A Statistical Process Control Case Study: Thomas K. Ross, PHD
A Statistical Process Control Case Study: Thomas K. Ross, PHD
A Statistical Process Control Case Study: Thomas K. Ross, PHD
O
Statistical process control (SPC) charts ne of Edward Deming’s quality princi-
can be applied to a wide number of health ples is to make quality everyone’s job.1
care applications, yet widespread use has not Health care organizations have always as-
occurred. The greatest obstacle preventing sumed that quality is everyone’s job and
wider use is the lack of quality management everyone knows the difference between quality care
and substandard care. Yet health care by relying on
training that health care workers receive. The
individualistic definitions of quality without perfor-
technical nature of the SPC guarantees that
mance evaluation systems has never made quality
without explicit instruction this technique
everyone’s job. Consequently, quality management
will not come into widespread use. Reviews is left to quality improvement director, department,
of health care quality management texts and/or committee, fails to achieve widespread sup-
inform the reader that SPC charts should be port in the organization, and produces few tangible
used to improve delivery processes and improvements in processes or outcomes.
outcomes often without discussing how they Implementing Deming’s principle requires em-
are created. Conversely, medical research ployees to understand what quality is, be capable of
frequently reports the improved outcomes identifying substandard performance, and have the
achieved after analyzing SPC charts. This authority to make changes that will improve perfor-
article is targeted between these 2 positions: mance. According to Deming, it is the responsibility
it reviews the SPC technique and presents a of management to suffuse this quality focus into their
organizations. Of course, this is the problem in health
tool and data so readers can construct SPC
care; other industries have developed rigorous sys-
charts. After tackling the case, it is hoped
tems to monitor quality and modify processes when
that the readers will collect their own data
output does not meet expectations. Berwick bluntly
and apply the same technique to improve addressed this problem; doctors, nurses, reception-
processes in their own organization. ists, and other health care workers are not trained in
or capable of changing how they work.2
Health care has been implementing quality im-
provement programs for decades but efforts to cross
into the fertile fields of quality management always
seem to flounder. The inability to apply the quality
management techniques of other industries is due to
the more complex nature of health care but part of
the problem is the way health care has approached
quality management. Berwick recognizes that health
221
222 QU A LITY MA N A G EM EN T IN HEA LTH CA RE /VO LU M E 15, ISSU E 4, OCTO BER –DECEM BER 2006
care workers have not been given the tools to sys- of quality management tools. Without understand-
tematically incorporate quality measurement in their ing quality management techniques and the ends to
work. which they can be applied, it is understandable that
The following case was developed to demonstrate health care providers would think these tools have
how statistical process control (SPC) can be applied little to offer them.
to health care processes. The SPC is a tool that de- Health care quality management texts inform the
fines what quality is, how performance is measured, reader that SPC charts should be used to improve
and when investigation and possibly correction must delivery processes and outcomes often without dis-
occur. The SPC, once understood, empowers the doc- cussing how they are created. Conversely, medical
tor, nurse, department manager, and other health research frequently reports the improved outcomes
care workers to enter into the quality management achieved after analyzing SPC charts. This article is
process on a more equal basis with those trained targeted between these 2 positions: it reviews the SPC
in quality management techniques. The initial audi- technique and presents a tool and an application so
ence for this article is health care managers, but as readers can learn how to construct SPC charts and
Deming and Berwick note, improvement occurs only offers suggestions as to when they can be used to im-
when all employees understand and embrace these prove health care delivery processes.
quality improvement techniques. Getting managers A case study is developed and followed to con-
and opinion leaders on board is only the first step clusion to demonstrate one potential application of
to institutionalizing quality practices throughout the the SPC to health care. The intent is to provide the
organization. readers with a basic understanding of how to build
This article does not assume that the reader is SPC charts and encourage them to access and analyze
versed in statistics or quality management tech- the case data using Excel to evaluate performance on
niques, it does however assume familiarity with their own. The case question is: are medicines de-
spreadsheet software. The goal is to demonstrate that livered in a timely manner to maximize medical ef-
quality improvement techniques can be used produc- fectiveness and patient comfort? The control charts
tively by all health care workers. At the end of this developed are used to draw conclusions regarding
case it is hoped that the reader will be able to identify the operation of the system, explore potential rea-
a use for the SPC in their area of responsibility, create sons for the performance observed, and speculate
SPC charts using Excel, and interpret these charts to on system changes that could improve the timeli-
improve outcomes. ness of medicine delivery. After working through the
case, it is hoped that the reader will be motivated
QUALITY IMPROVEMENT IN to apply the principles and techniques to their own
HEALTH CARE work.
