How to design a case report form (CRF)

You design a case report form by working backwards from the protocol. List what each endpoint needs, decide at which visit you collect it, then build one field per data point with the checks that keep it clean. The tour above shows the build in Datacapt. The steps below work in any EDC.

What your CRF has to capture

Your CRF holds every observation the protocol asks for, per participant and per visit. Your statistician analyses it, your monitors verify it against source and inspectors read its audit trail. Leave a data point out and nobody can analyse it. Add a field nobody needs and your sites pay for it in entry time and queries.

Most CRFs share the same core forms: eligibility, demographics, medical history, vital signs, concomitant medications, adverse events, study treatment and end of study. Your endpoints add the forms specific to your trial.

How to design a CRF in 8 steps

1. Start from the schedule of assessments. Take the protocol table that lists each visit and each assessment. Every cell becomes a form or a field. Mark the primary and secondary endpoints so you know which fields carry the analysis.
2. List each data point once. Collect the date of birth and let the EDC calculate age. Record a lab value with its unit in one place. Duplicate fields create discrepancies your team has to query later.
3. Pick the question type that matches the data. Use numeric fields for measurements, dates for events and coded lists for categories. Keep free text for the rare answer nothing else fits, since your team cannot count or code it without cleaning it first.
4. Name variables and set units the same way across forms. Short, consistent names such as the CDASH-style VSORRES for a vital sign result save your data manager hours at export. Show units and expected formats next to the question, where the site user reads them.
5. Write edit checks as you build. Add range checks on measurements, required flags on key fields and cross-field checks where two answers must agree, such as a visit date after the consent date. The site sees the warning at entry and fixes the value before your team raises a query.
6. Hide what does not apply. Use visibility conditions: show the pregnancy test for participants of childbearing potential, and show the adverse event form when the site answers yes to "Any adverse event?". Sites complete shorter forms faster.
7. Test with a dummy participant. Enter a fake participant through every visit, including the paths that trigger each check. Ask a site coordinator to do the same and fix what confuses them before the first real patient.
8. Review, sign off and version. Have the principal investigator, the data manager and the statistician sign off the CRF against the protocol. When the protocol changes, publish a new CRF version and keep the previous one in the audit trail.

See the build in Datacapt

The 3-minute tour at the top of this page runs the same build on demo data. You will:

• Open the participant list, where each participant shows their status and progress in real time.
• Pick from up to 25 question types, including numeric, date, single and multiple choice, calculated fields and conditional logic.
• Set the question title, variable name and answer options, and watch the form update as you type.
• Add a participant to the study to test the form.
• Read the study dashboard with progress tracking and the KPIs your team follows.

Already have the protocol as a document? You can ask an AI agent connected to Datacapt through MCP to draft the forms into a draft study, then review them before anything goes live. Read how the four levels of AI agent access compare.

Mistakes that turn into queries

• Data nobody will analyse. Every extra field adds entry time at every visit for every participant.
• Free text where a list would do. "Headache, mild" and "mild headache" become two values your coder has to reconcile.
• Hard stops on plausible values. Block an unusual but real blood pressure and a tired coordinator enters a false one. Use a warning and ask the site to confirm.
• Units left to the site. Weight in pounds at one site and kilograms at another means a cleaning job before any analysis. Fix the unit, or ask for it next to the value.
• No test run. Skip it and your first real patient becomes your user acceptance test.

What regulators expect from an eCRF

ICH E6(R3) asks sponsors to build quality into the trial design and to focus on the data that decide whether the results hold. For the CRF, that means fewer fields, each one justified by the protocol.

If you submit to the FDA, 21 CFR Part 11 applies to your EDC: audit trails, electronic signatures and access controls. In Europe, the GDPR governs how you process participant data. Read what regulatory compliance means for an EDC.

💡 Datacapt records every field change in the audit trail and hosts study data on certified health-data infrastructure (HDS). See Security & privacy.

Keep reading

• What a case report form is, and how it differs from an eCRF
• Designing an eCRF: steps and best practices
• Every EDC feature in Datacapt
• All tours in the Demo Center