Huang’s Startup Cuts Through Opaque Insurance Practices

Ninety-two percent of Americans are covered by health insurance, allowing patients to access medical care that could otherwise put them into debt.  However, the staggering complexity of medical claims processing leads to widespread burdens on healthcare providers, ranging from financial strain to staff burnout.

“When a healthcare organization treats patients covered by an insurance payer (company), it must navigate dozens of distinct payer contracts, fee schedules, and guidelines,” explained Jian Huang, a professor in the Min H. Kao Department of Electrical Engineering and Computer Science (EECS). “In fact, even working within a single payer-provider agreement, healthcare revenue specialists routinely face claim denial and underpayment patterns that are impossible to track manually.”

This overwhelming complexity combines with a constant influx of new claims data to create enormous workloads, contributing to exceptionally high staff turnover rates—as high as 20 to 30 percent of healthcare revenue specialists per year.

Huang is an expert in the large-scale visualization of raw data using artificial intelligence (AI). His research group pioneered Visualization as a Service (VaaS), a sophisticated cloud-based method that can help large organizations discover, share, and use deep data insights to make advantageous business decisions. He and two of his former students, Tanner Hobson (PhD/EECS, ’23) and James Osborne (BS/EECS, ’11), realized that VaaS could transform healthcare finance by casting light on hidden patterns in health claim data—benefitting healthcare workers and patients alike.

Thanks to support from the University of Tennessee Chancellor’s Innovation Fund and early investment from the UT Research Foundation, Huang, Hobson, and Osborne created a privacy law-compliant VaaS tool that can help healthcare organizations to optimize claim processing.

Their startup company, VisualizAI, launched its commercial platform in March; one month later, Huang was invited to participate in the highly selective Harvard Business School Foundry—a program supporting deep-tech ventures with high potential impact.

Huang also shared his team’s insights into using AI to parse dense claim processing data as a featured speaker at the national AI Readiness for Medicare Effectiveness conference in Washington, DC, in July.

“VisualizAI is fully aligned with the University of Tennessee’s mission to perform highly visible research that also makes truly positive changes,” Huang said. “If hospitals can’t financially survive, patients suffer.”

Claim Processing and Code Complexity

When a patient enters a medical facility, they describe their symptoms in plain language. Doctors and nurses translate those symptoms and their own analysis into diagnostic codes—often building on their own experience or preference to decide between closely related codes—then translate their treatment decisions into procedure codes. The coded information is entered into an Electronic Data Interchange file (837 file) and sent to the insurance payer.

The payer’s automated system then evaluates whether the procedure and diagnosis follow the payer’s standards, adding payment adjustment codes or other changes as needed. The adjusted payment gets sent back to the medical organization along with an Electronic Remittance Advice file (835 file) annotating how much money was sent for each claimed procedure. Much of the time, the 835 will include underpayments (payments below the contractually allowed amount for a procedure). Enough underpayments strain medical providers, hampering patient access to medical services—sometimes in areas with the most significant unmet needs.

The organization and proper use of diagnostic, procedure, adjustment, and other codes is immensely complex, with guidelines that change frequently. In the case of severe underpayment or misfiling, a healthcare practice can file an appeal, cycling the claim back through the practice and payer’s claims processing systems.

Though the constant, overlapping changes are impossible for a human to tease apart, individual healthcare revenue specialists may learn patterns for certain payers that can help them file claims more smoothly in the future.

“Unfortunately, if that know-how resides in people’s heads only, that kind of precious knowledge rarely spreads across the entire organization,” Huang said.

A Scalable Solution

VisualizAI leverages AI and the specialized VaaS architecture Huang’s group developed to compile 837 and 835 file data streams across an organization. The platform automatically identifies patterns in claim processing that matter to the healthcare provider’s bottom line—transforming disconnected patterns previously only recognized by individual revenue specialists into what Huang calls “institutional intelligence.”

With deeper insight into the flow of claims processing between a healthcare provider and each payer, clinical teams and healthcare administrators can achieve a higher rate of payment collection while putting less pressure on each revenue specialist. With healthcare providers in a better financial position, patients can also benefit from greater quality and availability of care.

“With VisualizAI, healthcare leaders can understand exactly how payers adjudicate their claims, coach their operational teams accordingly, and improve their claims processing techniques in real time,” Huang said. “By translating dense claim processing data into clear, explainable next steps, our technology cuts through the ‘paralysis of analysis’ and transforms big claims datasets from a hindrance to an asset.”

Contact

Izzie Gall (egall4@utk.edu)