FAQ

Frequently Asked Questions for STARR-OMOP

Below is a list of frequently asked questions for STARR-OMOP v5.4.

How do I know which event is represented in each row

Question

How do I know which events are represented by a single row in each one of the tables (e.g. condition_occurrence)? I only see numbers and dates but no other information.

Explanation

The OMOP CDM tables use concept IDs to represent clinical events. These concept IDs are numeric identifiers that map to human-readable terms in the concept table. To understand what a row represents, you need to join the clinical event table with the concept table.

In the condition_occurrence table, the condition_concept_id field contains the standardized concept representing the condition. To get the actual condition name, you join with the concept table.

Let’s demonstrate how to retrieve condition information with their human-readable names:

from google.cloud import bigquery

project_id = 'som-rit-starr-training'
bq_client = bigquery.Client(project_id)

1

Importing Python BigQuery package to interact with BigQuery.

2

Setting the project ID for query execution.

3

Creating the BigQuery Client to run queries.

query = """
SELECT
    co.person_id,
    co.condition_concept_id,
    c.concept_name,
    c.vocabulary_id,
    co.condition_start_date
FROM
    `som-rit-starr-training.starr_omop_cdm5_deid_1pcent_lite_2022_10_30.condition_occurrence` AS co
INNER JOIN
    `som-rit-starr-training.starr_omop_cdm5_deid_1pcent_lite_2022_10_30.concept` AS c
    ON co.condition_concept_id = c.concept_id
    AND c.vocabulary_id = 'SNOMED'
LIMIT 3
"""

df = bq_client.query(query).to_dataframe()

1

The person_id links to the patient who has this condition.

2

The condition_concept_id is the numeric code representing the condition.

3

The concept_name from the concept table provides the human-readable condition name.

4

The vocabulary_id indicates which standardized vocabulary this concept comes from (e.g., SNOMED).

5

The condition_start_date indicates when the condition was diagnosed or observed.

6

The condition_occurrence table contains all condition records.

7

The concept table contains the mapping from concept IDs to readable names.

8

Joining on condition_concept_id = concept_id links the numeric code to its meaning.

9

Select only the SNOMED concepts. They are the majority of the concepts in the condition_occurrence table.

10

Execute the query and convert results to a pandas DataFrame.

Now let’s check the table

You can apply this same pattern to any OMOP-CDM table that contains concept IDs. For example:

• In drug_exposure, join drug_concept_id with the concept table
• In procedure_occurrence, join procedure_concept_id with the concept table
• In measurement, join measurement_concept_id with the concept table

Conclusion

OMOP CDM tables store standardized numeric concept IDs to represent clinical events. To see what these numbers mean, you must join with the concept table using the appropriate concept_id field. This design allows for standardized, interoperable healthcare data analysis while maintaining semantic meaning through the vocabulary tables. The concept table acts as a central reference that translates all numeric codes into human-readable descriptions.