SmarterDx raises $50 million to fill in clinical documentation gaps with AI

The company uses its algorithm to make sure that 100% of charts are 100% accurate

Most clinical documentation and review is still done manually, often by clinical documentation improvement nurses, whose job it is to help patch gaps. This all leads to human errors, and even small errors or omissions in clinical documentation can lead to missed, rejected, or underpaid claims, resulting in less revenue for a hospital. 

"The average hospitalized patient who's in the hospital for five days, they have 20 to 30 different diagnoses that are managed by half a dozen different clinical teams involving labs, medications, imaging, and all sorts of stuff," said Dr. Michael Gao, co-founder and CEO of SmarterDx, a company using artificial intelligence to automate clinical documentation.

"The likelihood that somebody misses a single thing, not even in terms of treatment, but just in terms of the corresponding documentation around that statement, is actually pretty high. And so, plugging that hole, and getting into 100% accuracy across 100% of charts, is where we felt the opportunity was."

Now the company is planning on incorporating more data into its AI, so it announced a $50 million B funding round on Tuesday led by Transformation Capital, along with continued investments from Bessemer Venture Partners, Flare Capital Partners and Floodgate Fund. This round brings its total funding to $71 million to date.

The idea for SmarterDx came to Gao when we was working at as a physician at New York Presbyterian, where he also helped lead AI for the health system.

"What I noticed is that when we did things like mortality review or outlier review or certain types of quality review, manually, after existing software and existing technology, that there were still inaccuracies in documentation that would be spotted," he said.

"I realized that if, after all of these vendors, there are still things left on the table, then perhaps there are ways to leverage new developments in AI to better capture these findings, and better help hospitals capture these findings."

SmarterDx taps into clinical data from a hospital's EHR, as well as financial data, which may potentially from a separate billing system, to determine is something is missing.

For example, if a patient has a blood pressure of 180 over 100, and they get prescribed a high blood pressure medication, even if the doctor forgot to write the words “high blood pressure” somewhere on the chart, SmarterDx's AI can guess that high blood pressure was something that was being treated. Conversely, if there is documentation, but there's just no corresponding evidence, the AI can detect inaccuracies in that direction at well. 

"We take data that corresponds to the care delivered. If you imagine the labs, medications, the orders, the vital signs, and so on and so forth, we use that fingerprint to understand what diagnoses were being treated during that patient's visit to the hospital," Gao explained.

What SmarterDx doesn't do is actually make any changes to the patient record itself: it supplies recommendations to nurses and then, subsequently, to the doctors, who use their own clinical judgment to  whether they want to make the change or not. 

Currently, the software is being deployed at about 15 health systems, including top 10 US News and World Report academic medical centers, small independent community hospitals, and relatively large health system chains.  

From the perspective of a hospital, this type of solution can mean the difference between being paid and not being paid: if care was provided but it wasn't written down then it's hard to argue that a payer should pay you for that care because you don't have the corresponding evidence around the care that was provided. 

After using SmarterDx, hospitals see an average increase on the order of $2 million per 10,000 discharges, which can easily be over $10 million for a health system. 

"As the average hospital has an operating margin of 1.6%, which means that they're living paycheck to paycheck, effectively, and so this revenue that we're finding, which is fair reimbursement for care that they're delivering, helps them invest in people, invest in more MRI machines, and make the requisite investments to expand the care that they're able to deliver as well," said Gao. 

Health systems, meanwhile, get better insights into where to allocate their resources. So, while, as they otherwise wouldn't have a perfect inventory of the care that's being provided to patients, which makes it very difficult to know where to put their money. 

SmarterDx also helps improve productivity, allowing doctors and nurses to operate at the top of their licenses, while also decreasing burnout by taking away menial tasks. 

"Oftentimes, people equate burnout with working hard but I don't think that's quite the right equation. Everybody who joins healthcare knows that they want to work hard and want to provide great care for patients. Burnout is more when people are doing work that is low value," Gao explained.

"These clinical documentation nurses who are reviewing records and looking for gaps, when they're using their clinical judgment, and identifying sophisticated patterns to help doctors be more accurate and more complete, that's great; when they're scrolling through thousands of labs, and hundreds of medications, and just reading through each value one at a time, that's what causes more burnout. And so, by automating the boring stuff, and allowing the person to practice at the top of their license, and to focus on what they've been trained to do and what their years of experience allows them to do effectively, that's how to decrease burnout."

The new funding will be used to drive product innovation that the company says will create additional value for health systems. That includes adding new clinical and financial data integrations, such as outpatient data. while refining its algorithms that identify new revenue and quality opportunities.

"Getting this additional data and these additional integration points will help us create more algorithms to cover more area within the hospital and to, at the end of the day, create more value for our customers," said Gao.

It will also go toward supporting scale as the company, including building out its team, as it expands from 15 health systems to what the company hopes will be hundreds in the future, so it can improve a health system's processes in the future as well.

"At the end of the day, it really is about 100% of accuracy in 100% of charts. I can certainly appreciate that that sounds administrative or, even banal, to some degree, but these days, in the information age that we live in, the accuracy of the data that the hospitals have underpins everything we do," Gao said.

"You would not be able to run McDonald's if each McDonald's store could not give you accurate reporting of what their customers were buying. And so, fixing this information layer, and making it 100% accurate, is what will allow hospitals to actually improve their operations, to track the effects of their care, and improve their delivery and allocation of resources for that care. It's about converting what is, today, messy clinical data into accurate and actionable insights that the hospital can really act on."

(Image source: smarterdx.com)