Discrepany Management_Katalyst HLS

Discrepancy Management
Presenter: Priya Gehlawat

Icons Used
Questions
Demonstration
Hands on
Exercise
Coding
Standards
A Welcome
Break
Tools
2
ReferenceTest Your
Understanding
Contacts

Introduction:
The ultimate goal of clinical data management (CDM) is to complete every study
with a dataset that accurately represents data captured in the study.
Discrepancy Management is the process of identifying and managing potential
problems with data collected during a study.
OVERVIEW
3

After completing this chapter you will be able to understand:
 Definition of Discrepancy
 Sources of Discrepancy
 Discrepancy Examples
 Goal of Discrepancy Management
 Types of Discrepancy
– System generated Discrepancies
– Electronically generated Discrepancies
– Manual Discrepancies
 Activities involved in CDM
 Discrepancy Management Process
– Flowchart
– Description – Paper & EDC study
 Good Practices
OBJECTIVES
4

• In an EDC study the electronic discrepancies generation can be set to an
online or an offline mode
• The turn around time to action a discrepancy from the time of generation is 2-
3 days
• A single discrepant data or open query can lead to database unlock and
can change the fate of the product
DO YOU KNOW
5

Discrepancies are ‘Inconsistencies’ found in clinical trial data which need to be
corrected as per study protocol.
Discrepancy is defined as any data point that may be:
 Inaccurate
 Illogical
 Incomplete
 Missing
 In violation of protocol-specific rules and conventions
DEFINITION OF DISCREPANCY
6

• A Data Entry error e.g. Mandatory fields missing
• An Investigator incorrectly reporting a value e.g. Lab values
• An incorrect attribute on a question e.g. Range (an Upper Bound that is too low)
SOURCES OF DISCREPANCIES
7
Discrepancy

SOURCES OF
DISCREPANCY…CONTD
8
At the Site
 Site personnel trial conduct error
 Site personnel transcription error
 Site equipment error
 Human error in reading equipment or
print out
 Inadequate instructions given to the
subject
 Subject does not follow trial conduct
instructions
 Subject completes questionnaire
incorrectly or provides incorrect or
incomplete answers to questions
 Data captured incorrectly on the source
 Data entry errors
 Fraud
At the Data Center
 Data entry errors
 Electronic Data Acquisition errors
 Incorrect database updates based on
data clarification form or query
 Programming error in user interface or
database or data manipulations

• Empty fields
• Unrelated items
• Incorrect Range
• Discrete value checks
• One value greater/less than/equal to another
• Dates not in logical sequence
• Inconsistent header information
• Any missing visits or pages
• Visits not in compliance with protocol
• Inclusion/exclusion criteria not met
• Lab values not in normal ranges
DISCREPANCIES - EXAMPLES
9

• Discrepancy Management also known as ‘Query Management’ is the process of
identifying and managing potential problems with data collected during a study
• It is a process of cleaning subject data in the Clinical Data Management System (CDMS)
that includes manual checks and programmed procedures (Edit checks)
• Discrepancies that cannot be corrected in-house and require clarification from the site
are queried by raising Data Clarification Form (DCF) / Query Request Form
• Resolution of discrepancies may involve:
—Updating data resulting from data entry errors
—Updating data issues according to Study Assumptions example SEC (Self –Evident
Correction), Client specifications etc
— Sending Data issues to the Investigator for clarification or correction
DISCREPANCY MANAGEMENT
10

Goal of Discrepancy Management:
The ultimate goal of clinical data management (CDM) is
to complete every study with a dataset that accurately
represents data captured in the study.
GOAL OF DISCREPANCY
MANAGEMENT
11

CLASSIFICATION OF
DISCREPANCIES
12
Manual review of data
and CRFs by clinical or
data management
Edit checks
programmed to
check discrepancies
of data by the data
management system or
entry application
Computerized checks
or analysis by data
management or
biostatistics using
external systems
Discrepancies can be broadly classified as:
1. Edit check -System/Computerized checks in the Clinical data management
system.
2. Manual Checks – Manually generated Checks.
3. External System Checks - Checks created externally in systems like SAS and
other tools.

