In short
- Run your own engine, such as Mirth Connect, when you have many custom HL7 v2 interfaces, a team that can operate servers handling PHI, and a need to control every mapping; buy an integration platform when you need maintained connections to many EHRs and would rather pay per connection than staff an interface team.
- NextGen Healthcare moved Mirth Connect to a single commercial, proprietary license with version 4.6 in 2025, so self-hosting now means licensing 4.6 or later, staying on the last open-source release 4.5.2, or using a community fork such as Open Integration Engine.
- Either way, someone must meet the HIPAA Security Rule's technical safeguards for the systems that move PHI, and when a platform does it for you, it needs a BAA.
Run your own integration engine when you have many custom interfaces, a team that can operate servers handling PHI around the clock, and a need to control every mapping. Buy an integration platform when you need maintained connections to many EHRs through one API and would rather pay per connection than staff an interface team.
This comparison is for CTOs and integration leads deciding how to move data between their product and the EHRs, labs and payers it works with, and it covers what Mirth Connect's 2025 licensing change means for self-hosting. For a comparison of specific platforms, see our look at Redox, Health Gorilla and 1upHealth.
What does an integration engine do that a platform does not?
An integration engine moves and reshapes messages. The Mirth Connect source repository describes channels, which are message pathways that users build and manage, with filtering, transformation, extraction and routing, translating message standards into the one your system understands. You design each channel, you decide every mapping, and the engine runs on infrastructure you control.
An integration platform sells the result of that work. It maintains connections to many EHRs, translates their data into its own model, and exposes one API to your product. You do not see or own the mappings, and you depend on the platform to support each EHR, data type and write your product needs.
- Engine: you build every interface, you own every mapping, you run the servers.
- Platform: the vendor builds and runs the connections, you build against its API, you pay for each connection or by volume.
- Both: you still need each health system's approval for what your product reads and writes.
| Question | Self-hosted engine | Integration platform |
|---|---|---|
| Who builds interfaces | Your team, channel by channel | The vendor builds and runs connections |
| Mappings | You see and own every one | You do not see or own them |
| Servers | You run them | The vendor runs them |
| How you pay | License, people and operations | Per connection or by volume |
| HIPAA safeguards | You meet them | The vendor, under a BAA |
| Best fit | Many custom v2 feeds and an operations team | Many EHR vendors and a small team |
| Health system approval | Still needed | Still needed |
Why does HL7 v2 push teams toward an engine?
Because v2 interfaces are rarely alike. The HL7 FHIR comparison describes v2 as widely adopted and built from reusable segments in which most data elements are optional, extended with Z-segments whose meaning is opaque without the sender's explanation. It says consistent mapping rules between v2 and FHIR are impractical, so mappings are done implementation by implementation, and one v2 message can become tens or even hundreds of FHIR resources.
If your product connects to many sites that each send slightly different v2 feeds, the work is in those per-site mappings. An engine lets your team write and test them directly. A platform may handle them for you, but only for the feeds and fields it supports.
What changed with Mirth Connect's license?
In March 2025, NextGen Healthcare announced it was moving Mirth Connect from a dual license, open source and commercial, to a single commercial and proprietary licensing model starting with Mirth Connect 4.6. NextGen listed three options for open-source users: stay on their current version, upgrade to the last published open-source version, 4.5.2, or become licensed users and upgrade to 4.6.
- NextGen said the GitHub source code, discussion boards and historical community threads would remain available.
- It named three tiers, Enterprise, Gold and Platinum, all including access to its new Mirth Command Center, and said Command Center is not available in the European Union or the United Kingdom.
- It described Mirth Connect for Research as a complimentary license for approved, active research studies affiliated with a non-profit institution.
- The announcement did not publish prices, so licensing cost has to come from NextGen directly.
The code in the public repository is released under the Mozilla Public License 2.0. A community project, Open Integration Engine, describes itself as an open source fork of the now closed-source Mirth Connect, licensed under MPL 2.0 with community governance. Check any fork's release activity and security fixes before you rely on it for PHI.
