel regression model of absolute standardized difference onto covariates was 0.04, suggesting that 4% of the variance in this VVV was attributable to these measured covariates and the nested structure. The results did not change substantially by stage of hypertension.

If an antihypertensive medication reduced a patient’s SBP by 10 mm Hg, given the observed VVV, clinicians would expect to observe no reduction in patient’s SBP at the follow-up visit 25.2% of the time. Clinicians would expect to observe a SBP reduction of <5 mm Hg at the follow-up visit 36.9% of the time. It would take ~4 separate visits to be 80% certain that the true BP had decreased by 10 mm Hg, assuming BP measurements in these visits were independent.

Discussion

In this study of real-world outpatient BP VVV, we demonstrate marked variability that would largely overwhelm the detection of treatment effects between 2 visits. The mean absolute change between 2 consecutive visits was about 12 mm Hg and this was consistent across patient subgroups. Less than 5% of VVV in BP values was attributable to patient characteristics. Given that antihypertensive drugs tend to reduce BP less than this variation, patients and their physicians would routinely not be able to determine whether there is a treatment response based on measurements at a follow-up visit. Moreover, for patients in a decision-making zone of BP, the VVV was sufficiently large to have many patients pass in and out of the treatment range. The VVV has received attention because of its prognostic importance, but it also has immense implications for hypertension treatment management. Understanding the VVV in real-world practice is a critical first step toward developing the necessary strategies to support decision-making. Our findings highlight the importance of using ambulatory or HBPM or novel cuffless wearable technologies, many of which are in development, to provide more robust out-of-office BP evaluation.

This study extends the literature in several ways. We introduce the evaluation of VVV in consecutive visit dyads. The rationale is that clinicians commonly compare BP to the previous visit when evaluating an intervention, even as they may also look at them in the context of other prior measurements. Our analysis provides evidence of marked variability in real-world office-based BP measurements using large, contemporary data from a large health system. It also demonstrates more clearly the challenges VVV may pose for diagnosis, treatment, and monitoring of patients with hypertension based on BP readings in an outpatient setting. Multiple previous studies have demonstrated VVV, and its prognostic significance, using clinical trial data and data from longitudinal cohort studies.7,9,11,12,21,22 For example, by conducting secondary analysis of the data from the ALLHAT trial, Muntner et al11 showed large variation in the degree of VVV across all measurements among study participants and that higher VVV of SBP was associated with increased risk of cardiovascular disease and mortality.

A notable finding of this study was that the mean VVV of real-world BP measurements was similar to that reported in the middle quintile of ALLHAT trial and the quintiles of VVV were similar in the 2 studies (8.7–11.0 mm Hg in ALLHAT compared with 10.6 mm Hg in our study). It might have been expected that VVV in the trial is less than what would have been observed in the real-world setting with standardization of measurement. Although it is challenging to perform a direct comparison between the studies, the fact that the VVV is largely similar suggests that greater standardization of clinician-measured BP may not decrease this variation further.

The cause of VVV is likely a result of many factors. Factors associated with arterial stiffness, including older age, female sex, and higher pulse pressure, have been shown to be associated with VVV suggesting that alterations in vascular function may contribute to greater VVV.10,22 However, in this study, we show that much of the VVV in the real-world population is not related to patient factors, suggesting the need for a broad strategy to standardize BP measurements in every patient. VVV might also be the result of seasonal climatic changes, patient adherence, and response to antihypertensive treatment in treated patients, or may reflect the inconsistencies in the timing, context, measurement device, and setting of BP measurement.23 Human-error in BP measurement due to use of inappropriate cuff size, inappropriate position, and inadequate resting time prior to measurement may also contribute to VVV in BP values.4 The bottom line, though, is that unless we can reduce the noise, we are unlikely to be able to use episodic outpatient measurements to make clinical decisions and improve long-term BP stabilization.

