Garbage In, Garbage Out

The question came up at a recent meeting:

“How many atrial fibrillation ablations did I perform last year?”

It was a seemingly simple request. One that our vast array of computer technology and Electronic Medical Record should be able to conquer in a few microseconds.

So after an administrator talked to the responsible Information Technology officer and they searched and scanned the appropriate clinical and billing files, a day later came the answer:

Exactly one.

My swollen legs can attest to the inaccuracy of this result.

Yet with the speed of the hospital intranet, the results were broadcast as gospel in an e-mail to those who needed to know. Decisions were being made. Projections for upcoming years decided.

Which lead me to ponder the obvious.

How helpful will our electronic medical records be in supporting clinical care if doctors, rather than administrators and hospital information technology geniuses, cannot access and utilize clinically relevant data?

It’s the classic case of garbage in, garbage out.

But of course our policy wonks have already anticipated this criticism. They understand the health care records are, by and large, free text. They know that computers don't work well with the nuances of the written sentence. So to compensate for this short-coming, they have plans to implement the Mother of All Coding Schemes to permit more accurate data queries, licensed and sanctioned by the American Medical Association, called ICD-10®. Our government needs this scheme, it was argued, to determine if the government is getting an adequate return on their health care investment dollar.

ICD-10®, it is promised, will provide for unprecedented data queries of electronic data, housed in data silos so large that even they will be the envy of Google. Once these highly specific codes are used, we are told, we’ll really be able to forecast clinical trends, pricing trends, and define value-added components of our health care delivery in America. We will be able to determine which patients need our most urgent clinical interaction, purchase the equipment for that interaction in real-time, and assure it’s delivered “just-in-time” to the hospital loading dock, saving countless dollars.

So they say.

Sadly, real life gets in the way of such promises. We see this all the time in clinical medicine. “Mr. Jones, if you just stop smoking and take these medications …”, only to see Mr. Jones “forget.” Coding schemes are like this when applied to free text databases. The coded information is only as good as the people capable of classifying and sorting clinical care into pre-specified code bins accurately, and to date, there has never been a prospective evaluation to see if people really can achieve the promised results of this tactic. Which is why doctors will be the ones who will be required to either code the visit, under penalty of fraud if we're inaccurate, or teach others to code the visit with the same accuracy required to assure "accuracy." One only has to look at the labyrinthine nature of the new coding to see where it falls flat.

But there is other more ominous problem with the EMR data queries that goes far beyond the codes: the software that uses those codes. Hospitals and doctors, you see, don’t have access to the software. It’s owned by proprietary companies whose employees are already overworked to add the next bell and whistle to their software to stay ahead of the competition in the Great Race to become the great National Electronic Medical Record System sanctioned by Congress. No one really knows if the software is programmed correctly, they just assume it is. If a doctor takes issue with results reported by the Great EMR System using the World’s Most Powerful Coding Scheme, the hospital administration is rendered impotent by the need to ask, pretty please, Mr. EMR vendor if he could check their results. To which, of, course, the EMR vendor adds the request to their ever-growing queue of similar requests with promises like “well, we’ll work on it when we have time, but changing things would affect ALL of the hospitals with which we have contracts so we’ll determine if the request is important enough to pursue first.”

God forbid doctors be allowed to design and review their own clinically-relevant data queries.

So there you have it.

What they're calling "clinically-relevant health care data queries" in the era of our new Electronic Medical Records. Data queries decided by others in the name of the greater good yet often completely useless to those of us providing the clinical care, yet full of great hope and promise for a brighter health care future...

... at least until a more accurate coding scheme with greater clinical relevance called ICD-11® comes along.

-Wes