Living registries vs traditional registries: a different kind of evidence infrastructure

Living registries continuously ingest and update real-world clinical data. Traditional registries cannot. Here is what that difference means for pharma evidence.

A living registry is a registry that continuously ingests structured and unstructured clinical data, dynamically updates patient cohorts as new records arrive, and produces evidence that evolves alongside the population it observes. A traditional registry, by contrast, fixes its data schema before enrolment begins, captures a defined set of variables, and closes.

That distinction is not a matter of execution quality. It is architectural. Traditional disease registries were built to capture a moment in a disease’s history. Modern pharma evidence generation requires infrastructure that sustains an evidence engine across years – not a snapshot that answers one question and then sits still.

Why registries became a cornerstone of evidence generation

Disease registries earned their place in clinical research for a clear reason. As explored in how disease registries work in clinical research, they gave researchers something clinical trials structurally cannot provide: longitudinal, real-world patient data collected outside the constraints of a controlled protocol. A registry could follow patients across years, capture how a disease actually progresses in practice, and document how treatments perform once they reach the diversity of real clinical populations.

This mattered because trials answer narrow questions well and broad questions poorly. Regulators, payers, and clinicians needed a source of evidence that reflected the patients trials exclude and the outcomes trials do not have time to observe. Traditional registries filled that role with genuine scientific value – a well-run registry from the past two decades has informed treatment guidelines, safety monitoring, and reimbursement decisions that would not exist otherwise.

None of what follows disputes that contribution. The question this post asks is different: was the traditional registry model built for the pace, diversity, and regulatory demands that pharma evidence generation now requires? The answer has structural implications that go beyond any single registry’s execution.

The structural limits of the traditional model

Traditional registries are built around a fixed data schema, defined before enrolment begins and rarely revisited afterwards. That schema determines every variable the registry will ever capture. If a new research question emerges two years into a registry’s life and requires a variable the schema did not anticipate, the registry cannot answer it. The data was never collected.

Manual curation compounds this constraint. Coordinators enter what the schema asks for, from the records they can access, in the time available. The registry does not evolve as clinical practice evolves; it captures a point-in-time design and holds that design fixed for the life of the evidence generation effort.

An organisation running five distinct evidence programmes is not running one infrastructure five times more efficiently. It is running five separate infrastructures, each rebuilt from the ground up, each carrying its own fixed costs and its own multi-year lag before it produces anything.

This is not a critique of any individual registry’s design or execution. It is a description of what a fixed-schema, single-purpose architecture necessarily produces when the evidence questions pharma needs to answer keep changing faster than the registry can be rebuilt to answer them.

What ‘living’ actually means in a registry context

A living registry continuously ingests structured and unstructured clinical data, dynamically updates patient cohorts as new records arrive, and produces evidence that evolves alongside the population it observes.

Contrast that with the snapshot logic of a traditional registry: enrolment closes, the schema locks, and the dataset represents a fixed moment that grows stale the day after collection ends. A living registry has no such closing date. As clinical documentation is generated at participating sites, the registry ingests it, extracts the relevant clinical concepts, and updates the cohort in near real time. The registry a researcher queries in year three reflects the clinical reality of year three, not an assumption frozen in year one.

What makes continuous ingestion meaningful is a data layer built on clinical NLP capable of reading unstructured narrative and converting it into structured, auditable variables at the same pace the clinical record itself is generated. Savana cNLP (EHRead), trained exclusively on real EHRs across seven languages, is the extraction engine that makes this conversion regulatory-grade rather than approximate.

Evidence that scales across the pharma lifecycle

The distinction between a living registry and a traditional one is not that the former is a better version of the latter. It is a different kind of asset. A fixed-schema registry answers one evidence question and then requires a new build to answer the next. A living registry accumulates evidence that compounds across the entire pharma lifecycle.

Consider how that continuity actually plays out across pharma phases:

Regulatory-grade as a non-negotiable baseline

Dynamism can sound, to a sceptical reader, like a trade-off against rigour. It is worth addressing that concern directly.

A living registry must meet the same standards of traceability, reproducibility, and auditability that any registry intended for regulatory or HTA use is required to meet – arguably higher standards, given that continuous data ingestion introduces more points where governance could fail if it were not designed in from the outset. As detailed in what regulatory-grade RWE actually requires from FDA and EMA, evidence must be pre-specified, reproducible by an independent party, and traceable to its clinical source. None of these requirements relax because a registry updates continuously rather than at fixed intervals.

Dynamism and rigour are not in tension. Governance-by-design is precisely what makes continuous evidence generation compatible with regulatory scrutiny rather than a liability against it.

The network question: where does the data come from?

A living registry is only as good as the clinical network feeding it. Continuous ingestion and dynamic cohort updating are architectural properties; they require a source of continuous clinical data to operate on, and that source is a federated network of hospitals with harmonised data standards and privacy-compliant infrastructure already in place.

Without that network, the living registry concept remains a design specification rather than an operating layer. The Smart Health Alliance (SHA) is the activation layer that makes this concrete: a pre-validated federated network spanning 280+ institutions across 15 countries, connected through infrastructure that keeps patient data at its source while making its clinical content available for evidence generation. Privacy compliance and technical readiness are built in – centres do not need to be onboarded from scratch for each evidence objective.

The decision pharma needs to make

The registry conversation in this industry has been asking the wrong question for some time. The question has not been which registry vendor to select for a given evidence objective. It should be what kind of evidence operating layer an organisation is building for the next decade, because that choice compounds in ways a single-objective decision does not.

Pharma organisations that continue commissioning single-use, fixed-schema registries are not simply choosing a slower method for any individual evidence objective. They are accumulating what might be called evidence debt: each registry that cannot be reused, extended, or federated into a broader dataset represents years of clinical documentation that will need to be recreated – at full cost and full timeline – the next time a related question arises. That debt does not appear on a single evidence objective’s budget line. It appears in the cumulative cost of rebuilding, indefinitely, an infrastructure that a living design would only have to build once.

This is what Savana NGR was built to resolve: not a better version of the traditional registry, but the operating layer that treats evidence generation as a sustained capability rather than a series of disconnected data collection exercises. Whether an organisation builds that capability with Savana or elsewhere, the underlying decision is the same one this post has argued for throughout. It is an infrastructure decision, not a procurement decision, and it deserves to be evaluated as one.