Skip to main content
    Healthcare Software Integration Patterns That Work in UK Care
    Integration

    Healthcare Software Integration Patterns That Work in UK Care

    The integration patterns we actually use to connect Salesforce, care management, finance, rostering, and NHS-facing systems in UK healthcare.

    Digitalised Solutions Team
    May 29, 2026
    8 min read

    Why integration is the centre of gravity

    In UK healthcare and care, the difference between a software estate that runs the operation and one that fights it almost always comes down to integration. Applications are commodities. The patterns that move data between them — and the discipline behind those patterns — are what determines reporting reliability, compliance evidence, and staff time.

    The patterns we actually use

    There are a small number of integration patterns that recur across well-run estates. Pick deliberately; do not let the integration emerge by accident.

    1. Point-to-point — use sparingly

    Two systems talk directly. Cheap to build, expensive to maintain. Acceptable for a single high-value integration where neither side will change frequently. Avoid as the default; estates with more than a handful of point-to-point links become impossible to evolve.

    2. Hub-and-spoke via Salesforce

    Salesforce often becomes the de facto hub because it integrates with almost everything. Workable when Salesforce is genuinely the system of truth for the data being routed; problematic when it is being used as a passthrough for clinical data it does not own.

    3. Middleware / iPaaS

    A dedicated integration platform (MuleSoft, Boomi, Workato, Azure Integration Services, and others) sits between the applications. The right answer once you have more than three or four integrations, or any compliance-sensitive flows. The middleware layer is also where you enforce identifier mapping, retry logic, and audit trail.

    4. Event-driven with platform events

    Instead of polling, systems publish events ("new referral", "placement confirmed", "medicines refused") and subscribers react. The pattern of choice for workflows that need to feel real-time to operations. Salesforce Platform Events, change data capture, and webhook patterns from care systems make this practical in UK estates today.

    5. Federated identity and master data

    Not glamorous, but the highest-leverage pattern. One canonical identifier per entity (service user, staff member, location), propagated everywhere, with a documented authority per attribute. Without this, everything else is built on sand.

    NHS-specific integration

    For organisations interacting with the NHS, two further considerations apply.

    NHS Number as the canonical patient identifier. Capture it, validate it via PDS where you have access, and use it consistently. ODS codes do the same job for organisations and sites.

    FHIR UK Core and standards-based APIs. GP Connect, the National Record Locator, the Personal Demographics Service, and the broader NHS API landscape are increasingly FHIR-based. New integrations should default to FHIR UK Core profiles rather than bespoke message formats.

    What good integration design looks like

    A short list that separates resilient integrations from fragile ones.

  1. **One direction of write per field.** Reads can happen anywhere; writes flow from the system of truth only.
  2. **Idempotent operations.** Replaying the same event must not create duplicates.
  3. **Explicit identifier mapping** between every pair of systems.
  4. **Retry, dead-letter, and alerting** baked into the middleware layer.
  5. **Audit trail** of every integration event, retained against your data retention policy.
  6. **Versioned contracts** so that a vendor upgrade does not silently break a downstream system.
  7. **A clinical safety case** under DCB0129 / DCB0160 for any integration touching medicines, care, or clinical decisions.
  8. Common antipatterns

  9. Salesforce flows that call out to external systems synchronously in user-facing transactions.
  10. Nightly batch jobs running operational workflows that users expect to be real-time.
  11. Reporting layers built by joining live application tables with no canonical model.
  12. Integration logic hidden inside individual record triggers, invisible to anyone reviewing the estate.
  13. No owner for the integration layer itself — only owners for the applications it connects.
  14. How to move from where you are to a defensible pattern

    Realistic sequencing:

    1. Map every existing integration, including the ones nobody owns.

    2. Decide system of truth per data domain.

    3. Pick a target pattern — usually middleware with event-driven flows for the high-value workflows.

    4. Migrate the highest-risk or highest-value integrations first.

    5. Decommission point-to-point links as the middleware patterns mature.

    6. Document and assign ownership of the integration layer.

    Where to go next

    The Integration Architecture page sets out how we deliver this work in UK healthcare engagements, and the Healthcare Software Solutions hub gives the wider context across CRM, compliance, and reporting. For Salesforce-specific stabilisation work that often precedes integration, see CRM Stabilisation.

    Healthcare software decisions

    Turn research into a practical review of CRM, CQC evidence, reporting, integrations, and operational risk.