HL7: What You Don’t Know Can Hurt You
How many times do hospital IS/IT directors come across this scenario?
Your hospital has just signed on the dotted line for a new EMR. It’s meant to streamline clinician workflows, facilitate improved quality of care, and help meet all-important Meaningful Use targets. And all you have to do is interface it.
You’ve got a budget, a timeline, an interface engine, a couple of developers, a couple of analysts, a pile of legacy systems to integrate and a mid-sized mountain of data. Sounds like you’re all set.
But are you really? What about the “unknown unknowns”? As you’ve probably experienced, in HL7 interfacing, what you don’t know really can hurt you – especially when you don’t know you don’t know it.
12-Month HL7 Interfacing Timeline?
Let’s say you’ve budgeted 12 months for interfacing implementation, based on past experience and a rough scope of the project, with a little flex built in for contingencies. You’ve completed a integration site survey and the vendor has provided a broad interface spec. Your analysts have collected 50-odd samples of HL7 messages and extrapolated a few dozen data types that they’re pretty sure cover all your bases. They’ve started testing interfaces by trial and error…testing and tweaking, testing and tweaking…and then they discover that the system under implementation not only needs unexpected test result categories but also provides lab results in a set of unexpected formats. Your lab interface needs a complete overhaul.
So you add another week to the project, and another line to the budget.
Eventually you’ve sorted that one out, but now it turns out that you need a list of pharmacy codes. You contact your Director of Pharmacy and she assigns an analyst to pull the codes. A couple of weeks go by, and the questions go back and forth. Could they pull the codes? Do you have them? Can you go ahead without them? Are you sure?
This happens five times, with other systems. Add another four weeks.
At night, you start having long, troubled dreams of sorting through haystacks for needles.
18 Months Later…
You’re 2 weeks out from the go-live. Validation testing is coming down to the final wire. You’ve cleared multiple checklists, and you hope you’re ready to launch. But you know there’s still work to do. Clinicians expect the interfacing to work like clockwork. This EMR and the new integration are supposed to make their lives easier. The last thing you need is a series of troubleshooting all-nighters, an endless support bottleneck, and unhappy nurses.
So you test some more, until you’re sure. You’ll get it done. That’s what you do. But if there were some way you could uncover those “unknown unknowns” before you started, well, your job would be a lot easier.
Your turn: What’s the worst “unknown unknown” you’ve ever faced down in an HL7 interface implementation?