Blood pressure reductions: correcting for regression to the mean

Area of Science:

• Cardiovascular Health
• Biostatistics
• Public Health Interventions

Background:

• Blood pressure (BP) reductions in health programs are typically measured by comparing pre- and post-participation readings.
• Interpreting these BP changes is complicated by regression to the mean, a statistical artifact where high initial measurements tend to decrease naturally.

Purpose of the Study:

• To develop a statistical model for predicting and adjusting for regression to the mean in BP measurements.
• To accurately determine the net reduction in BP attributable to screening and treatment programs.

Main Methods:

• Developed a statistical model estimating the regression effect based on measurement error and baseline BP differences.
• Applied the model to data from community screening programs and the Hypertension Detection and Follow-up Program.
• Subtracted the estimated regression effect from observed BP declines to calculate net reductions.

Main Results:

• In screening programs, the observed mean diastolic BP decline was 7 mm Hg; adjusted for regression to the mean, the net decline was 2 mm Hg.
• The model's predicted net reductions in the Hypertension Detection and Follow-up Program closely matched independent measurements (within 0.1 mm Hg).

Conclusions:

• A statistical model can accurately quantify and adjust for regression to the mean in BP studies.
• Net BP reductions from health programs are often significantly smaller than crude observed declines.
• Accurate assessment of hypertension interventions requires accounting for this statistical artifact.