You are right. The challenge is to implement web services without much help from vendors. But that challange is what makes it interesting (fun).

Some big pluses include: being in control of your own destiny (mostly) with regard to changing systems on the backend, centralized security and auditing of data access is easier, easier to bring new systems online because the data definitions are more clear and easier to understand, and the process of implementing SOA had a beneficial transformative effect on our IT silos, which are now significantly less silo’ed.

The hard parts, though, were: defining what your SOA is (SOA governance), achieving consistent performance across a dozen different vendor systems (with different contracts, different life cycles, etc.), and getting the vendors to go along with the idea at all since most are totally ignorant (although they’ll talk about in their marketing stuff ad nauseam). Due to our environment, we do have some systems in a CDR because the vendor just couldn’t cope. We are also making an “operational data store”, which is essentially a cache of heavily used clinical data, to enable some really high performance data access systems that our basic system just can’t handle without incurring significant expense.

Cost-wise, we wound up cheaper with the SOA, particularly over time, but the calculation is highly institution- and environment-dependent. We started in a best-of-breed place with a good dev shop, so we were in a prime position. Somebody trying to do it at an outsourced Epic shop should have his/her head examined.

If the vendors were on board with this concept, it would be pretty easy, but unfortunately most just aren’t, and the rest are just paying lip service to it.

Actually I’d prefer to discuss the pros and cons of the virtual database conceptual model versus the CDR-centric conceptual model. Can we agree they are both viable?

Also the Epic numbers in your example are for EVERYTHING, not just an Epic EMR. They include the Hardware – Servers, Backup Servers, Reporting Servers, maybe Citrix plus thin clients or thick workstations, Microsoft licenses, the interfaces, connecting to patient monitors and streaming all that new data somewhere etc. Plus all the staff the customer is going to use on the project, any Epic installers, any consultants, etc. Plus any content they buy from third parties, any tools (Business Objects) that they don’t already have, etc.

You just have to do an All In number.

My bet is that over a comparable time period, when you add up everything that goes behind that Epic number and compare it to what MD Anderson spent on all its comparable items you’d likely find they spent even more. You can’t pick and choose which costs. You have to do an real comparative number.