The trusted source for
healthcare information and
Critical Path Network
Control charts: Valuable tools if you know how to use
Not every process is an appropriate target
Control charts, quality tools that can help tighten the focus on process variations, increasingly are gaining acceptance among some health care quality professionals. In fact, a number of Joint Commission on Accreditation of Healthcare Organizations requirements specifically mention the use of control charts.
However, as with many quality tools adopted from industry, the learning curve can be fraught with confusion and misconceptions.
A tool of statistical process control (SPC — the data analysis method of quality improvement), a control chart is a graph that "provides a pictorial representation of what you measure over a period of time and allows you to identify when special causes of variation are active in your process,"1 according to author D. Lynn Kelly.
"What a control chart does for you is tell you whether, over time, the process is varying," points out Steve David, MBA, president and CEO of SkyMark, a Pittsburgh-based software manufacturer. "Every process is bound to have some variation, like when you sign your name. This is a common-cause variation; I like to call it an expected variation. But when somebody bumps your elbow when you sign, that’s a special cause, and it will show up in the data." (For examples of health care common-cause and special-cause variations, see box, below.)
"If you want to improve a process, the first thing you want to do is insulate it from those special causes," he continues. "In a hospital, if you track patient falls, you want to make sure that people know that you know when somebody is likely to be fall-prone; i.e., that you have the kinds of systems in place that insulate your process from special causes. In Phase I, you recognize special causes and try to eliminate them. But in Phase II, once you have variation within expected limits, you try to narrow it further to get a process that minimizes common-cause variation. That way, you get a nice, predictable process."
One of the challenges of working with control charts is understanding what they tell you. For instance, just because a process is in control doesn’t necessarily mean you are providing optimal health care. "Don’t assume that just because a process is in control it is a good process," notes Patrice L. Spath, RHIT, a health care quality consultant with Brown-Spath & Associates in Forest Grove, OR. "Say you’re looking at wait times in the emergency department. The way you admit patients creates a certain wait time. You can plot that, and it may be stable. That doesn’t necessarily mean that’s the best way to do it."
"You may be in control with a 100% mortality rate. That will make you predictable but certainly not satisfactory," adds Marilyn Hart, PhD, professor of business at the University of Wisconsin, Oshkosh, who lectures and writes about health care and SPC. She and her husband Robert Hart are co-authors of Statistical Process Control for Health Care (Pacific Grove, CA: Duxbury; 2002).
This does not mean that control charts are bad or misleading tools; it means that, like many tools, they have limits and recognizing those limits will help you use them properly. "One of their limits is that you can have a nice, predictable, stable process that is terrible health care," David says. "Let’s say your target C-section rate is 12% and you are running merrily along at 28% to 30%; that is not good health care. Typically, a control chart doesn’t have and shouldn’t have a target line on it — it just answers the question, Are we in control?’"
Considering the limitations and benefits of control charts, how do you know when to use them? "You don’t get a control chart and then see where to hang it," Hart warns. "You start with a process that needs improvement, then you go to a decision point, such as, do we need a control chart, or do we need a tally sheet?" The first step, she advises, is to select a process to study by posing these questions:
"Only study processes that you can and will take corrective action on when problems or opportunities to improve are discovered," Hart says. "Otherwise, people lose faith in the system."
Once you have selected the process to study, she continues, you should use a control chart in these cases:
As new data arise, they immediately are plotted on the chart with the previously established limits to see if the new data reflect the common-cause system or whether they are due to a special cause of variation. If they are due to a special cause and the cause was detrimental to the process, it must be investigated and corrective action taken. If the cause was an improvement, it should be made standard operating procedure.
"You seldom, if ever, start immediately with a control chart," adds Hart. "You may, for instance, use a tally sheet and then a pareto chart to first identify the part of the process you may wish to put a control chart on. (See a list of alternative tools and when to use them.)
In a hospital setting, there are a number of practical reasons and specific processes that would seem to lend themselves to control charting.
"A lot of the things the Joint Commission requires are specific and well-defined — i.e., the number of patient falls or the number of times you use restraints," David notes. "Or things like the number of readmissions within a certain time of discharge or the amount of time it takes patients who present with chest pain to be dosed with thrombolytics. Control charts are one good way of charting quality and timeliness indicators."