One of the major difficulties in advancing health
One of the greatest obstacles to quality improve- care quality is the lack of specificity in defining health
ment in health care is the belief of providers that care processes, establishing performance standards,
quality management tools developed in other indus- and measuring compliance with standards after they
tries are not relevant to health care services. Not only are defined. Attempts to improve operations and out-
are the initiatives of other industries deemed irrele- comes are difficult if not impossible when standards
vant to health care, many providers hold that quality and measures are ill-defined or absent. Efforts to de-
improvement mechanisms developed in other health fine medical processes continue to be hotly debated,
care organizations are not applicable to the institu- and opponents argue that there is too much variation
tion they work in. This belief is the direct result of the in medical practice to establish one way of treating
second factor, and health care providers are not famil- patients and dictating to physicians will not improve
iar with the purpose, operation, and interpretation patient care and may hurt patients.3
A Statistical Process Control Case Study 223
On the other hand, countless research has been The SPC is a broader approach that continuously
done to define proper medical practice; much of this examines processes to identify undesirable trends in
work is amendable to and should be tracked by the performance and institute corrective action before
SPC to improve health care outcomes for patients. harm arises. Fasting and Gisvold used the SPC in
McGlynn et al found that patients received 54.9% of anesthesia to study a range of adverse events that are
recommended care, more important than their con- less severe and more frequent than sentinel events
clusion was their technical supplement that provided that contained the potential for harm. As a result
the indicators of what should occur and when it of their study, they were able to reduce anesthesia
should occur for 30 conditions.4 accidents.6 The SPC complements root cause analysis
Other industries recognize that poor performance and extends process improvement efforts by adding
is the result of variation, that is, deviations from how ongoing monitoring of a broad range of events to qual-
things should be done that adversely impact out- ity management. JCAHO (LD.5.2) notes that sentinel
comes. Controlling variation is the key to process event analysis is reactive and does not meet the intent
improvement. Health care providers are arguably of the JCAHO patient safety standard.7 The SPC is not
saddled with more sources of variation than pro- concerned with particular cases but rather with the
ducers of other goods and services but the issue re- ongoing function of systems and thus is a proactive
mains; is there too much variation in treatment and approach to operations that monitors performance to
will a more systematic approach to health care de- detect changes in system performance before prob-
livery processes improve outcomes? The SPC is a lems arise.
tool health care workers can use to determine when
variation is a routine part of patient care, neces- The 6 steps of SPC
sary and beneficial for patients and the health care The goal of the SPC as envisioned by Walter
system, and when variation is excessive and poten- Shewhart is to determine when a system is out of con-
tially harmful. Armed with this information, it is the trol and requires adjustment to improve its output.
duty of health care providers to reduce inappropriate An in-control system simply indicates that it is oper-
variation. ating close to its historical performance; this perfor-
mance, however, may fail to meet generally accepted
IMPROVING HEALTH CARE standards or customer expectations. A system that
is out of control indicates performance that is sig-
Root cause analysis or routine monitoring? nificantly different from historical performance. The
Current quality control efforts revolve around root goals of the SPC are to identify when performance
cause analysis, that is, a comprehensive examina- deviates sufficiently to endanger quality and improve
tion of an adverse event with the goal of reducing in-control performance to improve outcomes.
harm to patients by preventing the reoccurrence of Shewhart’s second goal was to devise a quality
the event. JCAHO-accredited health care organiza- monitoring and improvement system that could be
tions must conduct a root cause analysis focusing operated effectively by workers whose expertise lies
on systems and processes when a sentinel event oc- in areas other than statistics.8 Such a system required
curs. Root cause analysis is limited by its initiating clear signals to indicate when system performance
cause; it is undertaken after an adverse event has oc- fluctuates too much and be easy to use. Shewhart cre-
curred and examines only those processes that were ated his system in a manufacturing environment for
part of the care episode. In arguing for a broader ap- those trained in engineering and, as will be demon-
proach it has been noted “when multiple sources of strated, it is as applicable for service industries and
variation are present, isolated observations provide those trained in medical sciences. The SPC is de-
insufficient information on which to base objective signed for ease of use; the technique can be described
decision making.”5 and completed in 6 steps.
224 QU A LITY MA N A G EM EN T IN HEA LTH CA RE /VO LU M E 15, ISSU E 4, OCTO BER –DECEM BER 2006
The SPC can be used to monitor the behavior of tem is operating, the first statistic is the mean. The
any system that produces outputs that can be mea- mean is the measure of central tendency and re-
sured numerically. The SPC can monitor the percent- ports “average” performance. The mean is used to
age of defects in a sample (P charts), the number of de- determine whether performance meets the desired
fects in a sample (c charts), or output characteristics standard. The mean for a sample is calculated as
(average and variance, X̄ and R charts) among other follows: X̄ = xi /n, the sum of the sampled ob-
things. These applications require employees to rou- servations divided by the sample size. If Excel is
tinely collect simple performance measures to moni- used to record data, the mean is calculated as fol-
tor quality: percent defective, number of defects, the lows: =AVERAGE(range of observations). In medica-
average weight or length of a product, time required tion management the mean can be used to determine
to deliver a product or service, and the variance in whether patients received the specified dose—larger
these characteristics. The case uses X̄ and R charts or smaller dosage is a problem. Likewise, science and
since the concern is the interval between prescribed common sense dictate that the most effective inter-
medication time and the actual delivery of medicines ventions are delivered proximate to a medical event;
to patients. An earlier article in this journal provides the case seeks to determine whether medicine is de-
a concise explanation of when the different control livered to patients within (plus or minus) 2 hours of
charts should be used.9 the prescribed time.