The implications of the study are that it is difficult to detect treatment change in routine practice, which commonly involves having patients return for a BP check. The VVV can obscure meaningful changes following interventions. For example, if an antihypertensive medication might be expected to reduce SBP by 10 mm Hg, a responder might be expected to have a reduction <5 mm Hg in 37% of the time given the VVV. In fact, it would take ~4 separate visits to be 80% certain that the real BP had declined by 10 mm Hg. As such, the common practice of episodic, office-based measurement may be highly insensitive in detecting the types of changes that might be expected with interventions.

There are several options to address the problem. Current clinical decisions about titration of antihypertensive medications are often made based on 1 or 2 office-based BP measurements. Given the large VVV we found this study, some people may argue that they should not be used to evaluate BP and that out of office, more continuous monitoring should be used instead. Another view is that office BP measurements should not be used alone, but should be used in combination with out-of-office measurements, such as ABPM or HBPM recommended by the American Heart Association and US Preventive Services Task Force, as a better method for diagnosing hypertension and evaluating the impact of antihypertensive treatment.3,5 A recent systematic review has shown that ABPM and HBPM have higher sensitivity and specificity in detecting hypertension when comparing with office measurements of BP.1 The gap in using these methods lies in their implementation, which is mainly limited by the technology (cumbersome, expensive, and inconvenient), cost, access to testing, time needed to instruct patients on HBPM protocol, and provider adherence to the correct testing protocol.6 To overcome these gaps in implementation may require improvements in technology, reimbursement, access to testing, provider education, and adoption of clinical decision support. For example, policy changes may be needed to increase incentives for out-of-office testing, especially for ABPM, which is not currently accessible to many patients in the US. To overcome provider concerns about insufficient time to teach patients how to conduct HBPM, community health workers could be trained to teach patients. Barriers related to knowledge of the recent United States Preventive Services Task Force recommendations and how to order ABPM testing could be overcome by clinical decision support tools integrated with the electronic health record that provide guideline recommendations to providers and facilitate test ordering. When ABPM or HBPM is not available, using an average of many office BP measurements may reduce within-patient variability and improve management decisions.24 Artificial intelligence can be used to denoise the variation,25 but it still requires more BP measures. Having people return at same time of day and day of week and taking seasonal changes into account may also potentially reduce some variation. Finally, with the introduction of novel wearable cuffless BP measuring devices, such solutions may be on the horizon.26 Compared with traditional BP monitoring devices, cuffless wearable BP technologies can provide very large numbers of readings for days and months without the cuff-induced discomfort. Although the reliability and clinical usefulness of such technologies still need to be proved, cuffless wearable BP technologies hold the great promise to improve BP screening, monitoring, and management.

Limitations

This study has several potential limitations. First, although BP values in YNHHS were primarily measured by automated sphygmomanometers (OMRON Healthcare, Inc), the use of that device across different sites was not assessed and it is possible that there were legacy devices. Our study focused on information that is available to the clinicians to make management decisions. And in clinic, clinicians often do not have access to information about which device was used for prior measurements. We showed the mean VVV of real-world BP measurements was similar to that reported in clinical trials, suggesting the use of different BP devices in real-world practice is not a major source of VVV. Second, in the data available to us, only 1 BP measurement was documented for each encounter, which could be either a limitation of current practice or a reflection of a limitations in current standards of documentation. Therefore, we were unable to determine how often BP were taken 2 or more times and their impacts on VVV. Third, we did not consider the timing, context, or seasonal changes of BP measurements while estimating VVV. Fourth, treatment of hypertension was ascertained based on medication prescription data and we did not have data on patient adherence to medication. As previous studies reported that ~50% of the patients with hypertension do not take medications as prescribed,27 we might overestimate the true posttreatment VVV. Fourth, when calculating the patient- and dyad-level measures of VVV, patients who had few visits were weighted equally with the patients who had many visits.28 This may result in over or underestimate of the true VVV in the study population. Finally, these data reflect the population catered by the YNHHS and may not necessarily be generalizable to other populations.

Conclusions

There was marked VVV in real-world office-based BP measurements. This finding highlights the challenges VVV pose for diagnosis, treatment, and monitoring of patients with hypertension based on BP readings in outpatient settings, suggesting the need to go beyond episodic clinic evaluation.

