Cross-form checks

Which listings to upload, what each check looks for, the visit schedule.

Cross-form checks look at several listings at once: adverse events, concomitant medication, visits and the subject list (consent, randomisation, end of study, death). These checks are hard to program in most EDCs because they span forms, so data managers usually do them by hand in Excel before every data review or interim lock.

What to upload

Upload up to 8 CSV or XLSX files, or one workbook with a sheet per listing. Hidden sheets are skipped. Each listing is recognised by its column names: CDASH and SDTM names (AETERM, AESTDAT, CMTRT, CMINDC, VISDAT, RFICDTC, DTHDAT...), typical Rave, REDCap and OpenClinica export names, and Russian names such as «Дата визита» or «Препарат». The result shows which column took which role. A listing that is not recognised is listed with its columns, so you can rename them and upload again.

Subject numbers are compared without separators and prefixes: "101-001", "101 001" and "S-101-001" are the same subject.

The checks

• AE treated, but no medication: the AE says treatment was given, and no medication links its AE number or starts between the AE onset (minus the day window) and its end.
• Medication for an AE, but no such AE: the indication says "adverse event" or the record links an AE number that the subject does not have.
• Before informed consent: a visit or AE start before the consent date. An AE before consent is often medical history entered on the wrong form.
• After end of study: a visit, AE or medication that starts after the end-of-study or death date.
• Ongoing or ending after death: an AE or medication still ongoing, or ending after the death date. The fatal AE usually needs outcome "fatal" and an end date on the date of death.
• End date before start date in AEs and medication.
• Duplicates: the same AE term and onset, the same drug and start, the same visit twice or two visits on one date for one subject.
• Visit out of window: needs the visit schedule, written as Screening=-14±14; Day 1=0; Week 2=14±3; Week 4=28-2+5. Days count from randomisation when the subject list has it, otherwise from the subject's first visit. Visits not in the schedule are listed and not checked.
• Subject not in the subject list: a subject who has AEs, medication or visits but no row in the subject list.

A check runs only when its listings are there; the result says which checks were not run and why. Findings about consent, end of study and death are marked critical.

Working the list

The Excel report has a summary, a sheet with all findings and a sheet per check, with empty Status and Comment columns for your notes. Each row names the listing and its row number, so you can find the record and raise the query in your EDC.

Try it

Example files: subjects.csv, ae.csv, cm.csv, visits.csv. Upload all four files together and type the schedule Screening=-14±14; Day 1=0; Week 2=14±3; Week 4=28±3. Every check finds something: an AE before consent, an AE with treatment but no drug, a drug for an AE that does not exist, events after end of study and after death, duplicates and visits out of window.

Sources

• CDISC CDASH: the AE, CM, SV, DS and DM variable names the portal recognises.
• SCDM, Good Clinical Data Management Practices: data validation and edit check chapters.