Example: FHIR to Cohort

This pattern starts from FHIR resources, extracts tables, then builds a cohort for analysis.

The example uses public HAPI FHIR data and intentionally keeps extraction simple. In a real project, you would usually replace the basic groupby logic with study-specific feature definitions, terminology filters, date windows, and endpoint rules.

For precise REST field selection and site-specific preprocessing functions, read FHIR Paths And Custom Extraction next.

What To Check Before Scaling

Confirm resource references use the same patient identifier format.
Limit pages while developing queries.
Save raw extracts before aggregating features.
Validate that features are measured before the outcome window.
Keep implementation-guide-specific extraction code versioned.

python

from fyron import FHIRRestClient
from fyron.cohort import join_cohort_tables, validate_cohort_table

client = FHIRRestClient("https://hapi.fhir.org/baseR4")

patients = client.search_df("Patient", params={"_count": 100}, max_pages=2)
conditions = client.search_df("Condition", params={"_count": 100}, max_pages=2)
observations = client.search_df("Observation", params={"_count": 100}, max_pages=2)

features = observations.groupby("subject.reference").agg(
    n_observations=("id", "count"),
).reset_index()

outcomes = conditions.groupby("subject.reference").agg(
    n_conditions=("id", "count"),
).reset_index()

cohort = join_cohort_tables(
    features,
    outcomes,
    on="subject.reference",
    how="inner",
)

validate_cohort_table(
    cohort,
    required_columns=["subject.reference", "n_observations", "n_conditions"],
)

FHIR schemas vary by server and implementation guide. Keep extraction logic explicit and version-controlled.