The second statistic, range (R), provides a mea-
Step 1: Data collection sure of the variance in the sample. Range is the high-
Data collection is the most time consuming part of est value in the sample minus the lowest value and
the SPC process. A manager wanting to assess per- is used to assess the variability in performance. Ex-
formance using the SPC must determine the desired cel can be directed to scan a series of numbers and
sample size, how often samples are to be drawn, identify the maximum and minimum values and per-
procedures to ensure that the samples are random, form the subtraction by the command: =MAX(range
and who is responsible for data collection. Managers of observations)-MIN(range of observations).
must realize that they face a trade-off between the The need for both statistics is demonstrated by
cost of collecting data (sample size and sampling fre- assuming that the average turnaround time for a labo-
quency) and the accuracy of information obtained ratory test is 60 minutes. Those awaiting results pre-
from the collected data. Larger samples typically pro- fer a process in which 50% of tests are available
duce more accurate information but are more costly in 65 minutes and the remaining 50% in 55 min-
and time consuming to collect. The events sampled utes to a process where one half of results are avail-
must be randomly drawn to ensure that the sample able in 10 minutes and the other half take 110 min-
is not biased; that is, the information gained from utes. Average turnaround time, X̄ , for each process is
the sample must be representative of the larger pop- 60 minutes but the first process has a 10-minute range
ulation for it to accurately assess the performance of (65–55 minutes) while the second has a 100-minute
the nonsampled phenomena. The procedures and re- range (110–10 minutes). Lack of predictability in the
sponsibility for data collection should ensure that the second process is a problem. The person awaiting re-
data collector has no incentive to collect either favor- sults from the second process does not know whether
able or nonfavorable performance. After the data are he or she will receive them quickly or have to wait
collected, it should be recorded in a spreadsheet. almost 2 hours. Processes with the narrowest range
are generally superior; the process with a 10-minute
Step 2: Calculation of descriptive statistics range provides users with a clear idea of when re-
After the data are recorded, descriptive statistics sults will be available and allows them to schedule
must be calculated. Descriptive statistics provide the work accordingly. The variance in the second process
information necessary to understand how the sys- can lead to inefficiency if delays in the availability of
A Statistical Process Control Case Study 225
sample size (the number of observations in the sam- mean and range, and charting the values for new
ple) rather than by the number of samples drawn. samples.
The product of the control chart factor and the aver-
age range is added (or subtracted) from the historical Step 5: Interpreting performance
performance (the average of the averages) to deter- Step 5 is the examination of actual and expected
mine the upper (or lower) control limit. For example, performance to determine whether the process is in
when samples of 25 are drawn, the average output of control or out of control. An in-control or stable pro-
a process is expected to routinely fluctuate around cess is one where actual performance, X̄ or R, falls
its historical performance by ±15.3% of its average within the control limits with data points lying on
range (±0.153 ∗ R̄). Investigation is required when- either side of the centerline, without exhibiting a pat-
ever the average for a sample varies by more than tern. Figure 3 presents 6 configurations to look for
15.3% of the average range. when examining control charts. Chart 1 shows an in-
The upper and lower limits for the R chart also control process in which data points do not breach
require 2 formulas: the upper or lower control limits and are randomly
UCL :D2 ∗ R̄ distributed around the centerline. Chart 2 presents
the classic out-of-control process where multiple
LCL :D1 ∗ R̄
sample values (samples 2, 5, and 8) breach the es-
D1 and D2 are control chart factors and, similar to A, tablished control limits. Charts 3–6 show no samples
these factors produce tighter control limits as sam- breaching the control limits but all have recogniz-
ple size increases. When samples of 10 are drawn, able patterns that indicate instability in the process
a sample range could be up to 77.7% above or be- or a systemic change. Sample values fall only on one
low the average range (1.000 ± 0.541 ∗ R̄) before in- side of the centerline in chart 3, an upward trend be-
vestigation is warranted. Using a larger sample of 25, ginning in period 5 is evident on chart 4, a cyclical
acceptable variation is reduced to 54.1% above or be- pattern occurs on chart 5, and the lack of variance
low the average range (1.000 ± 0.541 ∗ R̄). It is only starting in period 6 on chart 6 suggests that none of
when a sample range exceeds the control chart limits these systems are operating as they have in the past.
