Table 2.
Problem | Example | Potential Solutions |
---|---|---|
Data quality and validation | Selecting measurement of interest for a clinical trial when multiple measurements are available (e.g., laboratory data) Inaccurate information in EHRs Coding errors |
Specific parameters (e.g., using date or time windows) stated in protocol or operating procedures for extracting data from EHR into eCRF Use codes linked to reimbursement, which have greater likelihood of reliability Stakeholder collaboration to develop validation methodology Stakeholder collaboration to contribute data for EHR validation studies |
Complete data capture | Clinical endpoints SAEs Problematic in multiple-payer systems Death |
Develop standards for data sharing and privacy Explore linking EHRs to national death registries |
Heterogeneity among systems | Multiple different vendors within a given country or region Inconfigurable systems Lack of flexible architecture Lack of common data fields, data definitions, and difficulty with data mapping Incomplete data capture Missing fields of interest (i.e. relevant to some diseases but not others) Inability to link systems (i.e. different patient identifiers) |
Commit resources to harmonization efforts Form working group with representation from all stakeholders to develop consensus agreement on a common set of data variables to be included in all systems |
System knowledge | Inadequate understanding of database and its structure Researchers may not understand limitations of database |
Transparency Develop and maintain data standards and operations manuals Report strengths, limitations, and nuances of databases in primary manuscripts Informatics training for investigators |
EHR electronic health record, SAE serious adverse event