Digit Preference in Office Blood Pressure Measurements, United States 2015–2019

Abstract

BACKGROUND

Blood pressure (BP) measurement error may lead to under- or overtreatment of hypertension. One common source of error is terminal digit preference, most often a terminal digit of “0.” The objective was to evaluate national trends in terminal digit preference in office BP measurements among adults with treated hypertension.

METHODS

Data were from IQVIA’s National Disease and Therapeutic Index, a nationally representative, serial cross-sectional survey of office-based physicians. The analysis included office visits from 2015 to 2019 among adults aged ≥18 years receiving antihypertensive treatment. Annual trends were examined in the percent of systolic and diastolic BP measurements ending in zero by patient sex, age, and race/ethnicity, physician specialty, and first or subsequent hypertension treatment visit.

RESULTS

From 2015 to 2019, there were ~60 million hypertension treatment visits annually (unweighted N: 5,585–9,085). There was a decrease in the percent of visits with systolic (41.7%–37.7%) or diastolic (42.7%–37.8%) BP recordings ending in zero. Trends were similar by patient characteristics. However, a greater proportion of measurements ended in zero among patients aged ≥80 (vs. 15–59 or 60–79) years, first (vs. subsequent) treatment visits, visits to cardiologists (vs. primary care physicians), and visits with systolic BP ≥140 or diastolic BP ≥90 (vs. <140/90) mm Hg.

CONCLUSIONS

Despite modest improvement, terminal digit preference remains a common problem in office BP measurement in the United States. Without bias, 10%–20% of measurements are expected to end in zero. Reducing digit preference is a priority for improving BP measurement accuracy and hypertension management.

Accurate blood pressure (BP) measurement is central to the diagnosis and treatment of hypertension.¹ Although BP is measured at nearly every clinical encounter, accurate and reliable assessment of BP is challenging. BP is inherently variable, and variability is further increased when health care providers do not follow standardized measurement procedures.¹ Recommended BP measurement procedures can be difficult to implement in routine clinical practice resulting in inaccurate assessment of BP.^2,3 Measurement error can lead to under- or overtreatment of patients with hypertension.⁴

One common form of measurement error is terminal digit bias, or a preference by the observer to round measurements to a specific end digit, most commonly zero.⁵ In ambulatory care, BP is assessed with either a manual device using an auscultatory approach or an electronic device using an oscillometric approach.¹ Without terminal digit bias, approximately 20% of measurements with manual devices, where typically only even digits are recorded, and 10% of measurements with automated devices, where any end digit can be recorded, are expected to end in zero. However, previous clinic-based studies in the United States illustrate terminal digit preference is common.^6–9 Additional research shows patients at practices with higher terminal digit preference are less likely to have an antihypertensive prescription⁶ and more likely to experience an adverse cardiovascular event.¹⁰

Terminal digit preference may have decreased in recent years. The phasing out of mercury devices, concerns about measurement error with manual BP measurement, and discrepancies between in- and out-of-office BPs have reduced the use of auscultatory BP measurements in favor of automated BP measurements.¹ Though automated BP devices do not eliminate measurement error, they may reduce terminal digit preference. Several clinic-based studies have documented a reduction in terminal digit preference after the introduction of automated BP devices.^10,11 Additionally, national and global cardiovascular health initiatives emphasize proper BP measurement technique.^12–16

In this context, trends in terminal digit preference were evaluated in a nationally representative sample of office BP measurements among adults with treated hypertension. The hypothesis was digit preference would be reduced over time.

METHODS

Data source

Data were from IQVIA’s National Disease and Therapeutic Index (NDTI), which provides nationally representative data on the patterns and treatment of disease encountered by office-based physicians in the continental United States. NDTI data have been used in other studies examining ambulatory care.^17,18 The NDTI uses the American Medical Association and American Osteopathic Association master lists to sample approximately 4,000 physicians quarterly, stratified by geographic region and specialty. For each participating physician, 2 random consecutive workdays are sampled. For each patient seen, the physician reports all diagnoses and all prescription and nonprescription medications newly ordered or previously ordered and continued. The dataset does not include physician or clinic identifiers to maintain confidentiality. Data are weighted to estimate all patient contacts on all workdays for all US office-based physicians.

