Handling of Partial Dates in EDC Studies at Roche

1
Handling of Partial Dates in EDC Studies at Roche

• Conor Eustace, Roche PDOD, Clinical Programming,
  Basel
• Macro User Group, Liverpool 2005

2
Purpose

• To detail the requirements for partial date
  handling in EDC studies at Roche
• To explain the challenges posed by the RocKET
  environment for partial date handling
• To describe the solution adopted

3
Partial Date Handling in Oracle Clinical

• Because Oracle Clinical is an integral part of
  the RocKET environment, existing procedures for
  partial date handling in Oracle Clinical drove
  expectations for handling of partial dates in EDC
  studies
• Oracle Clinical provides robust functionality for
  entering and storing partial dates
• For a date whose expected format is dd-mon-yyyy,
  the date can be saved even if either the day part
  or the day part and month part can be omitted
• In data entry mode partial dates would be
  displayed as shown below

4
Storage and reporting of partial dates in Oracle
Clinical at Roche

• Two questions are created in Oracle Clinical for
  each date
• A Date type question whose SAS name ends in dc
  stores the partial date as entered
• A Char type derived question whose SAS name ends
  in dt which is derived based on the following
  rule
• If a full date is entered save the full date
• If a partial date with only the date missing use
  15 as the day
• If a partial date with only the month use 30-Jun
  as the month and year

Example
5
Requirement for partial date handling

• Our partners in statistics required that the
  same information be available for partial dates
  in EDC studies as was already the case for paper
  studies running in Oracle Clinical, i.e.
• Should be possible to tell if the full date was
  entered or if only a partial date was available
• If a partial date is entered, it should be
  possible to tell if just the date part is missing
  or if both the date and month part are missing
  this requirement is fulfilled by the presence of
  the dc question (e.g. trbegdc)
• Whether the date is full or partial a complete
  format date should be available for analysis
• If the date is partial, estimations are
  performed as follows
• just the day part missing use 15
• both the day and the month parts are missing use
  30-Jun
• - This requirement is fulfilled by the derived
  dt question (e.g. trbegdt)

6
Challenges

• Fulfilling these requirements presented some
  challenges
• In the RocKET environment date type questions are
  defined in the Oracle Clinical GLIB and mapped
  using Clinical Gateway to Macro.
• Oracle Clinical Date type questions are therefore
  mapped to Date/time type questions in Macro
• Macro Date/time type questions cannot accept
  partial dates
• Therefore no simple way of getting all the
  required information back to Oracle Clinical
  across Clinical Gateway

7
Solution adopted (1)

• It was agreed that the Study Management teams
  would review all dates in the study DCS (data
  collection specification) and identify (by
  marking with a symbol) dates for which a
  partial date might be an acceptable response if
  the full date was unavailable

• All remaining dates were then considered dates
  for which a partial date is never an acceptable
  response

8
Solution adopted (2)

• A flag question (NCMPL_YN) was created in the
  Oracle Clinical GLIB. The NCMPL_YN question has
  a DVG attached with just the YES option active
• Each date identified by the SMT on the DCS as
  potentially having a partial date response, was
  associated to an occurrence of the partial date
  flag question (NCMPL_YN) in the Oracle Clinical
  DCM

9
Solution adopted (3)

• When the study definition is pushed to Macro,
  each date identified as potentially being partial
  will have a partial date flag question associated
  
• If the site does not know the exact day they are
  instructed to enter 15 for the day and tick the
  Partial date? question
• If the site only knows the year they are
  instructed to enter 30-Jun for the day and month
  and tick the Partial date? question

10
Solution adopted (3)

• This approach ensures our partners in statistics
  receive the same information as before
• The Partial date? question indicates that only a
  partial date was available
• If the Partial date? question is ticked the
  statistician can tell if the day part only was
  missing or if both day and month part were
  missing on the basis of whether 15 was entered
  for the day or 30-June was entered for the day
  and month.
• A full format date is always available for
  analysis (based on the estimation rule if the
  date is partial)

11
Technical Implementation of Solution in Macro

• To make this workaround function properly some
  additional code needs to be added to the Macro
  study definition

At eForm level Only Collect if Conditions
need to be added to each Partial date? question.
This ensures the Partial date? question only
becomes active if either 15 is entered for the
day or 30-Jun is entered for the day and month
12
Dealing with Future dates (1)

• If we do not know the precise day of a date in
  the current month, our partial date convention
  tells us to enter 15 for the day
• What if todays date is 13/05/2005 ?
• Entering 15/05/2005 would mean entering a date in
  the future
• It is a Roche standard to program Reject
  conditions to reject dates in the future
• The situation is similar if both the day and
  month part are unknown for an event in the
  current year the convention of entering
  30/06/2005 could lead to us entering a date in
  the future
• Special Reject and Warning validations were added
  to each date question associated with the Partial
  date? question to deal with this situation

13
Dealing with Future dates (2)

• Warn If validation
• Validation term mevalue isafter datenow
• Validation message
• 'The date is in the future. If this is because
  you have estimated the day or day-month part
  according to the partial date convention, please
  overrule this warning message using ''Date is a
  partial date'' as the reason for overrule, and
  tick the ''Partial'' tick box next to the date
  field. Otherwise please click ''save and close''
  and enter the correct date.'

14
Dealing with Future dates (3)

• Reject if validation
• Validation term

• Validation message 'Date '
  format_date(mevalue, 'dd/mm/yyyy') '
  occurs in the future. Please check.'

15
Dealing with Future dates (4)

• Note that
• The order in which the Reject and Warning are
  coded is important.
• The reject must be evaluated before the warning
• In Macro the most recently saved validation
  executes first, therefore good practice is to
  code the Warning, save then code the Reject

16
Writing validations against partial date
questions

• If partial dates are included in date comparisons
  it is possible (although perhaps not that likely)
  that validation messages may fire incorrectly
• For example, the day and month for a disease
  start date is not known.
• 30/06/2004 is entered and the Partial date?
  question is ticked according to the convention
• If the stop date is known precisely and entered
  as 29/06/2004, a check in Macro to see if start
  date is before stop date will fire
• To avoid the possibility of the sites getting
  such incorrect warning messages, we have
  decided to deal with all date comparisons offline
  in Oracle Clinical

17
Conclusions

• The workarounds described provide an effective
  way of dealing with partial dates
• Advantages
• All the necessary information is delivered for
  statistical analysis
• The SMT gives careful consideration to date
  handling at the design phase
• Seems to be simplest solution which fills our
  requirements
• Disadvantages
• Additional programming and testing of Arezzo
  expressions is required
• The sites need to follow a partial date rule
• Some modifications to existing statistical
  programming may also be necessary

18
Questions

• Thank you for your attention
• Any questions ?