STATISTICAL THINKING
The early data were taken usually as three to four readings per day. Beginning Jan ‘95 most data were taken once/day in the morning except during periods of high blood glucose, when more readings were taken. In general, analysis of the days with multiple readings failed to show any specific pattern in blood glucose levels throughout the day; thus, the data are graphed all together.
Benefits
There is clearly still plenty of room for improvement in controlling the complex process of Carolyn’s blood glucose. However, we have already obtained many benefits, most of which we believe can be directly attributed to our use of Statistical Thinking with control charts:
1. We met our objective of safely getting down to one blood glucose test per day. This was accomplished because we successfully reduced the variation in her “process” making it possible to “see” and predict within limits what was actually going on with her blood glucose.
2. We learned much about the causes of Carolyn’s high blood glucose (different foods, exercise, illness, infections, and especially emotional stress.)
3. Carolyn is taking less insulin now than when we started in June ‘94 and she also has lower blood glucose (current insulin = 26 units; June ‘94 = 36 units.) My theory is that her body has now gotten used to lower blood glucose and is now more robust against fluctuations. In fact, Carolyn used to feel jittery when she was at near normal blood glucose levels; she no longer feels that way.
4. Carolyn was sick less often during the winters of 1994-1995 and 1995-present, compared to ‘93-94. (Winter ‘94-95: sick four times. 1995-present: sick two times. Winter ‘93-94: No hard data but memory recalls a frequency of an illness every one-three weeks.) In addition Carolyn has been much more energetic. High blood glucose tended to make her lethargic, whereas with lower blood glucose she clearly has had more energy.
5. A successful control strategy was developed for appropriately correcting high blood glucose with minimal over-control.
6. The time between insulin reactions has improved, that is, the time between these reactions increased when compared to where it was when we first started to control the blood glucose. (Figure 14, Note: log of timebetween-reactions was used due to the highly skewed nature of the time data.)
STATISTICAL THINKING
7. Carolyn has had a significant drop in her Glycosylated Hemoglobin (GH) levels since we started this program. (Figure 15). This test is an estimate of the long-term average blood glucose over approximately the last two months prior to performing the test. It is an indicator of future complications due to diabetes. Recent studies have indicated that people with GH levels below 8.1 percent have significantly lower risk of kidney disease.
8. Carolyn was anticipating having a second laser surgery on her eye in Nov. ‘94; however, the eye exam in November indicated no need to have the surgery.
9. We have also developed some additional ideas for further improvement of our control of Carolyn’s diabetes. One thought to reduce the chance of insulin reactions in the night is to have Carolyn take a small amount of food (a sugar tablet) in the early morning (e.g. 2 or 3 am.) Of course, that idea is not very popular with the subject and has been rejected at this time.
Lessons Learned
Beyond the tangible benefits of the control plan for Carolyn’s health, we have also learned or reaffirmed some additional valuable lessons regarding Statistical Thinking and science. These include:
1. A process has no regard for the “specifications”; it just does what it is capable of doing. There is no better evidence than this case of diabetes. The insulin regulatory system controls blood glucose to within +/- 25 mg/dl. The best we’ve been able to get Carolyn’s variation is +/- 104 mg/dl. Tampering only increased the variation.
2. Statistical Thinking should be used to help us develop theories as well as to test them. To do so we should remember to use all of our scientific/mathematical background to try to explain the patterns in data.
3. Chaos is alive and well and living in diabetes. Recognition of the presence of conditions in some systems that can result in chaos can help us be conscious of the importance of working to create robust processes which are less sensitive to sources of variation.
4. The real value of control charts and Statistical Thinking is to help us learn about our processes. It is a serious fallacy to avoid introducing control charts to a process due to lack of adequate knowledge of how to control it. Failure to introduce the charts essentially guaranties that one will continue to be ignorant of how to control the process.
5. The human body is a marvelous creation that is extremely robust
STATISTICAL THINKING
While I was preparing this case study my wife pretty well summed up our results when she said “I don’t know what the data say about whether or not my diabetes improved, but I can tell you for sure I know it worked because I feel a lot better now than I did before we started!” As my friend, Ken Kotnour, once said in quoting Dr. Deming: ” the customer doesn’t always know what they need, but they will treasure it if you give it to them”.
“I don’t know what the data say about whether or not my diabetes improved, but I can tell you for sure I know it worked because I feel a lot better now than I did before we started!”