Guide · eCRF designUpdated September 28, 2026

How to design an eCRF that produces clean data

Most data queries are designed in, not entered in. Twelve practical rules for turning a protocol into case report forms that sites complete correctly the first time, with examples from real form builds.

  • Field types that fit the data
  • Checks at entry, not at lock
  • Draft, test, approve

Free sandbox · No credit card · 21 CFR Part 11 aligned

Concomitant medications · CM-004
Subject 002-0009 · Log formDraft

Medication name

CMTRT
Ibuprofen

Dose per administration

CMDOSE
400mg

Frequency

CMDOSFRQ
Three times daily (TID)

Start date

CMSTDAT
UNK-MAY-2026

Ongoing?

CMONGO
YesNo
One concept per field, units fixed

The short version

  • Collect only what the protocol and analysis plan need. Every extra field costs site time, monitoring time and cleaning time.
  • One concept per field, with a question type that matches the data: numbers as numbers, dates as dates, categories as coded choices. Free text is the exception.
  • Follow CDASH naming conventions so the hand-off to SDTM programmers is predictable.
  • Put data checks at the point of entry (skip logic, range checks, calculated fields) so problems are fixed while the participant is still in the clinic.
  • Draft, test with realistic sample data, then approve. In Capture only approved forms reach the casebook.

Principles

Start from the analysis, not from the paper form you used last time

An eCRF exists to collect the data the protocol requires, in a form the statistical analysis plan can use. That sounds obvious, but many forms are built by copying the previous study's forms and adding to them. The result is a casebook where sites fill in fields nobody analyses, and monitors verify data nobody needs.

A better starting point is the schedule of activities and the endpoints. For each assessment in the schedule, ask what data points the analysis needs, at which visits, in what units. Those become your fields. Anything else should earn its place: a safety reason, a regulatory requirement, or a clear operational use. The clinical trial protocol template shows how a well-structured protocol makes this mapping straightforward.

Rule 1: one concept per field

A field called "blood pressure" with a value of "120/80" is two measurements in one box, which means it cannot be range-checked or analysed without parsing. Split it into systolic and diastolic. The same applies to dose and unit, date and time, and anything a statistician will want to count separately.

Rule 2: match the question type to the data

Capture offers ten question types: short text, number, linear scale, slider, single choice, multiple choice, dropdown, date, time and long text, plus table sections for repeating rows. Use numbers for anything measured, single choice for mutually exclusive categories, multiple choice only when more than one answer can genuinely apply, and dates for dates. Keep long text for comments that nobody will analyse.

Rule 3: fix the units

Put the unit in the field definition rather than asking sites to type it. If sites work in different units (pounds and kilograms, mg/dL and mmol/L), decide on one and convert at source, or add a unit selector and convert in analysis. Mixed units in one field are one of the most expensive problems to clean up.

Rule 4: use standard names

CDASH, the CDISC standard for data collection, defines variable names and controlled terminology for common domains such as demographics, vital signs, adverse events and concomitant medications. Using CDASH-style names (AETERM, VSORRES, CMTRT) makes forms familiar to data managers and simplifies mapping to SDTM. Capture's templates carry an OID and an SDTM domain label on every form.

Number
Scale
Slider
Date
Date and time
Free text
Single choice
Multiple choice
Dropdown
Yes / No
Haematology (table section)
AnalyteResultUnitRef. range
Haemoglobin13.8g/dL13.0 to 17.0
White blood cells6.210^9/L4.0 to 10.0
Platelets41210^9/L150 to 400

Choosing field types

Which question type for which data

DataQuestion typeWhyExample
Measured valueNumber with fixed unit and decimalsRange checks and analysis without parsingWeight 81.4 kg
Category, one answerSingle choice or dropdownCoded values, no spelling variantsSex, outcome of AE
Category, several answersMultiple choiceEach option exported as its own flagSeriousness criteria
Rating scaleLinear scaleAnchored, coded 0 to nPHQ-9 items
Continuous self-ratingSlider with anchorsVisual analogue scales with no preset valueVAS pain 0 to 100
DateDate, partial dates allowed where neededConsistent format, partial dates preservedAE onset UNK-JUN-2026
Clock timeTime24-hour, unambiguousBedtime 23:30
Derived valueCalculated fieldNever disagrees with its inputsBMI, eGFR, QTcF
Repeating entriesTable section or log formOne row per event, own audit trailCon meds, lab analytes
Free commentLong textOnly where analysis does not need itInvestigator comment

Form structure

Visit forms, log forms and the casebook

Data collected at a scheduled visit (vital signs, questionnaires, physical exam) belongs on visit forms placed on the schedule. Data that happens on its own timeline (adverse events, concomitant medications, protocol deviations) belongs on log forms with repeating rows. Mixing the two, for example asking "any new AEs?" as a visit form question and then recording details elsewhere, creates duplicates and reconciliation work.

Rule 5: keep forms short and single-purpose

A form should cover one assessment. Separate forms can be scheduled independently, reused across studies, verified with different SDV settings and locked separately. A 200-field "Visit 3" form can do none of that.

Rule 6: use skip logic instead of "if yes" instructions

If a follow-up question applies only to some answers, show it only for those answers. Conditional branching removes blank fields that monitors have to query ("was this left empty on purpose?") and keeps forms shorter for sites.

Rule 7: record "not done" explicitly

A blank field is ambiguous. Capture exports distinguish not done (ND), not available (NA), visit not done (VND) and not applicable from a genuinely missing answer, so data managers can tell an assessment that was skipped from one that was forgotten.

