Category: Data & AI Infrastructure
Analyses that inform internal strategy get by on convenience data — but evidence meant for payers, HTA bodies, or regulators must survive methodological scrutiny it was often never designed for.
There is a hard line between real-world analysis that informs a slide deck and real-world evidence that withstands external review. Payers, HTA bodies, and regulators increasingly accept real-world evidence — but on their terms: pre-specified protocols, fit-for-purpose data with documented provenance and quality, transparent methods, and reproducible results. Studies assembled on convenience data with post-hoc definitions routinely fail that bar. Evidence-grade platforms and services exist to close the gap, pairing curated data whose quality is documented with the methodological rigor external reviewers demand.
The difference is mostly decided before any analysis runs. Decision-grade work starts from a protocol that pre-specifies the question, cohort definitions, outcomes, and statistical approach; it uses data whose fitness for the purpose is documented — provenance, completeness for the required variables, validation of key definitions; and it is executed so the result can be reproduced, with data versions and code under control. Platforms supporting this work contribute on both fronts: data assets with published quality characterization, and tooling that enforces protocol discipline — versioned cohort definitions, audit trails from raw data to final estimate, and documentation generated as a byproduct of the workflow rather than reconstructed afterward. What no platform can supply is the judgment about whether real-world evidence fits the decision at all; some questions still require a trial.
Ask for evidence about the evidence: whether studies built on the platform have actually been used in submissions to payers, HTA bodies, or regulators, and what the reviewers' data-quality questions were. Ask how the platform documents data fitness for a given protocol — the credible answer is a per-study assessment against the variables and follow-up your protocol requires, not a generic quality certificate. On the tooling side, test reproducibility concretely: can the platform re-execute a study against the exact data version used originally, and does the audit trail connect every published number back through code and cohort logic to source data? If reproduction requires heroics, the platform documents work; it doesn't do it.
A sensible first project is one where the evidence need is real but the stakes allow learning — commonly an external comparator analysis or a payer-facing burden-of-illness study rather than a registrational use. Running it under full protocol discipline, with the target audience's evidentiary standards explicitly in view, surfaces the gaps — in data fitness, in documentation, in internal process — while they are still correctable, and produces a template the organization can reuse when the higher-stakes need arrives.
Increasingly yes, for the right questions — natural history, external comparators, safety characterization, and support for effectiveness in specific circumstances — and formal guidance now exists in major jurisdictions. Acceptance turns on data fitness and methodological transparency for the specific use, never on real-world evidence as a category.
Recurring themes in external review are data that cannot support the required variables or follow-up, definitions chosen after seeing the data, unaddressed confounding, and results that cannot be reproduced or traced. Most of these are decided by study design and data selection — before analysis, not during it.
Sometimes, but fitness is judged per question: a dataset adequate for hypothesis generation may lack the validated outcomes or continuous observation a comparative-effectiveness claim requires. Decision-grade work starts by assessing the data against the protocol's specific demands, not by assuming yesterday's dataset transfers.
The capabilities stack rather than compete: platforms contribute characterized data and reproducibility tooling, research organizations contribute methodology and regulatory experience, and internal teams own the question and the strategy. What matters is that someone accountable holds the full chain — question, protocol, data fitness, execution, documentation — to the standard of the decision it serves.
Panalgo, a Norstella company, launched LinQNotes and Ella AI to speed real-world data analysis that supports market access and commercial decision-making.
Vendor: Norstella · press release (Norstella) · 2025-05-14