Health Reports
From BpTRU to OMRON: The impact of changing automated blood pressure measurement devices on adult population estimates of blood pressure and hypertension
by Tracey Bushnik, Heather Gilmour, Vincent Mak and Anne Mather
DOI: https://www.doi.org/10.25318/82-003-x202401000001-eng
Abstract
Background
In Cycle 7 (2022), the Canadian Health Measures Survey (CHMS) introduced the OMRON (OM) IntelliSense HEM-907XL blood pressure (BP) monitor after using the BpTRU (BT) BPM-300 BP monitor for six cycles. This study assessed differences between adult BP values measured by both devices and whether equations could be developed to compare BP measurements taken using the two devices.
Data and methods
In Cycle 6 (2018 to 2019) of the CHMS, BP was measured using BT and OM devices. Between-device estimates of systolic BP (SBP), diastolic BP (DBP), and hypertension prevalence were compared for 1,072 adults aged 18 to 79 years. Sex, age, body mass index (BMI), and the use of antihypertensive medication were examined in linear regression models to predict SBP and DBP values measured with OM based on those measured with BT.
Results
Average SBP measured with OM was 6 millimetres of mercury (mmHg) higher than average SBP measured with BT, and average DBP measured with OM was 2 mmHg lower than DBP measured with BT. Hypertension prevalence based on OM readings was 35.4%, compared with 34.0% based on BT readings. Between-device BP differences varied in magnitude by sex, age group, and BMI category. Average model-predicted estimates of BP were comparable to measured estimates, but predicted values were lower at higher levels of BP.
Interpretation
Switching BP monitors will substantively affect population estimates of adult SBP and DBP but may have less impact on estimates of adult hypertension prevalence. The prediction equations proposed in this study can be applied to adult BP data from cycles 1 to 6 of the CHMS for comparison with BP measurements taken in Cycle 7, with some caveats. The impact of changing to the OM monitor in Cycle 7 should be acknowledged when reporting estimates of adult BP based on the CHMS.
Keywords
Blood pressure measurement, crossover study, automated sphygmomanometers, hypertension, prevalence.
Authors
Tracey Bushnik and Heather Gilmour are with the Health Analysis Division at Statistics Canada. Vincent Mak is with the Centre for Direct Health Measures at Statistics Canada. Anne Mather is with the Social Statistics Methods Division at Statistics Canada.
What is already known on this subject?
- Hypertension affects close to one in four adult Canadians and is a risk factor for chronic disease such as cardiovascular, cerebrovascular, and renovascular diseases.
- Blood pressure (BP) measurement can vary across methods and devices. Thus, ongoing surveillance via population health surveys may be hampered when the measurement device is changed.
- In Cycle 7, the Canadian Health Measures Survey (CHMS) introduced the OMRON (OM) IntelliSense HEM-907XL BP monitor after six cycles of using the BpTRU (BT) BPM-300 BP monitor.
What does this study add?
- This study compares BP measurements taken by the OM and BT devices on a sample of adult individuals aged 18 to 79 years during a single visit to the mobile examination centre in Cycle 6 of the CHMS.
- Average systolic BP (SBP) measured with OM was 6 millimetres of mercury (mmHg) higher than average SBP measured with BT, and average diastolic BP (DBP) measured with OM was 2 mmHg lower than DBP with BT in the study sample.
- Average hypertension prevalence based on BP measured with OM was 1.4 percentage points higher than hypertension prevalence based on BP measured with BT.
- The prediction equations proposed in this study can be applied to adult BP data from cycles 1 to 6 for comparison with BP measurements in Cycle 7, with some caveats.
- The change in BP monitor will substantively affect population estimates of adult SBP and DBP but may have less impact on estimates of adult hypertension prevalence. The impact of changing to OM in Cycle 7 should be acknowledged when reporting estimates of adult BP based on the CHMS.
Introduction
Hypertension is common in Canada, affecting almost one-quarter of the adult population aged 20 to 79 years.Note 1 It is an important modifiable risk factor for chronic disease and associated with increased risk of morbidity and mortality.Note 2 Ongoing surveillance via population health surveys is necessary to monitor and assess the population burden of hypertension in Canada.
Approximately one in five Canadian adults with hypertension is unaware of their condition,Note 1 and self-reported blood pressure (BP) tends to underestimate the prevalence of hypertension.Note 3 Since 2007, the Canadian Health Measures Survey (CHMS) has used validated, standardized automated BP measurementNote 4 to monitor systolic BP (SBP), diastolic BP (DBP), and the prevalence of hypertension in the population.
The BpTRU (BT) BPM-300 BP monitor was used for cycles 1 to 6 (2007 to 2019) of the CHMS for BP measurement. However, a replacement device was needed after BT ceased operations in 2017. After consultation with experts and other national health surveys, the validated OMRON (OM) IntelliSense HEM-907XL BP monitorNote 5 was chosen for BP measurement starting in Cycle 7. Because BP measurement can vary across methods and devices,Note 6, Note 7, Note 8 a crossover study was implemented in Cycle 6 to compare measurements taken by the two devices on the same individuals aged 6 to 79 years.
This analysis estimates SBP, DBP, and hypertension prevalence based on measurements from BT and OM devices and compares differences across selected characteristics for adults aged 18 to 79 years. It also assesses the feasibility of developing prediction equations so BP data from earlier CHMS cycles can be compared with BP estimates from Cycle 7 onward. The crossover study data for children and youth aged 6 to 17 years were analyzed with the same objectives, and the results have been published separately.Note 9
Data and methods
Data
Data are from Cycle 6 of the CHMS, a nationally representative cross-sectional survey that collects questionnaire and directly measured health information from community-dwelling individuals aged 3 to 79 years living in the 10 provinces. People living in the three territories or on reserves or Indigenous settlements in the provinces, the institutionalized population, residents of certain remote regions, and full-time members of the Canadian Forces are excluded (about 4% of the Canadian population). The CHMS involves an in-person household interview and a subsequent visit to a mobile examination centre (MEC). The household interview gathers general demographic and socioeconomic data and detailed information on health, nutrition, medication use, and lifestyle. At the MEC, direct physical measurements are taken, including BP, height, and weight. More information about the CHMS is available online.Note 10
Crossover study protocol
Cycle 6 of the CHMS (January 2018 to December 2019) collected automated digital BP measurements from all respondents using the BT BPM-300 BP monitor (BpTRU Medical Devices Limited, Coquitlam, British Columbia). The BT monitor takes six readings, with a one-minute interval between the beginning of each consecutive measurement, and averages the last five to determine SBP and DBP levels.Note 11 Crossover study respondents also had their BP measured using the OM IntelliSense HEM-907XL BP monitor (OMRON Healthcare Incorporated, Vernon Hills, Illinois). Average mode was used, in which the BT monitor takes three readings, each one minute apart, and averages all three to determine SBP and DBP levels.Note 12
The last eight sites in Cycle 6 were available for the crossover study. The number of sites and the target sample sizes were selected to ensure that differences in the devices could be measured for each age group independently: ages 6 to 19 years, two sites (n=154); ages 20 to 39 years, four sites (n=210); ages 40 to 59 years, six sites (n=338); and ages 60 to 79 years, eight sites (n=491). Respondents in the crossover study were randomly assigned to have their measurements taken with BT or OM first, based on whether their randomly generated identification number was even or odd.