Judy Homa-Lowry, RN, MS, president of Homa-Lowry Healthcare Consulting in Metamora, MI, agrees. "In terms of hospitals, one of the most common places to start would be to measure quality control processes; there is a Joint Commission requirement to do that," she says. "So it makes sense to do what you have to do based on regulatory standards that look at patient safety. In terms of looking at things as basic as refrigerator checks, it’s a good tool.
"Many times when we visit organizations, people know they have to monitor refrigerators," she continues. "Now, in the lab or pharmacy, that may not be an issue, but in a patient unit, where you store food and meds in separate refrigerators, the collection of those data is not always 100%. Second, when you do begin to look at those data, most people just look at whether things are being done or not, but there should also be some review of ranges."
The generally accepted range of variance in control charts is three sigma, or three standard deviations from the mean; that’s approximately three times as large as what is allowable in the Six Sigma process but, nonetheless, would indicate that more than 97% of the variations are within the acceptable range.
"You could simply measure the temperature with a thermometer, and that will give you a number," adds Brian Lowry, Homa-Lowry’s husband, and vice president of Technical Services at Curtis Metal Finishing Co., in Sterling Heights, MI. "If the meds are not being allowed to freeze, you might try to maintain the temperature above 32° and below 46°. If you simply checked the refrigerator on a given day to make sure the temperature was between those two, and it was, that would point you to a chart that only says good’ or bad.’ But if you looked at it and recorded the temperature on every shift, you’d get three readings a day and might see if temperatures are swinging wildly from day to day. If you averaged out all the different temperatures, your average reading could be fine, but if they swung wildly from day to day, that could point to a problem."
Basically, a control chart will give you an aggregate picture of a global issue, Spath says. "Let’s say you are measuring nosocomial pneumonia rates. If all you knew was that the rate went up, you would not know the why; you would then use a control chart to monitor the stability of performance over time. It might be performance of an outcome, or it might be specific steps in a process."
Returning to the pneumonia example, Spath notes that "the number of cases could be plotted on a control chart and it would be a measure of outcomes. You then also could plot specific process steps that impact whether someone gets pneumonia, such as the percentage of patients who ambulate within 24 hours post-op, or who sit at the edge of the bed within 12 hours post-op. The analog in manufacturing would be studying each step in making the widget vs. how much the completed widget conforms to standards."
Control charts are used most effectively when you are aware of what they can and cannot do, and when they work best. "Control charts give the most information when they are kept on a single-stream process," Hart cautions. "Therefore, do not mix data from different shifts, different surgeons, etc., on the same control chart until after they [have been] kept on their own control chart, found to be in control [stable over time], and found to be similar by use of a control chart with rational subgroups."
The best control charts are done in real time, she says. "If you really want to improve processes, once you’ve identified the process, collect some data to start. You gather the data to plot your limits; you may or may not be in control. If you are not, you may project them into the future as trial limits, so when I get my points, I plot them immediately and know if I am inside the limits. As new data arise, they are immediately plotted on the chart with the previously established limits to see if the new data reflect the common-cause system, or whether they are due to a special cause of variation."
If you know right away that you have a special-cause variation, "you have a much better chance of finding it," Hart notes. "If, however, I plot a chart and see that three weeks ago I was out of control, that’s a different story." Finally, she advises, "Look at your data before, then make the change, then look again to make sure you did change."
Spath reiterates that the goal of process improvement is not met merely by charting the output from your current processes and keeping that output stable. "That’s a mistake people are making," she says. "They have wonderful control charts that show everything is stable. But until you benchmark your results with another organization, you may not really know how well you’re doing."
Another challenge is that there are many different types of control charts, and it is important to know when to use which chart. "Some people learn how to use one control chart, and they use it for everything," Spath notes. "Or they have software that let you plug numbers in, and whatever they give them, they use. Other software has decision trees that help them along."
The bottom line is, control charts can help you identify a problem or problems in your processes; you still have the hard work of improving that process ahead of you. Or, as Spath says, "A control chart lets you know when to get excited — when a process has become unstable, and when you need to dig into the cause of it."
1. Kelly L. How to Use Control Charts for Health Care. Milwaukee: ASQ Quality Press; 1999.