that investigation and potential corrective action are Charts 2–6 suggest that the process has changed;
required. it is the manager’s job to identify if and why perfor-
mance has changed and the impact of this change
on patients. SPC charts may record positive or nega-
Step 4: Graphing performance tive changes to system performance. Once the change
Steps 2 and 3 provide all the necessary informa- and its impact are understood, managers need to in-
tion to graph actual performance, the average, and stitutionalize those changes that improve outcomes
the range between the highest and lowest values, or initiate corrective action for those that reduce the
against historical and expected performance. Sam- effectiveness of the health care process.
ple averages (or ranges) are graphed as XY charts
Step 6: Investigation
with the x-axis defining when the sample was col-
lected (reported in a chronological order) and the Steps 1 through 5 are necessary to identify when a
y-axis recording the value of the sample average or process should be reviewed, control charts clearly in-
range. The centerline (CL), X̄¯ or R̄, and the upper dicate when a limit is exceeded or the starting point
and lower control limits are also graphed to provide of a trend but they do not identify what has changed
the baselines against which actual performance is or the impact of the change on patients. Step 6 is the
measured. Routine monitoring of performance, after most challenging part of the process; does breach-
the control limits are established, requires the rela- ing a control limit or an identifiable pattern indi-
tively simple task of collecting data, calculating the cate that the process is out of control and requires
A Statistical Process Control Case Study 227
Figure 3. Control chart examples. UCL indicates upper control limit; CL, centerline; and LCL, lower control limit.
correction (or does it signal improved performance)? tative samples are drawn. A particular sample may
Correspondingly, in the absence of limit breaches or include a disproportionate number of high (or low)
trend, is a process performing at a high enough level? values and breach an upper (or lower) control limit,
The answers are not straightforward, and the nature indicating a process change when no change has oc-
of sampling ensures that occasionally nonrepresen- curred. The first task of an employee, after a control
228 QU A LITY MA N A G EM EN T IN HEA LTH CA RE /VO LU M E 15, ISSU E 4, OCTO BER –DECEM BER 2006
limit is exceeded, is to determine whether the system time or day of service?12 The next section applies the
is truly out of control or whether a nonrepresentative 6-step process to a set of hypothetical data to illus-
sample has been drawn. trate the SPC technique. The readers are encouraged
When a sample exceeds a control limit (Fig 3, to calculate the descriptive statistics and control lim-
chart 2), the first step in the investigation is to in- its and create the control charts for themselves.
crease the sample size to determine whether the out-
of-range result holds as more observations are exam- A MEDICATION MANAGEMENT CASE
ined. If the expanded sample produces a mean or
range that falls within the control limits, the man- Effective and high-quality medical care requires
ager can assume that the process is in control, re- that medicines be administered on a timely basis,
sume monitoring, and avoid tinkering with a stable and the performance standard expected in this case
system. One of the primary goals of the SPC is to fo- is that medicines should be administered within
cus employee effort on areas that require correction 2 hours of the prescribed time. Rather than exam-
by eliminating processes that do not need attention. ining every dose delivered, the SPC allows the use of
If the out-of-range results persist after the sample a sample of a small number of drug administrations
size is increased, the employee can assume that non- to determine whether the system is in control, is it
representative sampling did not produce the control meeting the 2-hour window, or is it out of compli-
limit breach and the harder task of determining why ance? A 100% sample may not be particularly infor-
performance has changed arises. Breaches of con- mative as it is unlikely if 100% of drugs are delivered
trol limits are designed to be rarely occurring events within 2 hours since natural variation is at work, that
so employees will not spend significant amounts of is, patients have the right to refuse medication and do
time or efforts investigating trivial variation in per- so, the patient may be receiving other treatments and
formance. The SPC can maximize the effectiveness be unavailable for medication, etc. A 100% sample
of improvement efforts by concentrating employee would also be arduous, if not impossible, to collect
efforts toward assignable variation and controllable given that a 500-bed hospital may dispense 160,000
causes and away from stable processes. medications in a month. The 6-step SPC technique
Given the rarity of breaches of limits or patterns described above is followed to analyze medication
demonstrated on charts 3–6 in Figure 3 that indi- management.
