Statistical Process Control Charts for Clinical Users

Quality improvement in perioperative care focuses on teams, close to the point of care, iteratively testing changes to a process within a local setting. To learn whether they are successful, the teams need timely feedback, signaling if their changes do or do not result in improvement. This feedback most frequently occurs by defining and collecting key process and outcome measures and plotting them over time. For example, care teams testing approaches to reducing surgical site infections (SSIs) may gauge success using short-term changes in process measures, including reliable administration of presurgical antibiotics and the longer-term outcome of SSI.

Teams need to apply a rigorous method to inform them if changes from one data point to the next are simply chance or whether the data indicate that meaningful improvement is occurring. Without such a method, overreaction to one data point may occur. For example, 1 month with a low rate of SSIs does not necessarily suggest ongoing and sustained performance at this new level.

Based on Shewhart's Theory of Variation, statistical process control (SPC) chart methodology is standard practice for visualizing, interpreting, and learning from process and outcome improvement measures, plotted over time. Shewhart described two types of variation in data: “common cause”—innate to the system, like random week-to-week variation—and “special cause,” which is not part of the routine system, signaling something important changed and possibly improved. In general, once there are at least 20 data points, SPC methods plot improvement measures over time, with a centerline (typically the mean) and upper and lower limits. In health care, five “control chart rules” ( Fig. 28.1 ) signal if the data are consistent with common (random) or special cause (something has changed) variations.

Open full size image

Rules for signaling special cause. In total, more than five rules have been developed for signaling special cause variation; however, in health care, the five rules described are most commonly used. SPC , statistical process control. ^a Sigma (σ) is a measure of variation between successive data points. Each type of SPC chart has a specific method for calculating σ. σ differs from a standard deviation by measuring the variability of a process over time instead of the overall variation in the static distribution in the data.