Verifying Clinical Data

Patient-reported vs. clinician verified data

Most patient registries start with self reported data because it is scalable and low-cost. But for many research and regulatory purposes, verified clinical data is required.

What typically needs verification

Diagnosis confirmation

For rare diseases, self reported diagnosis is unreliable. Many participants have the wrong diagnosis; others have multiple diagnoses and may report a different primary. For research purposes, and especially for clinical trial recruitment, diagnosis must be verified against medical records or genetic testing.

Methods:

• Medical record review by a study coordinator or clinician
• Genetic test report review (for genetic diseases)
• Clinician attestation (treating physician confirms diagnosis)
• Death certificate review (for mortality data)

Genetic/molecular data

Mutation or variant data should be extracted directly from clinical genetic test reports, not transcribed by participants. Participants frequently report variant details incorrectly.

Build a document upload feature so participants can upload their genetic test report, and have a study coordinator extract the variant data into standardized fields (HGVS notation, gene symbol, variant type).

Key clinical measurements

For outcomes analysis, key measurements (functional scores, lab values, imaging findings) should be extracted from the medical record or entered by the treating clinician, not relying on participant memory.

Data quality framework

Edit checks and validation rules

Build validation into your data collection forms:

• Range checks (age cannot be negative; weight cannot be 500kg)
• Logic checks (symptom onset cannot be before birth date)
• Completeness checks (flag records with missing required fields)
• Consistency checks (diagnosis date should precede treatment start date)

Missing data management

Design your analysis plan before data collection, including a missing data strategy:

• What percentage of missingness in a field triggers exclusion from analysis?
• Will you use imputation? Which methods?
• How will you handle "not applicable" vs. "unknown" vs. genuinely missing?

Duplicate detection

Rare disease communities are small. The same participant may enroll multiple times, or enroll in multiple registries. Build duplicate detection:

• Match on date of birth, sex, and diagnosis date
• Consider a unique participant identifier (with appropriate privacy protections)
• Coordinate with other registries in your disease space

Audit trail

Maintain a complete audit trail of:

• Who entered each data element
• When data was entered and modified
• What changes were made
• Source documentation for verified data

This is required for regulatory submissions and is good research practice.

Study coordinator verification workflow

For registries with clinical site participation, a typical verification workflow:

1. Participant enrolls and completes self report
2. Participant signs medical records release authorization
3. Study coordinator at clinical site receives notification
4. Coordinator reviews medical record and completes clinical data form
5. Discrepancies between patient-reported and record-sourced data are flagged
6. Data manager reviews and resolves discrepancies
7. Record is marked as "clinician verified"

Data quality metrics to track

• Completeness rate: % of required fields with non-missing values, by element and by time period
• Timeliness: Median days from enrollment to complete baseline data
• Verification rate: % of enrolled participants with clinician verified diagnosis
• Inconsistency rate: % of records with at least one logic or range check failure
• Attrition rate: % of participants who complete each follow up assessment

Report these metrics to your SAB quarterly.

Key resources

• AHRQ Registry User's Guide Chapter 5: Data Collection
• FDA Data Standards for Clinical Trials
• CDISC CDASH (Clinical Data Acquisition Standards Harmonization)
• TransCelerate Data Standards

← Module 11 | Module 13: Recruitment Strategies →