Case Studies/Testimonials

Clinical Notes Mining Service (CNMS)

Posted by Bhaskar Bhattarai on June 10, 2014

1. Executive Summary

Raw text in clinical notes originating from a physician or nurse is invaluable. A significant quantity of meaningful clinical content is buried in the free text. Due to the unstructured nature of these notes, the conventional data-mining methods cannot be directly used. Many healthcare organizations are struggling to make sense of unstructured clinical notes and are losing out on the full potential of newly adopted Electronic Medical Records (EMR). The clinical details from these text complements the structured details and needs to be harnessed for decision-support, patient care and qualitative analytics. Deerwalk presents Clinical Notes Mining Service (CNMS), a feature designed to extract clinical elements from unstructured notes.

 

2. Solution

The critical challenge in building a Natural Language Processing (NLP) based system adapted to healthcare is to identify word or group of words of clinical significance; and to accurately map the identified textual segments into a hierarchical semantic network, into groupers, into codes, etc.


Making sense of unstructured clinical notes

2.1. IDENTIFY
Anne, a nurse, is speaking with a Member over phone and has Deerwalk's Everest application open to type in the conversation notes. Between sizeable pauses in typing, Everest automatically sends the notes to Deerwalk's Clinical Notes Mining Service.

The CNMS is an NLP based stack adapted to healthcare knowledge sources. It runs through the incoming text identifying word or expressions that could potentially be related to diagnosis or medication. It then sends a response to Everest with a list of matching terms and their associated codes.

2.2. CONFIRM
At real-time, Anne will see the list of matching terms on the same screen where she is typing the notes. She can then selectively mark the true positive matches. When she saves the notes, the marked diagnosis and medications will be automatically saved as well.
Clinical details that otherwise would have been lost is brought into structural realm and can now be utilized in existing data-mining processes and analytic reporting.

2.3. LEARN
After Anne has saved the notes, Everest sends Anne's selection along with the original response back to CNMS, where upon receipt, it is persisted to build a training set. As Anne continues to use this feature on Everest, CNMS will continue to improve.

3. Conclusion

Deerwalk's Clinical Notes Mining Service alleviates the burdens of healthcare companies when it comes to extracting structured details from clinical notes. It is an NLP based system that will automate the process of mapping diagnosis and medication codes from EMR clinical text. The clinical information thus obtained provides a more complete representation of patient's medical record. Apart from the obvious benefits of automation, this latest offering seriously augments the myriad services offered by Deerwalk's existing products in cost-effective care-coordination and incisive quality of care analysis.

Download PDF