How can statistics and data analysis help improve outcomes in healthcare? William H. Woodall, professor of statistics at Virginia Tech, has been focused on that question for over ten years. A
renowned expert on quality control, he has published over 150 refereed papers, most on various aspects of process monitoring. His healthcare-related publications include “Plotting Basic Control Charts: Tutorial Notes for Healthcare Practitioners,” “Use of Control Charts in Health-Care and Public-Health Surveillance,” and “Dynamic Probability Control Limits for Risk-adjusted Bernoulli CUSUM Charts.” In 2015, he published a paper in collaboration with Professor Stefan Steiner of the University of Waterloo and surgeon Dr. Sandy Fogel, “The Monitoring and Improvement of Surgical Outcome Quality.” Minitab asked Dr. Woodall to share some of his experiences and thoughts about the role data analysis can play in making health care better.
Is quality improvement different for healthcare than for other industries?
There are complications when you’re dealing with people, because patients themselves are going to vary quite a bit. So in some cases the situation is more complicated than it is in manufacturing, but it certainly can be done—and the fact that improving healthcare quality means saving and improving lives makes it much more important and interesting. The use of data-driven quality improvement methods is much more recent in healthcare than in other kinds of organizations, but that just means there’s a lot of potential. As emphasized by the
Institute for Healthcare Improvement, one must have data before and after any process improvement initiative in order to check the level of success. That’s fundamental—without data, one doesn’t even know the current level of performance. Also, many processes in hospitals are very similar to processes in industry, and those processes can be improved using more standard approaches.
Some still look at medicine as more of an art than a science, but I think that group of people is shrinking in size. I think it would be helpful to have a course on process improvement taught in every medical school.
What is your sense of the sentiment about data analysis and quality improvement in healthcare?
There is a growing demand for the use of statistics in healthcare quality improvement. Interest is very high, and
The Joint Commission, which certifies hospitals, has quality improvement requirements. It is my impression, however, that healthcare providers are often much less comfortable with data and statistical analysis than, say, manufacturing engineers. A lot of people get into healthcare so they can work with people, not numbers, and my experience is that some healthcare professionals are less prepared to deal with analyzing data than would be desirable. For example, I’ve seen where some practitioners have taken data on the amount of time spent in a waiting room or other similar measurements, which are continuous data, and convert them into binary pass/fail data based on some upper specification limit. They don’t realize that in making that conversion, they’re actually losing a lot of valuable information in their data. Although many statisticians are involved in clinical trials, there’s a need for more statisticians and industrial engineers to collaborate with practitioners on projects and data analysis for quality improvement.
Are the data analysis needs of healthcare quality practitioners unique?
Well, the kind of data that a healthcare professional is likely to see isn’t always the same kind of data one would see in a manufacturing plant, so if healthcare professionals are taught to improve healthcare quality using manufacturing quality data as illustrations, one can understand why they’d get the impression that the methods don’t apply to them. It can be a different set of problems! In healthcare, there’s a focus on the use of attribute data, and on time-related responses on how long it takes to do this or that task. Many projects are focused on trying to reduce the time between surgeries, the time between hospital-acquired infections, the time to receive test results, and similar metrics. There’s a lot of binary data—you know, patients either survive 30 days following surgery, or they don’t. That means you’re not going to find as many applications requiring the use of normal distributions in healthcare, and so someone learning to do quality improvement in healthcare might not want to start with the same methods one would learn for improving quality in a manufacturing plant. I’d start with the methods for analyzing attribute data and time-related data.
How did you get involved with the application of data analysis in healthcare?
I’ve studied process improvement and quality control in industry for more than 30 years, but I’ve been working on healthcare-related monitoring for over ten years. I began gravitating toward health-related quality improvement primarily through the influence of Professor Stefan Steiner of the University of Waterloo, who had been working with physicians in the U.K. on monitoring surgical outcome quality. We then spent several years studying and developing public health surveillance methods. In the past few years, I’ve worked with Dr. Sandy Fogel, a surgeon at Carilion Clinic in Roanoke, Virginia. He is a
National Surgical Quality Improvement Program (NSQIP) champion, whose job is to improve surgical outcome quality. By the comparison of his risk-adjusted outcomes with those of other NSQIP hospitals, Dr. Fogel was able to identify areas that needed improvement. He then worked with his colleagues to implement best practices that reduced the overall mortality rate after surgery and the rate of surgical site infections. We published the results of his work as a case study last year. I recently analyzed data collected by Dr. Fogel, showing that his process improvement initiatives have reduced the average length of stay after surgery. This is tied to the
Enhanced Recovery after Surgery Program.