A health measures specialist (HMS) or a medical radiation technologist (MRT) trained in BP measurement explained the protocol of the crossover study to the respondent at the MEC. The HMS or MRT measured the respondent’s mid-arm circumference for cuff placement and selected the appropriate cuff size for each BP monitor. The cuff for the monitor to be used first was then placed on the respondent’s arm. If the BT monitor was used first, the respondent was left alone for a five-minute rest period, after which the HMS or MRT returned to the room. The HMS or MRT then pressed start, remained for the first reading, then left the room for approximately five minutes while the BT monitor took the remaining five readings. If the OM monitor was used first, the HMS or MRT pressed start to activate the five-minute timer on the machine then left and returned to the room after approximately nine minutes. During this time, the OM monitor automatically took three readings following the five-minute rest period. No additional five-minute rest period was used after the first BP monitor, and the HMS or MRT did not speak to the respondent or move the respondent’s position while changing the cuff and starting the protocol for the second BP monitor.
Age group targets were surpassed during collection, resulting in 1,341 crossover study respondents with valid BP measurements from OM and BT devices. Those aged 18 to 79 years were eligible for this analysis, a total of 1,104 respondents. Less than 3% of eligible respondents were excluded from the analysis for the following reasons: having a difference between their average BP reading from BT and OM devices over three standard deviations from the overall mean difference (SBP and DBP, n=1; SBP only, n=6; and DBP only, n=13), being pregnant (n=4), or having insufficient data to derive body mass index (BMI) (n=8). The final analytical sample was 1,072.
Measures and definitions
BP: For the BT device, mean SBP and DBP were estimated from at least one and up to five valid measurements; less than 3% of mean SBP and DBP values were based on fewer than five measurements. For the OM device, mean SBP and DBP were estimated from at least one and up to three valid measurements; less than 1% of mean values were based on fewer than three measurements.
Antihypertensive medication use: Current medications were recorded during the household and MEC interviews, and these were assigned to codes from the Anatomical Therapeutic Chemical (ATC) classification system, corresponding to beta blockers (ATC codes in category C07, excluding C07AA07, C07AA12, and C07AG02), agents acting on the renin-angiotensin system (ATC codes in category C09), thiazide diuretics (ATC codes in category C03, excluding C03BA08 and C03CA01), calcium channel antagonists (ATC codes in category C08), and other antihypertensive agents (ATC codes in category C02, excluding C02KX01).
Hypertension: Individuals were classified as hypertensive if they had a mean SBP of 135 millimetres of mercury (mmHg) or higher, had a mean DBP of 85 mmHg or higher, or reported using antihypertensive medication in the past month. These SBP and DBP thresholds are consistent with those recommended for defining hypertension using automated BP measurements in clinical settings.Note 13
Height: Individuals’ height was measured to the nearest 0.1 cm using a ProScale M150 digital stadiometer (Accurate Technology Incorporated, Fletcher, United States).
Weight: Individuals’ weight was measured to the nearest 0.1 kg with a Mettler Toledo VLC with Panther Plus terminal scale (Mettler Toledo Canada, Mississauga, Canada).
BMI: Individuals’ BMI was calculated as their measured weight in kg divided by the square of their measured height in m. Three categories of BMI were examined: neither overweight nor obese (BMI less than 25 kg/m2), overweight (BMI of 25 kg/m2 to less than 30 kg/m2), or obese (BMI of 30 kg/m2 or higher).
Covariates
Sex at birth (male or female) and age in years were reported at the visit to the MEC.
Analytical techniques
Mean SBP and DBP, mean differences, hypertension prevalence, and prevalence differences were estimated from BT and OM measurements across selected characteristics, stratified by sex. Mean SBP and DBP estimates were rounded to the nearest integer. The frequency distribution of absolute between-device differences in SBP and DBP was estimated by sex. Sex-specific Bland–Altman plots illustrated device differences in measured SBP and DBP, stratified by hypertension status. Intraclass correlation coefficients (ICCs) quantified the degree of correlation and agreement between the two sets of measurements for each stratum. ICC estimates and their 95% confidence intervals (CIs) were calculated based on a k=1, consistency, two-way mixed-effects model.Note 14
The feasibility of developing prediction equations to compare BP measurements based on BT readings with those based on OM readings was assessed. Overall and sex-specific scatterplots and histograms were used to examine the correlation between and the distribution of SBP and DBP from both devices according to age group, BMI category, and antihypertensive medication use. The REG procedure in SAS was used to estimate separate SBP and DBP base models with SBP (or DBP) measured with OM as the outcome and SBP (or DBP) measured with BT as the sole independent variable. Sex and an interaction between sex and the BT value were then tested in each model to assess the need for sex-specific models. Age in years and BMI in kg/m2 and their polynomials (quadratic, cubic) and antihypertensive medication use were added sequentially to each base model. Model fit statistics, including likelihood ratio tests of nested models, residual plots, and predicted R-squared values, were evaluated. Sex-specific SBP and one overall DBP equation—base and with covariates (full)—were generated from the final selected models. The HPREG procedure with forward selection and the Schwarz Bayesian information criterion specified for model fit were then used to confirm the model selection. Regression coefficients were truncated to five decimal places, and the predicted R-squared was rounded to two decimal places. Sex-specific probability density plots of predicted versus measured SBP and DBP stratified by hypertension status were generated. Differences between predicted and measured mean SBP and DBP (paired t-tests with statistical significance set at p < 0.05) and hypertension prevalence (McNemar’s tests with statistical significance set at p < 0.05) were also assessed by sex, age group, BMI category, and antihypertensive medication use.
BP is highly variable,Note 15 and consecutive BP measurements tend to regress to the meanNote 16 (e.g., low initial BP values tend to rise, while higher values tend to fall). In a sensitivity analysis, device order—whether the BT device was used first or not—was added as a covariate in each final model, and the model fit statistics and predicted values produced by these adjusted models were assessed.
To evaluate whether crossover sample differences translated into population-level differences, weighted estimates of SBP, DBP, and hypertension prevalence based on BT measurements, OM measurements, and OM predictions were calculated. These weighted estimates were generated using survey weights and bootstrap weights of respondents in the crossover sample that were calibrated on the full Cycle 6 population by age group (18 to 19 years, 20 to 39 years, 40 to 59 years, and 60 to 79 years) and sex.
All analyses were conducted in SAS Enterprise Guide 9.4 and SAS-callable SUDAAN 11.0.3 except for the ICCs and the probability density plots. The psych packageNote 17 was used to generate the ICCs and the ggplot2, ggpubr, and scales packagesNote 18, Note 19, Note 20 were used to generate the density plots in RStudio 4.1.3.
Results
OMRON versus BpTRU
The average age of the study sample was 53 years. Slightly over half were women (n=576) or under the age of 60 (n=610), two-thirds were overweight or obese (n=709), and about one in four (n=260) were taking antihypertensive medication (Table 1). Overall mean SBP measured with OM was 6 mmHg higher than that measured with BT. Mean DBP measured with OM was 2 mmHg lower than that measured with BT. The average between-device differences across characteristics ranged from +2 to +11 mmHg for SBP and from -4 to 0 mmHg for DBP. The mean difference in SBP between the two devices was larger for men (7 mmHg) than for women (4 mmHg). It decreased for both sexes with increasing age and increasing BMI but varied little according to antihypertensive medication use. The difference in DBP between the devices was similar across sexes, varied somewhat by age group, and decreased with increasing body mass, mainly for women.
