GENEVA OS

GENEVA OS® (Generative Evidence Acceleration Operating System) is a data infrastructure platform developed by Atropos Health, described as the first operating system for healthcare evidence generation. It is designed to convert healthcare data into personalized, publication-grade and regulatory-grade evidence rapidly, accurately, and transparently. The platform is intended for health systems, life sciences organizations, clinicians, and researchers who need to generate real-world evidence from their own data without moving or exposing sensitive patient information.

GENEVA OS installs directly within a customer's existing cloud environment, behind their firewall, and operates in a federated model — meaning no data transfer is required. It supports cloud-independent deployment and is compatible with existing cloud infrastructure, with the goal of reducing data management costs while maintaining security and data possession.

Core Architecture and Data Model

• Converts medical data into an in-memory database that stores patient timeline objects, representing the full longitudinal record of each patient from birth to death where available.
• Uses a flat data model, eliminating the need to join objects as in traditional relational databases.
• Patient timeline objects enable temporal queries that are 50x faster and 30x cheaper compared to conventional approaches.
• Overall query run time is reported as 273x faster than standard methods.
• Installs 100% behind the customer's firewall, with sensitive patient data remaining in place.

Temporal Query Language (TQL) and Advanced Cohort Engine (ACE)

• GENEVA OS includes a proprietary domain-specific language called TQL (Temporal Query Language), purpose-built for searching large medical datasets for patient features occurring in a specific sequence or timeframe.
• TQL queries are shorter, faster to generate, and faster to run than equivalent SQL queries.
• TQL uses knowledge graphs so that medical concepts can be queried directly, without requiring specific knowledge of the underlying dataset structure.
• The same TQL queries can be run across multiple similarly structured datasets simultaneously.
• TQL commands are designed to mimic the way physicians think, making it straightforward to translate clinical questions in conversational format into executable queries.
• The Advanced Cohort Engine (ACE) is the component that uses TQL to search and extract patient data from longitudinal patient-oriented records.
• ACE was published in March 2021 (Callahan A, Polony V, Posada J, Banda J, Gombar S, Shah N. PMID: 33712854).

Federated Nodal Deidentification and Privacy

• GENEVA OS supports federated multi-nodal queries, allowing cross-node queries without requiring data transfer between nodes.
• Includes safe harbor encoding and nodal patient deidentification with query time interval encoding.
• Members of the Atropos Evidence™ Network can access longitudinal patient data from participating network sources, producing deidentified patient records that include death information, while data holders retain possession and security of their data.
• The federated architecture enables clinicians and researchers to obtain a more complete view of patient health history with each query.

Analytics Pipeline and Study Design

• Query outputs from TQL and ACE feed into an automated analytics pipeline that generates user-ready reports and key study artifacts in seconds.
• Supports templated observational study designs and advanced statistical methods.
• Designed to produce both publication-grade and regulatory-grade studies.
• Supports multiple user types and experience levels, enabling high-quality evidence generation across varied workflows.

GENEVA OS is the foundational data layer underlying Atropos Health's broader portfolio of evidence-generation tools, including ChatRWD®, Green Button®, Forge®, and the Atropos Evidence™ Agent. It operates as part of the Atropos Evidence™ Network and integrates with the Alexandria® medical evidence content library. Deployment is cloud-independent and federated, with installation managed in coordination with the customer's existing cloud provider.