The goal of the Data Quality Dashboard (DQD) project is to design and develop an open-source tool to expose and evaluate observational data quality.
This package will run a series of data quality checks against an OMOP CDM instance (currently supports v5.3.1 and v5.2.2). It systematically runs the checks, evaluates the checks against some pre-specified threshold, and then communicates what was done in a transparent and easily understandable way.
The quality checks were organized according to the Kahn Framework1 which uses a system of categories and contexts that represent stratgies for assessing data quality. For an introduction to the kahn framework please click here.
Using this framework, the Data Quality Dashboard takes a systematic-based approach to running data quality checks. Instead of writing thousands of individual checks, we use “data quality check types”. These “check types” are more general, parameterized data quality checks into which OMOP tables, fields, and concepts can be substituted to represent a singular data quality idea. For example, one check type might be written as
The number and percent of records with a value in the cdmFieldName field of the cdmTableName table less than plausibleValueLow.
This would be considered an atemporal plausibility verification check because we are looking for implausibly low values in some field based on internal knowledge. We can use this check type to substitute in values for cdmFieldName, cdmTableName, and plausibleValueLow to create a unique data quality check. If we apply it to PERSON.YEAR_OF_BIRTH here is how that might look:
The number and percent of records with a value in the year_of_birth field of the PERSON table less than 1850.
And, since it is parameterized, we can similarly apply it to DRUG_EXPOSURE.days_supply:
The number and percent of records with a value in the days_supply field of the DRUG_EXPOSURE table less than 0.
Version 1 of the tool includes 20 different check types organized into Kahn contexts and categories. Additionally, each data quality check type is considered either a table check, field check, or concept-level check. Table-level checks are those evaluating the table at a high-level without reference to individual fields, or those that span multiple event tables. These include checks making sure required tables are present or that at least some of the people in the PERSON table have records in the event tables. Field-level checks are those related to specific fields in a table. The majority of the check types in version 1 are field-level checks. These include checks evaluating primary key relationship and those investigating if the concepts in a field conform to the specified domain. Concept-level checks are related to individual concepts. These include checks looking for gender-specific concepts in persons of the wrong gender and plausible values for measurement-unit pairs. For a detailed description and definition of each check type please click here.
After systematically applying the 20 check types to an OMOP CDM version approximately 3,351 individual data quality checks are resolved, run against the database, and evaluated based on a pre-specified threshold. The R package then creates a json object that is read into an RShiny application to view the results.
DataQualityDashboard is licensed under Apache License 2.0
This project is supported in part through the National Science Foundation grant IIS 1251151.
1 Kahn, M.G., et al., A Harmonized Data Quality Assessment Terminology and Framework for the Secondary Use of Electronic Health Record Data. EGEMS (Wash DC), 2016. 4(1): p. 1244. ↩︎