Cracking the Code of Medical Coding
How Nym Restores Trust in Revenue Cycle Management — Recovering Billions in Lost Income for Providers and Insurance Companies and Reducing Healthcare Costs
By Amihai Neiderman, CEO and Co-Founder, Nym Health
That healthcare is confusing, opaque, and expensive is a broadly discussed problem so far with few promising solutions. Last year alone, 71 percent of patients reported frustration with their healthcare experiences. What is less publicized is how frustrated medical practitioners and even insurance companies also feel with the clunky healthcare model, rife with poor incentives and inefficiencies that impede healthcare delivery and jack up costs. As a patient, it’s easy to have little compassion for hospitals or medical or insurance behemoths. But the reality is that solving their frustrations is an important way to improve the experience and meaningfully reduce costs for patients.
Adam Rimon, my former colleague in the Israeli Intelligence Corps, and I co-founded Nym Health to do just that. Our entry point is revenue cycle management (RCM), with a targeted focus on automating the historically byzantine process of medical coding that produces tens of billions of dollars of waste annually. It all started when I happened into a massive problem begging for an autonomous solution.
A Chance Discovery & a Problem In Desperate Need of Solution
When my wife was in medical school, she was doing clinical research that required poring over thousands of electronic medical records (EMR) to find patients who had the same surgery profile. Without a tool to search all EMRs simultaneously, her work was manual, slow, and cumbersome.
As a computer engineer trained to automate whatever could be automated, I was mystified. I served in unit 8200 of the Israeli Intelligence Corps (akin to the NSA in the US) for ten years. Part of our charge was to make intelligence efforts as efficient, and therefore autonomous, as possible. During a conversation with my wife about how she and other physicians at her hospital use EMR systems, she introduced me to an intriguing process problem whose potential automated solution could positively impact everyone in that value chain, and that was the problem of medical coding.
Healthcare Revenue Cycle Management: A Convoluted Process With No Clear Winners
All physicians and nurses are required to take notes in every medical encounter. This, for example, includes what patients say about their overall health, their medical history — covering previous diagnoses, surgeries and hospitalizations, among other things — and the medications they take regularly. These notes must then be translated into alphanumeric codes representing diagnoses and procedures, each of which comes with a set price tag. The insurance company analyzes these codes to determine how much they will reimburse providers and how much they will bill the patient.
Sounds simple — but it is anything but. Converting physician notes, often filled with technical terminology and context, into codes requires understanding the intricate classification system managed by the World Health Organization. The list of codes — which presently includes nearly 200,000, each representing procedures or diagnoses — is always growing in number and nuance, putting additional pressure on physicians to find the coding “needle” in a massive haystack.
What would already take days, assigning medical codes became even more unwieldy and high stakes to physicians. Increasingly, providers began outsourcing coding to overseas medical coders. While outsourcing takes this headache off their plates, it puts the work even further from the medical experts who can glean context from physicians notes to arrive at specific codes. This amounts to a significant error rate.
Such errors unleash a host of issues. In the atypical event that an insurance company will accept all codes for a claim, issuing the reimbursement to physicians can take months. In the more typical scenario, an insurance company will challenge or reject codes, leading to a lengthy reconciliation where physicians must revisit their notes and provide explanation of selected codes. At best, these delays snarl a provider’s revenue cycle, which can impact cash flow and payouts. Far worse, their reimbursements are reduced or denied, amounting to potentially serious financial challenges. Every year, doctors and hospitals lose $30 billion in unpaid and delayed reimbursements.
But the stakes can even be higher than lost revenue. A pattern of incorrectly coding medical interactions can raise the red flag of fraud, which is punishable by crippling fines or even prison. Yet, a gray area does exist, such that coders can strategically select certain codes over others to increase the likelihood of reimbursement — a practice born in reaction to a common belief that insurance companies unfairly reduce or deny far too many reimbursements.