The analysis included all office visits from 1 January 2015 to 31 December 2019 among adults aged ≥18 years (unweighted N, range: 108,195–190,217). Hypertension treatment visits were defined as those with a recorded diagnosis of hypertension, and at which hypertension treatment occurred. Hypertension diagnoses were identified using survey-specific diagnostic codes. Each record of new or continued drug therapy was linked to the relevant diagnosis. Analyses of hypertension treatment visits were restricted to those visits with systolic and diastolic BP measurements recorded (unweighted N, range: 5,585–9,085). The proportion of hypertension treatment visits with missing systolic or diastolic BP measurements was between 1.3% and 2.7% during the study timeframe.

BP measurement and recording

The BP measurements used in the present analysis are based on values recorded on the survey form for each patient visit. The type of BP measurement (manual vs. automated) is not ascertained. NDTI survey procedures do not instruct the physician or individual taking the BP measurement to use a specific protocol.

Patient and physician characteristics

Physicians record patient demographic characteristics for each encounter on the survey form, including sex, age, and race/ethnicity. Each visit is characterized as a first or subsequent treatment visit, based on whether the physician has seen the patient previously for this episode of the diagnosis. Physician specialty is obtained from the American Medical Association and American Osteopathic Association master lists.

Statistical analyses

Characteristics of all adult office-based visits and hypertension treatment visits were examined from 2015 to 2019. Next, annual trends were examined in the number and proportion of hypertension treatment visits with recorded systolic and diastolic BP measurements with a terminal digit “0.” Trends were examined overall and by patient characteristics and physician specialty. Patient characteristics included sex (male or female), age (18–59, 60–79, or ≥80 years), and race/ethnicity (categorized as White, Black, Hispanic, Asian, or other based on categories listed on the survey form). Patient visits were categorized as first or subsequent treatment visits. Physician-reported specialty was categorized as primary care (family practice, general practice, or internal medicine), cardiology, or other.

The frequencies of specific BP values were plotted using histograms to visualize the extent of terminal digit preference. The distributions of systolic and diastolic BP from 2019 are included as examples of the patterns observed.

Trends were examined in the proportion of BP values with a terminal zero by BP level. The measurement ending in 0 was considered the midpoint of the interval including all possible end digits; for 0 to represent the middle of the interval one half of the measurements ending in 5 at the lower and upper bound of the interval were included.¹⁹ For example, to calculate the proportion of systolic BP measurements of 140 mm Hg in the interval from 135 to 145 mm Hg, half the number of BP measurements of 135 and 145 mm Hg were included in the denominator.

Trends were also examined in the proportion of BP values immediately above and below thresholds for hypertension control recommended by clinical practice guidelines in use during the study period.^3,20,21 Some previous studies have found evidence of a preference for values just below target BPs.^19,22 Trends were examined in the proportion of measurements within ± 2 units of 130, 140, and 150 mm Hg systolic BP, and within ± 2 units of 80 and 90 mm Hg diastolic BP.¹⁹

Next, trends were examined in the proportion of systolic and diastolic BP measurements with other end digits. Additionally, the proportion of measurements ending in 1, 3, 7, and 9 in each year was calculated. While there is some evidence observers round measurements to 5,²³ measurements ending in 1, 3, 7, or 9, are likely only obtained using automated devices.

Data were analyzed using IQVIA’s web-based customer portal. All analyses included weights to provide estimates nationally representative of office visits among US adults aged ≥18 with hypertension treatment in the years 2015–2019.

RESULTS

From 2015 to 2019, there were over 900 million office-based visits per year by US adults (Supplementary Table S1 online). Between 2015 and 2019, there were approximately 60 million hypertension treatment visits annually with systolic and diastolic BP recorded (Table 1). The majority of hypertension treatment visits occurred among adults aged 60–79 years, were for subsequent hypertension treatment, and among primary care physicians. In all years, nearly 60% of visits had a systolic BP measurement <140 mm Hg, and nearly 80% of visits had a diastolic BP measurement <90 mm Hg.