In EDC discrepancy can be classified as :
 Discrepancy identified after manual review
 Discrepancy identified by System
Discrepancies identified after manual review –
 Identified during manual review of the data example – Reconciliation of Lab data,
Safety data etc
 It cannot be programmed
 Example of Manual discrepancy –
Diabetes is recorded in Medical History however medication name for diabetes have
not been recorded on the Conmed (Concomitant Medication) module
Discrepancies identified by System –
 Discrepancies identified by the system with the help of Edit checks
 These are programmed
 Example – Pregnancy have been marked “YES” however the SEX is recorded as
“MALE”
DISCREPANCY CLASSIFICATION
BASED ON ITS ORIGIN
13

System generated discrepancies are those that are generated when a response entered in
the database does not meet the attributes/ Validation Procedures specified for a given field
during study build –
• These are discrepancies that relate to –
– single data point – Univariate discrepancy.
– multiple data-point – Multivariate discrepancy.
They are automatically created during Data Entry
• They are caused due to the following errors on the CRF:
—Missing DataValues
—Incomplete Data
—Out of Range Data
—Wrong Data-type
—Discrete Value Groups assignment
SYSTEM GENERATED
DISCREPANCIES
14

• Example:
SYSTEM GENERATED
DISCREPANCIES : EXAMPLE
15
Gender M F Height cm
A Response missing for
Gender or entered other than
M or F
A Response for Height in
inches when expected in cm

• Discrepancies that are generated by Validation Procedures which are programmed by an
Edit Check programmer.
• These are generated either after a Batch Validation session , when the procedure is
manually executed or auto generated post data entry.
• These discrepancies relate to a multiple data-point/questions/modules on the CRF.
• Examples -
ELECTRONICALLY/SYSTEM
GENERATED DISCREPANCIES
16
Range
Checks
• Designed to identify statistical outliers eg - Hb value
should be between 12 – 14mm of Hg
Consistency
checks
• Designed to identify potential data errors by checking
sequential order of dates, corresponding events eg -AE
stop date/time is before AE start date/time

• They are caused due to the following errors on the CRF-Inconsistent data
• If “Normal” is ticked and response is provided for “If Abnormal specify”
ELECTRONICALLY GENERATED
DISCREPANCIES : EXAMPLE
17

• Data managers may manually review CRF data looking for odd patterns or
inconsistencies that have not been programmed into the edit checks.
• These discrepancies may relate to one or more data-points/questions
• They are created during Data Entry (Comments) and Data Review (Add Manual)
• Based on manual review of data ,Queries or DCFs are raised in the clinical
database or sent to the site and tracked for resolution.
MANUAL DISCREPANCIES
18

• They are caused due to the following errors on the CRF: − Illegible Data
MANUAL DISCREPANCIES :
EXAMPLE
19
If response is provided in free-form that is illegible

IDENTIFYING DATA DISCREPANCIES:
TYPES
20
Range
Checks
• Designed to identify
statistical outliers
Consistency
Checks
• Designed to identify potential data errors by checking
sequential order of dates, corresponding events
Manual
Checks
• Specified when a degree of manual review needs to take place in order to determine
if a query needs to be sent to the site
Discrepant
Data Listings
• Specified when a check is not programmatically feasible due to either the structure of
the database (comparing items in tables), the need to check external data (such as lab
data) against CRF data or the check involves the review or interpretation of free text

IDENTIFYING DATA DISCREPANCIES :
EXAMPLE
21
Range
Checks
• Hb value should be
between 12 – 14mm of
Hg
Consistency
Checks
• AE stop date/time is before AE start date/time
Manual
Checks
• AE 1 (Vomiting) with start date (5-Jan-09) & stop date(6-Jan-09) seems to be duplicate of
AE 3 (Vomiting ) with start date (5-Jan-09) & stop date(6-Jan-09). Kindly verify and
amend the CRF as applicable
Discrepant Data
Listings
• Visit 1 Lab sample date (1-Feb-09) as per CRF however in Central Lab report, it is dated 11-
Feb-09. Kindly verify and update the CRF as appropriate

