Over the past few years I've helped a handful of firms build longitudinal records for their members and patients. We've called these comprehensive or longitudinal and focused them on patients or people or humans, but the overall goal is the same: to craft as complete a view of an individual as possible in order to serve them better. Which sounds a lot like what data warehouses have been doing since the 90s.
For the successful ones, it's not. Here's why:
I spent the week at HIMSS talking with colleagues and vendors about what's next in Health IT and nearly everything is pointing to the explosion of data liquidity across the health landscape. The 21st Century Cures Act requirements paired with private demand has created a sea change in APIs and interoperability. I'm super excited. I love this stuff.
But it's a new world for data professionals. We’ve been working with mostly trustworthy data from within a closed system: our company and partners. We have some quality issues, but nobody's actively malicious. And the data has a reasonably discoverable provenance.
All of this is reflected in our data lake/warehouse/mart thinking. Limited, known sources become trusted tables with some ETL magic applied. Query away.
The LHRs we're building today however are entering into this new world where data is coming from every direction, from everyone who has a point of view on the situation:
Doctors and other providers
Payers and their value added partners
Pharmacists and PBMs
Home health and skilled nursing
Uber who drove the patient in, and of course
The patient and their family and friends.
Wild.
I boil it down to two keys:
LHRs can't just present data. They must answer questions.
You can't leave trust in the data to users. You must help them decide.
I'll write more about each of these; there's lots to unpack. What I want to impress now is that our lake, warehouse, and mart methods aren't adequate. If you're building a Member Data Warehouse and calling it an LHR, you're going to be sad. If you can't give a single straight answer to "what's the person's BMI?" or a reasonably confident list of "what medications is this person taking?", you're a data provider, not an LHR.
Stay tuned!