Build vs Buy for Healthcare Integration: The Real Numbers
Every healthcare integration project hits this fork: deploy your own engine (MedHook, Mirth, HAPI FHIR) or pay a managed platform (Redox, Health Gorilla) to handle it. The managed pitch is compelling — data flowing in days, no servers to run, EHR connections pre-built. And for your first integration, it is usually right.
The question is what happens at integration number 5, 10, or 20.
The real cost at different scales
These numbers are based on what we have seen across multiple engagements. Managed platform pricing varies by vendor and negotiation, but the shape of the curve is consistent.
| Scale | Managed platform (monthly) | Self-hosted engine (monthly) | Delta |
|---|---|---|---|
| 3 connections, 50K msgs/mo | $2,000-4,000 | $200-500 (cloud infra) | Managed is 4-10x more |
| 10 connections, 300K msgs/mo | $6,000-12,000 | $400-800 | Managed is 8-15x more |
| 20 connections, 1M msgs/mo | $15,000-30,000 | $600-1,200 | Managed is 15-25x more |
| 30 connections, 3M msgs/mo | $25,000-50,000 | $800-1,500 | Managed is 20-33x more |
Self-hosted costs are infrastructure only (compute, storage, networking). They exclude engineering time for setup and operations, which is the hidden cost that makes the comparison honest.
The engineering cost to self-host:
- Initial setup: 1-2 weeks of a DevOps-capable engineer
- Per-interface configuration: 1-5 days depending on complexity
- Ongoing operations: 2-4 hours/week for monitoring, updates, and troubleshooting
- At $150-200/hour fully loaded, that is roughly $1,500-3,500/month in engineering time
Even including engineering time, self-hosted breaks even with managed around 5-8 connections. Above that, the gap widens every month.
What managed gives you (and what it takes)
Speed to first integration
This is the real advantage. Redox has pre-built connections to 50+ EHR platforms. When a hospital partner says "we need data flowing in two weeks," a managed platform can meet that timeline. Self-hosted requires deploying infrastructure, configuring the EHR adapter, testing the connection, and handling credential management. If you miss the hospital's pilot deadline, the speed savings of self-hosted are irrelevant.
No infrastructure responsibility
Someone else pages when the integration layer goes down at 3 AM. Someone else updates when an EHR vendor changes their API. For a five-person startup where every engineer is building product, that operational burden matters.
Data passes through a third-party cloud
This is the trade-off most teams underweight. PHI transits through the managed platform's infrastructure, which means a BAA with the platform vendor, trust in their security posture, and a third-party dependency for your most sensitive data flow. This is acceptable for most startups. It is a non-starter for government healthcare, military health, organizations with strict data residency requirements, or any contract that prohibits third-party PHI processing.
Vendor lock-in is real
Your application code targets the managed platform's API. Migrating away means rewriting your integration layer. We have done this migration for multiple clients who outgrew their managed platform, and it is consistently a 4-8 week project depending on the number of connections and the complexity of the data model mapping.
You lose access to the raw message
Managed platforms normalize data into their standard model. This is convenient until you need a vendor-specific HL7v2 Z-segment, a non-standard FHIR extension, or a custom transform that the platform's normalization strips out. We have seen this become a blocker for clinical decision support applications that depend on site-specific data elements.
The hybrid path
The most pragmatic approach we have seen:
Phase 1 (0-12 months): Use a managed platform for your first 3-5 EHR connections. Focus engineering time on your product, not integration infrastructure. Accept the per-connection cost as the price of speed.
Phase 2 (12-24 months): When message volumes grow and per-connection costs become significant, deploy a self-hosted engine alongside the managed platform. Route your highest-volume connections through the self-hosted engine first — that is where the cost savings are largest.
Phase 3 (24+ months): Migrate remaining connections to self-hosted as your team builds integration expertise. Keep the managed platform for long-tail, low-volume connections where the per-connection cost is below the engineering cost of self-hosting.
This trades short-term efficiency for long-term cost control without risking the early product timeline. The transition from Phase 1 to Phase 2 is where most teams call us.
The questions that decide it
- Do you have someone who can manage Docker containers and monitor an integration engine? If no, start managed.
- Does your contract or compliance posture prohibit third-party PHI processing? If yes, you must self-host.
- How many EHR connections will you have in 18 months? If more than 8, model the cost difference now.
- How much does your team cost? If you have a $200K/year integration engineer on staff already, self-hosted is almost always cheaper. If you would need to hire one, factor that into the comparison.
We have built both ways — self-hosted engines and managed platform integrations — and helped teams transition between them. If you are trying to model the cost for your specific volumes and timeline, we can walk through it with you.