Problem to Solve: The Statistic Datatype is not able to express that a statistic is related to a category for a categorical variable.

FHIR resource element(s) affected: Statistic.category

Use case: An example is reporting the baseline measure of "Ethnicity (NIH/OMB)" in 3 categories (Hispanic or Latino, Not Hispanic or Latino, Unknown or Not Reported). Without the change this would require creating 3 separate EvidenceVariable Resources (one for each category) and 3 separate Evidence Resources (to report the proportion in each category). With the change, this is achieved with 1 EvidenceVariable Resource (containing 3 category entries) and 1 Evidence Resource (containing 3 statistics each associated with a specific category).

Best practices: Reporting categorical statistics together with the same Evidence Resource is preferred to keep the dataset tightly related, e.g. to confirm the values across all categories add up to exactly 100% of the expected total.

Recommended approach – change to FHIR schema:

• Add Statistic.category with cardinality 0..1 and datatype string.
• Short description = "associated category for categorical variable"
• Definition = "When the measured variable is handled categorically, the category element is used to define which category the statistic is reporting."
• Comments = "Simple strings can be used for descriptive purposes. Exact matching to EvidenceVariable.category.name for the Evidence.variableDefinition[variableRole=measuredVariable].observed=Reference(EvidenceVariable) could facilitate validation within datasets."

Importance/Tradeoffs: High importance because many variables are categorical, especially for baseline measures, and especially common throughout ClinicalTrials.gov for age, sex, gender, race, ethnicity, and region. Added value is great efficiency where the number of resources needed to express categorical statistics is reduced by > 50%. There is no backwards compatibility concern because no existing elements are being removed or modified.  No clear tradeoff problem identified.