A health system or laboratory can make its data discoverable for research without copying its archive to an outside party. The integration method should be chosen to fit three things: how the organization’s systems and teams already work, its compliance posture, and its IT team’s security requirements. The method follows the institution, not the other way around.

What does “discoverable without moving it” mean?

Discoverable means a qualified researcher can learn whether relevant data or specimens exist, at an aggregate level, before anything is released. Custody means the institution keeps control of the records and decides what leaves, to whom, and under what terms.

These two goals are compatible. A researcher’s first question is usually whether a cohort exists, not a request for patient-level records. That question can often be answered with a count or a description of available attributes. Patient-level release is a separate, later decision that goes through the institution’s own review.

What are the common ways to connect institutional data?

Most approaches fall into two broad models.

A federated model leaves the data where it is. A federated network sends questions to each site and returns governed results, such as counts, rather than copies of the underlying records. Each institution runs the query against its own systems and controls what comes back.

An extract model copies a defined subset of data to another environment. An extract is a one-time or recurring export of selected fields, usually de-identified, prepared for a specific use.

Many arrangements combine the two. Discovery runs federated, and an extract is prepared only after a specific request is approved. The Healthcare Partners page describes how Biobase approaches federation: the network carries queries and governed results, not copies of the archive.

How do federated and extract approaches compare?

Each model has trade-offs that an institution should weigh against its own constraints.

  • Custody. Federated access keeps records inside the institution. An extract places a copy elsewhere, so the receiving environment’s controls matter as much as the institution’s.
  • Freshness. A federated query reads current data. An extract reflects the data at the time it was produced.
  • Effort. A federated connection requires local infrastructure that can answer queries. An extract requires preparing, validating, and transferring files for each use.
  • Review. Both models need a defined review step before patient-level information is released. The difference is when that review happens and what it covers.

Neither model is correct for every institution. The right choice depends on the three factors below.

What should decide the integration method?

Organizational needs

Start with how systems and teams work today. Clinical data typically sits across several systems: the electronic health record (EHR), the laboratory information system (LIS), which tracks specimens and results, the picture archiving and communication system (PACS), which stores images, biobank inventories, and billing. Identify which of these hold the information researchers ask for and who maintains each one.

Data formats vary too. HL7 and FHIR are messaging standards for exchanging clinical data. Exports may also arrive as flat files, PDFs, or proprietary LIMS formats. An integration that requires a new data warehouse or a migration project places the burden on the institution. Biobase connects to the systems an institution already runs and normalizes varied formats into a shared semantic layer, as the Platform page describes.

Compliance posture

De-identification is the removal or transformation of information that could identify an individual. Under HIPAA, covered entities can de-identify data through the Safe Harbor method or through Expert Determination. HHS guidance on de-identification explains both.

Decide which method your institution requires and where it should run. Then define the review steps: who approves a release, what documentation is needed, and how institutional review board (IRB) approval or exemption is handled for a given study. An integration should enforce those steps, not depend on someone remembering them.

IT and security requirements

The IT team decides how external systems may connect. Its questions usually cover network access, authentication, encryption in transit and at rest, logging, and who can see what. Many institutions run a formal security review before any connection. The HIPAA Security Rule sets the baseline safeguards for electronic protected health information.

Bring the IT team in at the start. Their preferred connection pattern often determines which integration methods are practical.

What should an institution ask any research data partner?

These questions apply to any partner, including Biobase.

  1. Does patient-level data leave our environment during discovery, or only aggregate results?
  2. Who decides what is discoverable, to whom, and under what terms?
  3. Where does de-identification run, and which method does it use?
  4. Is every access logged, and can we review the audit trail?
  5. Can we trace the lineage of a released dataset back to its source records?
  6. Which steps require our approval before a release?
  7. What does our team need to build, maintain, or migrate?
  8. How are costs recovered, and how is value from the data shared?

Clear answers to these questions make a security and compliance review shorter. Vague answers are a reason to pause.

How does Biobase approach integration?

Biobase matches the integration method to the institution’s organizational needs, compliance posture, and IT team’s requirements. Its federated architecture is designed so the institution keeps custody and decides what is discoverable and under what terms. De-identification, lineage, and audit trails are enforced in the pipeline, so access is traceable and releases are reviewable.

Before any of that, an institution can start with a collection-level inventory. The collection preparation guide describes what to record. For how connected data moves through discovery, linking, and delivery, see from archive to research-ready cohort.