The Challenge of Negation in Quality Measurement

One of the later discussions today in HL7 Structured Documents was how to resolve the issue of not having a way to represent certain events (encounter, procedure, supply) that did not occur.  These events are relevant in QRDA because there are now some quality measures which require being able to count numerators or denominators based on events that might not have occurred (perhaps for some reason).  For example, patient was not supplied this medication because it was out of stock.

One proposal was to allow use of the negationInd RIM attribute as an extension in these classes.  I'm vehemently opposed to this for several reasons.  Regardless of whether this extension would show up only in QRDA documents, the possibility that it could wind up back in the patient data stream as an input for providing care is reasonable.  Understanding how software mistakes happen, this creates a situation where one system could easily "misunderstand" the data, because it was never designed to do so, and at the time the design, it would never have expected its understanding to have to change. So a correctly developed system which found itself ingesting data that may have come from a quality management pathway, could wind up introducing incorrect data into the patient record.

Structured Documents is still working on how to resolve this problem.