DM workspace

Specification check

What the check finds in an eCRF specification and how to read the report.

Upload the eCRF specification of a study built in another EDC. The portal finds the mistakes that usually surface only in UAT or after go-live, when they are expensive to fix.

Which specifications it reads

Medidata Rave ALS (Architect Loader Spreadsheet): sheets Forms, Fields, DataDictionaryEntries, Checks, CheckSteps, CheckActions.
REDCap data dictionary in CSV or Excel.
Any Excel spec with a fields sheet (Form and Field columns) and, when present, sheets of code lists (Codelist, Code, Decode), forms and edit checks (Check ID, Field, Query text). Column names are recognised in English, Russian and French.

What it finds

Critical errors break the build or the data: a field twice on a form, a choice field without a code list, a code list that is not defined, a repeated code, a reversed range (minimum above maximum), a check on a field or form that is not in the spec, a repeated check ID, REDCap logic that refers to a missing variable.
Important ones to fix before UAT: no question text or no translation into one of the languages, a check without query text, a form without fields, a field on a form missing from the list of forms, the same field with different type or code list on different forms, spaces at the start or end of a code, invalid characters in a field name, an identifier not marked “Identifier” in REDCap.
Info: a field name longer than 8 characters (does not fit SDTM and SAS XPT v5), a numeric field without a range, an unused code list, a question that looks like a direct identifier (name, initials, phone, passport). Collecting such data must be allowed by the protocol, the informed consent and the personal data law.

Every finding gives the sheet and row so it is easy to find in the source file. The Excel report has a summary, all findings and a table of forms.

Example

Press “Check the Excel example”. The training spec DEMO-301 has planted mistakes: WEIGHT twice, AESEV without a code list, a reversed DIABP range (140–30), check AE001 on the missing field AEENDAT, check AE002 without query text, code “3” twice in RACE, the unused code list SEV and more. The REDCap example shows a reference to the missing variable [smoking] and a phone field not marked as an identifier. Both files can be downloaded.

Sources

Field names in the example follow CDISC CDASH.
Specifications and edit checks: SCDM, Good Clinical Data Management Practices.

Nothing is stored. The file is read in memory and discarded; the log keeps only the number of fields and findings. The check does not replace spec review and UAT: it catches mechanical errors, people check the meaning.