Underestimation of Blood Pressure and Stroke Risk by Manual Blood Pressure Measurement

This study demonstrates that manual blood pressure measurements systematically underestimate both blood pressure levels and the associated risk of incident stroke compared to automated devices, potentially leading to inadequate hypertension management.

The Gist

Imagine you are trying to measure the temperature of a pot of soup to see if it's hot enough to eat. You have two ways to do this:

1. The Robot Thermometer (Automated): You stick in a digital probe, press a button, and it gives you a precise number like "142 degrees." It doesn't care what it thinks the number should be; it just tells you the truth.
2. The Human Guess (Manual): You stick in a thermometer, but instead of reading the exact number, you glance at it and think, "Hmm, that looks like it's right around 140." So, you write down "140," even if it's actually 142.

This study is all about how doctors measure blood pressure, and it found that the "Human Guess" method is secretly making us feel safer than we actually are.

The "Round-Off" Problem

When doctors measure blood pressure by hand (using a stethoscope and a squeeze-bulb cuff), they often fall into a habit called "Terminal Digit Preference."

Think of it like this: If a doctor hears a heartbeat that sounds like it's at 123, their brain might lazily round it down to 120 because it's a nice, round number. If it's 127, they might round it up to 130. They do this so often that if you look at thousands of hand-measured blood pressures, you'll see a massive spike in numbers ending in zero (like 120, 130, 140).

The study found that when doctors use automated machines, the numbers are spread out evenly (121, 122, 123, etc.). But when they use manual methods, the numbers are "clumped" at the zeros. This rounding trick makes the blood pressure look lower than it really is.

The "Optimistic Illusion"

The researchers compared two huge groups of patients:

• Group A: Patients whose doctors used the new, automated machines.
• Group B: Patients whose doctors used the old-school manual method.

They found that Group B (Manual) had blood pressure readings that were, on average, 6 points lower than Group A. It's as if the manual method was wearing "rose-colored glasses," telling the doctor, "Don't worry, the pressure is fine!" when it was actually a bit high.

Because the numbers looked lower, fewer people in the manual group were diagnosed with high blood pressure (hypertension), and fewer were put on medication. The doctors thought, "Oh, they are under control," but they were actually sitting on a ticking time bomb.

The Real-World Consequence: The Storm

To see if this "optimistic illusion" actually mattered, the researchers looked at who had a stroke (a brain attack caused by high blood pressure).

Here is the scary part: For the exact same blood pressure reading, the manual group was at higher risk of having a stroke.

Imagine two people:

• Person A has a reading of 125 from a machine.
• Person B has a reading of 120 from a human doctor.

The study found that Person B (the 120 reading) was actually in as much danger as Person A (the 125 reading). The manual method had "hidden" 5 points of danger. It's like driving a car where the speedometer is broken and always reads 5 mph slower than you are actually going. You think you are driving safely, but you are actually speeding, and the crash is more likely to happen.

The Takeaway

This paper is a wake-up call. It suggests that for decades, many patients have been told their blood pressure is "normal" simply because their doctor was rounding numbers down.

• The Old Way: A doctor listens, guesses, rounds to the nearest ten, and says, "You're fine."
• The New Way: A machine measures, gives the exact number, and says, "You actually need a little more help."

The study concludes that we need to switch to automated machines in doctors' offices. It's not just about being fancy; it's about getting the real numbers so we can treat high blood pressure correctly and prevent strokes. If we keep using the "round-off" method, we might be missing the warning signs until it's too late.

Technical Summary

1. Problem Statement

Hypertension is a leading modifiable risk factor for cardiovascular disease, particularly stroke. While clinical guidelines (e.g., 2025 multi-society guidelines) recommend the use of automated oscillometric devices to minimize observer bias, manual auscultatory methods remain widely used in primary care.

• The Core Issue: Manual measurement is prone to observer bias, specifically terminal digit preference (TDP), where clinicians round blood pressure (BP) readings to the nearest even number (often ending in "0").
• The Consequence: This bias leads to systematically lower recorded BP values, resulting in the underdiagnosis of hypertension, undertreatment, and a potential underestimation of the associated risk of adverse outcomes like stroke.
• Knowledge Gap: There is a lack of large-scale, real-world evidence quantifying the magnitude of this underestimation and its direct impact on longitudinal clinical outcomes (specifically incident stroke) when comparing manual vs. automated modalities.

2. Methodology

The study utilized de-identified data from the Optum Labs Data Warehouse (OLDW), covering over 350 million enrollees.