Table 1.

Characteristics of office-based hypertension treatment visits—National Disease and Therapeutic Index, 2015–2019

	2015	2016	2017	2018	2019
Raw number of visits, N	9,085	7,198	5,585	5,690	6,374
Total visits, N in millions	60.7	58.7	63.7	67.7	68.5
Sex, %
Male	51.9	52.7	53.3	54.0	55.7
Female	48.1	47.3	46.7	46.0	44.3
Age (years), %
<60	41.2	40.4	38.5	38.0	36.8
60–79	48.2	49.0	50.0	50.0	50.4
≥80	10.6	10.6	11.6	12.1	12.8
Race/ethnicity, %
White	71.2	69.9	68.4	68.8	67.4
Black	17.2	17.5	17.4	16.7	17.9
Hispanic	5.7	6.0	6.3	6.7	7.2
Asian	5.2	5.5	5.5	6.1	6.3
Other	0.6	1.1	2.4	1.6	1.2
Visit type, %
First treatment visit	16.9	17.3	18.0	20.3	22.3
Subsequent treatment visit	83.1	82.7	82.0	79.7	77.7
Physician specialty, %
Cardiology	9.8	10.2	10.0	10.0	9.4
Primary care	74.3	72.3	71.6	71.2	73.4
Other	15.9	17.6	18.4	18.8	17.2
Blood pressure level, %
SBP <140 mm Hg	61.6	59.6	58.8	59.4	57.0
DBP <90 mm Hg	78.5	77.4	78.4	77.5	76.5

Primary care defined as family practice, general practice, or internal medicine. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure. In 2015 and 2016, <1% of weighted visits were missing information on patient sex or race/ethnicity. In 2015, 5.0% of visits were missing information on first or subsequent visit and in 2016, 3.2% of visits were missing information on first or subsequent visit.

Trends in terminal digit preference overall and by subgroup

In 2015, the proportion of visits with systolic (41.7%) and diastolic (42.7%) BP measurements with a terminal zero were higher than expected based on chance (Table 2). In 2019, 37.7% of systolic and 37.8% of diastolic BP recordings ended in zero.

Table 2.

Percentage of office-based hypertension treatment visits with blood pressure measurements with a terminal digit zero

	Systolic blood pressure					Diastolic blood pressure
	2015	2016	2017	2018	2019	2015	2016	2017	2018	2019
Overall, %	41.7	39.0	38.0	38.1	37.7	42.7	40.9	38.5	39.4	37.8
Sex, %
Male	42.2	39.1	37.3	39.1	38.4	41.5	40.8	37.4	39.5	35.5
Female	41.0	38.9	38.8	36.8	36.8	43.7	41.0	39.8	39.3	39.7
Age (years), %
<60	41.0	39.3	37.0	38.5	35.4	41.7	39.0	36.8	37.5	37.0
60–79	42.0	38.5	38.3	37.3	38.7	42.9	41.5	38.7	39.6	38.4
≥80	42.9	40.0	39.5	39.9	40.1	45.5	45.7	43.8	44.4	37.9
Race/ethnicity, %
White	42.1	39.5	38.5	39.4	37.6	42.5	41.1	38.9	40.2	38.0
Black	39.6	36.2	37.4	33.9	36.9	41.1	37.6	37.0	37.1	35.2
Hispanic	42.1	39.5	38.7	35.9	39.3	44.2	44.0	39.1	37.8	39.2
Asian	41.7	38.7	37.3	36.7	39.9	46.4	44.4	40.5	40.3	41.3
Other	36.1	44.8	27.8	36.8	35.6	47.2	44.5	33.6	33.7	42.5
Visit type, %
First treatment visit	45.2	39.9	42.1	42.9	42.1	45.4	39.5	40.2	42.3	40.4
Subsequent treatment visit	41.5	38.9	37.1	36.8	36.4	42.6	41.2	38.2	38.7	37.1
Physician specialty, %
Cardiology	45.5	44.2	43.2	44.5	43.0	49.4	50.9	50.3	51.0	47.7
Primary care	42.1	38.5	37.4	38.4	37.2	42.8	41.1	37.0	39.0	37.4
Other	37.6	37.9	37.5	33.5	36.8	38.1	34.6	38.1	34.7	34.1
Blood pressure level, %
SBP <140 mm Hg	36.6	33.7	33.0	32.6	32.6	—	—	—	—
SBP ≥140 mm Hg	49.8	46.8	45.0	46.1	44.4	—	—	—	—
DBP <90 mm Hg	—	—	—	—		39.4	36.9	33.5	34.7	33.3
DBP ≥90 mm Hg	—	—	—	—		54.7	54.8	56.7	55.6	52.7