cate processes are not functioning as they have in
the past, the manager’s and employees’ job is to pin- Step 1: Collect data
point the cause(s) of change and enact corrective ac- A random sample of 50 medications was drawn
tion if the change has impaired outcomes. The SPC for the day, evening, and night shifts every day for
enhances employees’ ability to pinpoint causes by a month; a total of 4650 observations (50 medica-
providing an early detection system to identify per- tions × 3 shifts per day × 31 days) and recorded in
formance that is inconsistent with historical perfor- a spreadsheet. The sample size of 50 was arbitrarily
mance. Prompt identification of change may enhance determined; it is hoped that a sample size of this mag-
employees’ ability to pinpoint the cause of change nitude would persuade skeptical employees that the
while the causes of change are still fresh in their data were valid. As the validity of the SPC process
mind. is demonstrated, the sample size could and should
The uses to which the SPC can be applied in health be reduced. Technical note: with a sample size of
care are numerous: are generally accepted standards 50, an X̄ and SD (standard deviation) chart is recom-
of care being followed, is health care provided dif- mended but this case will use the more understand-
ferently to different populations, are waiting times able R chart.9 A copy of this data can be obtained at
appropriate, and does performance vary with the per- http://personal.ecu.edu/rossth/QMHCv15i4.xls and
sonnel delivering care, the location of service, or the pasted into Excel.
A Statistical Process Control Case Study 229
Step 2: Calculate descriptive statistics investigating the delivery processes on these shifts?
The SPC will answer these questions; at this point
The first sample (Day 1, Day Shift, Monday) pro- the high mean and wide range suggest that desired
duced an average delivery time of 111.38 minutes, performance is not being achieved.
=AVERAGE(B5:B54). The difference between actual
and prescribed administration time was recorded as Step 3: Calculate control limits
absolute values so medicines administered prior to √
and after the prescribed time are both recorded as X̄ UCL : 115 + ((0.75 × 1/ 50) × 65) = 121.9 minutes
positive values. Once the Excel formula is entered, Centerline (the mean) = 115.0 minutes
√
it can be copied to the remaining columns (through X̄ LCL : 115 − ((0.75 × 1/ 50) × 65) = 108.1 minutes
column CP) to calculate average administration time R UCL : (1.55 − (0.0015 × 50)) × 65 = 95.9 minutes
for each shift each day. The performance of the first Centerline(therange) = 65.0 minutes
sample can be contrasted against the most timely av-
R LCL : (0.45 + (0.001 × 50)) × 65 = 32.5 minutes
erage delivery time of 100.12 minutes (#52), and the
least timely, 134.44 minutes (#71). The average time Enter the X̄ control limits and centerline directly be-
between the prescribed medication time and the ac- low the calculation of the average medication time in
tual administration of medicines for all 93 samples the spreadsheet and copy across all columns. Simi-
(3 shifts per day × 31 days), X̄¯ , is 115 minutes. The larly, the R chart limits and centerline should be en-
Excel formula is =AVERAGE(B56:CP56). The center- tered and copied below the range for each sample.
line is thus established at 115 minutes on the basis The control limits indicate whether the medication
of historical performance, as stated earlier an indus- process is stable and subject to only natural varia-
try average or patient expectation, if known, could be tion, and average medication time for a sample of 50
used to establish expected performance. should fluctuate between 108 and 122 minutes. Sim-
Sample #1 has a range of 98 minutes between ilarly, the range between the most and least on-time
the most on-time delivery of medicine and the least delivery should vary from 33 to 96 minutes. If these
timely delivery, =MAX(B5:B54)-MIN(B5:B54). This thresholds are exceeded, the SPC indicates an unsta-
formula must again be copied across to column CP ble process or a potential change in performance that
to calculate the range for each shift each day. Sample requires investigation. Breaches of the upper limit in
#1’s range of 98 minutes can be contrasted with the this case indicate deterioration in performance (less
low of 33 minutes (#41) and the high of 100 (#10). The timely administration of medicines) while down-
average time between the most on-time and the least ward breaches may indicate positive changes in the
on-time administration for the 93 samples is 65 min- process and improved performance.
utes. The average and the range indicate that average
medication administration time is 115 ± 65 minutes Step 4: Graph actual and expected performance
or actual administration of medicine ranges from 82.5 Once the averages and ranges for each sample have
to 147.5 minutes (115 ± 65/2) before or after the pre- been calculated and the upper and lower control
scribed time. limits and centerlines are entered, Excel can create
A cursory review of performance, based on the de- control charts through the INSERT function. After
scriptive statistics, provides a manager with a good INSERT is selected, the user selects CHART and LINE
idea of where she or he should devote her or his at- (type of chart) and enters the desired data range for
tention. For example, the highest mean delivery time the X̄ chart, the data range entered must include the
occurs on the Monday through Friday evening shifts mean for an X̄ chart (or the range for the R chart), the
and the greatest variance occurs on the day shifts dur- upper and lower control limits, and the centerline.
ing the week. Is this performance acceptable? Should In Figure 4, the x-axis reports the sample number,
the manager devote his or her time and energy to 1 though 93.
230 QU A LITY MA N A G EM EN T IN HEA LTH CA RE /VO LU M E 15, ISSU E 4, OCTO BER –DECEM BER 2006
differences provide valuable information to under- month, 1 though 31, rather than the sample number
stand system performance; which shifts provide the since only 1 shift per day is graphed.
most on-time delivery of medicines, are there differ-
ences in performance related to the day of the week, Day shift
and why is end-of-month performance different from Medications are routinely delivered in less than
that in the rest of the month? 115 minutes on the day shift, and there are only 5
(out of 31) samples in which the medication time
Step 6: Investigate when indicated and fix as
is above the centerline, that is, the historical aver-
appropriate
age. Performance according to the X̄ chart meets the
Control charts do not judge performance, in this established standard yet the R chart raises concern.
case, both charts indicate that investigation is re- Although average administration time is within the
quired and suggest that the medication management established control limits, the variability in perfor-
system is out of control. The UCL on the X̄ chart mance is troubling. The upper limit is 95.9 minutes;
shows that medicines are routinely not administered 12 samples breached this limit, suggesting a unifor-
within the desired 2-hour window. Having received mity problem. The large range indicates that indi-
this signal, it is the responsibility of managers and vidual patients routinely receive their medications
employees to determine whether the system requires beyond the desired 2-hour window, thus violating
fixing. The first question that should be asked is: the established standard. The greatest variation oc-
what factor(s) prevents the timely administration of curs Monday through Friday while performance on
medicines? This is an open-ended question but the Saturday and Sunday is more uniform (the 2 data
SPC can make identifying the cause(s) easier by exam- points clustering around 40 on the 6th, 7th, 13th,
ining performance on different days or shifts. Diag- 14th . . . days of the month). What factors are differ-
nosing and improving a system are easier when sub- ent between weekdays and weekends, which could
standard performance can be isolated to a particular account for this difference in performance?
shift, day, or unit; that is, can differences be detected This finding demonstrates the need for both con-
between shifts, days, or units with high performance trol charts; the X̄ chart may show performance within
and those with low performance? control limits but performance may be unacceptable
if the range is large. Assuming that one half of pa-
DIAGNOSING THE PROBLEM tients received their medication within 1 hour and
the other half in 3 hours of the prescribed time, the X̄
Given that medication times are failing to meet the chart would report an acceptable average of 2 hours
desired standard, what is wrong? Are there unique but the average of 2 hours with a 2-hour range clearly
factors occurring on different shifts or days that pre- indicates that medications are not delivered within
vent prompt administration of medicines? At this the 2-hour goal.
point the manager could decide to draw a larger sam-
Evening shift
ple to see whether the results persist, but we will
assume that they are valid and proceed to diagnosing Analysis of the X̄ chart for the evening shift shows
the problem. that the Monday through Friday shifts routinely pro-
Sorting the data allows the performance of indi- duce samples above the UCL. Obviously, this should
vidual shifts or days to be graphed against the estab- be the chief concern of employees, and graphing
lished control limits. If the data are sorted by shifts the performance of the second shift against histori-
(use the Excel SORT function: select DATA, SORT, cal performance reveals no samples falling under the
OPTIONS, ORIENTATION, LEFT TO RIGHT, row 4, mean performance time of 115 minutes and 23 sam-
day, evening, and night shifts), the charts in Figure 5 ples breaking the UCL. The R chart reinforces the
can be created. The x-axis now reports the day of the concern as it shows a difference between the most
232 QU A LITY MA N A G EM EN T IN HEA LTH CA RE /VO LU M E 15, ISSU E 4, OCTO BER –DECEM BER 2006
Figure 5. Performance by shift. UCL indicates upper control limit; LCL, lower control limit.
A Statistical Process Control Case Study 233
on-time and least timely delivery of medicines is ity of the evening weekday shift to deliver medicines
80 minutes. Twenty observations are above the cen- within 2 hours of their prescribed time and the vary-
terline and these are in consecutive runs of 5, again ing performance of the day shift. There appears to be
emphasizing that employees working the evening at least 1 factor that accounts for the lack of on-time
shift Monday through Friday are less likely to de- administration of the evening shift and wide range in
liver medications at the prescribed time and are less delivery times on the day shift during the week, given
uniform in the delivery of medication than employ- the performance demonstrated on Saturday and
ees working on the weekend. As asked in the anal- Sunday. The difference between weekday and week-
ysis of the day shift, what factors are at work that end performance may be due to patient volume,
account for the different performance of the evening staffing, or the assignment of duties.
shift between weekdays and the weekend? After iso-
lating evening shift performance, it is clear that per- Comparing performance across shifts
formance on this shift is failing to achieve the stan- The discussion above was based on analyzing per-
dards set and further investigation is needed. formance on a shift; analyzing performance across
shifts indicates that there are 1 or more factors that
Night shift
explain the different performance of the day, evening,
The night shift has the promptest medication times and night shifts. Average medication time clusters
with the majority of samples falling below the LCL around 110 minutes on the day shift, 125 minutes
and only 3 samples produce an average above the on the evening shift, and 105 minutes on the night
centerline. Breaking the lower limit may indicate su- shift. The night shift provides medicines 20 minutes
perior performance but it could also reflect a bro- closer to their prescribed times than the evening shift
ken reporting system—data may not be recorded or and 5 minutes closer than the day shift. Similarly,
recorded properly.13 The 3 samples above the center- there are pronounced differences in the range be-
line were also above the UCL and occurred on the tween the shortest and longest administration times,
last 3 days of the month. This may be the result of a 100 minutes on days, 80 minutes on evenings, and
system change that requires rectification or perhaps 60 minutes on nights. The night shift demonstrates
it was due to an unplanned absence and performance consistently superior performance as in more timely
will return to its previous level with the return of the delivery of medicines and more uniform delivery
absent employee. The R chart shows very consistent times than the other 2 shifts. Obviously the reasons
performance on the night shift, its range falls below for the differences in performance across shifts must
the centerline indicating superior performance com- be understood, whether it is wholly or partially due
pared with the day and evening shifts. Once again, to the distribution of nursing and pharmacy duties
the pattern reflects strings of 5 and 2 showing that the between shifts, to improve performance.
weekend shifts are more consistent in their delivery The stratification of the control charts by shift
times than those of the Monday through Friday shifts. makes it apparent that performance improvement ef-
It is interesting to note that the end-of-month change forts should begin on the evening shift. Managers and
that indicated a movement away from the prescribed employees should embrace process improvement as
time reduced the variation in performance. an ongoing process (ie, continuous quality improve-
While on-time delivery of medication is the goal of ment [CQI]); hence, over time the goal should be ad-
the organization, the lack of consistency between the ministration of medicines at or at least closer to the
shifts indicates process instability. Unstable in this prescribed across all shifts and days but the first step
case means an inability to predict outcomes; the orga- begins by analyzing the process on the shift with the
nization cannot predict when medications will be de- widest gap between desired and actual performance.
livered given the differing performance across shifts. The CQI is an ongoing process of striving for bet-
The manager should explore the reason(s) for inabil- ter performance; once an organization improves its
234 QU A LITY MA N A G EM EN T IN HEA LTH CA RE /VO LU M E 15, ISSU E 4, OCTO BER –DECEM BER 2006
performance and SPC indicates these improvements After the major causes are identified, employees
are stable, it should attempt to reduce its control should explore the issues within each major cause
limits and institute tighter control over processes to to identify if, why, and how it impacts the deliv-
produce even better patient outcomes. An institu- ery of medicines. For example, examining staffing
tion that successfully implements the SPC can ex- as a major potential cause affecting the timeliness
pect better patient outcomes, higher patient and em- of medication may lead the QI team to investigate
ployee satisfaction that accompany better outcomes, staffing levels, employee qualifications and training,
and more effective use of resources. productivity of personnel, or any numbers of staffing
issues. Examining caseload may lead the team to in-
vestigate whether the number of patients and/or the
Identifying the major causes of a problem intensity of care required affects the timeliness of
The first task in improving outcomes is to iden- drug delivery. Similarly, questions of process design
tify an area for improvement; in this case, the SPC may explore job assignments and scheduling (ad-
was used to recognize the untimely delivery of med- missions, surgeries, ancillary tests, discharges, and
ications on the evening shift on weekdays. The next housekeeping duties), while pharmacy issues may
task is to explore and ultimately identify the poten- include the delivery of medications from the phar-
tial causes of substandard performance. Why is per- macy, medication errors (dosage and/or type), illeg-
formance failing to meet expectations? Fishbone or ibility of orders, and adverse drug reactions and/or
cause and effect charts are often employed to explore contraindications.
the potential causes of performance problems. Fish- In this case the reasons for off-schedule medication
bone charts begin the exploration process by identify- may arise from too few employees and too many pa-
ing the major reasons why unacceptable performance tients, communication problems, poor oversight, or
could occur and then exploring each reason to iden- any other factor suggested on the fishbone diagram.
tify the specific organization practices that could con- Any or all of these issues (plus others not identified)
tribute to the problem. Off-schedule medication may could impact the timeliness of the medication man-
be the result of 4 or more major causes, for brevity 4 agement process.
causes; staffing, caseload, process design, and phar- The role of manager, quality or process improve-
macy, are identified in Figure 6. ment director, and other members of the health
A Statistical Process Control Case Study 235
care delivery process is to evaluate and eliminate The SPC compliments and extends current health
the causes identified on the fishbone diagram until care efforts to improve health care processes and out-
the most probable factor(s) is identified and correc- comes. Analysis of the case demonstrates that the SPC
tive action taken. The identification process may in- can provide a wealth of information to understand
volve reaching consensus among the involved parties how current processes are performing and a basis to
or running investigations and tests. institute improvement. More important than the in-
After the most likely cause is identified and correc- formation generated however is how instilling this
tive action taken, the team should determine whether way of thinking into employees will change how they
the action produced the intended effect, and were approach their work. The discussion shows that there
medications delivered more timely? If timeliness can be multiple causes for substandard performance,
does not improve, then further review is required. these causes are not revealed by the SPC but rather
If corrective action is successful in improving perfor- provide employees with a starting point to apply
mance, outcomes must continue to be monitored to their analytical and problem-solving skills. Employ-
ensure that the improvement is not lost over time and ees will determine how successful the quality im-
as a baseline for further improvement thus initiating provement process is and they must see themselves
a CQI cycle. as stakeholders in the process. We know that health
The SPC improves management by setting clear care workers are committed to improving the health
performance standards for employees, establishing of their patients; the SPC is simply a tool to assist
a consistent evaluation standard for managers and them in these efforts by quantifying performance and
employees to use, and providing a tool to moni- signaling when a process has changed sufficiently to
tor processes when managers are absent; that is, the impact outcomes.
head nurse has the ability to monitor evening and The SPC is a tool specifically designed for those
night processes in spite of the fact that she or he not trained in management science or statistics to
may generally work in the day shift (or weekend improve the quality of their work and remains un-
performance when she or he works Monday through derutilized in the health care field. Quality improve-
Friday). ment in health care is not going to be the result of
a top-down approach but will become only an inte-
CONCLUSION gral part of the health care delivery system when all
employees embrace quality improvement tools. The
Improving patient care is a formidable task, which first step is to convince health care workers that qual-
should not be hampered by a lack of or misunder- ity improvement tools are relevant to their work and
standing of quality management techniques. This ar- that they can measure performance themselves. It is
ticle reviewed the SPC technique, demonstrated how only when employees begin to measure and analyze
a widely available spreadsheet package can be used their performance that we can expect to see ongoing
to record and analyze data, and analyzed an SPC case. and widespread improvement in health care delivery
Readers were encouraged to access the case data, per- processes and outcomes.
form their own calculations, create their own graphs
to complete the case, and consider how a comparable
process could be used to monitor and improve care REFERENCES
in their organization. This case demonstrates that the 1. Lighter DE, Fair DC. Quality Management in Health Care. 2nd
SPC can be used anytime timeliness is a factor af- ed. Sudbury, Mass: Jones and Bartlett Publishers; 2004.
fecting medical outcomes and/or patient satisfaction 2. Galvin R. Deficiency of will and ambition: a conversation with
Donald Berwick. Health Affairs. 2005;24(Web Exclusive):W5-
such as the timeliness of discharge, procedure time,
1–W5-9.
test turnaround time, registration time, waiting time, 3. Timmermans S, Maulk A. The promises and pitfalls of evidence
and etcetera. based medicine; non-adherence to practice guidelines remains
236 QU A LITY MA N A G EM EN T IN HEA LTH CA RE /VO LU M E 15, ISSU E 4, OCTO BER –DECEM BER 2006
the major barrier to the successful practice of evidence-based 8. Humble C. Caveats regarding the use of control charts. Infect
medicine. Health Aff. 2005;24:18–28. Control Hosp Epidemiol. 1998;19(11):865–868.
4. McGlynn EA, Asch SM, Adams J, et al. The quality of health 9. Amin S. Control charts 101: a guide to health care applications.
care delivered to adults in the United States. N Engl J Med. Qual Manag Health Care. 2001;9(3):1–27.
2003;348(26):2635–2645. 10. Shewhart W. Economic Control of Quality of Manufactured
5. Laffel G, Blumenthal D. The case for using industrial qual- Product. New York, NY: D Van Nostrand; 1931.
ity management science in health care organizations. JAMA. 11. Gaither N, Frazier G. Operations Management. Mason, Ohio:
1989;262(20):2869–2873. Southwestern Thomson Learning; 2002.
6. Fasting S, Gisvold S. Statistical process control methods allow 12. Benneyan J. Statistical quality control methods in infection
the analysis and improvement of anesthesia care. Can J Anesth. control and hospital epidemiology, part 1: introduction and
2003;50:767–774. basic theory. Infect Control Hosp Epidemiol. 1998;19(3):194–
7. 2001 JCAHO Standards in Support of Patient Safety and Med- 214.
ical/Health Care Error Reduction. Available at: http://www. 13. Finison LJ, Finison KS, Bliersbach CM. The use of control charts
dcha.org/JCAHORevision.htm. Accessed April 10, 2006. to improve healthcare quality. J Health Qual. 1993;15(1):9–23.