When Processes Moonlight as Trends
It took a little more than five years of studying W. Edwards Deming's message before I finally got it. Until then, I could apply statistics to a manufacturing environment on "product" issues; however, I would then attend meetings like the following on administrative issues, "forget" my statistics, and either get caught up in the discussion of "trends" suggested by the analysis below, or even generate such an analysis if asked.
This is a true story and typical of what I see played out in boardroom after boardroom--comparing the past 12 months of performance to the previous 12 months (usually via two bar graphs with some kind of trend line). The following is a report from an emergency-medicine environment and focuses on a particular activity, in this case responding to cardiac arrests.
The report stated:
"We're running a slightly higher number of cardiac arrests per month. The total amount of cardiac arrests has risen from a mean of 21.75 (June 94–May 95) to 22.92 (June 95–May 96). This is an increase of 14 cardiac arrests in the last 12 months."
My experience is that governing boards always have the vague objective of "knowing how we're doing," which is then passed on to an analyst who gladly and dutifully complies by diving into organizational databases. These vague objectives lead to vague analyses, which always seem to lead to vague conclusions and vague follow-up actions.
The report continued:
"Next, we interpreted the data relating to Vfib Cardiac Arrests…. This could be significant to our outcome, and… indicates a need for more sophisticated statistical analysis. It was already shown that the number of cardiac arrests has increased by a mean of 1.17 per month. Now we're adding to that increase a decrease of times we're seeing Vfib as the initial rhythm. From June 1994 to May 1995, we arrived on scene to find Vfib as the initial rhythm with an overall mean of 6.75 times. That gave us a capture rate of 32.03 percent. This last year, June 1995–May 1996, we're arriving to find Vfib as the initial rhythm with an overall mean of 5.92, and a capture rate of 25.81 percent. This obviously means that during the last year, we've responded to more cardiac arrests and found them in more advanced stages of arrest."
Let's test out these conclusions. I was handed this report to use as an example in a seminar and intuited that "total arrests responded to" is an important organizational number. It's also clear that the proportion of these arrests in "Vfib" is important. In the latter case, I had to create it (for example, 6/94 is six out of 18, or 33.33%). Sorry to horrify p-chart purists again, but I then "plotted the dots" to produce two run charts as shown above.
As you see in both cases, there's neither a "trend of seven" nor any "run of eight consecutive points either all above or below the median." So, the conclusion is that nothing has changed in two years--which everyone in the room would have concluded in about 15 seconds had the data been presented in this way and had they known the simple statistical theory of run charts. Moreover, it could have saved precious meeting minutes of empty discussion and hours of subsequent searches as people looked for explanations--and found them.
As espoused in previous columns and once again demonstrated, never underestimate the power of looking at a situation as a "process" and "plotting the dots"--even in administrative situations different from your day job.
Davis Balestracci is a member of the American Society for Quality and the Association for Quality and Participation. He previously served as chair of the statistics division of ASQ. His book, Quality Improvement: Practical Applications for Medical Group Practice (Center for Research in Ambulatory Health Care Administration, 1994), is in its second edition. Visit his Web site at www.dbharmony.com.