3M and CodeRyte

Today, 3M HIS announced it has acquired CodeRyte. As division scientist, this is exciting news from a technology perspective. Here’s my take on what it means for NLP:

The distinctive feature of CodeRyte’s technology is its strong statistical NLP capability. As I discussed in the 3M white paper, Auto-coding and Natural Language Processing, statistical machine learning systems offer the possibility of significant accuracy improvements in data-rich environments, compared to traditional rules-based approaches. A good example is in the outpatient coding environment, where data volumes are large, and CodeRyte has a history of performing very well compared to other systems.

There are multiple ways to boost the accuracy of statistical NLP, and the combination of 3M and CodeRyte will allow us to pursue several paths to the direct benefit of end users.  For example, the combination will allow us to:

  1. Increase the range of expertise. 3M’s depth of expertise and experience in coding and reimbursement will provide the opportunity to broaden CodeRyte’s reach to specialties it has not yet entered.
  1. Boost the accuracy of data. A statistical computer-assisted coding (CAC) engine can learn by observing data from multiple sources. By combining CodeRyte’s statistical NLP with 3M’s expertise and substantial investment in deriving the right codes for each patient record, providers will see higher CAC accuracies.
  1. Improve the statistical features. A statistical system can relate features in the input documents to assigned codes. The better the choice of features, the better the result from the statistical system. 3M’s three decades of experience and investment in coding offer the chance for deeper, more robust feature selection.

The combination of CodeRyte’s NLP with 3M’s clinical content and expertise brings many exciting opportunities for innovation in the application of NLP to health care. I look forward to discussing many of the advances that will come from the combination of our capabilities in future blogs.

Richard Wolniewicz is Division Scientist, Natural Language Processing, for 3M Health Information Systems.

3 responses to “3M and CodeRyte

  1. Per one Physician

    I understand that in terms of relationships, CodeRyte is an acquisition and Nuance is a partnership. But how does 3M explain these relationships — having two, different NLPs with which to use?

    • Richard Wolniewicz

      A common misconception of NLP systems is that they are a monolithic black box. An NLP system is an inherently complex combination of many subsystems and separate models unified together to produce a consistent result. For instance, the IBM Watson system has many dozens of independent sub-models for reasoning about time, spacial relationships, people and organizations, etc. In answering a single Jeopardy! questions, thousands of parallel processes ran many separate models, producing dozens to hundreds of possible results, and separate models assessed the confidence of each to produce a final answer.

      In the same way, 3M’s NLP is a combination of multiple sub-models into a single coherent whole. Nuance’s CLU brings deep health-care-specific rules-based NLP, which can surface complex health care concepts, both supporting CAC for complex coding situations where there is not sufficient data to train a statistical model, while also providing better features for boosting statistical models. Likewise, CodeRyte’s Confidence Assessment capability can score the CAC outputs to facilitate better coder workflows and can be applied directly to any NLP outputs, including those from CLU.

      The acquisition of CodeRyte enables us to deepen the integration between Nuance, CodeRyte, and 3M NLP technologies to continue improving a single, comprehensive health care NLP system across the continuum of care.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s