| n | Mean SBP (mmHg) | Mean DBP (mmHg) | Hypertension prevalence | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OM | BT | Difference (OM - BT) |
OM | BT | Difference (OM - BT) |
OM | BT | Difference (OM - BT) |
||
| Total | 1,072 | 120 | 115 | 6 | 70 | 72 | -2 | 35.4 | 34.0 | 1.4Table 1 Note † |
| Sex | ||||||||||
| Male | 496 | 123 | 116 | 7 | 71 | 74 | -3 | 40.5 | 39.3 | 1.2 |
| Female | 576 | 118 | 114 | 4 | 69 | 71 | -2 | 30.9 | 29.3 | 1.6 |
| Age group | ||||||||||
| 18 to 39 years | 233 | 112 | 105 | 8 | 67 | 69 | -2 | 4.3 | 4.3 | 0.0 |
| 40 to 59 years | 377 | 118 | 112 | 5 | 71 | 74 | -2 | 26.0 | 26.8 | -0.8 |
| 60 to 64 years | 171 | 125 | 120 | 5 | 73 | 75 | -2 | 52.6 | 50.3 | 2.3 |
| 65 to 69 years | 105 | 126 | 120 | 6 | 71 | 73 | -2 | 48.6 | 41.9 | 6.7Table 1 Note † |
| 70 to 74 years | 114 | 129 | 123 | 6 | 68 | 71 | -2 | 64.9 | 60.5 | 4.4Table 1 Note † |
| 75 to 79 years | 72 | 129 | 126 | 3 | 67 | 70 | -3 | 77.8 | 75.0 | 2.8 |
| Men | ||||||||||
| 18 to 39 years | 93 | 118 | 107 | 11 | 66 | 69 | -3 | 5.4 | Note F: too unreliable to be published | Note ...: not applicable |
| 40 to 59 years | 177 | 122 | 116 | 7 | 74 | 76 | -3 | 32.8 | 33.9 | -1.1 |
| 60 to 64 years | 92 | 125 | 118 | 7 | 72 | 75 | -3 | 56.5 | 54.3 | 2.2 |
| 65 to 69 years | 56 | 126 | 119 | 7 | 71 | 75 | -3 | 51.8 | 50.0 | 1.8 |
| 70 to 74 years | 54 | 126 | 119 | 8 | 67 | 71 | -3 | 68.5 | 64.8 | 3.7 |
| 75 to 79 years | 24 | 126 | 123 | 3Table 1 Note ‡ | 66 | 71 | -4 | 83.3 | 79.2 | 4.2 |
| Women | ||||||||||
| 18 to 39 years | 140 | 109 | 103 | 6 | 67 | 68 | -2 | 3.6 | 5.0 | -1.4 |
| 40 to 59 years | 200 | 113 | 109 | 4 | 69 | 71 | -2 | 20.0 | 20.5 | -0.5 |
| 60 to 64 years | 79 | 126 | 122 | 4 | 74 | 75 | -1 | 48.1 | 45.6 | 2.5 |
| 65 to 69 years | 49 | 126 | 122 | 4 | 70 | 72 | -2 | 44.9 | 32.7 | 12.2Table 1 Note † |
| 70 to 74 years | 60 | 132 | 127 | 5 | 69 | 71 | -1 | 61.7 | 56.7 | 5.0 |
| 75 to 79 years | 48 | 130 | 127 | 3 | 67 | 69 | -2 | 75.0 | 72.9 | 2.1 |
| Body mass index category | ||||||||||
| Neither overweight nor obese | 363 | 116 | 108 | 7 | 65 | 69 | -3 | 21.2 | 19.3 | 1.9 |
| Overweight | 406 | 122 | 117 | 6 | 71 | 73 | -2 | 38.4 | 36.2 | 2.2 |
| Obese | 303 | 124 | 120 | 4 | 74 | 75 | -1 | 48.2 | 48.5 | -0.3 |
| Men | ||||||||||
| Neither overweight nor obese | 122 | 120 | 110 | 10 | 66 | 69 | -4 | 26.2 | 24.6 | 1.6 |
| Overweight | 212 | 123 | 116 | 7 | 71 | 74 | -3 | 39.6 | 38.2 | 1.4 |
| Obese | 162 | 125 | 119 | 6 | 74 | 76 | -2 | 52.5 | 51.9 | 0.6 |
| Women | ||||||||||
| Neither overweight nor obese | 241 | 114 | 108 | 6 | 65 | 68 | -3 | 18.7 | 16.6 | 2.1 |
| Overweight | 194 | 121 | 117 | 4 | 71 | 73 | -1 | 37.1 | 34.0 | 3.1 |
| Obese | 141 | 122 | 121 | 2 | 73 | 74 | 0Table 1 Note ‡ | 43.3 | 44.7 | -1.4 |
| Antihypertensive medication use | ||||||||||
| Yes | 260 | 128 | 122 | 6 | 71 | 73 | -2 | 100.0 | 100.0 | 0.0 |
| No | 812 | 118 | 112 | 6 | 70 | 72 | -2 | 14.7 | 12.8 | 1.8Table 1 Note † |
| Men | ||||||||||
| Yes | 143 | 126 | 119 | 8 | 71 | 73 | -3 | 100.0 | 100.0 | 0.0 |
| No | 353 | 122 | 114 | 7 | 71 | 74 | -3 | 16.4 | 14.7 | 1.7 |
| Women | ||||||||||
| Yes | 117 | 129 | 125 | 4 | 71 | 72 | -1 | 100.0 | 100.0 | 0.0 |
| No | 459 | 116 | 111 | 4 | 69 | 71 | -2 | 13.3 | 11.3 | 2.0 |
|
... not applicable F too unreliable to be published
Source: Statistics Canada, Canadian Health Measures Survey crossover study, 2018 to 2019. |
||||||||||
About 44% of participants had a between-device absolute difference in SBP of 5 mmHg or less, and 80% had an absolute difference in DBP of 5 mmHg or less (Chart 1). Absolute between-device agreement differences in SBP were within 5 mmHg for 52% of women versus 34% of men (data not shown). Differences in DBP were within 5 mmHg for 83% of women versus 76% of men (data not shown).
Description of Chart 1
| Absolute difference (millimetres of mercury) | Systolic blood pressure |
|---|---|
| percent | |
| 0 to 2 | 17.6 |
| 3 to 5 | 26.3 |
| 6 to 10 | 33.3 |
| 11 to 15 | 17 |
| 16 or more | 5.8 |
| Diastolic blood pressure | |
| 0 to 2 | 40.8 |
| 3 to 5 | 38.9 |
| 6 to 10 | 18.8 |
| 11 to 15 | 1.6 |
| Source: Statistics Canada, Canadian Health Measures Survey crossover study, 2018 to 2019. | |
For men with hypertension, the difference between the BT and OM measurements decreased with increasing SBP (Figure 1-A). For men without hypertension (Figure 1-C) and women with or without hypertension (figures 2-A and 2-C), the difference between BT and OM remained relatively stable regardless of SBP level. For both sexes, the difference between BT and OM decreased with increasing DBP regardless of hypertension status (figures 1-B, 1-D, 2-B, and 2-D). The ICCs for those with hypertension ranged from 0.93 (95% CI: 0.90, 0.94) to 0.95 (95% CI: 0.93, 0.96), indicating excellent agreement and correlation between the two devices,Note 14 and from 0.80 (95% CI: 0.76, 0.84) to 0.90 (95% CI: 0.88, 0.92) for those without hypertension (good to excellent agreement and correlation).
