Which evidence risks should I test before selecting a real-world data source?
Population coverage and representativeness, observational breadth, linkage, provenance, and adaptability should be assessed as separate considerations. A composite score can obscure the difference between a limitation that depends on analytical assumptions and a structural gap that requires new evidence.
For each dimension, identify the risks associated with your research, and confirm whether your prospective sources preserve the necessary response pathways.
uMed’s five-test model consolidates recurring considerations across FDA, MHRA, EMA and NICE guidance and the Duke-Margolis framework for pre-commitment source selection. Ongoing fitness for purpose applies those considerations to the options available for extending the evidence base after source commitment.
How can I anticipate future evidence needs?
Evidence teams cannot predict every question that may emerge, but designing for maximum adaptability is neither necessary nor economical for every programme. Instead, teams should aim to identify the programme developments most likely to create new evidence requirements and establish whether the selected source could accommodate them.
A future evidence requirement should influence source selection when:
- It is plausible. Programme stage, anticipated regulatory and reimbursement decisions, emerging hypotheses and planned lifecycle evidence can indicate how the research may develop.
- It is decision-critical. Without the relevant variable, population or observation period, the evidence may not support interpretation, submission, reimbursement or the next stage of research.
- It would be difficult to address later. Meeting the requirement could involve a new provider, consent pathway, protocol, linkage agreement, recruitment programme or data-access negotiation.
A requirement that meets all three criteria should influence the source choice or evidence design. Requirements that can be addressed later at reasonable cost warrant less consideration in initial design.
Adaptability matters most in exploratory research, staged regulatory or reimbursement programmes, and studies likely to require additional measures or follow-up. A static source may be sufficient when the question, variables and observation period are unlikely to change.
For studies intended to support a marketing application, a 2025 review by FDA staff recommends early engagement with third-party real-world data providers. Provider assessment therefore forms part of research design where regulators may need to verify source records, cohort-selection logic, derivations and controls underlying the dataset.
Accordingly, due diligence should establish the necessary records, identifiers, permissions and operating capabilities. Contracts can then assign access, responsibility, lead time and cost to each material response.
Which evidence gaps can be addressed with data analysis?
Analytical methods can assess uncertainty within observed data under explicit assumptions. However, they cannot recover a variable that was never measured, supply data from an absent population, or reconstruct an audit trail that was not retained. Structural gaps such as these require extension of the evidence base or, where provenance has been lost, may not be repairable after commitment.
Feasibility of extension depends on the shape of source data and the evidence infrastructure behind it. For each dimension, the assessment should identify plausible limitations, the conditions that would make them material, and whether the selected source could support the required extension.
How can structural gaps in real-world data be addressed?
Structural gaps in real-world data can be addressed through additional collection, linkage or recruitment, if those options remain available after source selection. A newly relevant measure may be collected from an existing cohort; an absent population may require a new identification and recruitment pathway.
uMed’s evidence infrastructure supports both pathways through pre-consented, EMR-linked patient registries and clinically validated recruitment across care networks. QMUL DNA shows how an existing cohort supported a newly relevant biomarker measure; PERSONAL COVID-BP shows how recruitment was adapted when the original pathway proved insufficient.
QMUL DNA: adding a new data modality
The AccessPD QMUL DNA sub-study recruited from an existing pre-consented cohort for non-invasive, at-home saliva collection. Genomic results were linked with electronic medical records and electronic patient-reported outcome endpoints.
The study consented 335 patients, completed data collection in 11 weeks and achieved 86% adherence. The existing cohort allowed the study to add a new data modality while retaining its connection to longitudinal clinical and patient-reported evidence.
PERSONAL COVID-BP: adapting recruitment when the original pathway stalled
uMed was brought into the PERSONAL COVID-BP study after its established recruitment pathway had stalled. A new patient-identification and engagement programme reached the enrolment target in 21 days—five months ahead of schedule.
Of the 1,757 patients engaged, 213 were referred and 73% of those referred enrolled. The study demonstrates how an alternative recruitment pathway can address a population-reach limitation that becomes apparent during delivery.
Read the PERSONAL COVID-BP case study.
What makes a real-world data source decision-ready?
A decision-ready source fits the present purpose, makes its limitations explicit and leaves a viable response to the plausible changes that would be hardest to address later.
Before committing, the evidence team should be able to state:
- What the source supports now.
- Which programme developments could expose its boundaries.
- Which limitations can be managed analytically.
- Which require linkage, additional collection or new reach.
- Whether the required capability will remain available after commitment.
If a plausible, decision-critical gap cannot be addressed, it may be necessary to reevaluate the data source or evidence design.
Discuss the evidence gaps in your planned study
If your source assessment identifies an unmeasured outcome, an underrepresented population or a need for further follow-up, talk to uMed about your study. Share the research question, target population and evidence gap so the discussion can focus on whether additional collection, linkage or recruitment could address it.
‍