Article Information

Supplemental Material

Supplemental Methods

Tables S1–S7

Figures S1

Nonstandard Abbreviations and Acronyms

ABPM

ambulatory blood pressure monitoring

BP

blood pressure

CV

coefficient of variation

HBPM

home blood pressure monitoring

SBP

systolic blood pressure

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

VIM

variability independent of the mean

VVV

visit-to-visit variability

YNHHS

Yale-New Haven Health System

Disclosures In the past 3 years, Dr Krumholz received expenses and/or personal fees from UnitedHealth, Element Science, Aetna, Reality Labs, Tesseract/4Catalyst, F-Prime, and Martin/Baughman Law Firm. He is a co-founder of Refactor Health and HugoHealth, and is associated with contracts, through Yale New Haven Hospital, from the Centers for Medicare & Medicaid Services and through Yale University from Johnson & Johnson. Google, and Pfizer. Dr Lu was supported by the National Heart, Lung, and Blood Institute (K12HL138037) and the Yale Center for Implementation Science for the submitted work. Dr Spatz receives grant funding from the Centers for Disease Control and Prevention (20042801-Sub01), National Institute on Minority Health and Health Disparities (U54MD010711-01), the U.S. Food and Drug Administration to support projects within the Yale-Mayo Clinic Center of Excellence in Regulatory Science and Innovation (CERSI, U01FD005938), the National Institute of Biomedical Imaging and Bioengineering (R01 EB028106-01), and from the National Heart, Lung, and Blood Institute (R01HL151240). Dr Mortazavi received expenses or personal fees from HugoHealth, as a consultant. Dr Khera receives support from the National Heart, Lung, and Blood Institute of the National Institutes of Health under award, 1K23HL153775, and is a founder of Evidence2Health, a precision health and digital health analytics platform. The other authors report no conflicts.

Footnotes

For Sources of Funding and Disclosures, see page 314.

This manuscript was sent to Dr Saket Girotra, MD, SM, Guest Editor, for review by expert referees, editorial decision, and final disposition.

Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/CIRCOUTCOMES.122.009258.

Correspondence to: Harlan M. Krumholz, MD, SM, 195 Church St, 5th Floor, Yale Center for Outcomes Research and Evaluation, New Haven, CT, 06510. Email