Have you encountered reluctance among professionals to implement quality methods and statistical tools in their work?
I believe that the primary obstacle to the greater use of data and statistical tools is the need for more education and training. Simple methods can solve probably the vast majority of the problems—but of course, simple is a relative term.
The primary obstacle to implementing changes is convincing people to change their habits. Getting anybody to change can be hard—they don’t want to have to change what they’re doing. People like Dr. Fogel have no reluctance in using statistical methods, but he says the biggest barrier to improving surgical outcome quality is getting surgeons to change what they do, because, like the rest of us, they’re creatures of habit. They like doing the same thing over and over, they’re very good at it, and they’re confident in what they’ve been doing. Most often it is the process, not the surgeons, needing improvement. But there’s another factor, which is that in a manufacturing company, a supervisor can often require change. In some hospitals, the doctors work as independent contractors, so they have to be convinced to change. In the paper we co-authored, Dr. Fogel said that it is best to make process changes behind the scenes, if possible, where the surgeons are not directly involved.
What advice would you give to someone who wants to use data to improve a quality issue?
First, give careful thought to what data are required. I would then advise plotting the data before doing any analysis. Any data collected over time should be plotted in time order. Finally, focus on using the simplest adequate statistical methods. It is important to convince people that the conclusions of your analysis are valid, so they need to understand the analysis. A fundamental principle of applied statistics is that more complicated is not necessarily better. Generally, people are not going to be convinced to change based on arguments they don’t understand. So one shouldn’t make statistical analyses more complicated than necessary.
Do you anticipate any future trends for QI and SPC in healthcare?
The healthcare industry is becoming more concerned with outcomes. There are different types of quality metrics, and although one needs both process and outcome measurements, outcome information is more valuable. A typical process measurement would be the percentage of emergency room patients displaying heart attack symptoms who get treated within a certain amount of time. These metrics are very useful, but they do not tell us everything we want to know. I could, for example, measure whether my students and I show up to and leave class on time, how often the students submit homework on time, etc., but that wouldn’t necessarily mean anyone learned anything. That’s what’s behind the emphasis on outcome measurements. What are the outcomes for the patients? Which patients survived 30 days after surgery? Which ones had surgical site infections within 30 days? Metrics based on outcomes can be more informative than process measurements, and the federal government is becoming more focused on assessing performance based on outcome measurements.
How can statistical software help healthcare professionals improve quality in their organizations?
Statistical software is clearly a necessity for understanding and analyzing data. I like and use Minitab because it is very easy to use and provides all of the methods I need in process improvement situations. I support the use of Six Sigma in healthcare, and I would encourage those needing statistical software to give Minitab their consideration. The user-friendliness of Minitab is especially important in healthcare.
To obtain any of the following papers, contact Dr. William Woodall at firstname.lastname@example.org:
Mohammed, M. A., Worthington, P., and Woodall, W. H. (2008), “Plotting Basic Control Charts: Tutorial Notes for Healthcare Practitioners”. Quality and Safety in Health Care 17, 137-145.
Steiner, S. H. and Woodall, W. H. (2016). “Debate: What is the Best Method to Monitor Surgical Performance?”, BMC Surgery. 16:15 DOI 10.1186/s12893-016-0131-8. Available at http://bmcsurg.biomedcentral.com/articles/10.1186/s12893-016-0131-8.
Woodall, W. H. (2006), “Use of Control Charts in Health-Care and Public-Health Surveillance” (with discussion), Journal of Quality Technology 38(2), 89-104.
Woodall, W. H., Fogel, S. L., and Steiner, S. H. (2015). “The Monitoring and Improvement of Surgical Outcome Quality”. Journal of Quality Technology 47(4), 383-399.
Zhang, X. and Woodall, W. H. (2015). “Dynamic Probability Control Limits for Risk-adjusted Bernoulli CUSUM Charts”. Statistics in Medicine 34, 3336-3348.
In unserem kostenlosen monatlichen E-Newsletter finden Sie die aktuellen Neuigkeiten rund um Minitab, Lernprogramme, Fallstudien, Statistiktipps und weitere nützliche Informationen.
Teaching Statistics and Quality, One Story at a Time: Q&A with Dr. Ron Lasky
Durch Ihre Nutzung dieser Website stimmen Sie zu, dass Cookies verwendet werden. Cookies dienen zu Analysezwecken und zum Bereitstellen personalisierter Inhalte. Lesen Sie unsere Richtlinien