A forum to discuss ideas, approaches, standards, and architecture to establish and support open interoperability among healthcare IT systems.

Monday, July 24, 2006

To What Extent Computable?

As organizations strive to create semantically-meaningful, interoperable systems, the push is always toward establishing "computable data." In informatics speak, computable data is a representation of data values in a form that can be machine processesed and reasoned upon. It is usually depicted in a code value from some formal terminology where semantic links are meaningful and support activities such as decision support.

Present systems suffer in part because they lack this computable data. Free text information is helpful and valuable to a caregiver, but does not allow for the computer to provide the assistance that would be possible if a more structured representation were available. For instance, support functions such as drug-order checking, drug-drug interactions, allergy checking, and so on each require computable data to perform.

I have a big concern, however, that the pendulum has "swung" too far to the other side, where organizations are seeking solutions involving almost exclusively computable data. Computable data comes at a cost, and the "garbage-in, garbage-out" scenario applies here in spades. Without measures in place to validate data content, verify accuracy, and assure compliace to the valid codesets and terminologies in use within an organization, computable data is of no value.

Computable data comes at a cost. In addition to keeping up on data entry and assuring quality, the terminologies themselves must be maintained and kept current. Without quick response terminology maintenance, staff "in the trenches" will quickly become disenchanted and force-fit incorrect coded values for purposes other than what they are intended.

So what is the 'middle ground"? I am very supportive of computable data. It is essential in achieving the potential of EHR systems. That said, we must be judicious and not try to codify everything. Computable data comes at a cost. When that investment is made in capturing and maintaining data that has impacts on maintaining or improving patient health, it is a wise investment. If there is limited or no payback as a result of the effort, it is not a wise investment.

I recommend the following as a start to identify and manage computable data. It is only a start, but hopefully can get some dialogue going within the industry:

> Determine the behaviours you hope to influence, and codify data with direct/indirect impacts
> Identify the information that are the indicators of the health status you wish to monitor
> Identify the information that is the predominant basis for care or care consistency
> Focus on the "20%" which has the "80%" impact
> Relate the terminology work to business purpose. If you can't identify a purpose, question the effort
> Consider secondary uses of data in these decisions (for instance, epidemiology)
> Consider phased deployments of this. Not everything needs to be computable on day-one (but there are interdependencies that must be considered with this approach)
> Recognize that using computable data is not a wholesale replacement for free text

0 Comments:

Post a Comment

<< Home