Assigning Standard OMOP Vocabulary Concepts to CIBMTR Form 4100
A comprehensive guide to tools, databases, and techniques for standardized data mapping
Key Highlights
- Comprehensive Use of OMOP CDM: Utilize the entire OMOP framework, including standardized vocabularies, domain-specific mapping, and data validation.
- Integration of Multiple Tools: Leverage tools like ATHENA, Usagi, and various ETL solutions to transform and load data seamlessly.
- Mapping and Validation Process: Apply both automated and manual mapping techniques supported by robust validation methods to ensure accuracy.
Understanding the Data
The process begins with a thorough examination of CIBMTR’s Form 4100 (available here). Understanding the fields, value sets, and the underlying data is critical. The form contains data on demographics, diagnoses, procedures, treatments, and potentially unique responses that require standardization for data interoperability.
Data Collection Breakdown
Field Identification
Start by reviewing each answer field in the form. Identify the clinical, demographic, and procedural data elements. Determine if fields represent conditions, medications, lab results, or procedures, as each category aligns with different domain vocabularies such as SNOMED CT for clinical conditions, RxNorm for medications, LOINC for laboratory tests, and ICD for diseases.
Defining Value Sets
Collect all potential values used in the form. This involves extracting pre-coded values and narrative responses. Having a clear compilation of these values is essential to ensure that all source concepts can be mapped to appropriate standard concepts.
Tools and Databases for OMOP Standardization
Key Tools
To map the answer fields effectively, you can leverage multiple tools and databases from the OHDSI community and broader data analytics world:
OMOP Standardized Vocabularies
The OMOP Standardized Vocabularies are the cornerstone of mapping. They incorporate extensive medical terminologies such as SNOMED CT, LOINC, and RxNorm. These vocabularies provide concept IDs, names, and coding systems that facilitate uniformity across diverse data sources.
ATHENA
ATHENA is a web-based tool provided by OHDSI that allows users to search and browse the standardized vocabularies. By entering source values from the CIBMTR form, you can identify corresponding standard concepts. ATHENA simplifies the task of data alignment, ensuring comprehensive coverage of clinical and procedural terms.
Usagi
Usagi is a specialized tool designed for mapping non-standardized codes to standard OMOP concepts using text similarity. It automates the mapping process by comparing the form’s descriptions with standardized vocabulary descriptions, suggesting the appropriate concept IDs for each term.
Data Integration and ETL Tools
For transforming and loading data into the OMOP Common Data Model (CDM), employing ETL (Extract, Transform, Load) tools is essential. Whether you use SQL-based platforms like PostgreSQL or utilize dedicated ETL software like Informatica PowerCenter, these tools ensure that the data is accurately transformed from its source form into the OMOP’s structured tables.
Mapping Techniques and Process
Developing a Vocabulary Mapping Strategy
A robust mapping strategy involves several stages, each crucial to ensuring the data’s integrity and interpretability once standardized:
1. Extract and Review Source Data
Begin by systematically scanning the CIBMTR form data. This process involves:
- Extracting source values for every field in the form.
- Understanding the context behind each value, for example, differentiating between primary and secondary diagnoses or various demographic identifiers.
- Listing out different response types to ensure nothing is overlooked when matching to a standardized concept.
2. Automated Mapping with Usagi
Leverage Usagi to perform an initial automated mapping from your identified source values. Usagi utilizes text similarity to compare source data descriptions with the standardized vocabulary listed in ATHENA and other OMOP tools. The output provides suggestions that need review for accuracy.
3. Manual Oversight and Custom Mapping
Although automation greatly speeds up the process, certain fields may require manual mapping:
- Some values may not have a direct standardized match in the available vocabularies.
- For such instances, manually review the best-fit options and consider creating custom mappings or additional descriptive metadata.
- This step is essential for ensuring that the mappings faithfully represent the clinical or procedural intent behind the original data.