Description of Figure 1
Figure 1 presents Bland–Altman plots of the differences in blood pressure measurements between the OMRON (OM) and BpTRU (BT) devices for men, by hypertension status. It has four panels, labelled A, B, C, and D. Each panel contains a scatterplot overlaid with a regression line. In each panel, the x-axis is the average blood pressure measurement from both devices in millimetres of mercury (mmHg), while the y-axis is the difference in blood pressure measurements in mmHg between the OM and BT devices. Panel A presents systolic blood pressure (SBP) for men with hypertension. The regression line shows that an approximate +10 mmHg between-device difference at an average SBP of 90 mmHg decreases to about a +4 mmHg difference as average SBP approaches 170 mmHg. Panel B presents diastolic blood pressure (DBP) for men with hypertension. The regression line shows that an approximate -5 mmHg between-device difference at an average DBP of 50 mmHg becomes close to 0 mmHg as average DBP approaches 100 mmHg. Panel C presents SBP for men without hypertension. The regression line shows that an approximate between-device difference of +7 mmHg at an average SBP of 90 mmHg remains unchanged as average SBP approaches 135 mmHg. Panel D presents DBP for men without hypertension. The regression line shows that an approximate -5 mmHg between-device difference at an average DBP of 50 mmHg gradually shrinks to -1 mmHg as average DBP approaches 85 mmHg.
Description of Figure 2
Figure 2 presents Bland–Altman plots of the differences in blood pressure measurements between the OMRON (OM) and BpTRU (BT) devices for women, by hypertension status. It has four panels, labelled A, B, C, and D. Each panel contains a scatterplot overlaid with a regression line. In each panel, the x-axis is the average blood pressure measurement from both devices in millimetres of mercury (mmHg), while the y-axis is the difference in blood pressure measurements in mmHg between the OM and BT devices. Panel A presents systolic blood pressure (SBP) for women with hypertension. The regression line shows that an approximate between-device difference of +4 mmHg at an average SBP of 90 mmHg remains unchanged as average SBP approaches 170 mmHg. Panel B presents diastolic blood pressure (DBP) for women with hypertension. The regression line shows that an approximate -4 mmHg between-device difference at an average DBP of 50 mmHg becomes close to 0 mmHg as average DBP approaches 100 mmHg. Panel C presents SBP for women without hypertension. The regression line shows that an approximate between-device difference of +5 mmHg at an average SBP of 90 mmHg remains relatively unchanged as average SBP approaches 135 mmHg. Panel D presents DBP for women without hypertension. The regression line shows that an approximate -5 mmHg between-device difference at an average DBP of 50 mmHg becomes 0 mmHg as average DBP approaches 85 mmHg.
Overall OM-based hypertension prevalence was 35.4%, an estimated 1.4 percentage points higher than BT-based prevalence (34.0%) (Table 1). The difference in the prevalence of hypertension between the two devices was not consistent across sexes, age groups, or BMI categories.
Predicted versus measured blood pressure
The prediction equations are presented in Table 2. The predicted R-squared for the base models ranged from 0.82 to 0.87 and for the full models from 0.84 to 0.88. The full SBP and DBP models performed better than the base models for both sexes across age groups, BMI categories, and antihypertensive medication use categories (Table 3). The full SBP and DBP models produced predicted mean values that were no different than measured mean values, apart from the predicted SBP among individuals aged 75 to 79 years (difference of +2 mmHg).
| Equations | Predicted R-squared |
|
|---|---|---|
| Men 18 to 79 years | ||
| SBP | ||
| Base model | OMP-SBP = 24.35235 + 0.85321×(BT-SBP) | 0.82 |
| Full model | OMP-SBP = 37.39698 + 0.90131×(BT-SBP) - 0.42297×(age) + 0.00313×(age2) - 0.22358×(bmi) + 2.58234×(Bpmed) | 0.84 |
| Women 18 to 79 years | ||
| SBP | ||
| Base model | OMP-SBP = 13.25715 + 0.92145×(BT-SBP) | 0.87 |
| Full model | OMP-SBP = 23.13778 + 0.93959×(BT-SBP) - 0.23370×(age) + 0.00238×(age2) - 0.25365×(bmi) | 0.88 |
| Men and women 18 to 79 years | ||
| DBP | ||
| Base model | OMP-DBP = -1.57930 + 0.99070×(BT-DBP) | 0.86 |
| Full model | OMP-DBP = -8.88739 + 0.97550×(BT-DBP) - 1.51526×(male) - 0.02247×(age) + 0.53322×(bmi) - 0.00559×(bmi2) | 0.88 |
|
Notes: SBP = systolic blood pressure, DBP = diastolic blood pressure, BT = BpTRU, OM = OMRON, age = age in years (continuous), bmi = body mass index (continuous), Bpmed = taking antihypertensive medication (1 = yes, 0 = no), and male (1 = yes, 0 = no). OMP indicates OM-predicted SBP and DBP. Source: Statistics Canada, Canadian Health Measures Survey crossover study, 2018 to 2019. |
||
| Mean SBP (mmHg) | Mean DBP (mmHg) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OM measured |
Base model | Full model | OM measured |
Base model | Full model | |||||
| OM predicted |
Difference (OM predicted - OM measured) |
OM predicted |
Difference (OM predicted - OM measured) |
OM predicted |
Difference (OM predicted - OM measured) |
OM predicted |
Difference (OM predicted - OM measured) |
|||
| Total | 120 | 120 | 0 | 120 | 0 | 70 | 70 | 0 | 70 | 0 |
| Sex | ||||||||||
| Male | 123 | 123 | 0 | 123 | 0 | 71 | 71 | 1Table 3 Note † | 71 | 0 |
| Female | 118 | 118 | 0 | 118 | 0 | 69 | 69 | -1Table 3 Note † | 69 | 0 |
| Age group | ||||||||||
| 18 to 39 years | 112 | 111 | -1Table 3 Note † | 112 | 0 | 67 | 66 | 0 | 67 | 0 |
| 40 to 59 years | 118 | 118 | 1Table 3 Note † | 118 | 0 | 71 | 71 | 0 | 71 | 0 |
| 60 to 64 years | 125 | 125 | 0 | 125 | 0 | 73 | 73 | 0 | 73 | 0 |
| 65 to 69 years | 126 | 126 | 0 | 125 | -1 | 71 | 71 | 0 | 71 | 0 |
| 70 to 74 years | 129 | 128 | -1 | 128 | -1 | 68 | 69 | 0 | 68 | 0 |
| 75 to 79 years | 129 | 130 | 1 | 131 | 2Table 3 Note † | 67 | 67 | 1 | 67 | 1 |
| Men | ||||||||||
| 18 to 39 years | 118 | 116 | -2Table 3 Note † | 118 | 0 | 66 | 67 | 0 | 66 | 0 |
| 40 to 59 years | 122 | 123 | 1 | 123 | 0 | 74 | 74 | 0 | 73 | 0 |
| 60 to 64 years | 125 | 125 | 0 | 124 | 0 | 72 | 73 | 1Table 3 Note † | 72 | 0 |
| 65 to 69 years | 126 | 126 | 0 | 125 | -1 | 71 | 72 | 1 | 71 | 0 |
| 70 to 74 years | 126 | 126 | -1 | 125 | -1 | 67 | 69 | 1Table 3 Note † | 68 | 0 |