Manual Review:
 The first manual review of data frequently takes place when the CRF is received. CRAs or
data managers go over the CRF before it is sent to data entry.
 During the entry process discrepancies may be identified manually when the data entry
operator cannot read a field or when the values from first entry and second entry differ.
 Also, linked discrepancies, i.e inconsistency of data collected between multiple data points
that are linked together are hard to identify automatically. Hence data managers have to
review this manually.
Eg- Medication is recorded for an Adverse event ;however, Adverse Event is not
recorded for the patient.
IDENTIFYING DISCREPANCY-
MANUAL REVIEW
22
Manual
Review

External System Checks:
IDENTIFYING DISCREPANCY-
EXTERNAL SYSTEM CHECKS
23
 The data management system may not be able to support checks across
patient records, visits, and pages (and even if it does, CDM staff may not have
expertise to write those checks), so other applications are used example – SAS
checks etc
 These applications may be run by data management specifically to check
data, or they may be used by other groups to review data and begin analysis.
 Discrepancies may become apparent from any of these reports, analysis, or
graphs.

EXAMPLE OF EXTERNAL SYSTEM
CHECKS - SAS
24
Description of Checks
Analysis dataset name is consistent as per Naming Convention. (Refer appropriate document)
Does the derived dataset include all the subjects from the raw dataset?
Check whether derived data (values and units) agree with the original (Raw) data
Check for all derived variable are calculated correctly
Length of the variables is same as specified in Specification Document?
Check if all the variables especially the derived ones are labeled properly in all the VADs.
Check if the number of visits is consistent with the number of visits in the study/SAP.
Ensure duplicate measurements are properly handled (e.g., SAP may instruct to use mean if more
than 1 measurement taken)
Sorting order of the variable is proper as specified in Specification Document?
None of the value are truncated
(Find longest string in raw data and check it hasn’t been truncated).

RESOLVING DISCREPANCY
25
Answers that resolve discrepancies may come from:
 Internal Groups:
 Data Management (Medical Coders,
Safety Team)
 Clinical Research Associates
 Investigator:
 Discrepancies that are sent back to the investigators for
resolution are called Queries or DCFs (data
clarifications forms)
Resolving Discrepancies

The value in
question may
be correct as is
An actual
measurement
may replace a
missing value
A corrected
value may
replace an
incorrect value
The value may
be wrong but
no corrected
value is
available
RESOLVING DISCREPANCY
26
Discrepancies have a variety of resolutions:

Discrepancies can be resolved in various ways:
 Update data that led to the discrepancy with correct data
 Update the data by applying a Data Entry guideline or Study definition attributes or by applying study
team decisions or by investigator authenticated changes) (Example: change ‘NINE’ to 9)
 Change the Study Definition attribute that created the discrepancy (e.g. increase the upper bound)
 Modify or retire the Validation Procedure that generated an invalid discrepancy (Procedures can be
modified or Retired according to need, throughout the study)
WAYS OF RESOLVING
DISCREPANCIES
27

 Accept the data as correct, and resolved
 For Example: Heart Rate is given ND (Not Done) flags Discrepancy, can be resolved as No
Action Required if it is same in CRF
 Accept the data as incorrect, but resolved
 For Example: Date of Collection of sample is different from the visit date, but investigator
confirms the same
WAYS OF RESOLVING
DISCREPANCIES (CONT.)
28

Paper
Database Updated by:
Data Manager
Style of query writing:
Detailed/Asking
EDC
Data base updated by:
Investigator
Style of query writing:
Instructing without
Leading
DIFFERENCE BETWEEN EDC
AND PAPER QUERIES
29