Primary care defined as family practice, general practice, or internal medicine. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure.

Trends in the proportion of systolic and diastolic BP measurements were similar by sex, age, and race/ethnicity. The proportion of visits with a systolic or diastolic BP measurement ending in zero was generally higher at first (vs. subsequent) treatment visits, visits to cardiologists (vs. primary care and other physicians), and at systolic BP ≥140 mm Hg or diastolic BP ≥90 mm Hg. Additionally, the proportion of systolic BP measurements ending in zero was higher among those aged ≥80 vs. those who were 18–59 or 60–79 years.

Distribution of BP values

Histograms of systolic and diastolic BP measurements from 2019 showed terminal digit zeroes were common across the full range of BP values (Figure 1). The most frequently recorded systolic BP value was 130 mm Hg (6.3 million visits), followed by 140 mm Hg (5.3 million visits), and the most frequently recorded diastolic BP value was 80 mm Hg (9.6 million visits).

Figure 1.

Distribution of systolic and diastolic blood pressure measurements—hypertension treatment visits 2019.

Trends in terminal digit preference by BP level

Within intervals, the proportion of measurements ending in zero was lowest around 140 mm Hg systolic and 90 mm Hg diastolic BP (Figure 2). The proportion of measurements in those intervals with a terminal zero generally decreased during the study period.

Figure 2.

Percentage of systolic and diastolic blood pressure measurements with an end-digit zero, by blood pressure level—hypertension treatment visits 2015–2019. Figures report the proportion of measurements ending in zero in the specified interval. Each interval includes one half of the measures ending in 5 at the upper and lower bounds. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure.

Trends in BP values around treatment thresholds

The percent of systolic BP measurements of 138–139 mm Hg was 1.2–1.8 percentage points higher than the percent of systolic BP measurements of 141–142 mm Hg (Table 3). These percentages remained stable. The proportions of systolic BP measurements 2 units below and above 130 mm Hg were similar and remained stable; a similar pattern was observed around 150 mm Hg.

Table 3.

Percentage of systolic and diastolic blood pressure measurements above and below guideline-recommended treatment goals

	2015	2016	2017	2018	2019
Systolic blood pressure (mm Hg)
128–129	4.8	4.3	3.9	4.1	4.3
130	11.2	9.3	9.6	9.0	9.2
131–132	4.3	4.2	4.1	4.5	4.2
138–139	4.3	4.4	4.8	4.8	4.4
140	9.0	8.2	8.5	7.7	7.7
141–142	2.9	2.7	3.1	3.0	3.2
148–149	2.0	2.4	1.8	2.1	2.6
150	4.2	4.5	4.4	4.8	4.5
151–152	1.3	1.0	1.2	1.4	1.2
Diastolic blood pressure (mm Hg)
78–79	6.0	6.1	7.0	6.5	6.9
80	17.4	16.0	14.4	15.0	14.0
81–82	5.8	6.1	5.5	6.1	5.6
88–89	4.4	4.5	5.2	5.7	5.2
90	7.8	8.4	8.2	8.0	7.8
91–92	2.3	2.3	2.2	2.7	2.9

Among hypertension treatment visits, 2015–2019.

Among diastolic BP measurements, the proportions of measurements 2 units below and above 80 mm Hg were similar and there was little change over time. The percent of diastolic BP measurements of 88–89 mm Hg was 2.0–3.0 percentage points higher than the percent of diastolic BP measurements of 91–92 mm Hg.