| 75 to 79 years | 126 | 130 | 3 | 130 | 3 | 66 | 68 | 2 | 68 | 1 |
| Women | ||||||||||
| 18 to 39 years | 109 | 108 | -1 | 109 | 0 | 67 | 66 | -1 | 67 | 0 |
| 40 to 59 years | 113 | 114 | 1Table 3 Note † | 113 | 0 | 69 | 69 | 0 | 70 | 0 |
| 60 to 64 years | 126 | 126 | 0 | 125 | -1 | 74 | 73 | -1Table 3 Note † | 74 | -1 |
| 65 to 69 years | 126 | 126 | 0 | 126 | 0 | 70 | 70 | -1 | 70 | 0 |
| 70 to 74 years | 132 | 130 | -2 | 131 | -1 | 69 | 68 | -1 | 69 | -1 |
| 75 to 79 years | 130 | 131 | 0 | 132 | 1 | 67 | 67 | 0 | 67 | 0 |
| Body mass index category | ||||||||||
| Neither overweight nor obese | 116 | 114 | -1Table 3 Note † | 116 | 0 | 65 | 66 | 1Table 3 Note † | 65 | 0 |
| Overweight | 122 | 122 | 0 | 122 | 0 | 71 | 71 | 0 | 71 | 0 |
| Obese | 124 | 125 | 2Table 3 Note † | 124 | 0 | 74 | 73 | -1Table 3 Note † | 74 | 0 |
| Men | ||||||||||
| Neither overweight nor obese | 120 | 118 | -2Table 3 Note † | 120 | 0 | 66 | 67 | 2Table 3 Note † | 65 | 0 |
| Overweight | 123 | 123 | 0 | 124 | 0 | 71 | 72 | 1Table 3 Note † | 71 | 0 |
| Obese | 125 | 126 | 1Table 3 Note † | 125 | 0 | 74 | 74 | 0 | 74 | 0 |
| Women | ||||||||||
| Neither overweight nor obese | 114 | 113 | -1Table 3 Note † | 114 | 0 | 65 | 66 | 1Table 3 Note † | 65 | 0 |
| Overweight | 121 | 121 | 0 | 121 | 0 | 71 | 70 | -1Table 3 Note † | 71 | 0 |
| Obese | 122 | 124 | 2Table 3 Note † | 122 | 0 | 73 | 71 | -2Table 3 Note † | 73 | 0 |
| Antihypertensive medication use | ||||||||||
| Yes | 128 | 127 | -1 | 127 | 0 | 71 | 71 | 0 | 71 | 0 |
| No | 118 | 118 | 0 | 118 | 0 | 70 | 70 | 0 | 70 | 0 |
| Men | ||||||||||
| Yes | 126 | 126 | -1 | 126 | 0 | 71 | 71 | 0 | 71 | 0 |
| No | 122 | 122 | 0 | 122 | 0 | 71 | 71 | 1Table 3 Note † | 71 | 0 |
| Women | ||||||||||
| Yes | 129 | 128 | -1 | 128 | -1 | 71 | 70 | -1Table 3 Note † | 71 | 0 |
| No | 116 | 116 | 0 | 116 | 0 | 69 | 68 | -1Table 3 Note † | 69 | 0 |
Source: Statistics Canada, Canadian Health Measures Survey crossover study, 2018 to 2019. |
||||||||||
Some differences were observed between predicted BP and measured BP across the BP distribution (figures 3 and 4). For example, among men with hypertension, predicted SBP was less likely to fall within 135 to 165 mmHg (Figure 3-A) and predicted DBP was less likely to fall within 85 to 100 mmHg (Figure 3-B) than measured BP. For women with hypertension, predicted SBP was less likely to fall within 135 to 150 mmHg (Figure 4-A), while predicted DBP was less likely to fall within 85 to 95 mmHg (Figure 4-B), compared with measured BP. For men without hypertension, there was a difference between predicted and measured BP along the entire distribution for SBP (Figure 3-C) but minimal difference between the two curves for DBP (Figure 3-D). For women without hypertension, predicted SBP was less likely to fall within 95 to 100 mmHg than measured SBP (Figure 4-C), and there was minimal difference between the two curves for DBP (Figure 4-D).
Description of Figure 3
Figure 3 presents probability density plots of the full model’s predicted versus measured blood pressure (BP) values among men using OMRON (OM), by hypertension status. It has four panels, labelled A, B, C, and D. Each panel contains two curved lines. The solid line depicts OM-measured BP values, while the dotted line depicts OM-predicted BP values. The area under each curve equals 1, representing the total probability of 1 for the full range of values in the distribution. The x-axis corresponds to the BP distribution in millimetres of mercury (mmHg), and the y-axis corresponds to a probability scale from 0.00 to 0.06, increasing at 0.01 intervals. Panel A presents systolic blood pressure (SBP) for men with hypertension. The difference in the two curves shows that predicted SBP was more likely to fall within 115 to 135 mmHg and less likely to fall within 135 to 165 mmHg than measured SBP. Panel B presents diastolic blood pressure (DBP) for men with hypertension. The difference in the two curves shows that predicted DBP was more likely to fall within 75 to 85 mmHg and less likely to fall within 85 to 100 mmHg than measured DBP. Panel C presents SBP for men without hypertension. There was a difference in the two curves along the entire distribution, with predicted SBP less likely to fall within 95 to 110 mmHg, more likely to fall within 110 to 125 mmHg, and less likely to fall within 125 to 135 mmHg than measured SBP. Panel D presents DBP for men without hypertension, with minimal differences between the curves along the entire DBP distribution.
Description of Figure 4
Figure 4 presents probability density plots of the full model’s predicted versus measured blood pressure (BP) values among women using OMRON (OM), by hypertension status. It has four panels, labelled A, B, C, and D. Each panel contains two curved lines. The solid line depicts OM-measured BP values, while the dotted line depicts OM-predicted BP values. The area under each curve equals 1, representing the total probability of 1 for the full range of values in the distribution. The x-axis corresponds to the BP distribution in millimetres of mercury (mmHg), and the y-axis corresponds to a probability scale from 0.00 to 0.06, increasing at 0.01 intervals. Panel A presents systolic blood pressure (SBP) for women with hypertension. The difference in the two curves shows that predicted SBP was more likely to fall within 120 to 135 mmHg and less likely to fall within 135 to 150 mmHg than measured SBP. Panel B presents diastolic blood pressure (DBP) for women with hypertension. The difference in the two curves shows that predicted DBP was more likely to fall within 70 to 80 mmHg and less likely to fall within 85 to 95 mmHg than measured DBP. Panel C presents SBP for women without hypertension, and the difference in the two curves shows that predicted SBP was less likely to fall within 95 to 100 mmHg than measured SBP. Panel D presents DBP for women without hypertension, with minimal differences between the curves along the entire DBP distribution.
The underprediction at higher levels of SBP resulted in predicted hypertension prevalence estimates from the full model being slightly lower than measured estimates across characteristics, with an average difference of -2.7 percentage points (Table 4). The exceptions were for both sexes in the oldest age group and for women with obesity where the average difference ranged from 0.0 to +2.1 percentage points.
