During this article, I will be asking you to
look at certain examples on our live
Statit piMD demo. You may want to log in
and register now at http://pimd.statit.com
in order to view the examples referenced. If
you have already registered, you will go directly
to the examples.
When we first start collecting data for a process
and plotting it on a control chart, we may find
that it is not in statistical control. Our first
task is to identify the special causes (also
referred to as assignable causes) and deal with
them. Our goal at this point is to get our process
to be predictable in that it is operating within
its random variation limits. For example, on
our Statit piMD live demo site, take a look
at the indicator PN: 2 Pneumonococcal Screening
and/or Vaccination under Competency->Core
Measures->PN. This chart shows a situation
where we do not yet have a predictable process.
The other indicators in this subgroup, PN:
3 and PN: 1, are in control, and thus are predictable
in that we would expect the process to operate
within the control limits.
This early process analysis helps with Montgomery's
first two points of improving productivity and
defect prevention. It is also part of Wheeler's
Process Trial Charts.
Once we have the chart in statistical control,
we can move on to monitoring the process. At
this point we are trying to maintain the status
quo; we are working to keep the process running
within the limits. Our process is predictable
and we can expect our non-compliance rate to
be stable. If we get a high measurement, we
will only make an adjustment if it has exceeded
the control limits. We look for special causes
when a point goes out of control and make necessary
adjustments. Now we are preventing unnecessary
process adjustment (Montgomery) and using the
charts as Process Adjustment Charts (Wheeler).
Many people stop here. They use the charts
only to maintain the status quo. They are using
the charts for Statistical Process Control (SPC).
But, in doing so, they miss a great opportunity
for improvement. We need to break into the area
of Statistical Process Improvement (SPI).
With a stable process, we may still have a
certain number of non-compliant events. But
what is the Cost of Quality for these? Would
it be a good idea to further improve the process?
Are we satisfied?
Again, we can use the charts to take improving
productivity (performance) and preventing non-compliance
to a higher level. Control charts will give
us some diagnostic information and some information
about process capability that will help us in
our search for process improvement. Process
capability studies give us insight into how
our process is performing in relation to specifications
and of what the process in the current state
is able to. They will indicate whether we need
to shift the mean or reduce the variation.
Now we are working to improve the quality of
the output of our process. How do we improve
the process? We need to either shift the mean
or reduce the variation. We may know what to
do simply because we know the process so well.
However, we may need to try an experiment because
the process is complex or we haven't learned
enough about the process. Identify a process
change, implement it and study the results.
This is where Wheeler's Process Trial Chart
comes in. You can draw a chart of this data
and compare it to the charts for previous baseline
data, or you can implement a phase change with
Statit.
An example of a phase chart is AMI-4 Adult
Smoking Cessation Advice/Counseling under
Competency->Core Measures->AMI. Here
the change we made in the process made a significant
improvement to the process by shifting the mean
in the positive direction.
In Statit piMD, the expert of the indicator
can add a phase. For example, Customer Relationships,
3rd Available Long Appt. - Hood River is
seeing some improvement. There may have been
a systemic change that precipitated the change
in the chart. If so, perhaps the expert would
want a phase introduced at the time of the change
in the process.
The expert would determine if the change had made
a significant difference in the positive direction.
If not, PDSA (Plan, Do, Study, Act) again. But
this is a critical use of the control charts:
moving from the status quo to process improvement.
Process improvement is particularly important
in the healthcare area.
Finally, the control charts can be used for
gaining in-depth analysis of the process. We
can use what Wheeler calls "Extended Monitoring
Charts" to chart several parameters of
the process to discover which parameters are
the best predictors of process performance.
This is important information. Too many charts
to monitor can sometimes be as bad as no charts.
Finding a special cause for a parameter that
has little effect on the process outcomes is
probably a waste of time. However, knowing which
parameters have the most effect will help us
decide which charts to have process owners and
operators monitor. Another example would be
to use a composite score to judge the process
as illustrated with the HF Composite indicator
at Competency-.Core Measures->HF.
In short, you can use control charts to:
Use Control charts to not only control your
processes, but improve them as well.
References
1. Montgomery, D.C. (1997). Introduction
to Statistical Quality Control (3rd Ed).
New York: John Wiley and Sons, Inc.
2. Wheeler, D.J. Chambers, D.S. (1992). Understanding
Statistical Process Control, (2nd Ed). Knoxville,
TN: SPC.
