Managing Interoperability & Booming Data With ML, NLP, and AI

Good news!  Health IT vendors are producing better and better interoperability scenarios every day.  The bad news … interoperability is not finished, even with the established IT vendor platforms.  New sources of information are forming outside of these Health IT vendors’ purview at rates too fast to keep up with the creation of needed interfaces.  That, or the interfaces simply go uncreated because they’re too expensive for health systems to justify. 

Traditional sources of health data, though improving as noted, are not perfectly integrated and some silos remain across different organizations.  As such, they still generate a great deal of unstructured documentation:

• Patients seeking service from Health System B while a patient of Health System A may have to have their EMR record from Health System A’s EMR PDF’d and sent electronically or even have it physically printed and eFaxed to Health System B. 

• Patients receiving lab work from an independent lab have their lab results eFaxed to their Health System.

• Patients receiving provider care from an independent clinic have their appointment summaries eFaxed to their Health System.

Additional traditional sources where integration remains at arm’s length include:

• Patient Assist

• Insurance Claims

• Research

• Public Health

• Accounts Payable / Invoicing

Lastly there’s an explosion of new sources of “user generated” patient data:  Wearables.  Wearable categories include:

• Fitness and health trackers

• Smartwatches

• Smart rings

• Connected patch

Though there is a drive toward integrating wearables, many remain non-interfaced and the only means of incorporating the data into the medical record is through a “print and eFax” process.  

What all this means is that we have a problem that isn’t going away soon.

All these sources of documents where interfaces are non-existent means a substantial amount of information is reaching health systems in the form of unstructured documents.   Depending on health system size, this can mean millions to even hundreds of millions of pages per year (in the largest of health systems).  Getting the documentation and the data “attached” to the patient EMR entity is a time-consuming human-centric process.  This creates a tradeoff between the cost of input versus ease of access to information and, unfortunately, sometimes accuracy of information, which can impact patient care outcomes.

There’s good news.  Machine Learning (ML), Natural Language Processing (NLP), and Artificial Learning (AI) can play a key role in solving the problem.  By applying ML, NLP, and AI to the classification of documents, application of sub document types, indexing to patient medical record numbers (MRN), indexing to physician orders and encounter numbers, and even abstracting discrete data can become much more automated.  You can have an efficient, cost-effective process that presents accurate data to the provider in a format that is easily retrievable.

If you’re curious to know how Extract Systems utilizes ML, NLP, AI and other tools toward this end please send us a message or give us a call and we’d be happy to give you a personalized demonstration.