By Russ Gonnering at Brownstone dot org.

We, especially those engaged in the health professions, need to stop the relentless attempts to normalize all data to fit the nice bell-shaped distribution that we assume is how the world works. Sometimes it does, but sometimes it doesn't. We need to understand the words "It depends."

Two decades ago, I was a Six Sigma Blackbelt deeply immersed in the applications of statistical quality control to healthcare. Our efforts were primarily focused on the process, and as such, the way to improve the outcomes depended entirely on optimizing the processes of care. Our work was organized around the "DMAIC" wheel:

We would Define the problem and the process to improve it, Measure the process, Analyze it, take actions to Improve the process, and then make sure we could Control it. The Control would take the form of specialized charts that showed variation in our measures, such as this one on patient waiting time versus time of appointment:

There would be variation in any process, split into the normal random common cause variation (such as the background up and down motion of the data points around the center line) and special cause variation (like the points in the box). Voila! Masterful! It took a chart to show that patient wait time increased over the lunch hour!

I don't mean to be too sarcastic. There were many instances in which statistical process control led to significant improvements in patient care. For instance, we were able to reduce the time it took for a patient with chest pain from a blocked coronary artery to reach the Cath lab from 2 hours to 32 minutes. The problem came when we thought that everything could be improved this way.

Fresh from this success with the coronary blockage, we attempted to use the same techniques to reduce the time from abnormal mammogram to biopsy. When we started, that time was measured in weeks! Imagine the stress caused to the patient! We were able to get it down to 4 days…but the effort destroyed everything else in the pathology department, as we didn't have the structure to support the process.

More than a half-century ago, Avedis Donabedian understood that Outcome depends on a delicate dance between process and structure:

The structure that supports a process is more than bricks, mortar, and machines. It includes the intellectual capital of the professionals caring for the patient, the expectations and emotional state of the patient, the family structure, even the climate! This is the fallacy of believing that importing "best" practices will be the answer. The "best practices" of the Mayo Clinic work because of the interactions of all these elements.

It works well at the Mayo Clinic, but may, and often doesn't, work at other places. Indeed, even the Mayo Clinic recognizes that nuances of varying community needs must alter their care processes. What needs to be done is to use Critical Thinking skills to discover the "best practice" for the unique makeup of patients, professionals, and the system for each location.

We have empiric reasons to prove this is a viable approach. In 1990, Marian Zeitlin, Hossein Ghassemi, and Mohamed Mansour published Positive Deviance in Child Nutrition. A year later, Zeitlin followed this with a journal publication on the same topic. Both the book and the journal article noted that in impoverished countries, some children seemed to flourish (Positive Deviants) while others in the same situation did not.

The authors discovered that simple, sometimes overlooked factors such as supplementation with locally available nontraditional but high-quality foods, social interaction, and praise played monumental roles in achieving success.

By identifying these Positive Deviants and understanding what made them stand out, the differences could be applied to the larger population, with significant improvement. Importantly, these differences would only be applicable to individuals in the same micro-environment. It was the interacti...