Edit

  check   -­‐   a   critical   step   to   ensure   the   data  

quality  during  clinical  trials  
 
In   clinical   trial,   one   critical   task   is   to   ensure   that   the   data  
collected   or   data   entered   into   the   system   /   database   is   valid,  
correct,  and  logically  sound.    
This  task  requires  a  data  quality  plan  starting  from    
Designing  a  good  study  protocol  
Developing  efficient  case  report  forms  
Providing  clear  instructions  for  completing  case  report  
forms  (CRFs)  
Implementing  electronic  edit  checks    
Monitoring  the  study  data  /  source  data  verification    
Data  clarification  process    
Data  review  process.  (One  of  the  steps  is  to  implement  
the  electronic  edit  checks).    
Edit  check  is  a  program  instruction  or  subroutine  that  
tests  the  validity  of  input  in  a  data  entry  program.    
 
According   to   the   CDISC   clinical   research   glossary  from  
Applied  Clinical  Trials,  the  edit  check  is  defined  as:  
 
An   auditable   process,   usually   automated,   of   assessing   the  
content   of   a   data   field   against   its   expected   logical,   format,  
range,  or  other  properties  that  is  intended  to  reduce  error.    
 
NOTE:    
 
1. Time-­‐of-­‐entry  edit  checks  are  a  type  of  edit  check  that  is  
run   (executed)   at   the   time   data   are   first   captured   or  
transcribed   to   an   electronic   device   at   the   time   entry   is  
completed  of  each  field  or  group  of  fields  on  a  form.  
 
2. Back-­‐end  edit  checks  are  a  type  that  is  run  against  data  
that   has   been   entered   or   captured   electronically   and   has  
also  been  received  by  a  centralized  data  store.  
EDIT  CHECKS  –  19/05/16  
VINAY    
 
  
 
3. Electronic  edit  checks  allow  us  to  use  the  power  of  the  
computer   to   check   for   illogical,   incomplete   or  
inconsistent  data.    
 
In   clinical   trial,   one   of   the   most   important   tasks   facing  
clinical   data   management   personnel   is   to   produce   the  
electronic  Edit  Checks  specifications  for  a  study.  Developing  
the   electronic   edit   checks   specification   and   processing   the  
queries   that   result   from   them   -­‐-­‐   is   arguably   the   most   vital  
and  time-­‐consuming  data  cleaning  activity  data  management  
personnel   undertakes.   The   study   statistician   should   always  
participate   in   the   process   of   developing   the   electronic   edit  
checks   to   ensure   that   the   critical   edit   checks   are   included.  
Effectively   implementing   the   edit   check   can   prevent   the  
illogical,  incomplete,  or  inconsistent  data  from  entering  into  
the   data   capture   system   or   data   set,   which   will   make   the  
downstream  data  analyses  much  easier.    
 
There  are  two  types  of  edit  checks:    
 
Univariate   edit   checks   (include   range   checks):   these   are  
the   edit   checks   only   applicable   to   a   single   field   or   single  
variable.   For   example,   for   subject   weight,   we   can   set   up   an  
edit   check   to   ensure   that   the   extreme   or   unlikely   value   not  
to   be   entered.   Let’s   say   we   set   up   a   range   check   if   a   data  
entry   is   smaller   than   90   lb   or   greater   than   300   lb.   For   lung  
function   test,   we   may   set   up   an   edit   check   for   predicted  
FEV1   to   be   no   less   than   20%   because   it   is   unlikely   to   have  
someone   with   predicted   FEV1   <20%.   The   univariate   edit  
checks   are   usually   run   instantly   during   the   time   of   data  
entry.  
 
Multivariate   edit   checks   (also   called   aggregate   edit  
checks):  these  are  the  edit  checks  with  more  than  one  fields  
EDIT  CHECKS  –  19/05/16  
VINAY    
 
or   variables   involved.   These   edit   checks   cross   check   the  
entries   across   multiple   fields   /   variables   to   ensure   the   data  
is   logical   and   consistency.   For   example,   if   the   entry   on  
Gender  field  is  ‘Male’,  there  should  not  be  data  for  pregnancy  
test   result   field.   If   the   reason   for   subject   dropping   out   the  
study   is   entered   as   ‘adverse   events’,   there   should   be   a  
corresponding  entry  in  AE  data  set.  Statistician  can  provide  
great   inputs   in   identifying   the   multivariable   edit   checks.  
Some  multivariate  edit  checks  could  involve  the  complicated  
algorithm   and   take   considerable   time   to   run.   In   this  
situation,   the   multivariable   edit   checks   can   be   run   at   back-­‐
end  at  a  specified  interval  (for  example,  2  am  at  night).    
 
One  misunderstanding  is  to  think  that  all  data  issues  can  be  
resolved   by   implementing   the   edit   checks.   Edit   check   is   only  
one   of   the   steps   in   the   data   cleaning   process.   Also,   there  
should  be  balance  in  terms  of  the  number  of  edit  checks.  Too  
many   edit   checks   for   non-­‐critical   fields   could   be   very  
annoying   for   people   who   enter   the   data.   This   is   especially  
true   for   clinical   trials   using   electronic   data   capture   (EDC)  
where   the   data   entry   responsibility   is   delegated   to   the  
investigator  and  study  coordinators  who  may  lose  patient  if  
there  are  too  many  pop-­‐up  messages  during  the  data  entry.  
For   example,   if   the   telephone   number   needs   to   be   entered,  
an   edit   check   to   enforce   the   data   entry   to   follow   xxx-­‐xxx-­‐
xxxx   would   be   unnecessary   (xxxxxxxxxx   and   1xxxxxxxxxx  
should  also  be  accepted)  –  this  is  an  example  I  see  in  some  of  
the  web  forms  –  very  annoying).  
 
 

EDIT  CHECKS  –  19/05/16  

VINAY