A Robust, Reproducible Method For Evaluating The Suitability of Disparate Observational Databases for Pooled Analysis, Using The Omop Common Data Model

BACKGROUND: Data pooling - integration of patient-level data from different databases -- is used to increase sample size where individual databases are too small. Differences in data format and content, along with population heterogeneity, require careful evaluation to identify differences that may...

Full description

Saved in:
Bibliographic Details
Published in:Value in health 2017-10, Vol.20 (9), p.A776
Main Authors: Reisinger, SJ, McDonald, L, Carroll, R, O'Hara, DJ, Anstatt, D, Ramagopalan, S
Format: Article
Language:English
Subjects:
Analysis
Appropriateness
Clinical outcomes
Experts
Informatics
Mapping
Medical databases
Patients
Population characteristics
Public domain
Suitability
Transformation
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:BACKGROUND: Data pooling - integration of patient-level data from different databases -- is used to increase sample size where individual databases are too small. Differences in data format and content, along with population heterogeneity, require careful evaluation to identify differences that may interfere with interpretation of results of subsequent pooled analyses. Evaluation for pooling can be difficult, time-consuming, and is not reproducible across databases. Use of a Common Data Model (CDM) provides opportunities for development of efficient, reproducible methods to evaluate appropriateness of pooling across a wide variety of databases. METHODS: We developed a method, utilizing the Observational Medical Outcomes Partnership (OMOP) CDM to efficiently assess format, content, and appropriateness of disparate databases for pooling. It takes advantage of existing OMOP data transformation processes, documentation and programs available in the public domain through the Observation Heath Data Sciences and Informatics (OHDSI) Collaborative. Method summary: MAPPING: map databases into OMOP format. For many databases, mapping documentation is available through OHDSI. TRANSFORMATION: transform databases into OMOP format using existing OHDSI utilities or developing de-novo. ANALYSIS: OHDSI ACHILLES reports enable interactive exploration of patient and data characteristics for any OMOP-format database. EVALUATION: review by clinical experts to identify differences that may interfere with pooled results interpretation. PRACTICAL IMPLICATIONS: By design, OMOP data transformation normalizes data format and identifies content quality issues. OHDSI programs provide meaningfully comparable population characteristics, including diagnostic and treatment patterns, which can aid in understanding heterogeneity across disparate databases. De-duplication, if necessary, is facilitated by standardized-format data. CONCLUSIONS: Use of a CDM facilitates systematic evaluation of individual datasets for pooled analyses. This is especially useful for pooling data from different countries, where coding systems and practice patterns may vary. An additional benefit is that CDM-format data can be easily integrated / used for subsequent pooled analyses, assuming pooling is appropriate.
ISSN:1098-3015
1524-4733
DOI:10.1016/j.jval.2017.08.2239