| Hypertension prevalence (%) | |||||
|---|---|---|---|---|---|
| OM measured |
Base model | Full model | |||
| OM predicted |
Difference (OM predicted - OM measured) |
OM predicted |
Difference (OM predicted - OM measured) |
||
| Total | 35.4 | 32.6 | -2.7Table 4 Note † | 32.6 | -2.7Table 4 Note † |
| Sex | |||||
| Male | 40.5 | 37.9 | -2.6Table 4 Note † | 37.1 | -3.4Table 4 Note † |
| Female | 30.9 | 28.1 | -2.8Table 4 Note † | 28.8 | -2.1Table 4 Note † |
| Age group | |||||
| 18 to 39 years | 4.3 | 2.6 | -1.7Table 4 Note † | 3.0 | -1.3 |
| 40 to 59 years | 26.0 | 22.8 | -3.2Table 4 Note † | 23.3 | -2.7Table 4 Note † |
| 60 to 64 years | 52.6 | 50.9 | -1.8 | 49.7 | -2.9 |
| 65 to 69 years | 48.6 | 41.9 | -6.7Table 4 Note † | 41.9 | -6.7Table 4 Note † |
| 70 to 74 years | 64.9 | 62.3 | -2.6 | 60.5 | -4.4Table 4 Note † |
| 75 to 79 years | 77.8 | 77.8 | 0.0 | 79.2 | 1.4 |
| Men | |||||
| 18 to 39 years | 5.4 | Note F: too unreliable to be published | Note ...: not applicable | Note F: too unreliable to be published | Note ...: not applicable |
| 40 to 59 years | 32.8 | 29.4 | -3.4 | 29.9 | -2.8 |
| 60 to 64 years | 56.5 | 54.3 | -2.2 | 51.1 | -5.4 |
| 65 to 69 years | 51.8 | 48.2 | -3.6 | 48.2 | -3.6 |
| 70 to 74 years | 68.5 | 68.5 | 0.0 | 64.8 | -3.7 |
| 75 to 79 years | 83.3 | 83.3 | 0.0 | 83.3 | 0.0 |
| Women | |||||
| 18 to 39 years | 3.6 | Note F: too unreliable to be published | Note ...: not applicable | 3.6 | 0.0 |
| 40 to 59 years | 20.0 | 17.0 | -3.0Table 4 Note † | 17.5 | -2.5 |
| 60 to 64 years | 48.1 | 46.8 | -1.3 | 48.1 | 0.0 |
| 65 to 69 years | 44.9 | 34.7 | -10.2 | 34.7 | -10.2 |
| 70 to 74 years | 61.7 | 56.7 | -5.0 | 56.7 | -5.0 |
| 75 to 79 years | 75.0 | 75.0 | 0.0 | 77.1 | 2.1 |
| Body mass index category | |||||
| Neither overweight nor obese | 21.2 | 18.5 | -2.8Table 4 Note † | 18.7 | -2.5Table 4 Note † |
| Overweight | 38.4 | 34.7 | -3.7Table 4 Note † | 34.5 | -3.9Table 4 Note † |
| Obese | 48.2 | 46.9 | -1.3 | 46.9 | -1.3 |
| Men | |||||
| Neither overweight nor obese | 26.2 | 24.6 | -1.6 | 23.8 | -2.5 |
| Overweight | 39.6 | 36.8 | -2.8 | 35.8 | -3.8Table 4 Note † |
| Obese | 52.5 | 49.4 | -3.1 | 48.8 | -3.7 |
| Women | |||||
| Neither overweight nor obese | 18.7 | 15.4 | -3.3Table 4 Note † | 16.2 | -2.5 |
| Overweight | 37.1 | 32.5 | -4.6Table 4 Note † | 33.0 | -4.1Table 4 Note † |
| Obese | 43.3 | 44.0 | 0.7 | 44.7 | 1.4 |
| Antihypertensive medication use | |||||
| Yes | 100.0 | 100.0 | 0.0 | 100.0 | 0.0 |
| No | 14.7 | 11.1 | -3.6Table 4 Note † | 11.1 | -3.6Table 4 Note † |
| Men | |||||
| Yes | 100.0 | 100.0 | 0.0 | 100.0 | 0.0 |
| No | 16.4 | 12.7 | -3.7Table 4 Note † | 11.6 | -4.8Table 4 Note † |
| Women | |||||
| Yes | 100.0 | 100.0 | 0.0 | 100.0 | 0.0 |
| No | 13.3 | 9.8 | -3.5Table 4 Note † | 10.7 | -2.6Table 4 Note † |
|
... not applicable F too unreliable to be published
Source: Statistics Canada, Canadian Health Measures Survey crossover study, 2018 to 2019. |
|||||
Adjusting for device order in each model had little effect on predicted mean BP values or predicted hypertension prevalence, nor did it alter the underestimation of predicted values at higher levels of BP (data not shown).
Population-level differences
When weighted to represent the population of adults aged 18 to 79 years, differences between the devices were comparable in magnitude and direction to the unweighted estimates. At the population level, average SBP measured with OM was 7 mmHg higher (95% CI: 5, 8) than that measured with BT, and average DBP measured with OM was 2 mmHg lower (95% CI: -3, -2) than that measured with BT (Table 5). Weighted differences in SBP and DBP between predicted and measured values were also comparable to the unweighted estimates (data not shown).
| Measured mean SBP (mmHg) | Measured mean DBP (mmHg) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OM | BT | Difference (OM - BT) |
95% CI | OM | BT | Difference (OM - BT) |
95% CI | |||
| from | to | from | to | |||||||
| Total | 118 | 112 | 7 | 5 | 8 | 69 | 71 | -2 | -3 | -2 |
| Sex | ||||||||||
| Male | 122 | 113 | 8 | 6 | 10 | 70 | 73 | -3 | -4 | -2 |
| Female | 115 | 110 | 5 | 3 | 7 | 68 | 70 | -2 | -3 | -1 |
| Age group | ||||||||||
| 18 to 39 years | 113 | 104 | 9 | 7 | 11 | 66 | 67 | -2 | -3 | -1 |
| 40 to 59 years | 118 | 113 | 5 | 4 | 7 | 72 | 74 | -3 | -4 | -2 |
| 60 to 64 years | 122 | 117 | 5 | 3 | 7 | 73 | 75 | -2 | -3 | -2 |
| 65 to 69 years | 127 | 122 | 5 | 2 | 7 | 70 | 73 | -2 | -4 | -1 |
| 70 to 74 years | 130 | 125 | 5 | 2 | 8 | 70 | 73 | -3 | -5 | -2 |
| 75 to 79 years | 124 | 121 | 3 | 1 | 6 | 63 | 66 | -3 | -5 | -2 |
| Men | ||||||||||
| 18 to 39 years | 116 | 105 | 11 | 8 | 13 | 65 | 67 | -2 | -3 | -1 |
| 40 to 59 years | 124 | 117 | 7 | 5 | 9 | 74 | 77 | -3 | -5 | -1 |
| 60 to 64 years | 123 | 117 | 6 | 4 | 8 | 73 | 76 | -3 | -4 | -3 |
| 65 to 69 years | 129 | 122 | 7 | 5 | 9 | 74 | 77 | -3 | -5 | -1 |
| 70 to 74 years | 131 | 125 | 5 | -1 | 11 | 71 | 76 | -5 | -7 | -2 |
| 75 to 79 years | 127 | 124 | 3 | 0 | 6 | 62 | 66 | -4 | -5 | -3 |
| Women | ||||||||||
| 18 to 39 years | 110 | 103 | 8 | 4 | 11 | 66 | 68 | -1 | -3 | 1Table 5 Note ‡ |
| 40 to 59 years | 113 | 109 | 4 | 2 | 5 | 69 | 71 | -2 | -3 | -1 |
| 60 to 64 years | 121 | 117 | 4 | 1 | 7 | 73 | 74 | -1 | -2 | 0 |
| 65 to 69 years | 125 | 123 | 2 | -1 | 5Table 5 Note ‡ | 67 | 68 | -1 | -4 | 1Table 5 Note ‡ |
| 70 to 74 years | 130 | 125 | 5 | 4 | 7 | 68 | 70 | -2 | -3 | -1 |
| 75 to 79 years | 122 | 119 | 3 | -1 | 7Table 5 Note ‡ | 63 | 66 | -3 | -5 | -1 |
| Body mass index category | ||||||||||
| Neither overweight nor obese | 112 | 104 | 9 | 7 | 11 | 63 | 66 | -3 | -4 | -3 |
| Overweight | 121 | 115 | 6 | 5 | 7 | 71 | 73 | -2 | -3 | -1 |
| Obese | 123 | 119 | 5 | 3 | 7 | 75 | 76 | -1 | -2 | 0 |
| Men | ||||||||||
| Neither overweight nor obese | 117 | 105 | 12 | 8 | 16 | 62 | 66 | -4 | -5 | -2 |
| Overweight | 122 | 115 | 7 | 4 | 10 | 70 | 73 | -3 | -5 | -1 |
| Obese | 125 | 118 | 7 | 5 | 8 | 76 | 78 | -2 | -3 | -1 |
| Women | ||||||||||
| Neither overweight nor obese | 110 | 103 | 7 | 4 | 10 | 63 | 66 | -3 | -4 | -2 |
| Overweight | 120 | 115 | 5 | 3 | 6 | 72 | 73 | -1 | -3 | 0Table 5 Note ‡ |
| Obese | 120 | 119 | 1 | -3 | 5Table 5 Note ‡ | 72 | 73 | 0 | -3 | 2Table 5 Note ‡ |
| Antihypertensive medication use | ||||||||||
| Yes | 124 | 120 | 5 | 3 | 7 | 70 | 72 | -2 | -4 | 0 |
| No | 117 | 110 | 7 | 5 | 9 | 69 | 71 | -2 | -3 | -2 |
| Men | ||||||||||
| Yes | 125 | 118 | 7 | 6 | 9 | 71 | 72 | -2 | -4 | 1Table 5 Note ‡ |
| No | 121 | 112 | 9 | 6 | 11 | 70 | 73 | -3 | -4 | -2 |
| Women | ||||||||||
| Yes | 123 | 121 | 2 | -2 | 5Table 5 Note ‡ | 69 | 71 | -2 | -3 | -1 |
| No | 113 | 108 | 6 | 4 | 8 | 68 | 69 | -2 | -3 | 0 |
Source: Statistics Canada, Canadian Health Measures Survey crossover study, 2018 to 2019. |
||||||||||
Between-device differences in population-level hypertension prevalence were also comparable to the unweighted estimates, with an average +1.6 percentage point difference (95% CI: -1.3%, +4.6%) (Table 6). Between-device differences fluctuated across the characteristics but with considerable variability, as shown by their wide confidence intervals.
