At UMich, AI-based fraud, waste and abuse system aims to cut costs and protect patients


Healthcare costs have spiraled out of control in the United States. But studies have shown how as much as one-third of healthcare expenditures can be attributed to wasteful or fraudulent healthcare services.


There’s also evidence that underserved communities are at higher risk for receiving wasteful care – and exposure to unnecessary procedures that may potentially cause harm while increasing their financial burden from out-of-pocket payments.

“The Centers for Medicare and Medicaid Services, private payers and third party administrators have used various utilization management and payment integrity systems to analyze healthcare transactions in order to identify fraud, waste and abuse,” said Dr. Mohammed Saeed, cardiologist and researcher at the University of Michigan Medical School in Ann Arbor.

“However, conventional software platforms have thus far been unable to curtail the costs associated with fraud, waste and abuse,” he continued. “Among the problems, most conventional solutions are designed to analyze transactions for an average patient – who may not exist.”

Conventional payment integrity systems are imprecise and impersonal, and lack a deep and clinically nuanced understanding of the patient, the provider and the healthcare setting of the services being rendered, he added.

“The same procedure may be high-value and beneficial for one patient while being wasteful for a different patient,” he explained. “For example, a pacemaker implant for one patient may be highly beneficial with respect to improvement in quality of life, while providing little or no clinical benefit to another patient with different complex medical characteristics.

“Moreover, there are associated potential complications from a pacemaker implant and risk of post-procedure complications that patients face – thus, it is imperative that providers are dissuaded from rendering inappropriate procedures through use of ‘smarter’ systems,” he continued.

Healthcare transactional data also is inherently “noisy” and may include non-specific diagnoses and incorrect procedure codes – or missing data altogether, he added.

“The inconsistent fidelity of claims data and lack of personalization in conventional payment integrity platforms thus lead to errors in classifying fraud, waste and abuse services,” he stated. “Payment integrity systems that are rigid can incorrectly classify transactions as fraud, waste and abuse, which can lead to increased provider abrasion while introducing delays and inefficiencies in healthcare delivery.

“However, such systems also are vulnerable to evolving provider schemes designed to avoid detection and thus fail to prevent payments from going out for inappropriate care services,” he said.


IT vendor Health at Scale has created a special class of artificial intelligence-based algorithms designed to uncover high-confidence, low-value care that is missed by existing utilization management and payment integrity systems.

“Such algorithms have a deeper and more clinically nuanced characterization of each transaction by factoring in historical information about each patient while drilling down to each provider’s practice patterns; and they incorporate knowledge graphs with the most up-to-date evidence and guidelines regarding appropriateness of care,” Saeed explained.

The system also is optimized to avoid provider abrasion by focusing on a small fraction of providers that are responsible for the most egregious patterns of inappropriate care,” he continued.

Such algorithms are intended to be deployed in prior authorization and claims adjudication and payment workflows.

“These algorithms scan administrative claims and transactions across prior authorization, prepay and postpay workflows with the goal of stopping fraud, waste and abuse before care is delivered or payments are made,” Saeed noted. “The system also integrates with existing systems via API or standalone user interface.”


The University of Michigan Medical School does not have any hard results yet, but points to results Health at Scale has measured with a variety of its healthcare provider organization clients.

When a claim is passed through the vendor’s system, the response time is typically less than 200 milliseconds, Health at Scale reported. Across various deployments, the vendor is seeing between 3% and 7% of total spend that is missed by existing payment integrity tools that is then surfaced by systems for avoidance and recovery.

About 95% of the flagged activity that has been passed on to worker teams has been classified as fraud, waste or abuse. Health at Scale claims high accuracy because it applies multiple factors when reaching decisions about a claim, including historical information about each patient and provider, determination of whether care that is rendered is statistically similar to what other providers do, and factors in clinical nuance and evidence-based knowledge graphs.


“As a cardiologist, many patients visit my clinic seeking a second opinion regarding the potential benefit of an elective procedure recommended by an outside provider,” Saeed said. “Almost half the time, I reach the conclusion that the recommended procedure would likely not be beneficial or is the wrong type of procedure altogether.

“Various studies across other specialties have also found that patients often receive medically inappropriate elective procedures,” he noted. “Not only do these inappropriate procedures contribute to skyrocketing healthcare costs, but patients are exposed to risks of harm due to procedural complications or set patients up for cascades of inappropriate follow-up care.”

Healthcare organizations should pick a fraud, waste and abuse system that is simple to deploy at scale, and must consider their own personnel resources to understand whether other add-on services are needed to maximize impact, he advised.

“AI-based tools that can be seamlessly deployed and integrated with existing workflows can help organizations realize almost immediate ROI,” he said. “And I would recommend organizations think boldly about the prevention of inappropriate care services. Organizations should avoid unnecessary complex analyses, redundant pilot studies and tiny incremental changes that in effect do not change the status quo.

“In the long-term, I strongly believe that innovative and differentiated fraud, waste and abuse systems driven by AI will also impact provider practice patterns so that low-value care is less often rendered, and potentially shield patients from the harms of inappropriate or unnecessary care services,” he concluded.

Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him:
Healthcare IT News is a HIMSS Media publication.

Source link


Please enter your comment!
Please enter your name here