• The DCF (Data Clarification Form) provides a method to organize and report discrepancies
that need further Site review.
• Each DCF contains specific discrepancies for each unique investigator/patient combination.
• The specific discrepancies contained in a DCF are those that have Review/Resolution statuses
that meet the DCF inclusion criteria established by the user when DCFs are created.
• Each DCF has a corresponding physical DCF report that can be customized, printed and sent
to the Investigator for review and response.
DATA CLARIFICATION FORM :
DCF
30

Prepare
discrepancies
for DCF
Print and
Review DCFs
(CREATED)
Modify DCF as
necessary
Review
returned DCFs
Query desired
DCF
Include on next
DCF or return
to Investigator
Investigator
review and
return DCFs
Print DCF
Distribute printed
DCFs and change
DCF status to
SENT
Change DCF
status to
RECEIVED
Enough info
from INV to
close ?
Sent discrepancy
to resolve
View
associated
discrepancies
Resolve
discrepancy
Updated status to
received DCFs
to CLOSED
YES
NO
DATA CLARIFICATION FORM
PROCESS
31
No
Yes
Update the
database and the
status of the DCF
to CLOSED

CRF scanned
CRF image
loaded into DM
workflow
DE associate
enters data
DM runs edit
checks on pages
DM either
corrects the data
per SEC or sends
query to the site
Site answers DCFs
DM verifies
response and
updates the
database
DM freezes/locks
DB
WORKFLOW - PAPER STUDY
32

PROCESS (PAPER STUDY)
33
Step 1:
 The data manager will run a batch validation session as soon as one or more
pages have been double data entered on a regular basis.
 The Data Manager will also manually review the data listings
Step 2:
 All approved edit checks will be executed for all entered pages (without
comparison failures)
IF THEN
 Data does not meet the criteria  A discrepancy will be created in the
discrepancy database

Step 3:
 The Data Manager will check all discrepancies with the status ‘Data/DM review’
34
IF THEN
 A discrepancy reflects Principal
Investigator issues
 The status of the discrepancy will be
changed into ‘INV REVIEW’
 A discrepancy reflects Data Entry errors
(after reviewing CRF) and/or obvious errors
(as defined in the DMP)
 The data will be updated in the data
entry screen with a proper audit change
reason. The discrepancy will get a status
‘Resolved’
 A system generated discrepancy is
generated
 The Data Manager investigates the
discrepancy (see Step 4)

Step 4:
 System generated discrepancies
35
IF THEN
 The discrepancy is for a
LENGTH error
 Increase the length of the field.
 After a new batch validation the discrepancy will
close automatically by the system
list of values error
 The Data Manager investigates the discrepancy.
Set status to INV REVIEW
DATE or PARTIAL DATE error
 The Data Manager investigates the discrepancy.
Set status to INV REVIEW

• Step 5:
When a discrepancy will get status „RESOLVED‟, the Data Manager will make a resolution
comment in the discrepancy database
• Step 6:
A Data Manager can make a manual discrepancy with a status in the discrepancy
database. A manual discrepancy can be made as a result of reviewing the raw data listings
• Step 7:
36
IF THEN
 A discrepancy has the review status INV
REVIEW
 The Data Manager creates DCFs.
 The DCF gets status CREATED.

Step 8:
 All DCFs will be printed in pdf format
 The DCF gets status PRINTED
Step 9:
 The DCF will be sent to the site for resolution. The DCF will get the status SENT
Step 10:
 DCFs return to the Data Management Department within sponsor specified time
period after being released by the Data Manager
37

Step 11:
 When the DCFs are returned, the Data Manager changes the status of the DCF into
RECEIVED
Step 12:
 The Data Manager handles all DCFs and makes an update in the database if
necessary. The status will be changed to CLOSED
Step 13:
 The Data Manager will run Batch Validation Sessions, handle discrepancies, review
data listings, and handle DCFs on an ongoing basis, during the study until database
lock. At the moment of lock of the database, all discrepancies will have the status
CLOSED or RESOLVED
38

Sites enter data
into the EDC
Electronic edit
checks
automatically fire
Site resolves
queries
DM runs any
manual checks on
the data
DM enters manual
queries into EDC
Site addresses
the queries and
updates the
database
DM verifies
response
DM freezes/locks
the data
WORKFLOW IN EDC STUDY
39

