Research Data Quality
Before using a research dataset, check the agreed population for missing values, duplicate identifiers and broken references. Trace a finding to its resources and verify the correction.
Read each indicator against the population it measures
A quality result needs its denominator and coverage: missing data can make an indicator not evaluable.
Selected cohort
The agreed population and resource types.
Quality policy
Completeness, uniqueness and reference checks.
Measured data quality
Passed, evaluated and not-evaluable counts.
Trace a finding to its resources, correct the source and measure again.
Review what the selected research dataset contains, where it falls short, and what changed after a correction.
Read the indicator against its population
Identifier uniqueness is 10/12 in this twelve-Patient dataset. One recorded finding identifies the duplicate instead of treating a passing FHIR check as proof of dataset quality.
Trace the duplicate to both Patients
The finding lists the two affected Patients and the identifier field. Their source context and the applied quality rule stay attached to the investigation.
Review the result after the source correction
The completed recheck records 12/12 unique identifiers under the same quality policy. Review the measured population and the indicators that still have no applicable data.
After the source correction, identifier uniqueness improves from 10/12 to 12/12 under the same quality policy. Twenty other indicators remain not evaluable because their resource types are absent.
Related workflows and guides
Evaluate one research data delivery.
Bring a permitted test cohort, the profile versions, and a question about completeness, uniqueness, or references. Compare the findings with your current checks.