OHDSI

The OMOP CDM is the foundational open-source standard that transforms heterogeneous native electronic health records into a consistent, analytics-ready format. It standardizes the structure (tables) and content (vocabularies) of observational data.

• Person-centric design: All clinical events are linked to a unique person, enabling a complete longitudinal health record.
• Standardized Vocabularies: Maps local codes (ICD-10, CPT-4, NDC) to common concepts using SNOMED-CT, LOINC, and RxNorm.
• CDM v5.4: The current version supports 39 clinical data tables, including drug exposure, condition occurrence, and measurements.

ATLAS is a web-based, drag-and-drop interface that allows researchers to design computable phenotypes and execute full-scale analytical studies without writing SQL code. It democratizes access to complex observational research.

• Cohort Definition: Define inclusion criteria using clinical events, temporal logic, and demographic filters.
• Concept Set Editor: Search and select standardized medical concepts from the OMOP vocabulary to define exposures and outcomes.
• Study Packaging: Export entire study designs as executable JSON packages for distribution to remote data partners.

ACHILLES performs automated descriptive analysis of an OMOP-mapped database to generate a comprehensive profile of its contents and quality. It creates a visual dashboard of data density, population demographics, and adherence to the CDM specification.

• Data Quality Assessment: Runs over 3,300 automated checks against the Data Quality Dashboard (DQD) to identify missing values, implausible dates, and invalid concept codes.
• Heel Rules: Executes a specific set of checks focused on healthcare logic, such as verifying that male patients do not have female-specific procedures.
• Data Source Catalogue: Enables a federated network to understand the characteristics of each site's data without exposing patient-level records.

HADES (Health Analytics Data-to-Evidence Suite) is the collective set of open-source R packages that power the statistical analysis in OHDSI. It provides a standardized, reproducible pipeline from hypothesis to publication.

• CohortMethod: Implements advanced comparative effect estimation using propensity score matching, stratification, and weighting to minimize confounding.
• SelfControlledCaseSeries: Analyzes the association between transient exposures and acute outcomes using a patient-as-their-own-control design.
• PatientLevelPrediction: Builds and validates machine learning models for diagnostic and prognostic prediction on OMOP data.

ARES (Automated Regulatory and Evidence System) is a next-generation data quality and characterization tool that provides a more interactive and scalable interface than ACHILLES. It is designed for rigorous regulatory-grade evidence generation.

• DQD Integration: Directly embeds the Data Quality Dashboard to flag plausibility, conformance, and completeness issues with drill-down capabilities.
• Characterization Results: Generates detailed population-level summary statistics for any defined cohort, including index date misalignment checks.
• Network Studies: Supports the execution of standardized study packages across a federated network, with results aggregated and visualized centrally.

The OHDSI vocabulary infrastructure is the semantic backbone that enables interoperability. It maps over 8 million source codes from disparate health systems into a unified set of Standard Concepts, resolving terminological chaos.

• Core Terminologies: Integrates SNOMED CT for conditions, LOINC for measurements, RxNorm for drugs, and CPT-4/ICD-10-PCS for procedures.
• Concept Relationships: Defines hierarchical, lateral, and mapping relationships between concepts to support advanced querying and ancestor/descendant logic.
• Vocabulary Versioning: Maintains a strict version history, allowing studies to be reproduced with the exact vocabulary context used at the time of analysis.

The foundational open-source data standard that transforms heterogeneous clinical data into a consistent, standardized format. The OMOP CDM defines standardized tables for clinical data, standardized vocabularies (SNOMED, RxNorm, LOINC) for semantic interoperability, and standardized derived elements for analytics. This harmonization enables the systematic execution of distributed research studies across disparate electronic health record systems without requiring data centralization.

A web-based interface for designing computable phenotypes—machine-executable algorithms that identify patient cohorts from observational data. ATLAS allows researchers to define inclusion criteria using structured codes, laboratory results, and medication orders through a drag-and-drop interface. The tool generates standardized JSON cohort definitions that can be shared and executed consistently across any OMOP-mapped database, ensuring reproducible research.

An automated profiling tool that performs data quality assessment and descriptive statistics on OMOP-mapped databases. ACHILLES executes over 170 pre-computed analyses, generating visualizations of population distributions, data density, and vocabulary conformance. This allows researchers to rapidly evaluate the fitness-for-use of a clinical database before conducting observational studies, identifying missing data patterns and coding inconsistencies.

The Health Analytics Data-to-Evidence Suite is a collection of R packages providing a comprehensive statistical toolkit for observational research on OMOP data. HADES includes modules for:

OHDSI maintains a unified vocabulary system that maps over 100 source coding systems into a common semantic framework. Key terminologies include SNOMED CT for clinical findings and procedures, RxNorm for medications, LOINC for laboratory measurements, and ICD-10-CM for billing diagnoses. This vocabulary layer is the critical infrastructure that enables cross-institutional query federation by resolving semantic heterogeneity between different health systems' local coding practices.

OHDSI operates as a federated network where data partners maintain physical control of their patient-level data. Research is conducted via a distributed query engine that sends standardized analytical code to each site, executes it locally behind institutional firewalls, and returns only aggregated summary statistics or contingency tables. This architecture satisfies regulatory requirements under HIPAA and GDPR while enabling large-scale, multi-national observational studies across hundreds of millions of patient records.