Rule 8: allow partial dates where people cannot remember

Medical history and adverse event onset dates are often known only to the month or year. Accepting partial dates such as UNK-JUN-2026 avoids invented days that later contradict source documents.

Stage 1

Draft

Build and edit freely

Stage 2

Approved

Locked for live use

Stage 3

In the casebook

Only approved forms

Audit trail
  • SDV requirement changed on Systolic BP

    Data manager · Reason: critical safety value

  • Form approved: Vital Signs (VS) v2

    Study owner · Locked for live use

  • Answer changed: Weight 68.0 to 68.5 kg

    Site coordinator · Reason: transcription error

Rule 9 and 10

Check data where it is entered, and let the system do the maths

Edit checks catch errors while they can still be fixed from memory. Range checks with a high and low limit, and custom-value rules with a priority, raise an auto-query in the site action list as soon as a value breaks the rule. Calculated fields such as BMI, body surface area, eGFR (CKD-EPI 2021), creatinine clearance, QTc (Bazett and Fridericia) and age from date of birth are shown read-only during entry, so they are never typed or mistyped.

  • Range and custom-value edit checks with priority.
  • Auto-queries appear in the site's action list immediately.
  • Fourteen built-in clinical calculators.
  • Too many checks cause query fatigue: target the fields that matter.
Edit checks software for clinical trials
Vital signs · Systolic blood pressure

Edit checks / auto-queries

2
Auto-query

Type

Range High

Operator

Greater than

Value

180

Priority: High

Auto-query

Type

Range Low

Operator

Less than

Value

80

Priority: Normal

Query raised automatically

Value 192 violates limit (180). Please verify.

Rule 11

Draft with AI, decide as a human

Nobody needs to type 40 lab analytes by hand. Capture's AI form builder drafts questions or a lab table from a description or an uploaded CSV, XLSX, PDF or DOCX, including units and reference ranges split by sex and age where the source has them. Nothing is saved until a person reviews and applies it, and it only works on draft forms, so the design decisions stay with you.

  • Drafts questions and lab tables from a description or a file.
  • Unrecognised analytes flagged, never guessed.
  • Only works on draft forms; approved forms are locked.
CRF builder for clinical trials
Build this form with AI

Question form

CRF or ePRO

Lab table

Analytes, units, ranges

Vital signs form: BP, pulse, temperature, respiratory rate, weight
Upload CSV, XLSX, PDF or DOCX
14 questionsReview before applying
  • 1. Systolic blood pressure

    Required · Number · mmHg · 0 dp

    Number
  • 2. Diastolic blood pressure

    Required · Number · mmHg · 0 dp

    Number
  • 3. Pulse rate

    Required · Number · beats/min

    Number
  • 4. Body temperature

    Required · Number · °C · 1 dp

    Number
  • 5. Position during measurement

    Required · Sitting, Supine, Standing

    Single choice
  • + 9 more questions
Back Apply to form

Design a form, break it, then approve it

Build an eCRF in the free sandbox, enter test data that should fail your checks, and see what sites will see.

Design your first eCRF free

Rule 12: review before release

Draft, test, approve, submit

Forms move through a draft to approved lifecycle in Capture. Only approved forms appear in the casebook, and approved forms are locked for live use.

  1. 1

    Draft

    Build forms from the protocol, templates or AI drafts. Change anything while in draft.

  2. 2

    Test

    Create test subjects and enter realistic and deliberately wrong data to exercise every check and branch.

  3. 3

    Approve

    Approve each form when it matches the protocol. Approved forms are locked and identical at every site.

  4. 4

    Submit

    Export the blank eCRF PDF (cover page, form index, visit-by-form matrix) for the ethics or IRB package.

Design review

eCRF design review checklist

Every field traced to the protocol

Each field supports an endpoint, a safety requirement or a regulatory need.

No compound fields

One measurement or concept per field.

Units and decimals fixed

Defined on the field, not typed by sites.

Visit forms and log forms separated

AEs, con meds and deviations on logs with repeating rows.

Checks targeted

Edit checks on the data that matters, not on every field.

Completion guidelines written

Short instructions for anything a site could interpret two ways.

FAQ

Questions teams ask before they switch

Something not covered here? Ask us directly.

What makes a good eCRF?

It collects only the data the protocol needs, uses question types that match the data, applies checks at entry, and is tested with realistic data before release.

What is CDASH?

The Clinical Data Acquisition Standards Harmonization standard from CDISC. It defines standard fields and names for common data collection domains, which simplifies later mapping to SDTM.

Should I copy my paper CRF into an eCRF?

Use it as input, not as a template. Electronic forms can use skip logic, checks and calculated fields that paper cannot, so a direct copy misses most of the benefit. See paper CRF to eCRF conversion.

How many edit checks should an eCRF have?

Enough to catch the errors that matter for safety and endpoints. Too many checks create query fatigue at sites without improving the data that drives the analysis.

Who should review an eCRF before go-live?

Typically the data manager, a clinical representative, the statistician and someone from a site who will actually complete it.

Can AI design an eCRF?

AI can draft questions and lab tables from a description or file. People should still decide what the study collects and review every field before approval.

What do I give the ethics committee?

Usually a blank CRF. Capture exports a blank eCRF PDF with a cover page, form index, visit-by-form matrix and every approved form printed empty.

Design forms sites can complete on the first try

Free sandbox with every feature. No credit card, no sales call.

Design your first eCRF free