Step 1:
 As soon as an eCRF page is saved discrepancies will be created in case of any inconsistencies.
Step 2:
 Besides Step 1 the Data Manager will run a batch validation session. This can also result in
discrepancies in the eCRF.
Step 3:
 The Data Manager will also manually review the data listings. He/she can create manual
discrepancies in the eCRF in case there are inconsistencies.
Step 4:
 The site monitor also has the privilege to create manual discrepancies in the eCRF.
Step 5:
 Discrepancies have color coding (sponsor specific). For eg. Red could mean that the discrepancy
is for DM group to review. Yellow means it is for another group of users.
PROCESS (EDC STUDY)
40

Step 6:
 Each user has to focus on their discrepancies depending on the color
PROCESS (EDC STUDY)
41
IF THEN
 A discrepancy is generated
that can be resolved by a
correction of the entered
data
 Correct the data and save the page.
 The discrepancy will be closed by the system and is no
longer visible (in the eCRF)
 A discrepancy is generated
that cannot be resolved by a
correction, but should be
answered by another user
group
 Send the discrepancy to the appropriate user group. If
necessary, include a comment.
 The discrepancy review status will change accordingly.
Step 7:
At database lock , all discrepancies will have the status CLOSED or RESOLVED.

• It is filtered by Query status
EDC - SAMPLE REPORT
42

There are instances where unresolved discrepancies are accepted as it is. These are
appropriately documented and provided to the study team and statistician for analysis purposes.
Examples:
Weight not taken at V2, ‘Not Done’ can be entered by the Data Managers in at this visit as
agreed by study team.
Per protocol Lab sample is expected to be collected at V1, however if site confirms that the Lab
sample was not collected at this visit then this is acceptable as is as agreed by study team.
UNRESOLVED ACCEPTABLE
QUERIES
43

 When any good discrepancy management system is teamed with an effort to identify discrepancies starting
early in the data handling for a study and continuing throughout, it is possible to cut down the time to close
of study to within a few days of receipt of the last CRF.
 Ideally, a problem that has been resolved should not continue to show up in lists of outstanding discrepancies.
For e.g., if a discrepancy sent to a site asks them to investigate a value that is slightly high and they respond
that the value is in fact correct, then the discrepancy should not be raised again, even if the discrepancy
checks run over the same data.
 After edits are made, it is essential to assure that all cleaning rules are rerun over the data. It is very
common for updates to the data as part of a discrepancy resolution to cause some other problem with the
data!
GOOD PRACTICES
44

Associates will be grouped and handed out a paper CRF with multiple
discrepancies. The group will have to underline the discrepancies and specify a
query should be raised
LEND A HAND
46

• Name any four Discrepancy types?
• What do you mean by System generated discrepancy?
• Name 2 types of System generated discrepancy?
• What do you mean by Multivariate discrepancy?
• What does DCF & DQF stand for and for what they are used ?
• Two types of discrepancy, based on where it originates
TEST YOUR UNDERSTANDING
47

• Discrepancy Management is a vital vehicle in Clinical Trails to ensure:
— The integrity & quality of data
— That the collected data is complete and accurate so that result are correct
• Discrepancy management is a process of cleaning subject data in Clinical Data Management System
(CDMS), it includes
— Manual checks
— System /Programmed checks
— External System checks
• Every discrepancy does not indicate an error with the data, only that the data does not meet
expectations.
• Discrepancies are resolved by internal groups/investigators and tracked for closure.
• Discrepancy management helps in cleaning the data and gathers enough evidence for the deviations
observed in data.
SUMMARY
48

You have successfully completed -
Discrepancy Management

Discrepany Management_Katalyst HLS

More Related Content

What's hot (20)

Similar to Discrepany Management_Katalyst HLS (20)

More from Katalyst HLS (20)

Recently uploaded (20)

Discrepany Management_Katalyst HLS