Regression to the Mean in Quality Measurement

Regression to the mean is a statistical phenomenon that can lead quality professionals to draw incorrect conclusions about the effectiveness of their interventions. Understanding this concept helps you design more rigorous evaluations and avoid overestimating the impact of improvement efforts.

What Is Regression to the Mean

Regression to the mean occurs when extreme values on a first measurement tend to be followed by values closer to the average on subsequent measurements, regardless of any intervention. This happens because extreme values often include a component of random variation. On the next measurement, the random component is likely to be less extreme, pulling the value back toward the average. This is a mathematical inevitability when measurements include any degree of random error.

How It Misleads Quality Teams

Quality improvement teams often target areas or providers with the worst performance. If performance is measured, an intervention is implemented, and performance is measured again, any improvement may be partly or entirely due to regression to the mean rather than the intervention. For example, if a hospital's infection rate spikes in one quarter and a quality initiative is launched, the rate may naturally decrease in the following quarter simply because the spike included a random component. The team might credit the intervention for an improvement that would have occurred anyway.

Identifying Regression to the Mean

Suspect regression to the mean when improvement projects are initiated in response to unusually poor performance, when interventions target outliers identified by a single measurement, or when the magnitude of improvement seems disproportionately large relative to the effort invested. The phenomenon is more pronounced when sample sizes are small, measurement variability is high, or the selection criterion for intervention is based on a single data point.

Strategies for Mitigation

Use multiple baseline measurements rather than a single data point to identify true performance problems. Employ control charts that distinguish special cause from common cause variation before launching interventions. Include control groups when possible so that regression to the mean affects both groups equally. Use interrupted time series analysis to model pre-intervention trends. These approaches help separate genuine improvement from statistical artifacts and produce more credible evidence of intervention effectiveness.