There’s no doubt about it — the air is ringing with the sounds of vendors promising big things from big data, from population health to clinical support to management of bundled payments. But can they really offer these blessings? According to enterprise health IT architect Michael Planchart (known to many as @theEHRGuy), there’s a lot of snake oil sales going on.
In his experience, many of the experts on what he calls Big Bad Data either weren’t in healthcare or have never touched healthcare IT until the big data trend hit the industry. And they’re pitching the big data concept to providers that aren’t ready, he says:
- Most healthcare providers haven’t been collecting data in a consistent way with a sound data governance model.
- Most hospitals have paper charts that collect data in unstructured and disorganized ways.
- Most hospitals — he asserts — have spent millions or even billions of dollars on EMRs but have been unable to implement them properly. (And those that have succeeded have done so in “partial and mediocre ways,” he says.)
Given these obstacles, where is big data going to come from today? Probably not the right place, he writes:
Well, some geniuses from major software vendors thought they could get this data from the HL7 transactions that had been moving back and forth between systems. Yes, indeed. They used some sort of “aggregation” software to extract this data out of HL7 v2.x messages. What a disaster! Who in their sane mind would think that transactional near real time data could be used as the source for aggregated data?
As Planchart sees it, institutions need quality, pertinent, relevant and accurate data, not coarsely aggregated data from any of the sources hospitals and providers have. Instead of rushing into big data deals, he suggests that CIOs start collecting discrete, relevant and pertinent data within their EMRs, a move which will pay off over the next several years.
In the mean time, my colleague John Lynn suggests, it’s probably best to focus on “skinny data” – a big challenge in itself given how hard it can be to filter out data “noise” — rather than aggregate a bunch of high volume data from all directions.