• Study Design: Two distinct cohorts were analyzed:
  1. Cross-sectional Cohort (2024): To compare BP distributions and control rates.
  2. Longitudinal Cohort (2019–2024): To evaluate the association between BP levels and incident stroke.
• Classification of Measurement Modality: Since device type is not explicitly recorded in EHRs, the authors used terminal digit preference as a proxy:
  • Manual Group: Providers with <0.5% odd terminal digits (indicating heavy rounding to even numbers).
  • Automated Group: Providers with 45–55% odd terminal digits (indicating a uniform distribution of digits 0–9).
• Population:
  • 2024 Cohort: 776,209 patients; 336,634 successfully matched (1:1) via propensity score matching.
  • 2019 Cohort: 1.9 million patients; 686,482 matched for stroke analysis.
• Matching & Covariates: Propensity score matching (1:1 nearest neighbor, caliper 0.0005) was performed to balance groups on age, sex, race, insurance, BMI, smoking, diabetes, hypertension history, and medication use.
• Statistical Analysis:
  • Cross-sectional: T-tests for mean BP; Chi-squared tests for control thresholds (<130/80, <140/90).
  • Longitudinal: Cox proportional hazards regression with BP modeled as a cubic spline (5 knots) to assess the hazard ratio (HR) of incident stroke.
  • Negative Control: Incident vehicular accidents were analyzed to rule out residual confounding.

3. Key Contributions

• Novel Proxy Method: Validated the use of terminal digit distribution in administrative data to reliably distinguish between manual and automated measurement practices at the provider level.
• Quantification of Bias: Provided precise estimates of the "measurement gap" between manual and automated devices in a real-world setting.
• Outcome Correlation: Demonstrated that the same numerical BP reading carries a different risk profile depending on the measurement method, linking measurement bias directly to stroke risk.

4. Key Results

A. Blood Pressure Distribution and Control (2024 Cohort)

• Mean Systolic BP (SBP): The automated group had significantly higher mean SBP (131.7 mmHg) compared to the manual group (125.9 mmHg), a difference of 5.8 mmHg.
• Clinical Thresholds:
  • Patients meeting the <130/80 mmHg goal: 33.2% (Automated) vs. 38.8% (Manual).
  • Patients meeting the <140/90 mmHg goal: 61.2% (Automated) vs. 70.9% (Manual).
• Terminal Digit Analysis:
  • Automated: Even distribution of digits 0–9 (9–11% each).
  • Manual: 31.2% of readings ended in "0", confirming significant rounding bias.
• Sensitivity Analysis: Results remained consistent when analyzing the second recorded BP measurement in visits with multiple readings.

B. Association with Incident Stroke (2019 Cohort)

• Risk Discrepancy: For any given SBP level, the manual group had a higher risk of stroke compared to the automated group.
• Hazard Ratios:
  • An automated SBP of 125 mmHg had an HR of 1.16 (95% CI: 1.10–1.22) relative to a reference of 120 mmHg (automated).
  • A manual SBP of 120 mmHg had the same HR of 1.16 (95% CI: 1.09–1.22).
  • Interpretation: A manual reading of 120 mmHg carries the same stroke risk as an automated reading of 125 mmHg. The manual method effectively "hides" approximately 5 mmHg of risk.
• Negative Control: No significant difference was found in vehicular accidents between groups (HR 1.01), suggesting the stroke association is specific to BP measurement bias rather than general patient selection bias.

5. Significance and Implications

• Clinical Underestimation: Manual BP measurement systematically underestimates both the severity of hypertension and the associated cardiovascular risk. A patient with a "controlled" manual BP of 120/80 may actually have a risk profile equivalent to 125/85 measured automatically.
• Clinical Inertia: The underestimation likely contributes to clinical inertia, where providers fail to titrate antihypertensive medications because the recorded values appear lower than the true physiological risk.
• Public Health Impact: With a significant portion of providers still using manual methods, the transition to automated devices is critical not just for accuracy, but for preventing downstream stroke events.
• Guideline Adherence: The findings strongly support current guidelines recommending automated office BP measurement. As practices transition, there may be a temporary "rise" in hypertension diagnoses, which reflects improved detection rather than a worsening epidemic.

Conclusion: The study provides robust evidence that manual blood pressure measurement introduces a systematic bias that underestimates blood pressure levels and stroke risk by approximately 5 mmHg. Transitioning to automated measurement is essential for accurate diagnosis, appropriate treatment intensification, and the reduction of cardiovascular morbidity.