Error Detection and Correction in Data Collection
Julia Challinor, RN, PhDAssistant Adjunct Professor of NursingUniversity of California, San Francisco
INCTR annual meeting10-12 December 2005
Chennai, India
Data Audit
Questions about omissions and errors NO “white-out” ink
Typographical mistake? Due to poor training of the data managers for this study? Is the mistake significant to the findings? Does this site have more than average number of omissions
and errors?
Data Manager
What if YOU make an error? Data entry
The wrong value was inserted by hand NO erasure NO block coverage
Problems
Corrected
Problems
Corrected
ProblemsProblems
Lab Problems
Lab Problems
Lab Problems
More labs than spaces What to do? ADD MORE CRF lab pages…
Data Entry Error
Put a single line through the value, write the correct value and date and initial the change
Notify your data center or appropriate person Correct database
Error Correction14 mg 14 mg jc 4/5/03
17 mg
Finding Errors
It is essential that data entry is routinely verified Double data entry
Expensive Time consuming
Checking case report forms chosen at random Two data managers check each other’s data entry The principal investigator does a routine random check A member of the research team does a routine random check
Reporting Errors
Who needs to know the error occurred? Depends on the error
Hierarchy for reporting errors should be described in the study PROTOCOL The principal investigator needs to be kept
informed A regularly scheduled review of data entry
History and Trail
Make a written notation of omissions and errors that have been corrected
Monitors will not expect perfection But will need to be able to trace the omission or error for
clarification if needed It is not the responsibility of the data manager to
determine the severity of an omission or error This is the responsibility of the principal investigator and
the sponsoring agency among others
Humans
Data managers are humans Humans are not machines Humans make errors
Errors
It is important that errors are noted and a monitor can follow a trail to clarify any questions
A group of case study forms that are perfect are more suspect than a group with some corrections
“Red Flags”
Items that alert you to a potential error Test result value is significantly larger or smaller
compared to the last test for the patient A dose level or test result value is significantly
different for this patient than all other patients on same protocol
Protocols
KNOW your protocols Read the protocol Ask questions if you do not understand any part of
the protocol Review the protocol if you have a question on a
specific patient’s data Data Managers usually see all the results for all
the patients in a center on the same protocol Individual physicians do not
Recommendations
Internet based training program for clinical studies NIH has an elementary training at
http://ohsr.od.nih.gov/ St Jude Children’s Research Hospital
Free training site in English and Spanish http://www.cure4kids.org “Educating Clinical Staff in Clinical Research Data
Collection & Data Management