International Journal for Quality in Health Care 1998; Volume 10, Number I: pp.

69-73

Methodology matters — VIII

'Methodology Matters' is a series of intermittently appearing articles on methodology. Suggestions from readers of additional topics
they would like to see covered in the series are welcome.

Use and interpretation of statistical

quality control charts
JAMES C BENNEYAN
Department of Mechanical, Industrial and Manufacturing Engineering, 334 Snell Engineering Center, Northeastern University, Boston,

Downloaded from http://intqhc.oxfordjournals.org/ at Universidad Politécnica de Madrid on May 28, 2012

MA 02115, USA

Originally developed at Bell Laboratories by Dr Walter caused by regular sources within the process or its en-
Shewhart [1] in 1924 specifically to help detect statistical vironment, data exhibiting natural variation occur in pre-
changes in process quality, control charts have since become dictable and relatively common frequencies.
one of several primary tools of quality control and process Conversely, outcomes or in-process observations that have
improvement. Quality control charts are chronological graphs very small probabilities of occurrence, assuming only natural
of process data that, although based in statistical theory, are variability, usually represent deviations from the regular pro-
easy for practitioners to use and interpret. These charts cess. Such events suggest that the process fundamentally has
also can help users to develop an understanding of the changed, for better or worse, due to atypical unnatural
performance of a process and to evaluate any benefits or variability that should be traced to root assignable causes for
consequences of process interventions, complementing tra- management intervention. Examples of 'special causes' of
ditional methods by providing additional longitudinal in- unnatural variability might include changes in clinical pro-
formation that otherwise might not be detected [2]. cedures, skill degradation, equipment failure, new staff, and
This article provides a brief introduction to the use of statistical changes in population demographics, the rate of disease, or
quality control charts for analyzing, monitoring, and improving a patient's physiology.
health care processes. After an overview of key concepts, several Finally, the term 'statistical control' refers to the stability
examples illustrate control chart use and interpretation. The and predictability of a process over time and to the type of
article concludes with some common pitfalls to avoid and ref- variability that exists. A process that is completely stable over
erences for further exploration. Readers unfamiliar with the topic time exhibits only natural variability, with its regular random
should also read introductory materials on the principles of behavior remaining unchanged, and is referred to as being
quality management and the philosophies of the late quality in a state of statistical control. Conversely, a process that
pioneer, Dr W. E. Deming [3]. changes from its norm will exhibit unnatural variability and
is referred to as being out of statistical control. Note that it
is usually quite difficult to determine intuitively, without the
Key concepts: natural variability and aid of control charts, which type of variation exists and
therefore whether direct intervention ultimately would be
statistical control beneficial or harmful to process outcomes.
Most processes exhibit some variability, all of which can be
classified into one of two categories: 'natural' or 'unnatural'.
The natural variability of a process is the systemic variation Control chart format, use and
inherent as a regular part of the process. Some examples of interpretation
'common causes' of natural variability might include the time
of day, hospital census and case-mix, weight and physical Figure 1 shows the general format of a statistical control
condition of patients, other patient-to-patient differences, chart. Process data are collected at certain intervals over time
and varying behaviors and demographics. Because they are and formed into rational time-ordered subgroups (such as

Address correspondence t o James C Benneyan. Tel: ( + 1 ) 617 373 2975; Fax: ( + 1 ) 617 373 2921; E-mail:
benneyan@coe.neu.edu

69
Methodology Matters: J. C. Benneyan

Upper Control Limit (UCL)

Subgroup
Statistic
Center Line (CL)
(e.g., average,
standard deviation,
rate, number,
proportion)

Lower Control Limit (LCL)

I •I•I•I•1•I•I • • I •1 •I •I •I •I •I . 1 •
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 4 2 .

Downloaded from http://intqhc.oxfordjournals.org/ at Universidad Politécnica de Madrid on May 28, 2012