What are your self-hosting options now?
- License Mirth Connect 4.6 or later from NextGen and get its tooling and support relationship.
- Stay on 4.5.2, the last open-source release, and accept that you are responsible for its security from here on.
- Move to a community fork, such as Open Integration Engine, and judge it like any open-source dependency: releases, maintainers, security response.
- Use another engine, commercial or open source, and migrate your channels.
Staying on an old version of software that carries PHI is a decision your risk analysis has to cover. Write down why it is acceptable, what compensating controls you use, and when you will revisit it.
What does running your own engine really cost?
The license is the smallest part of the cost. Most of it goes to people and operations.
- Interface engineers who understand v2, FHIR and the systems on each end, and who can debug a failed message at night.
- Infrastructure: servers, databases, backups, failover and monitoring for message queues and errors.
- Security: patching the engine, its runtime and the operating system, plus network controls and encryption.
- Testing: a test environment per connected system, and a regression suite for every channel you change.
- Documentation: a specification for every interface, because the next engineer will need it.
A platform turns most of that into a subscription. You trade control and per-message cost for not having to run it.
Who meets the HIPAA safeguards in each model?
If you self-host, you do. The technical safeguards at 45 CFR 164.312 cover access control, including unique user identification and an emergency access procedure, with automatic logoff and encryption as addressable specifications. They also set standards for integrity, person or entity authentication, and transmission security against unauthorized access to PHI sent over a network. An engine sits on exactly those network paths, so it is in scope for all of them.
If a platform moves PHI for you, it is your business associate or subcontractor, and the contract has to say so. Under 45 CFR 164.504(e), a BAA must set permitted uses and disclosures, require appropriate safeguards and Security Rule compliance, require reporting of uses or disclosures the contract does not allow, and require subcontractors to agree to the same terms. Read the platform's BAA as carefully as its pricing.
What should you ask a platform vendor before you buy?
A platform demo shows the connections that work. Your questions should find the ones your product needs that do not. Ask for answers in writing, tied to your use case.
- Which of our customers' EHRs, and which versions, do you support today for the exact data we need?
- Which writes do you support, and into which places in each EHR?
- How do you handle a site-specific HL7 v2 feed with Z-segments or local codes?
- Can we see and export the mappings for our connections, and our data, if we leave?
- What is your pricing per connection, per message or per patient, and how does it change as we grow?
- Who signs the BAA, which subcontractors touch our PHI, and how do you report incidents?
- What happens to an in-flight message when one of your connections is down?
- Who does each health system talk to during go-live: you, us, or both?
Ask the same questions of yourself for the engine option. If your team cannot answer who handles a failed message at night or how a mapping change is tested, the engine's lower license cost will not hold up.
How do you decide?
- Many custom v2 feeds, a few large customers, and an operations team: an engine fits.
- Many EHR vendors, the same data from each, and a small team: a platform fits.
- Deep work in one EHR, such as launching inside the clinician's screen: neither replaces that vendor's own APIs.
- Mixed needs: a platform for EHR connectivity and an engine for lab, device or legacy feeds the platform does not cover.
Whichever you choose, keep your own data model between the integration layer and your product. That is what lets you move from an engine to a platform, or back, without rewriting features.
We have shipped Epic and athenahealth integrations, and for any engine or vendor platform we work from the vendor's documentation and sandbox. If you are deciding whether to run an engine or buy a platform, or what to do about an existing Mirth installation, talk to one of our engineers. Our free five-day technical audit ends with a written report on the options and their trade-offs for your interfaces.
Sources
- NextGen Healthcare: A new era for Mirth Connect (licensing change, March 2025)
- GitHub: nextgenhealthcare/connect (Mirth Connect source, MPL 2.0)
- GitHub: OpenIntegrationEngine/engine (open-source fork of Mirth Connect)
- HL7 FHIR R4: Comparison with HL7 v2
- 45 CFR 164.312: Technical safeguards (Cornell LII)
- 45 CFR 164.504: Business associate contracts (Cornell LII)