| |
 |
 |
|
 |
Page One Page Two Page Four Back to Newsletters |
| Volume VI, Issue 3 | July 2005 |
|
INFORMATICS FEATURE: The Role of Natural Language Processing in EMR Data Analysis |
|||
|
Implementation and adoption of
electronic medical records (EMRs)
is increasing, fueled in part by the
prospect that these systems will
facilitate quality improvement
through faster and more accurate
analyses of clinical data. A significant
amount of electronic data,
however, is unusable by One recent analysis concluded that of the information necessary to complete a comprehensive quality assessment of a health plan with a modern EMR, at most 50% could be obtained from commonly utilized coding schemes. Furthermore, structured data entry to the EMR has proven difficult for clinicians and this difficulty is amplified for preventive care activities, such as counseling about smoking cessation, which are based on the content of complex discussions between provider and patient. For complex conversations involving multiple topics, the task of creating complete and meaningful codes is simply impractical. |
However, multiple and complex aspects of the encounter are relatively easily captured by the clinician in free-text notes. One solution to this problem, utilizing so-called Natural Language Processing (NLP) technologies, would allow clinicians to capture relevant clinical information in a modality that comes easiest to them (natural language) while providing automatic structuring of information for reporting and post-entry processing. In the HIT2 project, we have developed a medical record classifier called "MediClass" that uses natural language processing to assess delivery of the "5 A's" of smoking cessation care (Ask, Advise, Assess, Assist, Arrange). The system has been deployed at four HMO’s in the CRN to enable a trial testing the effects of feedback to primary care clinicians regarding their delivery of the 5 A’s to smokers. Several manuscripts have been developed, and one recently published, reporting on the MediClass system design and its performance in assessing the 5 A’s. We found the system to perform with similar accuracy (and much less cost and time) to trained medical record abstractors. Systems such as MediClass show promise for addressing a wide variety of concerns in health care research and operations including care quality, disease surveillance, and adverse event detection. |
These systems can help bridge the gap between the promise and the realization of value in electronic medical records. To learn more, direct your inquiries to Brian.Hazlehurst@kpchr.org. -Brain Hazelhurst, KPNW
|
|
 |
|||
 |
 |
 |
 |
