A primary goal of the Observational Medical Outcomes Partnership is to assess the feasibility and utility of using observational data to identify and evaluate associations between drugs and health-related conditions.

The specific aims supporting this goal are:

• Define the pool of potential analytical methods to be used to identify drug-condition associations in observational data
• Assess the feasibility and utility of observational data for identifying associations between drugs and non-specified conditions
• Assess the feasibility and utility of observational data for monitoring Health Outcomes of Interest and evaluating the associations between drugs and Health Outcomes of Interest
• Assess the utility of natural history information generated from observational data in contributing to interpretation of observational analyses
• Assess how decision-makers can use observational analyses to support active monitoring of the effects of medicines post-approval

OMOP will realize this goal by achieving the following key objectives:

1. Define a representative pool of analytical methods that can be used to identify associations (hypothesis generation or strengthening) between drugs and health-related conditions, and test the usefulness and performance characteristics of these methods against known associations, in different types of observational data (administrative claims, inpatient and outpatient electronic health records) as well as specific databases within each type. Evaluation (hypothesis testing) of the associations is a secondary goal but not the prime focus of the OMOP effort.

2. Develop a set of tools for organizing access to and evaluating the usefulness of multiple observational datasets that will facilitate testing of these analytical approaches by research teams directly sponsored and funded by OMOP as well as via open collaboration with others in the research community.

3. Apply these tools and perform these analyses against a limited set of specific observational data sources that a) provide an immediate, practical test of the research objectives and b) represent both of the major types of currently usable observational data: administrative claims and electronic (provider-based) medical records.

4. For studies evaluating the performance of methods for monitoring Health Outcomes of Interest (HOI), target the analyses at a defined set of HOIs that have been chosen based on their recognized relevance and utility to the goals of OMOP, as derived from the Designated Medical Event (DME) list and qualified through literature searches and scientific consensus (already largely completed). These HOIs are constituted of specific pairs of drugs and health-related outcomes, focusing primarily on safety outcomes, but also including health benefits.

5. Study how information from analyses can be successfully used as part of the ongoing pharmacovigilance process and publish findings regarding their effective use in decision-making processes.

6. Create criteria that may assist in selecting databases most likely to be useful for drug safety surveillance, examine the spectrum of available data against those criteria, and choose at least some of the data for the studies above, based on these.

General Approach: Transparency in research; collaboration across stakeholders and sites

A key challenge of OMOP is supporting experiments that are as broadly useful as possible to overall efforts to improve drug safety surveillance and to the pharmacoepidemiology community, while still capable of being conducted within the limited budget and timeframe set for the OMOP pilot. To meet this challenge of being both efficient and collaborative, OMOP will establish both a Research Core and an Extended Research Consortium to conduct the research.

The OMOP Research Core.
This effort will be led by the three OMOP Principal Investigators, representing the pharmaceutical industry, academia, and FDA and assisted by a small research staff sourced from OMOP stakeholder organizations or retained on a consulting basis by OMOP. The Research Core will design, develop, and directly manage the execution of a set of research experiments against the core data sources referred to in objective #3 above, and broadly share findings and publish results as specified in the OMOP Publication Policy. While the Research Core may perform certain tasks directly themselves, respected research institutions in the field, including academic, government, and for-profit entities, under paid contracts or specific collaboration agreements with OMOP, and under the direction of the Research Core PIs, will conduct some of the work. These institutions will be chosen to execute specific pieces of the OMOP research protocols (experiments, analyses) via an open, competitive application and award process managed by the PIs and the OMOP Executive Director in compliance with the OMOP Grants and Contracts policy. All research materials from the Research Core will be made available to the public through open website and peer-reviewed publications. The activities of the Research Core will be managed by the PIs, with the support of the OMOP Executive Director and staff, and under review and oversight of the OMOP Advisory Boards.

Extended Research Consortium
OMOP will also encourage the participation of the broader scientific community in achieving its objectives. The research protocols, data models, database evaluation and quality assurance tools, analytical programs and findings generated by OMOP will be published or made publicly available as early as is feasible, allowing other researchers who are not part of or funded by the Research Core the opportunity to run specific OMOP protocols or variants on their own databases, develop parallel or complementary tools and approaches, and publish their results. OMOP will encourage sharing of results and tools developed in this way for the public benefit as a matter of course; external institutions whose aims are in alignment with OMOP’s, who can credibly and objectively perform relevant research, and who are willing to share tools, approaches, findings and other intellectual property developed as a result may be additionally recognized as members of an OMOP Extended Research Consortium. Consortium members may benefit from participation in joint publications, cross-citations and invitation to an annual Research Symposium, sponsored by OMOP, where members of the OMOP Research Core and Extended Consortium can share and discuss results. OMOP will seek input from the scientific community and Consortium members themselves on its structure and operation on an ongoing basis.

OVERVIEW OF PHASES OF RESEARCH PROGRAM

To address these aims, we will answer the following research questions, which are separated into phases to denote key project milestones:

Phase 1: Feasibility of data infrastructure

Research questions:
• Can we establish a consistent framework to use across disparate observational data sources?
• Does normalizing conditions in observational data improve identification of non-specified conditions?

Deliverables

• Common data model, and data transformation applications
• Common drug and condition vocabularies, and mapping tools
• Report: Comparison of condition vocabularies for observational screening
• Health Outcomes of Interest library
• Simulated dataset
• Report: Systems integration design and lessons learned

Phase 2: Feasibility of analyses

Research questions:
• Which identification methods are feasible within the current systems infrastructure?
• Can we establish standard data quality and characterization procedures to assess the viability of data sources for observational analyses?

Deliverables

• Report: Research Core Data quality summary
• Data quality assessment procedure
• Report: Feasibility of Identification methods
• Open-source library of method implementations
• Reference set of drug labeled events for screening studies
• Report: Health Outcomes of Interest natural history

Phase 3: Performance of Analyses

Research questions:
• What are the performance characteristics of each identification method in simulated data?
• What is the performance and consistency of each identification method for non-specified conditions across observational data sources and over time?
• What is the performance of each observational data source in identifying associations between drugs and non-specified conditions?
• What is the performance and consistency of each method in monitoring Health Outcomes of Interest?
• What is the performance of each observational data source in monitoring Health Outcomes of Interest?
• What is the performance of each observational data source in evaluating associations between drugs and Health Outcomes of Interest?

Deliverables:

• Open-source library of applications to conduct methodological research against common data model
• Report: Point-performance of identification methods in a simulated dataset
• Report: Point-performance of identification methods on non-specified conditions
• Report: Performance over time of identification methods on non-specified conditions
• Report: Consistency of identification methods on non-specified conditions
• Report: Concordance of identification methods and observational evaluation of Health Outcomes of Interest

Phase 4: Utility of Analyses and Process

Research questions:
• How does the performance of identifying associations in observational data differ from other surveillance approaches?
• How does natural history information from observational data contribute to a decision regarding the results of observational analysis?
• How do decision-makers interpret observational database analyses?

Deliverables

• Report: Utility of Natural History information
• Report: Efficiency of Identification: Comparison of observational data and other surveillance systems
• Report: Utility of observational screening and observational evaluation of Health Outcomes of Interest over time
• Report: Systems integration design and lessons learned
• Report: Partnership governance design and lessons learned

Additional detail is available about the following topics: