Skip to content
ushealthcloud

Life Sciences & Research

AI and compliant data infrastructure for medical research

Research-ready data and models across institutions and borders, with de-identification and GDPR-compliant data flows designed in from the first pipeline.

The problem

The data for your next study already exists, spread across sites and countries in formats that don’t match, and identifiable records can’t simply be copied around.

  1. 01

    Data that is hard to share

    Identifiable records can't move freely between sites, sponsors and countries, so studies stall at data access.

  2. 02

    Inconsistent definitions

    Each source codes diagnoses, procedures and outcomes differently, which undermines cohort and outcome definitions.

  3. 03

    Documentation burden

    Literature review, protocol drafting and eligibility screening take expert time away from the research itself.

Infographic: three sites code type 2 diabetes differently; each record is de-identified at its origin and mapped to one OMOP concept.

Same diagnosis, three sites, one definition

Site A · US hospitalICD-10-CM E11.9De-identifiedSite B · registryLocal code DM-02De-identifiedSite C · EU partnerSNOMED CT 44054006PseudonymizedOMOP concept201826Type 2 diabetesmellitus

Identifiers are removed where the data lives. Only the harmonized, de-identified record moves.

How it works

From scattered records to a research-ready cohort

Follow one multi-site study through a pipeline we build with your research team. Data is prepared where it lives, and investigators approve every cohort.

  1. 01System

    Sources are mapped

    Each site’s data is inventoried in place: what exists, what it contains and which rules apply to it.

    Source inventory

    • Sources: 3 sites · 2 countries
    • Formats: CSV, X12, FHIR
    • Legal basis: documented

    Data stays at its originData Science

  2. 02System

    Identifiers are removed

    Data is de-identified at the source using HIPAA Safe Harbor or Expert Determination, and EU data is pseudonymized before it moves.

    De-identification

    • Method: Expert Determination
    • Dates: shifted per patient
    • EU data: pseudonymized at origin

    Method documentedData Science

  3. 03System

    Definitions are harmonized

    Sources are mapped to the OMOP common data model, so diagnoses, drugs and outcomes mean the same thing at every site.

    OMOP mapping

    • Model: OMOP CDM
    • Vocabularies: ICD-10, RxNorm, LOINC
    • Unmapped codes: flagged for review

    Mappings versionedData Science

  4. 04AI agent

    Candidates are screened

    Research agents check candidates against the protocol’s eligibility criteria and explain why each one meets or fails them.

    Eligibility screen

    • Criteria: 12 inclusion · 8 exclusion
    • Candidates: ranked with reasons
    • Uncertain: sent to coordinator

    Reasons attached to every resultAI Solutions

  5. 05Person

    Investigators approve

    Investigators and coordinators confirm eligibility and approve the cohort. The agent never enrolls anyone.

    Cohort

    • Reviewed by: study coordinator
    • Cohort: approved
    • Changes: logged

    Investigator sign-offAI Solutions

  6. 06System

    Results can be reproduced

    Analyses run on versioned data with documented transformations, so an auditor or a new analyst can rerun them and get the same result.

    Analysis

    • Dataset: v1.4, frozen
    • Code: versioned
    • Results: reproducible

    Reproducible by designData Science

What changes

The same work for research teams, without the busywork or the blind spots

Today

  • Studies stall waiting for data access
  • Each site codes outcomes differently
  • Eligibility screened chart by chart
  • Identifiable records copied between teams

With UsHealthCloud

  • Data prepared where it lives
  • One common data model across sites
  • Candidates pre-screened with reasons
  • Only de-identified or pseudonymized data moves

How we help

Three services, one protected foundation

  • Data Science

    De-identification, OMOP/FHIR harmonization and real-world evidence analytics.

    Powers steps1236

    Learn more
  • AI Solutions

    Research agents for literature review, cohort screening and protocol drafting.

    Powers steps45

    Learn more
  • Cybersecurity

    Protection for shared research environments and the data moving between them.

    Learn more

Where teams start

Start with one workflow, measured against a baseline

  • Cohort discovery

    OMOP · EHR extracts

  • Trial eligibility screening

    Protocols · EHR data

  • Real-world evidence studies

    Claims · registries

  • Literature review

    Abstracts · full texts

  • Registry curation

    Registry exports

  • Cross-border data sharing

    Pseudonymized EU data

Rules that shape the work

Built for the regulations you answer to

  1. HIPAA de-identification

    Safe Harbor or Expert Determination decides when health data stops being PHI.

    Read the guide
  2. GDPR

    Health data is special category data, and transfers out of the EU need a documented mechanism.

    How we handle it
  3. The Common Rule (45 CFR 46)

    Protects human subjects in federally funded research, including review by an institutional review board.

FAQ

Questions we hear from compliance and data teams

  • Cross-borderCan you support studies with EU participants?

    Yes. Pipelines are designed for GDPR-compliant cross-border data flows, with data minimization and de-identification applied before data leaves its origin wherever possible.

  • StandardsDo you work with the OMOP common data model?

    Yes. We harmonize EHR and claims sources to OMOP and FHIR so cohort definitions are consistent across sites.

  • ModelsCan research agents work on our proprietary literature and protocols?

    Yes, within an environment you control, and only with the documents you choose to provide.

Next step

Bring us the study that is stuck at data access.

We'll map the sources, the rules that apply to each, and a path to research-ready data that never copies identifiable records around.

What the first call gives you

  1. 1A map of your sources and the rules on each
  2. 2The right de-identification method
  3. 3A path to one common data model

A short scoping call, then a written proposal. Please don't send patient information.