The Data Quality Dashboard
stops at the site boundary.

What the Data Quality Dashboard actually checks

The OHDSI Data Quality Dashboard is the standard tool for evaluating OMOP data quality at a site. It runs a systematic set of checks against a single OMOP instance and evaluates data across three categories:

• Completeness: whether expected data fields are populated and whether expected records are present for the patient population
• Conformance: whether data values follow OMOP conventions, including correct concept ID usage, value domains, and relational integrity
• Plausibility: whether values are clinically plausible given what else is known about a patient, including temporal plausibility and atemporal distributional checks

These checks are rigorous and valuable. A site that passes DQD has data that is structurally sound, correctly mapped, and internally plausible. That is a meaningful quality standard.

It is also entirely within a single site. The DQD was designed to evaluate one OMOP instance. It has no mechanism for comparing how that site's data relates to any other site's data.

What DQD does not check

The DQD does not evaluate whether two sites that both pass their individual DQD checks are semantically equivalent. It cannot, because it only sees one site at a time. The questions it does not answer include:

• Are the patients coded under concept X at Site A clinically similar to those coded under concept X at Site B?
• Do sites apply the same diagnostic confirmation thresholds before assigning a diagnosis code?
• Are the source codes that map to a given OMOP concept distributed similarly across sites, or do different sites rely on different underlying codes?
• Do co-occurrence patterns suggest that sites are capturing the same clinical population?
• Are treatment patterns and specialty involvement consistent, or do they reflect systematically different patient populations?

None of these questions have answers in a single-site DQD report. For a study that operates within a single site, that is fine. For a federated study that combines data across sites, these are the questions that determine whether the combination is valid.

DQD vs Vercori: what each tool addresses

Check	DQD	Vercori
Structural completeness within a site	✓ Yes	✓ Inherits
Value conformance to OMOP conventions	✓ Yes	✓ Inherits
Temporal and atemporal plausibility	✓ Yes	✓ Inherits
Cross-site semantic consistency	— Out of scope	★ Core function
Diagnostic confirmation variation	— Out of scope	★ Detected
Source code distribution comparison	— Out of scope	★ Measured
Cross-site co-occurrence analysis	— Out of scope	★ Measured
Submission-ready consistency documentation	— Out of scope	★ Produced
Tamper-evident reviewer audit log	— Out of scope	★ Produced

Why this distinction matters for federated study design

In a single-site study, DQD is sufficient as a data quality evaluation. The study operates within one institution's data, and the within-site quality questions are the right ones to ask.

In a federated study, data from multiple sites is combined. The validity of that combination depends on whether sites are semantically equivalent on the concepts being analyzed. That is not a question the DQD was designed to answer, and running DQD at each site and reporting that all sites passed does not address it.

The gap matters more as the number of sites and the clinical complexity of the study increase. A two-site study with well-characterized sites on a simple concept may have manageable cross-site risk. A ten-site study across diverse institution types on complex chronic conditions has substantial cross-site semantic risk that requires explicit evaluation and documentation.

Using both tools together

DQD and Vercori are designed to be complementary. DQD should run first, establishing that each site's data is structurally sound before cross-site comparison begins. Vercori is then designed to evaluate whether the structurally sound data across sites is semantically consistent.

Together, they are built to answer the full data quality question for a federated study: is each site's data correct within itself, and are the sites measuring the same clinical reality across the network?

The output of both evaluations should be documented and available for regulatory review. For submissions where real-world data quality is a material question, the combination of within-site structural quality documentation and cross-site semantic consistency documentation is the complete answer.