Trends in other terminal digits

Throughout the study, even terminal digits were more frequently reported than odd (Figure 3). Eight was the most common end digit after zero. Approximately 14% of systolic BP measurements ended in 8 in each survey year, while the proportion of diastolic BP measurements ending in 8 increased from 13.6% to 15.4%.

Figure 3.

Trends in systolic and diastolic blood pressure measurement end digits—hypertension treatment visits 2015–2019. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure.

Five was the most common odd terminal digit. The proportion of systolic BP measurements ending in 5 increased from 4.2% to 5.6%, while the proportion of diastolic BP measurements ending in 5 increased from 3.7% to 4.8%. There was little change from 2015 to 2019 in the proportion of systolic (6.4%–7.2%) or diastolic (6.3%–7.3%) BP measurements ending in 1, 3, 7, or 9.

DISCUSSION

In this serial, cross-sectional analysis of a nationally representative survey of office-based physicians, there were approximately 60 million visits for hypertension treatment annually from 2015 to 2019. In 2019, 37.7% of systolic and 37.8% of diastolic BP measurements recorded ended in zero. Terminal digit preference remains common, despite a decrease from 2015 to 2019 in the percent of systolic and diastolic BP measurements with a terminal zero. Additionally, there was a higher proportion of measurements with a terminal zero among adults aged ≥80 years, at first treatment visits, visits to cardiologists, and at higher levels of systolic and diastolic BP.

Several factors may have contributed to the decrease in terminal digit preference observed. First is the increased use of automated BP devices, though the survey does not include information on the devices used. In recent decades, as mercury devices have been eliminated and in recognition of the potential for error with the manual auscultatory technique, semi-automated and automated oscillometric BP devices have become increasingly used in clinical practice.^1,24

Second is an increased recognition and debate about BP measurement practices, especially in the context of the Systolic Blood Pressure Intervention Trial (SPRINT).²⁵ In 2015, SPRINT demonstrated a reduction in cardiovascular events and deaths treating adults with hypertension at increased cardiovascular risk to a systolic BP goal of <120 vs. <140 mm Hg.²⁵ The SPRINT protocol used automated BP measurements, including unattended automated BP measurement at some study sites.²⁶ The trial results subsequently informed BP treatment goals in the 2017 American College of Cardiology (ACC)/American Heart Association (AHA) high blood pressure guideline.³ While it is acknowledged the BPs achieved in SPRINT measured with automated devices using standardized procedures are on average lower than casual office measurements, the ensuing discussion about the trial’s measurement protocol may have influenced BP measurement procedures in practice.^24,27 Third is increased awareness of the importance of proper BP measurement technique as a result of global, national, and local initiatives to improve hypertension control.^12–16

Terminal digit preference has implications for patient care. Consistent with previous studies, terminal digit preference was higher at higher BPs among patients with hypertension.²⁸ One could argue rounding at BPs above guideline-recommended treatment goals is not particularly problematic because such patients would be treated regardless of their exact BP. However, rounding typically occurs down which could affect the intensity of pharmacologic treatment or use of other patient engagement strategies.^6,10 The lowest levels of terminal digit preference occurred near guideline-recommended thresholds for BP control (e.g., systolic BP <140 mm Hg or diastolic BP <90 mm Hg) and decreased over time in such intervals. It is possible providers obtain higher quality measurements when the BP value is likely to affect treatment decisions. However, it is unclear whether this is the case, and misclassification of hypertension control may negatively impact patient outcomes.

