It didn’t take long to realize that there was a gap in the system, when former intelligence officer Amihai Neiderman’s partner was manually reviewing thousands of medical records, in search of the one she needed for her research. “The unit in which I used to serve has so much data coming in that we have to rely on automation. Otherwise, we will drown. Everyone learns programming to automate their work. So, I told my wife, it seems to me that your manual work here is something we need to fix,” Neiderman recalled in his recent interview with AIMed. 

After gaining approval from the hospital for access to their electronic medical records (EMRs), Neiderman worked out a quick Natural Language Processing (NLP) solution to go through all the medical data to assist his partner with her research work. Almost immediately, Neiderman saw a door had opened for him. With the help of a close friend, a fellow intelligence officer with a computational linguistic background, the duo started what is today known as Nym Health. 

Autonomous Medical Coding 

Neiderman and his co-founder are both avid problem solvers, so they would not settle on a solution which tries to solve anything but the biggest challenge. That’s why Nym Health targets medical coding, an area that is labor-intensive and poses a huge hurdle to the present-day US healthcare space, as it is the only way hospitals can get reimbursed for treatments. Unlike a traditional artificial intelligence (AI)-machine/ deep learning-driven approach, which succumbs to a lack of transparency and inaccuracy, Nym Health focuses on the use of Natural Language Understanding (NLU) for their solution.

Amihai Neiderman (left) and Adam Rimon (right), co-founders of Nym Health.

NLU is an evolution of NLP. It is armed with medical knowledge and instead of processing a text, it works to understand it. “This is what computational linguistics does. It mimics what goes on in the brain when one is reading. While reading, we make sense of every single word and sentence, trying to build a narrative. Nym does the same,” Neiderman explained. After understanding the text of a meeting between patient and physician, the medical narrative is then reconstructed to extract insights, before the doctor’s notes are finally assigned with medical codes for billing. For image-based data, the team will concentrate on the physicians’ notes or research papers to interpret the images. 

Neiderman admitted that understanding physicians’ interpretations and trying to understand the narrative of the interaction is what the company spends its days on. Nym Health has hired a group of computational linguists and physicians with computer science backgrounds to handle jargon, clinical information, and decisions that are embedded within the text. 

Developing White Box AI

This June, Nym Health raised $6-million in seed funding led by Bessemer Venture Partners. Two weeks later, the company announced its partnership with HealthChannels, one of the largest providers of medical scribes and coders in the US, to develop autonomous clinical coding software, which provides an explanation for its billing rationale and creates an audit trail for reimbursement. Neiderman said Nym Health is very lucky to establish this collaboration, which is not just built on trust, but also a channel to showcase what they are able to achieve.

Nym-generated audit trail (i.e., reasoning behind the provided codes)

According to Neiderman, presently there is a person sitting next to a physician to take down notes, to save physicians’ time spent in front of the computer. There are about 50 million encounters of such human-centric tasks taking place per day. If the medical institution is keen on increasing its efficiency, they will have to hire thousands of coders to take over the responsibilities. With Nym Health, the entire pipeline will be automated with autonomous coding. This will promote a rapid scaling of businesses. 

In the long run, Neiderman expressed his interest in the creation of “explainable AI.” He said the system used by Nym Health explains all the processes and reasoning behind the codes provided – something that traditional NLP technologies are unable to do. So, the solution is spared from the transparency concern that surrounds many AI solutions today. “We are very open, we wish to form new partnerships with research centers or other tech companies, where we can further utilize our technologies,” Neiderman added. 

*

Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.