Echoing Paul Cerrato’s post on EHRs and Pay For Performance, which cites a study by Jonathan Weiner, et. al., at Johns Hopkins University titled New Paradigms for Measuring Clinical Performance Using Electronic Health Records, there are many shortcomings in current EHR support for a shift to Pay For Performance (P4P). Cerrato particularly calls out statistics showing very small fractions of ambulatory care encounters fully documented in EHRs and interoperable across providers – necessary to meaningfully impact quality of care across provider organizations.
There are two issues at work here: the capture and the use of data. When two provider EHRs cannot adequately share data for a single patient, this is an issue of the usage of the data. If the data cannot be effectively used across provider organizations, there is little chance of using the EHRs to drive improved quality, and thus succeeding at P4P.
To what extent that requires a difference in how data is captured is itself an interesting question. Certainly it helps those of us writing software if our users do our work for us and capture data in a structured format. However, as Cerrato mentions, there are a growing number of NLP capabilities for working directly with free-form clinical text and dictation rather than structured data. Cerrato mentions a few currently available commercial systems.
Looking to the future, I believe systems such as the VA’s ARC suggest that the necessary quality metrics will be derived directly from clinical text. I’ve previously argued that there are significant benefits to sticking with clinical text rather than making physicians switch to primarily structured data entry.
We have some ways to go with EHR implementations before we will have achieved the interoperability needed for P4P. But there is hope that changing the way all physicians document patient records need not be a barrier to getting there.
Richard Wolniewicz is Division Scientist, Natural Language Processing, at 3M Health Information Systems.