All told, RCM exists in an environment of deep mistrust between providers and insurance companies that feeds off itself: providers are often incentivized to code strategically in hopes of extracting more reimbursement, and insurance companies are often suspicious of the codes they receive. In turn, insurance companies frequently audit providers, which is a stressful, expensive, and time-consuming process for all parties that only further entrenches mistrust and resentment.
Unfortunately, the negative effects of this complicated web do not stop there. Eventually, lost revenue on all sides trickles down to the patient. To recoup unpaid reimbursements, providers will increase the cost of care. To recoup costs of audits, insurance companies will increase premiums. This amounts to a one-two punch for patients, who ultimately feel the brutal brunt of RCM inefficiencies directly in their wallets.
Why Other Companies Have Failed to Solve the RCM Problem
Healthcare RCM is but one inroads to discovering untold efficiency and financial loss in the greater healthcare industry. Yet it’s a facet of the system that holds enormous potential to stem the tide of waste and actually reverse it, to the tune of tens of billions in savings annually. It is no wonder that others have tried to tackle this problem with machine learning. Yet, to date, all have fallen short of a successful solution.
Entrants into RCM have taken a computer-assisted coding (CAC) approach, which is based on natural language processing (NLP) that analyzes physician notes to produce a list of codes. CAC tools interpret language at face value and therefore cannot capture the nuance in physician notes that contextualize and justify chosen codes. As a result, CAC’s error rates remain high, and it is also incapable of providing an “audit trail” to explain codes in the event of an audit, putting the onus back on providers to produce evidence. The net effect is little improvement in reimbursement rates and timelines and in insurance companies’ confidence in the codes they receive.
How Nym Cracked the Medical Coding Code
When Adam and I saw that many companies had failed to find an AI solution, we knew we had to take a different approach. Adam, who studied computational linguistics after his military service, used his expertise to pinpoint how and why NLP, in failing to understand a medical encounter, failed to accurately assign medical codes.
We knew that our technology could not be a standard off-the-shelf AI or NLP. Rather, it had to be built specifically to understand the explicit and implicit nature of medical terminology and physician notes. To achieve this, we hired a medical team to work in close partnership with our linguists and engineers. Instead of having free rein to build a typical AI, our engineers work in collaboration with linguists and doctors to bring unprecedented sophistication to our technology so that it is capable of interpreting the context of medical terms and physician notes.
We refer to our proprietary technology as Clinical Language Understanding (CLU) — custom-designed to read and analyze physician notes in about five seconds and with near-perfect precision. Not only does our CLU have a 98 percent coding accuracy rate, it also autonomously creates an audit trail of every medical encounter that gets coded. If an insurance company audits a Nym partner, all the partner has to do is log into our dashboard and share the auto-generated summary that explains why every code was selected.
The Invaluable Power of Trust
In the years since our founding, we’ve deployed our CLU across forty hospitals, including Geisinger, and several additional academic medical centers. Nym enables partners annually to save millions in lost revenue and operating expenses. Yet, perhaps the most invaluable impact of Nym is the one that cannot be quantified: trust. Several of our hospital partners have reported an elimination of audits since using Nym, which also means precipitous drops in insurance companies’ audit expenses.
Investors have also taken note. We are excited to announce our $25 million in additional funding, led by Addition. This latest raise follows our November 2020 Series A, bringing our total funding to $47.5 million. Addition joins existing investors GV, Dynamic Loop Capital, Tiger Global, Bessemer Venture Partners, and Lightspeed and investors Zach Weinberg and Nat Turner from Operator Partners. Thanks to our investors, Nym is actively scaling our team so that we can, in turn, scale our product for rapid adoption in the numerous hospitals currently on our waitlist, with more joining every day.
Financial waste in the healthcare industry costs more than dollars. It has a human toll, in the form of hassles that distract physicians from focusing on delivering care and contribute to increasingly prohibitive costs for patients. With each new partner, Nym can restore that much more trust, recoup millions for providers and insurance, and liberate physicians to focus on providing the best care they can to as many patients as possible.