REFERENCES

  • 1. Viera AJ, Yano Y, Lin F-C, Simel DL, Yun J, Dave G, Von Holle A, Viera LA, Shimbo D, Hardy ST, et al. Does this adult patient have hypertension?: the rational clinical examination systematic review.JAMA. 2021; 326:339–347. doi: 10.1001/jama.2021.4533CrossrefMedlineGoogle Scholar
  • 2. Parati G, Stergiou GS, Dolan E, Bilo G. Blood pressure variability: clinical relevance and application.J Clin Hypertens (Greenwich). 2018; 20:1133–1137. doi: 10.1111/jch.13304CrossrefMedlineGoogle Scholar
  • 3. Muntner P, Shimbo D, Carey RM, Charleston JB, Gaillard T, Misra S, Myers MG, Ogedegbe G, Schwartz JE, Townsend RR. Measurement of blood pressure in humans: a scientific statement from the American Heart Association.Hypertension. 2019; 73:e35–e66doi: 10.1161/HYP.0000000000000087LinkGoogle Scholar
  • 4. Kallioinen N, Hill A, Horswill MS, Ward HE, Watson MO. Sources of inaccuracy in the measurement of adult patients’ resting blood pressure in clinical settings: a systematic review.J Hypertens. 2017; 35:421–441. doi: 10.1097/hjh.0000000000001197CrossrefMedlineGoogle Scholar
  • 5. Krist AH, Davidson KW, Mangione CM, Cabana M, Caughey AB, Davis EM, Donahue KE, Doubeni CA, Kubik M, Li L. Screening for hypertension in adults: US Preventive Services Task Force reaffirmation recommendation statement.JAMA. 2021; 325:1650–1656doi: 10.1001/jama.2021.4987CrossrefMedlineGoogle Scholar
  • 6. Kronish IM, Kent S, Moise N, Shimbo D, Safford MM, Kynerd RE, O’Beirne R, Sullivan A, Muntner P. Barriers to conducting ambulatory and home blood pressure monitoring during hypertension screening in the United States.J Am Soc Hypertens. 2017; 11:573–580doi: 10.1016/j.jash.2017.06.012CrossrefMedlineGoogle Scholar
  • 7. Chang TI, Reboussin DM, Chertow GM, Cheung AK, Cushman WC, Kostis WJ, Parati G, Raj D, Riessen E, Shapiro B, et al. Visit-to-visit office blood pressure variability and cardiovascular outcomes in SPRINT (Systolic Blood Pressure Intervention Trial).Hypertension. 2017; 70:751–758. doi: 10.1161/hypertensionaha.117.09788LinkGoogle Scholar
  • 8. Diaz KM, Tanner RM, Falzon L, Levitan EB, Reynolds K, Shimbo D, Muntner P. Visit-to-visit variability of blood pressure and cardiovascular disease and all-cause mortality.Hypertension. 2014; 64:965–982. doi: 10.1161/hypertensionaha.114.03903LinkGoogle Scholar
  • 9. Hata J, Arima H, Rothwell PM, Woodward M, Zoungas S, Anderson C, Patel A, Neal B, Glasziou P, Hamet P, et al. Effects of visit-to-visit variability in systolic blood pressure on macrovascular and microvascular complications in patients with type 2 diabetes mellitus: the ADVANCE trial.Circulation. 2013; 128:1325–1334. doi: 10.1161/circulationaha.113.002717LinkGoogle Scholar
  • 10. Muntner P, Shimbo D, Tonelli M, Reynolds K, Arnett DK, Oparil S. The relationship between visit-to-visit variability in systolic blood pressure and all-cause mortality in the general population: findings from NHANES III, 1988 to 1994.Hypertension. 2011; 57:160–166. doi: 10.1161/hypertensionaha.110.162255LinkGoogle Scholar
  • 11. Muntner P, Whittle J, Lynch AI, Colantonio LD, Simpson LM, Einhorn PT, Levitan EB, Whelton PK, Cushman WC, Louis GT, et al. Visit-to-visit variability of blood pressure and coronary heart disease, stroke, heart failure, and mortality: a cohort study.Ann Intern Med. 2015; 163:329–338. doi: 10.7326/m14-2803CrossrefMedlineGoogle Scholar
  • 12. Nwabuo CC, Yano Y, Moreira HT, Appiah D, Vasconcellos HD, Aghaji QN, Viera A, Rana JS, Shah RV, Murthy VL, et al. Association between visit-to-visit blood pressure variability in early adulthood and myocardial structure and function in later life.JAMA Cardiol. 2020; 5:795–801. doi: 10.1001/jamacardio.2020.0799CrossrefMedlineGoogle Scholar
  • 13. Hara A, Thijs L, Asayama K, Jacobs L, Wang JG, Staessen JA. Randomised double-blind comparison of placebo and active drugs for effects on risks associated with blood pressure variability in the Systolic Hypertension in Europe trial.PLoS One. 2014; 9:e103169. doi: 10.1371/journal.pone.0103169CrossrefMedlineGoogle Scholar