Subgroup Number

Figure I General format of a quality control chart

by day or week). Some value of interest is calculated from each Table I Criteria for not being in a state of statistical control
subgroup soon after being collected, plotted in chronological
sequence on the chart, and then evaluated for typical versus Control chart out-of-control signals
statistically irregular behavior. At least 25 subgroups are
recommended to start a control chart. Any single subgroup value outside either control limit
Along with these subgroup values, three horizontal lines Eight consecutive subgroups on one particular side of the
are also calculated and plotted on the chart. These are called centerline (CL)
the centerline, the upper control limit, and the lower control Twelve of fourteen consecutive subgroups on one
limit. The centerline and control limits help define the in- particular side of the centerline
control central tendency and natural variability of the process, Three consecutive subgroups beyond 2 standard deviations
respectively. The centerline is almost always set equal to the on a particular side of the CL
arithmetic mean or expected value of the plotted statistic, so Five consecutive subgroups beyond 1 standard deviation
that approximately half of the subgroup values will fall on on a particular side of the CL
each side. The control limits are then usually set equal to the Thirteen consecutive subgroups within + 1 standard
centerline plus and minus three theoretical standard deviations deviation (on both sides) of the CL
of the plotted values. Six consecutive subgroups with either an increasing or
A process is considered to be in statistical control if all of decreasing trend
the plotted subgroup values are between the control limits Cyclical or periodic behavior
over a sufficient span of time. There also should be no
evidence of unusual behavior between the limits, such as
trends, cycles, visibly evident changes, and other forms of
non-random variation such as those defined by the additional
rules listed in Table 1. Note that unequal subgroup sizes the process is operating closer to some target value. Several
(such as a differing number of medications filled each week) health care examples of these uses can be found in reference [5].
result in varying control limits, although chart interpretation
is basically the same.
In this fashion, control charts are valuable for several Common types of statistical control
purposes throughout the process improvement cycle. These charts
uses are described and illustrated in greater detail elsewhere
[1,3-5] and include: Several common types of control charts exist, each con-
structed using slightly different formulas and each being
(1) testing for and establishing a state of statistical control;
appropriate for different types of data. The appropriate
(2) monitoring an in-control process for changes in process
equations for a given chart can be found in any good quality
and outcome quality;
control text [4,6,7]. Determining which specific type of control
(3) identifying, testing, and verifying process improvement
chart should be used is based on identifying the type of
opportunities. continuous or discrete data to be plotted. For example, three
The long-term objectives are to tighten the control limits by of the most common statistical distributions are the normal,
reducing process variation and to move the centerline so that binomial, and Poisson. While numerous other types are

70
 Use of quality control charts

?50I charts are used for monitoring the number and fraction of
£ 45 -• these, respectively, that have a particular characteristic of
- 4 0 - interest (e.g. a billing error, a late patient arrival, or a Cesarean
< 35 -r
c 30 -
section birth), where the probability of occurrence is reas-
2 25 - onably the same for each event. For example, Figure 3
1 20-- illustrates a p chart of the fraction of medication errors per
Q
15-: week (with non-constant control limits due to a varying
<5 10 -
number of medications per week). Note that these data reveal
H 1 1 11 -I 1 1 1 1 h—I 1- a gradually increasing trend in the error rate (criterion seven),
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 to the point where two subgroup values fall above the upper
control limit. Again, identifying and removing the root causes
of this unnatural variation is the first step in stabilizing and
100 improving this process.
90
80
Alternately, Poisson-based c and u charts often are ap-
propriate for count data for which no theoretical maximum

Downloaded from http://intqhc.oxfordjournals.org/ at Universidad Politécnica de Madrid on May 28, 2012

70
60 exists, such as the number of some particular type of oc-
50 currence per examined area, volume, or time period. Examples
40
30
can include the number of arrivals to an emergency room
20 per shift, telephone calls per hour, leukocytes per CBC, or
10 skin melanomas per patient. For example, Figure 4 illustrates
H—I 1 1 1 1 1 1 1 H -I 1 1 1 1 1- a u type of control chart for the number of patient falls per
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
month (with non-constant control limits now due to a varying
Subgroup Number
monthly census). Again, note the clear signals of atypical out-
Figure 2 A'and S control charts of laboratory beta-sub turn- of-control events (corresponding to the months 10/91,
around-time 1/93, and 8/93) in an otherwise stable and consistent process.
Further examples of each of the above types of control
charts can be found in the references cited. While these types
possible, usually one of the following can be reasonably are the most commonly used, they are based on certain
appropriate for describing many processes. assumptions that are usually but not always true. For example,
For continuous values that are normally distributed (i.e. g and h control charts [8] are appropriate in situations where
'bell-shaped'), both an X (pronounced 'x bar3) and S control the underlying count distribution is geometric, rather than
chart should be used together. These two charts monitor the Poisson. These charts also are useful in situations involving
process mean and standard deviation, respectively. Examples low rates by monitoring the number of dichotomous events
of this type of data might include recovery duration, times between outcomes, such as the number of surgeries between
to complete various activities, patient weights, lung capacity, postoperative infections [9]. As an example, Figure 5 illustrates
and intake and output. For example, Figure 2 illustrates X a g chart of die number of days between Clostridium difficile
and S control charts for the laboratory time to complete nosocomial infections. As can be seen via multiple violations
beta-sub pregnancy tests. The process is considered to be of the second and third criteria, for statistical control an
out of control if either chart exhibits out-of-control signals. increase in the infection rate appears to have occurred in the
In this case, note that both charts have out-of-control values vicinity of subgroup 34, with 18 of the next 22 subgroup
above the upper control limits. The root causes of these values — including a run of nine consecutive values — beneath
outcomes should be identified and removed in order to the centerline.
achieve a consistent and in-control process. Also note that
the chart later has a run of 10 consecutive values beneath
the centerline, signaling (via the second criterion in Table 1) Some final cautions
a now improved process.
For discrete data, the two most common pairs of control While ease of use certainly is valuable, the effects of some
charts are called np and p charts (for binomial distributions) common oversimplifications and misunderstandings about
and c and u charts (for Poisson distributions). Unlike the the use of control charts can be to decrease significandy
example given above, once the appropriate pair is identified either specificity or the ability to detect true process changes.
only one of the two charts should be used alone, rather than For example, periodic statistical errors include basing control
both together (due to statistical reasons beyond the current limits on the overall standard deviation of the data instead
scope of this article). The choice within a pair is largely of on the recommended formulas, not using three standard
preferential when all subgroups are the same size, whereas p deviation limits in almost all cases, and not accounting for
and u charts are preferred over np and c charts, respectively, significantly autocorrelated or multivariate processes. Other
when dealing with unequal subgroup sizes. Each of these pitfalls include not using at least 25—35 rational subgroups,
four charts is appropriate in the following situations. not taking sufficiendy large subgroups if process data are not
Given a certain number of dichotomous events, np and p reasonably bell-shaped, and failing to design charts with

