Ten years ago, Juergen Fritsch, chief scientist at Franklin, Tenn.-based M*Modal, was a postgraduate student at Carnegie Mellon, working on research for the government. “We were focused on a project eavesdropping calls," says Fritsch. "What emotional state [people] were in, and so on. We learned a lot about how people talk about things."
[See also: Speech recognition market poised for growth]
Since then, Fritsch and his colleagues have taken what they learned and developed this into natural language understanding (NLU) technology for healthcare. In a nutshell, NLU refers to a computerized system that can turn spoken and typed words in to structured data.
Fritsch spotlighted five areas where NLU is helping transform healthcare.
[See also: M*Modal to go private in $1.1B deal]
1. Conversational documentation. Clinical documentation is an important aspect of a physician’s work. They need to have their notes converted in to clinical information that is actionable, such as a call to prescribe medication. Normally this is done by inputting dictation or text to a computer system. An NLU system can automatically take this information and interpret the narrative documentation (a conversational description of the patient’s visit) to find actionable items.
Using NLU in this capacity can improve the efficiency of the documentation process. According to Fritsch, 60 percent of all clinical documentation is done in an unstructured form that a computer can’t understand. “Just by listening in on a physician conversation saying that they want to put a patient on a specific medication, a computer can take actions on that,” he says.
2. Computer-assisted coding (CAC). Once a doctor treats a patient, someone needs to look at the physician’s notes and assign a billing code in order for the to receive payment. Fritsch says CAC programs that use NLU “under the hood to understand the diseases and the diagnoses” can pull this information from unstructured physician notes and automatically assign a billing code.
“This helps the hospital do billing timely and efficiently, and it helps them get their money sooner,” says Fritsch. He adds that it can help greatly in the transition to implementing ICD-10 billing codes, as a NLU system can understand a conversational description of a diagnosis and check it against ICD-10 listings.
3. Clinical documentation improvement (CDI). Accurately documenting the condition of the patient and what procedures were performed “drives how much money [healthcare providers] get and how soon they get it,” says Fritsch. “If it’s not documented, it didn’t happen.” As procedures and billing codes become more complex, it is easier for a provider to accidentally leave a treatment undocumented, therefore losing the opportunity to receive payment. An NLU system can identify missing, incomplete or inconsistent information in a physician’s documentation.