Real-world data can strengthen oncology research, regulatory decisions, and clinical care. However, routine EHR information must undergo rigorous standardization, validation, and quality assessment before it can support reliable evidence. In this interview, Jessica Paulus, ScD, Vice President of Real-World Research at Ontada discusses applying Food & Drug Administration (FDA) and European Medicines Agency (EMA) data quality frameworks to a large oncology EHR-derived platform, exploring how continuous quality monitoring can help ensure that real-world data are fit for purpose.

Moving Quality From Subjectivity to Measurement

The FDA’s Q-CARD initiative and the EMA’s Data Quality Framework (DQF) reflect a broader shift toward measurable and reproducible approaches to data quality. Rather than evaluating data only at the end of an individual study, the frameworks encourage organizations to build quality processes into the underlying data platform.

“These frameworks really ask us to move quality firmly away from subjectivity and towards a very reproducible and measurable process,” Dr. Paulus said. They also “compel us to think about data quality, not as a one-study-at-a-time exercise, but more about an enterprise or a platform approach to think about study quality.”

Operationalizing those principles across oncology platforms is challenging because the data lifecycle may include structured EHR fields, physician notes, chart abstraction, artificial intelligence and natural language processing (NLP) outputs, claims, and mortality records. Each source brings distinct requirements for validation and interpretation.

For that reason, organizations must evaluate how data are collected, standardized, transformed, and maintained—not simply check variables at the conclusion of a study. “We have to continuously reassess the quality,” Dr. Paulus emphasized, because “care is changing so rapidly and information systems are changing so rapidly.”

Applying the Frameworks to ON.Genuity

Ontada applied the FDA and EMA principles to its ON.Genuity oncology real-world data platform, which includes information from approximately 4 million patients across more than 700 community oncology clinics. The platform incorporates structured EHR data, chart abstraction, NLP-derived variables, and linked external sources.

Rather than applying the frameworks to one dataset or study, the team applied them to the data-generating platform itself. This included standardizing more than 250 variables using fast healthcare interoperability resources (FHIR) and minimal common oncology data elements (mCODE) principles, maintaining metadata and lineage documentation, and validating clinically important concepts extracted from notes through NLP.

The team also externally validated overall survival using National Death Index mortality data in patients with metastatic pancreatic cancer. Because mortality is high in this population, the cohort provided a substantial number of events for assessing concordance between EHR-derived survival and an external mortality source.

The findings were encouraging. Data completeness exceeded 90% across important domains, including biomarkers, disease characteristics, and medication history. Of 26 standard edit checks, 21 identified no issues, while five found discrepancies affecting up to 1% of patients. Dr. Paulus noted that finding discrepancies is not necessarily a sign of failure: “It’s actually what’s supposed to happen. It’s evidence that the quality monitoring is working.”

Addressing these findings requires processes capable of detecting, identifying, and remediating problems. Logical conflicts, such as differing dates for metastatic disease across sources, may require clinical expertise rather than an automated correction. “Oftentimes, these can’t just be resolved algorithmically,” she said.

Maintaining Fitness for Purpose

High completeness alone does not determine whether data are appropriate for a particular research question. Instead, organizations must evaluate whether the variables most relevant to a specific decision are sufficiently complete and reliable.

“Incompleteness isn’t necessarily a guilty verdict about a data source,” Dr. Paulus explained. “You have to really consider what level of completeness or quality, for example, is needed for the decision at hand.” This concept of fitness for purpose is especially important as data are used across exploratory research, clinical decision making, payer questions, and regulatory submissions.

Ultimately, continuous quality governance can help real-world data complement clinical trials by providing evidence about patients who may not meet trial eligibility criteria because of age, comorbidities, or clinical complexity. “This data quality piece is so important to realize the potential of real-world data to bridge that gap for clinicians,” Dr. Paulus said.

Conclusion

Operationalizing FDA and EMA quality frameworks requires more than a final review of study variables. It calls for an enterprise-wide operating model that incorporates standardization, traceability, validation, discrepancy resolution, and continuous reassessment.

“There has been such incredible surge in enthusiasm and interest in real-world data, but to date, there really hasn't been strong enough evidence for real-world data to have that chair at the very front of the table,” Dr. Paulus concluded. “The work that we've talked about today is an incredibly critical point to try to bridge that gap, to move from enthusiasm to establishing a standard and showing how we've met the standard, and that's going to allow us to really lean on these data to inform those decisions.”

Key Takeaways

Routine EHR data must undergo standardization, validation, and traceability processes before supporting high-stakes research or regulatory decisions.
FDA and EMA frameworks encourage platform-level, measurable, and reproducible approaches to data quality.
Identifying discrepancies demonstrates that quality monitoring is functioning and enables targeted remediation.
Data quality should be judged according to fitness for purpose and the requirements of the decision being supported.
Continuously updated, high-quality real-world data can complement clinical trials by representing broader oncology patient populations.

About Dr. Paulus

Jessica Paulus, ScD, is the Vice President of Real-World Research at Ontada. She leads a team of scientists with expertise in leveraging real-world data networks to meaningfully inform clinical and regulatory decision making, and she provides scientific oversight to study design, analytic approaches, and dissemination strategies for real-world data–based studies.

Bridging the Regulatory Gap: Operationalizing Quality Frameworks in Oncology Real-World Data