| Measured | Predicted versus measured | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| OM | BT | Difference (OM - BT) |
95% CI | OM predicted (full model) |
Difference (OM predicted - OM measured) |
95% CI | |||
| from | to | from | to | ||||||
| Total | 28.3 | 26.7 | 1.6 | -1.3 | 4.6 | 25.5 | -2.8 | -5.5 | -0.1Table 6 Note † |
| Sex | |||||||||
| Male | 33.2 | 33.4 | -0.3 | -4.8 | 4.2 | 31.1 | -2.1 | -5.7 | 1.5 |
| Female | 23.5 | 20.0 | 3.5 | -0.3 | 7.4 | 20.0 | -3.6 | -7.4 | 0.3 |
| Age group | |||||||||
| 18 to 39 years | 5.1 | 5.4 | -0.3 | -1.4 | 0.9 | 4.8 | -0.3 | -0.9 | 0.4 |
| 40 to 59 years | 31.1 | 27.8 | 3.4 | -4.5 | 11.3 | 25.9 | -5.3 | -11.7 | 1.1 |
| 60 to 64 years | 45.4 | 47.4 | -2.0 | -13.2 | 9.3 | 43.0 | -2.4 | -6.1 | 1.3 |
| 65 to 69 years | 56.3 | 50.2 | 6.1 | -1.9 | 14.2 | 50.5 | -5.9 | -12.5 | 0.8 |
| 70 to 74 years | 65.1 | 61.4 | 3.7 | -0.7 | 8.0 | 61.4 | -3.7 | -8.0 | 0.7 |
| 75 to 79 years | 87.3 | 84.8 | 2.6 | -3.8 | 8.9 | 87.7 | 0.4 | -0.7 | 1.4 |
| Men | |||||||||
| 18 to 39 years | 8.5 | Note F: too unreliable to be published | Note ...: not applicable | Note ...: not applicable | Note ...: not applicable | Note F: too unreliable to be published | Note ...: not applicable | Note ...: not applicable | Note ...: not applicable |
| 40 to 59 years | 37.5 | 37.6 | -0.1 | -9.6 | 9.4 | 35.5 | -2.0 | -8.6 | 4.5 |
| 60 to 64 years | 52.1 | 57.6 | -5.6 | -29.1 | 18.0 | 44.8 | -7.3 | -20.4 | 5.8 |
| 65 to 69 years | 60.6 | 60.4 | 0.3 | -0.9 | 1.4 | 58.0 | -2.6 | -11.5 | 6.2 |
| 70 to 74 years | 78.2 | 73.0 | 5.1 | -5.7 | 15.9 | 73.0 | -5.1 | -15.9 | 5.7 |
| 75 to 79 years | 90.7 | 87.3 | 3.4 | -10.1 | 16.9 | 90.7 | 0.0 | Note ...: not applicable | Note ...: not applicable |
| Women | |||||||||
| 18 to 39 years | 1.7 | 1.9 | -0.2 | -0.8 | 0.4 | 1.7 | 0.0 | Note ...: not applicable | Note ...: not applicable |
| 40 to 59 years | 24.8 | 18.0 | 6.8 | -3.3 | 17.0 | 16.2 | -8.5 | -18.7 | 1.7 |
| 60 to 64 years | 39.3 | 38.0 | 1.3 | -1.8 | 4.4 | 41.4 | 2.1 | -3.3 | 7.5 |
| 65 to 69 years | 51.8 | 39.5 | 12.3 | -1.9 | 26.6 | 42.5 | -9.3 | -21.4 | 2.7 |
| 70 to 74 years | 52.3 | 50.1 | 2.2 | -1.4 | 5.8 | 50.1 | -2.2 | -5.8 | 1.4 |
| 75 to 79 years | 84.9 | 82.9 | 2.0 | -3.5 | 7.4 | 85.5 | 0.6 | -0.9 | 2.2 |
| Body mass index category | |||||||||
| Neither overweight nor obese | 11.6 | 11.4 | 0.2 | -3.9 | 4.3 | 10.4 | -1.2 | -3.9 | 1.6 |
| Overweight | 32.9 | 27.8 | 5.1 | -0.7 | 10.9 | 26.9 | -6.0 | -11.9 | -0.1Table 6 Note † |
| Obese | 45.8 | 46.6 | -0.9 | -6.5 | 4.8 | 44.9 | -0.9 | -5.0 | 3.3 |
| Men | |||||||||
| Neither overweight nor obese | 13.8 | 16.1 | -2.3 | -9.7 | 5.0 | 12.9 | -0.9 | -3.8 | 2.0 |
| Overweight | 36.5 | 34.5 | 2.0 | -3.1 | 7.2 | 33.3 | -3.3 | -7.7 | 1.2 |
| Obese | 45.4 | 46.5 | -1.1 | -9.9 | 7.8 | 43.7 | -1.7 | -8.4 | 4.9 |
| Women | |||||||||
| Neither overweight nor obese | 10.3 | 8.6 | 1.7 | -1.4 | 4.8 | 8.9 | -1.4 | -6.0 | 3.3 |
| Overweight | 28.9 | 20.4 | 8.5 | -3.6 | 20.7 | 19.8 | -9.1 | -21.0 | 2.8 |
| Obese | 46.3 | 46.9 | -0.6 | -1.7 | 0.5 | 46.9 | 0.6 | -0.5 | 1.7 |
| Antihypertensive medication use | |||||||||
| Yes | 100.0 | 100.0 | 0.0 | Note ...: not applicable | Note ...: not applicable | 100.0 | 0.0 | Note ...: not applicable | Note ...: not applicable |
| No | 13.7 | 11.7 | 2.0 | -2.1 | 6.1 | 10.3 | -3.4 | -7.0 | 0.3 |
| Men | |||||||||
| Yes | 100.0 | 100.0 | 0.0 | Note ...: not applicable | Note ...: not applicable | 100.0 | 0.0 | Note ...: not applicable | Note ...: not applicable |
| No | 17.0 | 17.3 | -0.3 | -7.0 | 6.3 | 14.4 | -2.6 | -8.2 | 3.1 |
| Women | |||||||||
| Yes | 100.0 | 100.0 | 0.0 | Note ...: not applicable | Note ...: not applicable | 100.0 | 0.0 | Note ...: not applicable | Note ...: not applicable |
| No | 10.6 | 6.5 | 4.1 | -0.7 | 9.0 | 6.5 | -4.2 | -9.0 | 0.7 |
|
... not applicable F too unreliable to be published
Source: Statistics Canada, Canadian Health Measures Survey crossover study, 2018 to 2019. |
|||||||||
Weighted differences between predicted and measured values were comparable in magnitude and direction to the unweighted estimates. The overall difference between predicted and measured BP in population prevalence was -2.8 percentage points (95% CI: -5.5%, -0.1%) (Table 6). Prevalence differences between predicted and measured BP fluctuated—with wide confidence intervals—across characteristics.
Discussion
The present analysis found systematic between-device differences in measured SBP and DBP but fewer between-device differences in hypertension prevalence in a sample of adults aged 18 to 79 years. Prediction equations to compare BT-based BP data with OM-based BP data were developed and assessed. The results suggest that these equations can be used as intended, but with some caveats.
In the study sample, mean SBP measured with OM was 6 mmHg higher than mean SBP measured with BT, and mean DBP measured with OM was 2 mmHg lower than mean DBP with BT. Between-device differences across characteristics ranged from an average of +2 to +11 mmHg for SBP and an average of -4 to -1 mmHg for DBP. Overall, these between-device differences are substantively higher than reported cycle-over-cycle absolute differences at the population level in average adult SBP and DBP. To date, published estimates for cycles 1 to 6 show these differences ranging from 2 to 3 mmHg for SBP and 0 to 2 mmHg for DBP.Note 21 In their study comparing BT with OM in 260 cardiovascular clinic patients—average age 66 years—Rinfret et al. reported no between-device difference in DBP but found that OM-measured SBP exceeded BT-measured SBP by almost 5 mmHg.7 Conversely, Myers et al. reported similar values for SBP and DBP between the BT and OM devices.Note 22 However, their sample of 100 individuals—average age 63 years—consisted of patients admitted to an ambulatory BP monitoring unit with an average SBP of 132 to 141 mmHg, depending on the device used. The present study also found that average between-device differences were lower at higher values of BP.
Average hypertension prevalence based on OM-measured BP was 1.4 percentage points higher (+1.2 for men and +1.6 for women) than BT-based prevalence in the study sample. When weighted to represent the population, the overall difference between OM and BT prevalence was +1.6 percentage points (95% CI: -1.3, +4.6). These percentage point differences fall within the range of variability of existing cycle-over-cycle differences in reported population estimates of adult hypertension prevalence.Note 23 There are two main reasons why the substantive between-device differences in measured SBP did not produce correspondingly large between-device differences in hypertension prevalence in this study. First, the observed between-device differences in SBP were smaller at higher values. Second, people were classified as hypertensive if their measured BP met the BP criteria for hypertension or if they reported taking antihypertensive medication, regardless of their measured BP. Among those classified as hypertensive in the current study, about 80% were taking antihypertensive medication, compared with 77% of the adult population in 2018 to 2019 (data not shown). Thus, only a relatively small proportion of hypertensive individuals are classified as such based solely on their measured BP.
Separate SBP and DBP prediction equations were developed to compare BT-based BP data with OM-based BP data. The models that contained additional covariates performed better than the base models. These full models produced sex-specific average predicted BP estimates that were comparable to measured BP across age groups, BMI categories, and antihypertensive medication use categories. The predicted BP values followed the distribution of measured BP, except at the upper levels of the BP distribution, where the predicted values were lower than measured ones. This resulted in slightly lower predicted hypertension prevalence estimates compared with those based on measured values. When weighted to represent the population, average overall predicted hypertension prevalence remained slightly lower than average measured prevalence.
The magnitude of the overall differences in predicted versus measured prevalence—unweighted and weighted—was within the range of variability of previously published population estimates of adult hypertension prevalence.Note 23 However, there were larger differences between predicted and measured prevalence for certain groups—e.g., women who were overweight. While not statistically outside the typical range of variability—owing to wide CIs—they were outside this range in terms of the magnitude of the point estimate difference. It is unclear whether a full cycle of CHMS BP data, with an increased sample size, would have produced equally large prevalence differences but with narrower CIs for certain groups—differences that would be considered substantive and statistically significant. Given this uncertainty, differences between predicted hypertension prevalence estimates from BP data in previous cycles and measured hypertension prevalence estimates in Cycle 7 onward should be interpreted with caution.
At the population-level, Emberson et al. predicted that an absolute reduction of 7 mmHg in SBP would result in a 16% reduction in first occurrences of major cardiovascular disease over 10 years.Note 24 At the individual level, a margin of 5 mmHg in SBP can affect the intensity of treatment of patients with hypertension.7 It follows that the magnitude of the difference in population estimates of SBP—and to a lesser extent DBP—resulting from the introduction of the OM device in Cycle 7 will be interpreted as a substantive change in the BP of Canadian adults. Thus, the impact of changing to the OM device should be acknowledged when estimates of adult BP from Cycle 7 onward are reported.
Strengths and limitations
This study has several strengths. BP, height, and weight were measured objectively using systematic methodologies. All respondents had their BP measured by trained staff following strict protocols during a single visit to the MEC, and device order was randomly assigned. The recommended set of prediction equations had predicted R-squared values ranging from 0.84 to 0.88, indicating good predictive power. Nevertheless, results of this study should be interpreted considering certain limitations. The results pertain to adults aged 18 to 79 years, and the equations are not generalizable to younger or older age groups. Cuff sizes were not concordant between the two devices, and different cuff sizes used on the same person may have affected the results.Note 7,Note 25 Limited study sample size reduced statistical power when examining and interpreting between-device differences in hypertension prevalence across multiple characteristics. The equations produced predicted BP values that were comparable to measured BP values on average, but with some underestimation at the upper end of the BP distribution. This should be noted when using the equations. Moreover, there was some uncertainty regarding the ability of the predicted BP values to produce adequate predicted hypertension prevalence estimates for certain groups. This too should be noted when using the equations. The crossover sample survey weight and bootstrap weights were generated by a simple calibration on the existing full population of Cycle 6 by age group and sex to show the potential impact of the device change on population estimates. The results may have been slightly different had the crossover sample weights been generated independently following all the standard steps in the CHMS weighting process.
Conclusion
This study of Canadian adults found systematic between-device differences in SBP and DBP measurements between the OM and the BT monitors for people aged 18 to 79 years. Consequently, adult BP data in Cycle 7 of the CHMS will be substantively different from adult BP data in previous cycles. The switch in BP devices and the resulting impact on BP data in Cycle 7 should be noted when reporting estimates of adult BP based on the CHMS. The prediction equations developed in this study can be applied to adult BP data from cycles 1 to 6 for comparison with BP measurements in Cycle 7, with the following caveats. First, the predicted values slightly underestimate the BP values at the upper end of the BP distribution. Second, hypertension prevalence estimates based on predicted BP values should be interpreted with caution.
Acknowledgments
The authors gratefully acknowledge the help of Joel Barnes, who provided the syntax to produce the probability density plots in RStudio.
- Date modified: