READING  SUPPLEMENT  
 
 
STATISTICAL  PROCESS  CONTROL  
 
Statistical   process   control   (SPC),   the   use   of   statistical   techniques   to   measure   change   in   systems,   is   one   method   of  
monitoring  quality  in  pharmacy  practice.  It  is  particularly  useful  because  a  key  determinant  of  quality  in  products  and  
services  is  consistency.  Statistical  analysis  can  be  utilized  to  improve  quality  by  identifying  inconsistency  in  systems,  
and  SPC  tools,  such  as  histograms,  Pareto  charts,  scatter  diagrams,  and  run  and  control  charts,  can  help  distinguish  
between  acceptable  and  unacceptable  inconsistencies  in  pharmacist  services.  
 
THE  RATIONALE  FOR  SPC  
 
Repeated   measurements   of   the   same   process   within   a   system   will   have   variable   outcomes   over   time.   Two   types   of  
variation  exist,  common-­‐cause  variation  and  special-­‐cause  variation.  
 
Common-­‐cause  variation  is  always  present  within  a  system  and  consists  of  modest  changes  that  occur  randomly.  In  a  
pharmacy,  this  type  of  variation  can  be  a  result  of  things  such  as  differences  in  individual  pharmacists  and  technicians,  
patient   populations,   situations,   and   chance.   Despite   the   myriad   of   factors   that   can   influence   variations   in   pharmacy  
practice,  over  time,  a  predictable  pattern  of  variation  emerges.  As  long  as  the  system  stays  the  same  (e.g.,  personnel  
and  workflow  processes),  common-­‐cause  variation  stays  the  same.  A  system  is  considered  to  be  in  a  state  of  statistical  
control  when  only  common-­‐cause  variation  is  present.  
 
In  contrast,  special-­‐cause   variation  occurs  when  there  is  an  interruption  to  the  regular  process.  This  may  be  due  to  a  
deliberate  event,  such  as  implementation  of  a  new  policy,  or  an  unexpected  occurrence,  such  as  pharmacy  technician  
calling  in  sick.  This  type  of  variation  could  be  desirable  or  undesirable.  For  example,  newly  hired  technician  will  help  
decrease  prescription  wait  times;  this  variation  would  be  desirable  and  deliberate.  On  the  other  hand,  if  a  pharmacy  
technician  will  call  in  sick,  there  will  be  an  undesirable  increase  in  prescription  wait  times.  Special-­‐cause  variation  is  
not  predictable,  and  it  fluctuates  over  time.  When  this  type  of  variation  is  present,  the  system  is  considered  to  be  out  
of  the  state  of  statistical  control.  
 
SPC  TOOLS  
 
Histogram  
A   histogram   is   a   graph   that   displays   frequency   distributions   for   unique   categories   of   measure.     Histograms   are   useful  
for  determining  the  overall  shape  of  the  data,  data  distribution,  and  variation  in  data.    For  example,  the  histogram  in  
Figure  1  shows  the  range  and  frequency  of  times  that  it  takes  for  new  prescriptions  to  be  ready  for  waiting  patients  at  
a   community   pharmacy.   It   appears   that   the   majority   of   prescriptions   were   ready   between   16   and   30   minutes   after  
receiving  the  prescription.  This  information  can  be  used  to  determine  if  the  range  is  acceptable  for  prescriptions  to  be  
ready.  It  may  also  prompt  an  investigation  to  figure  out  why  some  prescriptions  took  so  long  to  get  ready.    
 
Pareto  Chart  
A   Pareto   chart,   a   type   of   histogram,   categorizes   data   according   to   the   most   frequent   issues   on   the   left   to   the   least  
frequent  issues  on  the  right.  The  Pareto  principle,  states  that  80%  of  your  quality  output  comes  from  20%  of  what  you  
do,   is   the   foundation   for   the   chart.   Pareto   charts   can   assist   in   deciding   which   issues   should   be   emphasized   for  
problem-­‐solving,   thereby   having   a   larger   impact   on   quality   improvement.   The   Pareto   chart   in   Figure   2   reveals   that  
75%   of   dispensing   problems   can   be   attributed   to   two   categories,   too   long   to   fill   prescription   and   did   not   offer   to  
counsel.  Focusing  on  improvement  in  these  two  areas  is  likely  to  have  a  large  impact  on  quality  improvement  in  the  
dispensing  process.    
 
 
 
 
 

  1  
/rpramos2014      
  
60  

Number  of  new  prescriptions  

40  

30  

20  

10  

0  
5  to  10  

11  to  15  

16  to  20  

21  to  25  

26  to  30  

31  to  35  

36  to  40  

41  to  45  

46  to  50  

51  to  55  

56  to  60  

60+  
Minutes  
 
 
Figure  1.  Histogram:  time  in  minutes  for  new  prescriptions  to  be  filled  in  a  community  pharmacy.  
 
 
 
 
 
100  

80  

60  

40  

20  

0  
too  long  to  ^ill   did  not  offer  to   problem  with   out  of  stock   dispensing  
counsel   label   error  
 
 
Figure  2.  Pareto  chart:  types  of  dispensing  problems  
 
 
Scatter  Diagram  
Scatter  diagrams  show  patterns  in  data  and  relationships  between  two  variables  for  a  unit  analysis  (e.g.,  a  patient).  A  
statistical  program  can  be  used  to  plot  the  matching  variables  on  a  chart  and  conduct  a  regression  analysis  of  the  data  
on  the  chart.  Analysis  of  data  allows  a  regression  line  to  be  drawn  that  shows  the  relationship  between  variables  in  a  
scatter  diagram.  The  scatter  diagram  in  Figure  3  illustrates  how  overall  patient  satisfaction  relates  to  patient  loyalty  
for  each  patient  responding  to  a  survey.  In  this  instance,  there  is  a  positive  relationship  between  satisfaction  and  
loyalty.  
  2  
/rpramos2014      
 4  

3.5  

Overall  satisfaction  
3  

2.5   y  =  1.0973x  +  0.0657  

2  

1.5  

1  

0.5  

0  
0   0.5   1   1.5   2   2.5   3   3.5   4  
Pharmacy  loyalty  
 
 
Figure  3.  Scatter  diagram:  the  relationship  between  pharmacy  loyalty  and  overall  patient  satisfaction.  
 
 
Run  Chart  
A  run  chart  can  identify  performance  patterns  and  trends  by  showing  how  a  variable  changes  over  time.  In  Figure  4,  
the   percentage   of   patients   receiving   counseling   prior   to   discharge   from   a   hospital   was   stable   for   the   first   5   months  
and   then   began   to   increase.   Run   charts   can   also   be   used   to   make   comparisons   among   trends   by   using   multiple  
variables.    
 
 
45  
40  
35  
30  
Percent  

25  
20  
15  
10  
5  
0  
1   2   3   4   5   6   7   8   9   10   11   12  
Month  
 
 
Figure  4.  Run  chart:  percentage  of  patients  receiving  medication  counseling  prior  to  discharge  from  a  hospital.  
 
Control  chart  
A  control  chart  is  a  type  of  run  chart.  It  is  a  key  tool  in  SPC  and  serves  as  a  visual  aid  to  distinguish  between  common-­‐  
and  special-­‐cause  variations.  A  control  chart  is  comprised  of  a  series  of  measurements  over  time  and  three  horizontal  
lines  which  represent  the  upper  control  limit  (UCL),  mean,  and  lower  control  limit  (LCL).  The  solid  line  is  the  target  
value   of   the   historical   process   which   represents   the   mean,   and   the   two   dotted   parallel   lines   indicate   the   limits   within  
which   practically   all   observed   results   should   fall   as   long   as   the   process   is   under   normal   variation   (statistically  
  3  
/rpramos2014      
controlled).   Variation   that   occurs   between   the   UCL   and   LCL   is   considered   to   be   acceptable   and   will   not   trigger   a  
review  of  system.  Only  the  events  outside  the  acceptable  control  limits  are  pursued.  The  control  limits  are  typically  
set   at   3   SD   around   the   mean,   which   says   that   99.97%   of   measures   will   fall   within   the   control   limits.   This   approach  
allows  only  meaningful  changes  in  the  process  to  be  captured  and  greater  confidence  that  a  measure  observed  outside  
of  the  control  limits  is  really  due  to  special-­‐cause  variation.  
 
Figures  5  and  6  are  examples  of  control  charts  of  prescription  fill  times  for  two  pharmacies  in  the  same  chain.  Figure  5  
depicts  the  average  time  it  takes  for  a  prescription  to  be  filled  at  ABC  Pharmacy.  During  the  12-­‐month  period  shown,  
the   times   appear   to   be   under   control   because   they   fluctuate   around   the   mean   between   the   UCL   and   LCL.   On   the   other  
hand,  Figure  6  depicts  the  average  time  it  takes  for  a  prescription  to  be  filled  at  XYZ  Pharmacy.  The  chart  indicates  
that   delivery   times   for   the   8th   and   11th   months   of   the   year   are   out   of   control,   meaning   that   they   fall   way   outside   of  
what   is   expected   by   normal   forms   of   variability.   This   may   signify   a   problem   with   the   prescription   fill   process   and  
trigger  an  investigation  to  determine  the  cause  of  the  variation.  
 

UCL  
(32.9)  

Mean  
(20.1)  

LCL  
(7.3)  

 
 
Figure  5.  Control  chart:  average  time  in  minutes  for  a  new  prescription  to  be  filled  at  ABC  Pharmacy  
 
 
 
 
  UCL  
  (32.9)  
 
 
 
  Mean  
  (20.1)  
 
 
 
LCL  
 
(7.3)  
 
 
 
 
 
 
Figure  6.  Control  chart:  average  time  in  minutes  for  a  new  prescription  to  be  filled  at  XYZ  Pharmacy  
 
 
  4  
/rpramos2014      
There  are  two  basic  quality  control  charts  which  are  based  on  the  measurability  of  the  quality  characteristics,  namely:  
 
1. Attribute   Chart   –   this   is   the   chart   which   makes   use   of   discrete   data   classifying   the   number   of   items  
conforming   and   the   number   of   items   failing   to   conform   to   any   specified   requirements.   An   example   of   an  
attribute  chart  is  the  control  chart  for  fraction  defective  known  as  p  chart.  
2. Variable  chart  –  this  is  a  chart  using  actual  records  of  numerical  measurement  on  a  full  continuous  scale  such  
as  meter,  grams,  liter.  Examples  of  variable  charts  are  the  𝑋  and  𝑅  charts.  
 
 
Exercises:  
 
1. 20  sets  of  10  tablets  were  weighed  during  compression  at  30  min  intervals.  Calculate  LCL  and  UCL  and  find  
out  if  the  process  is  statistically  controlled.  
 
2
𝑋−  𝑋
a. Use  the  formula:        𝑠 =    
𝑛−1
 
Mean  of  10  tablets  (mg)  
373.9   374.1   370.0   367.9   374.1   365.1   368.0   371.1   366.9   372.5  
381.5   365.1   365.8   370.0   369.7   370.6   372.8   365.2   370.3   370.4  
 
𝑅
b. Using  the  table  below,  calculate  the  standard  deviation  using  the  formula:      𝛿 =    
𝐷
where  R  is  the  range  and  D  is  the  divisor  for  the  sample  size  used.  
 
Divisor  for  estimating  SD  from  the  range  
Sample  size  (n)   Divisor  (D)   Sample  size  (n)   Divisor  (D)  
2   1.1   9   3.0  
3   1.7   10   3.1  
4   2.1   25   3.5  
5   2.3   20   3.7  
6   2.7   30   4.0  
8   2.8   50   4.5  
 
2. The  volume  of  5  vials  was  determined  during  the  filling  of  an  injectable.  Determine  the  UCL  and  LCL  using  the  
formulas  below:  
    𝐿𝐶𝐿! =   𝑋 −   𝐴! 𝑅     𝑈𝐶𝐿! =   𝑋 +   𝐴! 𝑅  
    𝐿𝐶𝐿! =   𝐷! 𝑅     𝐿𝐶𝐿! =   𝐷! 𝑅  
 
Factors  for  estimating  3δ  
Sample  Size   Factors  for  𝑅  chart   Factors    
(n)   D3   D4   for  𝑋  chart  (A2  )  
2   0.00   3.27   1.88  
3   0.00   2.57   1.02  
4   0.00   2.28   0.73  
5   0.00   2.11   0.58  
6   0.00   2.00   0.48  
7   0.08   1.92   0.42  
8   0.14   1.86   0.37  
9   0.18   1.82   0.34  
10   0.22   1.78   0.31  
 
 
  5  
/rpramos2014      
 Data  for  construction  of  Control  Chart  for  Average  and  Range  
No.  of   Vol  (mL)  of  5  vials  every  30  mins.  
𝑋   R  
Insp.   A   B   C   D   E  
1   10.5   10.3   10.2   10.6   10.7      
2   10.4   10.0   10.3   10.2   10.1      
3   10.1   9.8   9.9   9.6   9.7      
4   9.8   9.8   10.0   10.2   10.1      
5   10.6   10.5   10.5   10.5   10.4      
6   10.7   10.5   10.6   10.7   10.9      
7   10.7   10.3   10.5   10.4   10.4      
8   10.8   10.2   10.7   10.8   10.9      
9   10.0   10.3   10.4   10.1   10.0      
10   10.0   10.0   10.0   10.4   10.3      
11   9.8   9.9   9.8   9.8   9.9      
12   10.7   10.6   9.6   10.9   10.5      
13   10.1   10.6   10.4   10.7   10.3      
14   10.5   10.9   10.5   10.6   10.4      
15   10.3   10.3   10.2   10.5   10.3      
 
a. Solve  for  𝑋  and  𝑋 ,  R  and  𝑅 .  
b. Solve  for  LCL  and  UCL.  
 
3. A  batch  of  ointment  was  filled  into  tubes  during  10  working  days.  500  tubes  were  filled  each  day.  The  
inspector  withdrew  a  random  sample  based  on  the  master  table  and  noted  the  number  of  leaking  tubes  
below.  
 
Day   No.  of  leaking  tubes  
 1   4  
2   6  
3   7  
4   5  
5   3  
6   1  
7   6  
8   3  
9   2  
10   0  
 
 
𝑑
a. Compute  for  fraction  defective  (p):            𝑝 =    
𝑛
where  d  is  the  number  of  defectives  found  and  n  is  the  number  of  units  inspected  
 
!
b. Compute  for  average  fraction  defective  (𝑝):          𝑝 =    
!
 
c. Calculate  the  LCL  and  the  UCL  through  the  following  formulas:  
 
!(!!  !) !(!!  !)
𝐿𝐶𝐿 =   𝑝 − 3   !
      𝑈𝐶𝐿 =   𝑝 + 3   !
 
 
d. Is  the  crimping  process  statistically  controlled?  
 
 
  6  
/rpramos2014      
References:  
 
 
Lerma,   Norma   V.   and   Marina   O.   Osi.   Drug   and   Cosmetic   Quality   Control   with   Instrumentation,   2nd   ed.   Manila,  
Philippines:  UST  Publishing  House,  1996.  
 
Warholak,  Terri  L.,  and  David  P.  Nau,  eds.  Quality  and  Safety  in  Pharmacy  Practice.  United  States:  The  McGraw-­‐Hill  
Companies,  Inc.,  2010.  

  7  
/rpramos2014