4. Building the Source-to-Concept Map
Once you have both automated and manual mappings ready, consolidate them into a comprehensive "Source_to_concept_map" table. This table will include:
| Source Code/Value | Source Description | Standard Concept ID | Concept Name | Vocabulary |
|---|---|---|---|---|
| Example Code 1 | Diagnosis Example | 123456 | Standard Diagnosis Term | SNOMED CT |
| Example Code 2 | Procedure Example | 789012 | Standard Procedure Term | LOINC |
This table serves as an audit trail, ensuring that each mapping is documented and can be revisited during data quality checks.
ETL Process Integration
After mapping, data needs to be transformed from its original format into the OMOP CDM. The ETL process involves:
Extraction
Retrieve the data from the CIBMTR data collection files. This step may involve connecting to legacy databases, flat files, or structured data formats. Transparency in this extraction ensures that no data is omitted.
Transformation
Transform the extracted data by applying your mapping rules. Here the Source_to_concept_map table is used to substitute the original field values with standardized concept IDs. This process also involves:
- Converting data types to match the OMOP CDM standards.
- Ensuring that each data record aligns with the appropriate domain (e.g., condition, measurement, drug exposure).
- Addressing any discrepancies such as missing data or non-standard responses.
Loading into OMOP Tables
Finally, load the transformed data into the corresponding OMOP tables (for instance, the condition_occurrence, drug_exposure, and procedure_occurrence tables). This standardized repository then allows for analysis using OMOP analytic tools.
Sample SQL Code for Mapping
A typical mapping operation might involve an SQL INSERT statement to capture relationships between source and standardized values. For example:
-- Mapping a source concept to a standard OMOP concept
INSERT INTO concept_relationship (
concept_id_1, -- Source concept ID
concept_id_2, -- Standard concept ID
relationship_id -- Relationship type (e.g., 'Maps to')
)
VALUES (
2000000001, -- Custom source concept ID
438067, -- Standard concept ID (e.g., a matched diagnosis)
'Maps to'
);
This SQL snippet illustrates the fundamental step in linking local data points with their standardized counterparts, ensuring traceability and auditability through the mapping process.
Data Validation, Quality Control, and Maintenance
Quality Assurance Measures
A crucial part of the standardization process is validating that the mappings accurately capture the original data’s intent. Implement robust data validation processes including:
- Automated tests: Use scripts and analytic queries to check for discrepancies or unmapped values.
- Manual audits: Regular reviews by domain experts to ensure that mappings match clinical meaning.
- Feedback loops: Provide systematic feedback to update the mapping rules when encountering new or unexpected values.
Ongoing Data Maintenance
Post-integration, it is necessary to maintain and update the mappings as new terminologies emerge or when the source data changes. This involves scheduling periodic reviews, refining the Source_to_concept_map, and engaging with the OHDSI community for updates to vocabularies.
Integrating Analytical Tools for Advanced Insights
Using Data Analytics Frameworks
Once the data resides in an OMOP format, data analytics tools facilitate further exploration and research:
- Leverage frameworks like Apache Spark or Python libraries (Pandas, NumPy) to handle and analyze large datasets.
- Advanced querying using SQL enables cross-sectional studies and comparative analysis among different patient groups.
- Graphical tools and reports can be generated to validate mapping consistency over time.
Integration with OHDSI Tools
The OMOP CDM’s standardized structure allows seamless integration with other OHDSI tools for deeper cohort analysis and observational studies. Tools like Atlas (for exploratory data analysis) further extend the capabilities once data standardization is complete.
Summary Table of Key Components
| Component | Description | Example Tool/Resource |
|---|---|---|
| Standardized Vocabularies | Core terminologies (SNOMED CT, LOINC, RxNorm) for mapping source data to common concepts. | OMOP Vocabularies |
| Mapping Tools | Automated and manual tools to match source values with standardized concepts. | Usagi, ATHENA |
| ETL Process | Extract, Transform, Load data to fit the OMOP Common Data Model. | SQL, Informatica, Python |
| Data Validation | Quality assurance through automated tests, expert reviews, and feedback loops. | Data audits, analytic queries |
References
- OHDSI Data Standardization - OHDSI
- OMOP Common Data Model FAQ - OHDSI
- Usagi Tool - OHDSI
- Understanding OMOP Basics - Research All of Us
- CIBMTR Form 4100 Documentation - CIBMTR
Recommended Queries for Further Exploration
Last updated March 7, 2025