The government’s initiative for meaningful use of electronic health records has fueled the demand for discrete, structured and actionable clinical data.

As part of this initiative, healthcare organizations are incentivized to leverage certified EHR systems to collect and report on several quality and safety metrics with the aim to use this data to ultimately improve the cost/quality equation in healthcare.

Such discrete data (unlike free-form narrative) can be used to facilitate consistent interoperability and exchange of information between systems and among providers within an integrated care continuum. It can also enable broader application of evidence-based medicine and clinical decision support systems, providing the basis for a transition from pay-for-reporting to pay-for-performance.

The current solutions proposed by EHR vendors utilize template-driven, point-and-click user interfaces as the main mechanism for structuring data within an EHR. In many cases, these solutions are being resisted by physicians because populating point-and-click templates via the keyboard and mouse is found to be less efficient than direct dictation.

Many doctors also believe point-and-click EHR templates limit physicians’ ability to capture the unique clinical story and to adequately document their medical decision-making process, which is unique to each patient encounter.

While the value of free-form narrative is clear, structured and discrete data is central to support organizational priorities, such as increased EHR adoption, achieving meaningful use to qualify for the government’s incentives, streamlining data capture and reporting to sustain quality initiatives, as well as to build the foundation for evidence-based medicine and clinical decision support initiatives.

The challenge to the industry is three pronged:
How to address physicians’ concerns for efficiency,
How to preserve the patient’s unique story, and
How to support organizational needs for structured data.

A successful solution must integrate into the physicians’ preferred workflows, streamline EHR data population, help accelerate EHR adoption and support compliance with CCHIT EHR certification criteria. In addition, such a solution must be cost effective, favoring technology as opposed to manual labor.

Advanced Natural Language Processing (NLP) technology holds the answer.

It can be used to automatically derive patient information from free-form, unstructured text and to produce actionable data that can be used to improve patient care and streamline workflow.

Well designed NLP systems are capable of analyzing “free-text” dictation, understanding the context and can tag the most important clinical data elements such as problems, social history, medications, allergies, and procedures.

Once tagged, these elements can then be used to populate the EHR and other clinical IT systems, enabling applications that range from retrospective analysis across large amounts of medical records to medical decision support at the point-of-care.

NLP can be integrated as part of the physicians’ preferred workflow environment, supporting both real-time and background dictation workflows.

For real-time workflow, NLP can work with speech recognition technologies to capture narrative dictation and enable the physician to interactively self-edit text and clinical facts as it appears on the EHR screen, but prior to committing information to the EHR. Whereas in the case of traditional background dictation/transcription workflow, NLP can be integrated into the workflow to process narrative and create rich clinical documents that include the corrected text and extracted clinical facts to the physician for final review and approval. Finalized documents can be committed to the EHR.

NLP technology has the potential to provide physicians the best of both worlds; it can support comprehensive documentation of care through narrative dictation that captures the uniqueness of each patient encounter and do so without the limitation of rigid documentation template. Unlike “blobs” of free-text, key patient data can be identified and automatically saved in the EHR and other clinical IT systems, where information can be analyzed, reported and used for better patient care.

Isam Habboush is the Natural Language Processing (NLP) product manager for Nuance Healthcare