OMOP established an open-source library of 10 Health Outcomes of Interest (HOI) definitions for use in observational studies. These ten HOIs are a subset of all conditions that are of importance due to their historical associations with drug toxicities, their medical significance, and/or public health implications. There is little consensus for best practice in defining HOIs in observational databases, as observational studies for the same outcome often use different definitions. In addition, data sources may vary in available data elements that can be used for definition (e.g., labs). Even in the context of available data elements, the use of many outcomes are based on limited support in validity and reliability information.
A summary is available of the operational definitions of the HOIs under study within OMOP. Where there was lack of consensus in the appropriate definition for an outcome, multiple alternative definitions were created; therefore an HOI may have more than one definition. The alternative definitions cover inclusive (broad) and specific (narrow) definitions, and make use of combinations of diagnoses with diagnostic procedures, therapeutic procedures, lab tests
and lab results.
HOI Library - click on each HOI to review systematic reviews of the literature, definitions, and programming code:
- Aplastic Anemia
- Acute Liver Injury
- GI Ulcer Hospitalization
- Hip Fracture
- Acute Myocardial Infarction
- Mortality after Myocardial Infarction
- Acute Renal Failure
The majority of the HOI definitions can be implemented with Regularized Identification of Cohorts (RICO). RICO is a procedure that standardizes patient cohort selection. To create HOIs, RICO is employed using cohort definitions are created using criteria that are specified in input parameters. Patients meeting the criteria are selected from the common data model. The majority of HOI definitions can be implemented using RICO. Two exceptions are the "Acute Myocardial Infarction Definition #4" and the "Mortality after MI Definition #4", which both involve application of the American College of Cardiology definition that includes a complicated set of criteria about troponin, CK, and EKG results. A stand-alone SAS procedure for identifying the AMI#4 cohort was developed, but is untested for lack of relevant EKG data in the central databases within the OMOP Research Lab. However, the SAS code is available to the broader OMOP research community to adapt to their data, at http://omop.fnih.org/PrepData. There is no SQL version of this script.
Another exception is Hip Fracture Definition #3, which involves text evidence from radiology reports which is not available for development and testing within the Research Lab. An implementation of this definition should be undertaken if the data source supports this kind of
Regularized Identification of Cohorts (RICO) - ProSanos Corporation
HOI process being conducted within the OMOP Project Phases
a. Perform systematic literature review for each Health Outcome of Interest
- Identify clinical diagnostic criteria and coding guidelines
- Extract operational definitions previously used in observational studies
- Describe any validation studies that measure case ascertainment performance (including sensitivity, specificity, positive predictive value)
- Synthesize national estimates of prevalence of disease (if available)
b. Draft HOI design document characterizing the pool of potential definitions to consider in observational data
- Submit HOI definition design document for public comment and review by Scientific Advisory Board
- Elicit feedback from experts in specific clinical domains
a. Develop software code to apply definitions to Research Core databases (on raw data and common data model)
b. Compare the prevalence of HOI in each observational data source with a national estimate or other published data
c. Conduct descriptive analysis of the ‘natural history’ of the HOI (including demographics, prior conditions, prior drug utilization) in each source
d. Validate HOIs in the data source if available
a. Integrate summary of OMOP analyses for each HOI
b. Draft recommended best practices for HOI definition and solicit public feedback
c. Publish standardized HOI summary document
The Implementation of the HOI Definition Process in OMOP
A Request for Proposal was disseminated and members of the OMOP research team, Scientific Advisory Board, and Executive Committee selected two independent research organizations after careful evaluation of proposals. Both organizations selected had extensive experience in systematic reviews of the literature to inform meta-analysis, guideline development, and evidence-based medicine reviews.
The OMOP research team defined the following process to be independently followed by the two HOI research organizations:
- Develop optimal search strategy to identify published manuscripts of studies of an HOI in observational datasets
- Identify the relevant literature of studies conducted in observational databases that would inform our definition of an HOI.
These were to include any studies reporting definitions, validation studies that measure case ascertainment performance (including sensitivity, specificity, positive predictive value), coding guidelines and clinical diagnostic guidelines
- For each paper, summarize results in an evidence table to help inform the final definition to be implemented in OMOP studies
No communication between the two HOI research organizations was permitted. The research organizations were also asked to identify and abstract clinical guidelines for a given HOI to help further inform the OMOP research team.
The OMOP research team, in collaboration with the two research organizations, developed the evidence table format. The OMOP researchers evaluated actual search strategies and the articles retrieved from each research team as the effort progressed to identify and correct obvious shortcomings in search strategy or results based mainly on relevant citations that were identified by the OMOP researchers but were not captured in the research organization searches. These article citations were provided to the research organization to help them identify gaps in their search strategies.
An analysis of the systematic reviews was conducted to assess the concordance between the reviews and to develop a composite search strategy that took advantage of the positive attributes and capture of each review. The goal of the composite search strategy to increase the capture of relevant observational database and validation studies and was based on the findings from the independent research organizations. The composite search strategy was implemented in PubMed, and all returned articles were abstracted to determine the relevance based on the original criteria provided to the independent reviewers. Studies conducted outside an observational database or that did not include a specific definition were intended to be excluded from the evidence tables. Relevant articles identified by designated members of the OMOP research team using the composite search strategy were compared to those studies identified by each of the two independent reviews. Comparisons between the research organization findings and the composite OMOP research team effort were undertaken to determine the effectiveness of the composite search strategy and the extent to which the findings between the two research organizations agreed.