Congratulations to The Californian for an excellent display of multiple data points in the April 28 article, "Hospital, doctor earn 'worse' mark," which focused on Bakersfield Heart Hospital and one of its surgeons, Dr. Sarabjit Purewal. All too often, published graphics connect too few data points. This unfortunate practice not only tells readers very little, it also can be innocently yet deceptively misleading, as in this instance.
The otherwise well-intended report by the California Office of Statewide Health Planning and Development is misleading. It reports only a single data point for hospitals and surgeons statewide. It compares each with the statewide average for only a single year.
This OSHPD report has fallen into the statistical trap of connecting too few data points to present what is actually happening. Moreover, logical conclusions most of us innocently would draw from these data can be totally incorrect.
I don't profess to be an expert in statistics. However, many in our local business sector work with a valuable management tool called a Control Chart for the purpose of understanding variation in systems and processes including mortality rates of hospitals.
To the credit of OSHPD, their other reports do include multiple data points. It's on these reports that The Californian's graphic is based. However, OSHPD included only "raw data," not in the form of a Control Chart.
Although construction of a Control Chart may seem complex, it's not. Its interpretation is even easier. Here's how it works:
If all data points are within the two (upper and lower) "control limits" in the Control Chart, then a system or process is understood to be under control. No remedial action is needed.
However, if one or more of the data points should fall outside either of these two control limits, each is construed to be out of control. Remedial action would be needed.
Where The Californian excelled (and the state failed) is including a graphic that displays data from the preceding eight years. The Californian's longer-term graphic permits an entirely different picture to emerge.
Using these statewide data in a Control Chart tells us that Bakersfield Heart Hospital was indeed "above the line" four times in the past eight years, as reported. However, the other two local hospitals similarly were "above the line" during this same eight-year period, one four times and the other five times.
That's the "bad news." This is the "good news." If you work with these same data based on local outcomes, an entirely different conclusion can be drawn. On this basis, each of the three hospitals is within the two control limits. In other words, no remedial action would seem to be required for any of the three.
Regional differences are not uncommon, especially in a state as large and diverse as California. Many already have criticized universal application of the Affordable Care Act (aka Obamacare) by correctly saying that a national "one size fits all" approach will not work.
Significant differences exist from one region to another. This is especially true in the nature and condition of patients treated, as the OSHPD report correctly indicates -- and as OSHPD properly accommodates in its data. Such variations would be expected to be even more pronounced on a national basis.
The lessons learned here are that variations in heart surgery mortality rates -- or data charted for any purpose -- first should be based on multiple data sets over a sufficient period of time. Second, these data need to be converted to a Control Chart to avoid honest yet misleading conclusions.
OSHPD should do so to be certain future reports will more appropriately help readers, and hospitals, better understand where remedial action is needed or not needed.
Working with inadequate data, inappropriately displayed, can lead to devastating interpretations -- not only for an institution and its physicians but also for its patients, who properly and justifiably trust hospitals and surgeons with their lives.
John Pryor is a risk management consultant in Bakersfield. For details on this Control Chart, email him at firstname.lastname@example.org. Another View presents a critical response to a previous editorial, column or news story.