In a study conducted in the United Kingdom, researchers documented an excess of diastolic BP measurements of “88,” as well as end-digit zeroes, and found women with recorded diastolic BP 88–89 mm Hg had a subsequently higher rate of mortality than those with diastolic BP 90–99 mm Hg.²² It appeared observers rounded down or remeasured BP until they obtained a more favorable value if they thought the patient did not need treatment.²² In other studies in the United Kingdom, there is evidence of increased preference for the digits just below pay-for-performance targets after their introduction (i.e., threshold bias).^19,28 While the proportion of diastolic BP measurements of 88–89 mm Hg was greater than the proportion of diastolic BP measurements of 91–92 mm Hg in the current study, there was otherwise no strong evidence of preference for values immediately below certain thresholds. It will be important to continue to monitor preference for terminal zeroes and values just below specific thresholds, as well as the resulting impact on patient treatment and outcomes.¹⁹

Results showing higher terminal digit preference among visits to cardiologists warrant further exploration. A limitation of this study is it is unknown who took the BP measurement recorded. The authors are not aware of other studies comparing terminal digit preference by specialty, but, terminal digit preference has been documented across a range of health care settings, including hypertension specialty clinics.⁸

In addition to factors related to BP measurement devices and procedures, this study highlights the importance of accurate BP recording. While there is reason to believe the use of automated devices increased during this time, the proportion of measurements ending in odd digits other than 5 did not change substantially. It is possible observers maintain a preference for even digits or intentionally round measurements from automated devices to zero; these factors are important to understand to address inaccuracies in BP recording.

The 2017 ACC/AHA high blood pressure guideline issues a strong recommendation for use of proper methods for accurate measurement and documentation of BP.³ The guideline also notes there is growing evidence supporting the use of automated office BP measurements.³ Automated office BP is similar to mean awake ambulatory BP and home BP, which are more predictive of cardiovascular events than casual office BP measurements.²⁹ The guideline stops short of recommending the use of automated devices in office settings. Meanwhile, the Canadian Hypertension Education Program recommends automated BP measurement as the preferred method of office measurement.³⁰ Indeed, terminal digit preference had decreased across Canadian primary care clinics following the adoption of automated office BP devices.¹⁰ However, terminal digit preference has not been eliminated, possibly due to continued use of manual devices even when automated devices are available,^10,31 or rounding of automated measurements. Regardless of the type of device used, it is important to train and retrain current and future health professionals in proper BP measurement technique and to ensure procedures can be incorporated into the clinical workflow.^2,32

This study has several limitations. First, there is no agreed upon “acceptable” percentage of terminal digit zeroes. While 10%–20% of measurements are expected to end in zero without bias, some have considered 10%–29% acceptable.³³ However, the percentage of BP measurements with a terminal zero was still in excess of this range. Second, the type of BP device and the measurement procedures used are unknown, including whether multiple readings were obtained. Thus, the contribution of increased use of automated devices to the decrease in terminal digit preference cannot be determined. Additionally, without knowing the type of device, it cannot be determined whether a terminal digit of 8 reflects the use of a manual device which is marked with only even numbers or a “preference” by the observer for a particular value. Third, it is unknown whether BP readings recorded on the NDTI survey are the same as those recorded in the patient’s medical record and used to inform clinical decision-making. Finally, clustering in measurement (or measurement error) which occurs when the same physician reports data for multiple patients could not be accounted for because the dataset does not link measurements to individual physicians or clinics.

This study has several strengths. To the authors’ knowledge, it is the first to document terminal digit preference in office visits nationally and provide estimates of trends over time. Additionally, NDTI data are available with a short lag time, which enables use of the data to monitor changes in near real time.

The present study has implications for clinical practice and research. It is possible to reduce terminal digit preference through the use of automated devices, clinical training and retraining, and clinical support.² US guidelines could more strongly recommend the use of automated BP devices in office settings. To better address the determinants of terminal digit preference, it would be useful to understand the BP devices currently used for hypertension screening and diagnosis, and to monitor the response to treatment.³¹ Finally, it will be important to monitor trends in terminal digit preference as clinical quality measures and pay-for-performance programs are implemented to ensure other biases do not occur.

This study highlights the need to improve the quality of BP measurement in office settings. Reducing digit preference is a priority and can improve the accuracy of BP measurement and hypertension management in the United States.

FUNDING

Kathryn E. Foti is supported by grant number T32 HL007024 from the National Heart, Lung, and Blood Institute, National Institutes of Health. The funder had no role in the conduct of the study or manuscript preparation.

DISCLOSURE

The authors declared no conflict of interest.