  • 14. Schutte R, Thijs L, Liu YP, Asayama K, Jin Y, Odili A, Gu YM, Kuznetsova T, Jacobs L, Staessen JA. Within-subject blood pressure level--not variability--predicts fatal and nonfatal outcomes in a general population.Hypertension. 2012; 60:1138–1147. doi: 10.1161/hypertensionaha.112.202143LinkGoogle Scholar
  • 15. Mena L, Pintos S, Queipo NV, Aizpurua JA, Maestre G, Sulbaran T. A reliable index for the prognostic significance of blood pressure variability.J Hypertens. 2005; 23:505–511. doi: 10.1097/01.hjh.0000160205.81652.5aCrossrefMedlineGoogle Scholar
  • 16. Ebinger JE, Driver M, Ouyang D, Botting P, Ji H, Rashid MA, Blyler CA, Bello NA, Rader F, Niiranen TJ, et al. Variability independent of mean blood pressure as a real-world measure of cardiovascular risk.EClinicalMedicine. 2022; 48:101442. doi: 10.1016/j.eclinm.2022.101442CrossrefMedlineGoogle Scholar
  • 17. Kenny DA, Kashy DA, Cook WL. Dyadic data analysis.Guilford Publications; 2020.Google Scholar
  • 18. Hedeker D, Mermelstein RJ, Demirtas H. An application of a mixed-effects location scale model for analysis of Ecological Momentary Assessment (EMA) data.Biometrics. 2008; 64:627–634. doi: 10.1111/j.1541-0420.2007.00924.xCrossrefMedlineGoogle Scholar
  • 19. Wu J, Kraja AT, Oberman A, Lewis CE, Ellison RC, Arnett DK, Heiss G, Lalouel JM, Turner ST, Hunt SC, et al. A summary of the effects of antihypertensive medications on measured blood pressure.Am J Hypertens. 2005; 18:935–942doi: 10.1016/j.amjhyper.2005.01.011CrossrefMedlineGoogle Scholar
  • 20. Von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP, Initiative S. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies.Int J Surg. 2014; 12:1495–1499doi: 10.1016/j.ijsu.2014.07.013CrossrefMedlineGoogle Scholar
  • 21. Clark D, Nicholls SJ, St John J, Elshazly MB, Ahmed HM, Khraishah H, Nissen SE, Puri R. Visit-to-visit blood pressure variability, coronary atheroma progression, and clinical outcomes.JAMA Cardiol. 2019; 4:437–443doi: 10.1001/jamacardio.2019.0751CrossrefMedlineGoogle Scholar
  • 22. Rothwell PM, Howard SC, Dolan E, O’Brien E, Dobson JE, Dahlöf B, Sever PS, Poulter NR. Prognostic significance of visit-to-visit variability, maximum systolic blood pressure, and episodic hypertension.Lancet. 2010; 375:895–905. doi: 10.1016/s0140-6736(10)60308-xCrossrefMedlineGoogle Scholar
  • 23. Parati G, Ochoa JE, Lombardi C, Bilo G. Assessment and management of blood-pressure variability.Nat Rev Cardiol. 2013; 10:143–155. doi: 10.1038/nrcardio.2013.1CrossrefMedlineGoogle Scholar
  • 24. Powers BJ, Olsen MK, Smith VA, Woolson RF, Bosworth HB, Oddone EZ. Measuring blood pressure for decision making and quality reporting: where and how many measures?.Ann Intern Med. 2011; 154:781–8, w-289-90. doi: 10.7326/0003-4819-154-12-201106210-00005CrossrefMedlineGoogle Scholar
  • 25. Chen S, Ji Z, Wu H, Xu Y. A non-invasive continuous blood pressure estimation approach based on machine learning.Sensors (Basel). 2019; 19:2585doi: 10.3390/s19112585CrossrefMedlineGoogle Scholar
  • 26. Kario K. Management of hypertension in the digital era: small wearable monitoring devices for remote blood pressure monitoring.Hypertension. 2020; 76:640–650. doi: 10.1161/hypertensionaha.120.14742LinkGoogle Scholar
  • 27. Abegaz TM, Shehab A, Gebreyohannes EA, Bhagavathula AS, Elnour AA. Nonadherence to antihypertensive drugs: a systematic review and meta-analysis.Medicine (Baltim). 2017; 96:e5641. doi: 10.1097/md.0000000000005641CrossrefMedlineGoogle Scholar
  • 28. Powers BJ, Olsen MK, Smith VA, Woolson RF, Bosworth HB, Oddone EZ. Measuring blood pressure for decision making and quality reporting: where and how many measures?.Ann Intern Med. 2011; 154:781–788. doi: 10.7326/0003-4819-154-12-201106210-00005CrossrefMedlineGoogle Scholar