Auto-encoder

The Auto-encoder solution allows companies to manage and classify free text terms captured during the clinical drug development process.  Auto-encoder acts as a global, central repository for managing these verbatims and corresponding dictionaries of terms.   Leveraging global thesauri contained in standardized dictionaries,  Auto-encoder facilitates the classification of verbatim terms using a variety of automated tools and functions contained within Auto-encoder:

• The Auto-encoder allows any fields to be coded from any dictionary automatically.  Standard commercial dictionaries are supported:
  • MedDRA
  • WHODrug

• Non-standard client dictionaries can also be defined and used.
• Multiple versions and views of the dictionaries can be created.
• Dictionaries are mapped to the eCRF pages at the item level.  Fields can be mapped as:
  • Mapped – this is the single field used as the lookup to the dictionary
  • Displayed – these fields are displayed on the coding screen and saved to the database
  • Not Displayed – these fields are not displayed on the coding screen but are still saved to the database.

• The Auto-encoder can restrict the amount of data to be processed by filtering at Site, Subject, Event and Page
• Autocoding is submitted as a batch job to allow other work to proceed.
• The Auto-encoder tries to match the Mapped field to the dictionary as follows:
  • Exact match
  • Modified string match – here extra spaces and illegal characters are removed

• If no records are returned the same process is used against a history, or thesaurus table.
• Once the coding job is complete the Coding Review screen can be reviewed which shows the results of the auto-encoding, and allows manual coding to be performed
• The review screen can be filtered to show specific results :
  • All
  • Uncoded
  • Proposed
  • Reviewed
  • Approved
  • Applied

• A manual code lookup can be performed for Uncoded and Proposed status.
• The status of the coding can then be marked as Reviewed then Approved (these are role based privileges)
• The coding can optionally be saved to the History table – particularly for uncoded records following a manual coding.
• The codings can then be updated to the database.  This is submitted as a job to allow other work to proceed.  Once completed the clinical database will be updated with the selected dictionary records.
• The Auto-encoder is fully integrated with TrialMaster EDC to allow for the automated and manual coding of adverse events and concomitant medications captured within EDC.