71
Methodology Matters: J. C. Benneyan

o
c 1
'• t.
&
2
Q.
F*^—» / \ / \ w-\ j \ i \ / \ I—H^—J—it—"T"
i V i V i i i V i 'i i i '» i i V i i • i i i V i i V i i \ i i i '• 1 i I i i \ \
10 15 Z0 25 30
Week Number

Downloaded from http://intqhc.oxfordjournals.org/ at Universidad Politécnica de Madrid on May 28, 2012

Figure 3 p control chart of fraction of medication errors per week

M > M I I I I t I I I I t I I I I I M I I I I I I I 1 I > I t I I I I I
1/91 4/91 7/91 10/91 1/92 4/92 7/92 10/92 1/93 4/93 7/93 10/93 1/94 4/94
Month

Figure 4 » control chart of patient falls per month

6-

i u
•i i l
15 30 45 75

Occurrence of Infection

Figure 5 £ control chart of nosocomial infection rate

desirable statistical properties (i.e. sensitivity and specificity). (especially in cases for which any of the above charts are
Common errors in control chart selection include failing more appropriate), and using standard control charts when
to identify an appropriate chart, using an X chart alone combining data from non-homogenous processes. Related
without an J" chart, overuse of the so-called 'individuals' chart implementation concerns include reacting to natural variability

72

J
 Use of quality control charts

('process tampering'), using control charts primarily to 'hold 4. Montgomery D. C. Introduction to Statistical Quality Control, 3rd edn.
the gains', and over-reliance on software. Although not the New York: Wiley, 1997.
current focus, additional information on these topics, related 5. Benneyan J. C. Statistical quality control methods in infection
quality control methods, and more advanced concepts can control. Infect. Contr. Hosp. Epidemiol, (in press).
be found in some of the listed references.
6. Gitlow H., Gitlow S., Oppenheim A., Oppenheim R. Tools and
Methods for the Improvement ofQuality. Homewood, IL: Irwin, 1989.
7. Banks J. Principles of Quality Control. New York: Wiley, 1981.
References
8. Benneyan J. C , Kaminsky F. C. Modeling discrete data in SPC:
1. Shewhart W. A. The Economic Control of Quality of Manufactured the g and h control charts. Am. Soc. Qual. Contr. (Ann. Qual. Congr.
Product. New York: D. Van Nostand and Co., 1931. Trans.) 1994; 32-42.
9. BenneyanJ. C. Design of statistical^ control charts for nosocomial
2. BenneyanJ. C , Kaminsky F. C. Another view on how to measure infection and other alternatives. In International Applied Statistics
health care quality. Qual. Progress 1995; 28: 120-124. in Medicine Conference Proceedings (in press).

Downloaded from http://intqhc.oxfordjournals.org/ at Universidad Politécnica de Madrid on May 28, 2012

3. Deming W. E. Quality, Productivity, and Competitive Position. Cam-
bridge, MA: Massachusetts Institute of Technology Center for
Advanced Engineering Studies, 1982. Received in revised form 25 July 1997

73