Currently, 18 percent of the GDP in the U.S. – $3.65 trillion – is spent on healthcare. At least $300 billion is lost every year to fraud, waste and abuse (FWA). With up to 10 percent of healthcare payments falling into this fraudulent spend, payers are turning to artificial intelligence (AI) to stop the bleeding.
Payers have long struggled to identify FWA, which increased during the pandemic due to laxer rules to hasten access to care.
A recent speaking session hosted by the AHIP Institute and Expo explored the benefits of AI for detecting fraud in the healthcare sector. Beth Griffin, VP, Healthcare Vertical, Mastercard, and Tim McBride, Director, Healthcare Product Development and Innovation, Mastercard, were joined by David Cusick, Principal of Milliman Inc. and Jessica Gay, Vice President and Co-Founder, Integrity Advantage.
During their session Preventing FWA, mitigating losses and improving efficiencies with AI, the panel discussed key ways that fraud affects healthcare organizations.
At the root of FWA is the fact that billing for healthcare is complex, multi-faceted and includes vast amounts of data. In a world where Medicare processes more than 4.5 million claims a day, it’s nearly impossible for humans to detect complex fraud without the help of AI.
Not all waste is the result of fraud. Claims may include billing errors – including incorrect or changed procedure codes – or contain unnecessary procedures by over-cautious practitioners. These also cost the healthcare system millions of dollars.
Here are five insights from the session, along with some additional research:
Insight 01 – Advanced AI models recognize complex criminal behavior
Source: Forbes magazine
Forbes magazine compared healthcare’s $300 billion annual FWA losses to other illegal activities, such as ransomware attacks and credit card fraud (see table above).
Healthcare fraud outweighs securities fraud by more than 650 percent, and credit card fraud by almost 1700 percent. It even overshadows crimes such as the Enron and Madoff scandals at $74 billion and $65 billion, respectively.
The FBI reported back in 2010 that “Estimates of fraudulent billings to health care programs, both public and private, are estimated between three and 10 percent of total health care expenditures.” Today, the problem continues to grow as law enforcement uncovers complex, organized crimes.
Money laundering contributes to the FWA issue
Public and private healthcare plans are now the top vehicle to launder dirty money in the U.S., according to the U.S. Treasury Department. The 2018 National Money Laundering Risk Assessment (NMLRA) reported that over $110 billion is processed annually, constituting one-third of all illicit funds laundered in the U.S.
COVID-related schemes
Fraudulent behavior follows current trends, and a global pandemic provides the perfect environment for fraudsters. In May 2021, 14 defendants were charged for their alleged participation in various healthcare fraud schemes that exploited the COVID-19 pandemic, resulting in more than $143 million in false billings.
“It’s clear fraudsters see the COVID-19 pandemic as a money-making opportunity,” said Christi A. Grimm, Deputy Inspector General for Investigations of the U.S. Department of Health and Human Service.
AI and complex cases
Artificial intelligence models identify healthcare claims fraud, prescription abuse, upcharges and many other FWA challenges.
Mastercard® Healthcare Solution’s AI creates an end‑to‑end suite of profiling and modeling capabilities that continuously adapt and improve results to recognize the patterns of complex fraud. This seamless combination of advanced AI tools delivers personalized decisions in milliseconds to payers, insurance companies, business leaders, or other entities. Models evolve at scale with their data, increase detection rates and decrease operational costs and false positives.
Insight 02 – Artificial intelligence drives healthcare SIU efficiencies
Operating a full-time team to create and maintain a rules-based healthcare FWA system is expensive. Legacy, rules-based systems are hard-coded and feature thousands of algorithms. These rules are self-limiting and lose value over time as fraudulent behaviors evolve. The research, analysis and work to update the rules are expensive and time-consuming.
According to Tim McBride, updating rules can take an average of 240 human hours over the lifetime of a legacy solution. Up to 40 percent of alerts investigated are false positives, he added, at a minimum investigation cost of $7,500 each.
Compare that to using a single advanced AI model that provides its own updates through real-time, continuous self-learning, adapting to the changing behavior it observes. It also learns through the final disposition of cases and claims, so it can identify valid claims and flag new behavior patterns that may indicate fraud.
Special investigations units (SIUs) can focus on real fraud
With a 20 times reduction in false positives, Mastercard’s AI flags only highly likely fraud alerts to investigators for follow-up. SIUs are more efficient and successful, focusing on suspicious activities and complex fraud schemes rather than chasing down false leads. AI saves both time and money.
Insight 03 – AI detects healthcare procedure code errors
Medical billing errors happen when an incorrect diagnostic or procedure code is used on a patient’s bill. Credit rating agency Equifax reported that hospital bills of more than $10,000 have an average error of $1,300.
Complex medical procedures are more difficult to break down, but some errors are as simple as billing for examining both legs when only one was examined. Others are intentionally upcoded when patients are billed 45 minutes for a short consultation, or unbundled, when practitioners bill for each separate step of a procedure rather than using the single code for the overall procedure.
High-risk healthcare providers
The Government Accountability Office (GAO) found that 62 percent of Medicaid fraud cases were attributable to providers and in 14 percent of the cases examined, beneficiaries were complicit. While the study was written in 2016, recent reports show that fraudulent billing, kickbacks and medical identity theft continue to dominate this area.
Milliman used Mastercard Healthcare Solutions’ AI to detect $239 million in historical provider fraud and procedure code errors for one of its health plan payers. “Mastercard’s AI provided the insights we were hoping for,” said David Cusick, Principal and Healthcare Technology Consultant at Milliman.
Drawing data from NCCI and public records
In legacy FWA solutions, payers’ teams must update the National Correct Coding Initiative (NCCI) codes manually, creating new rules and possibly deleting rules for prior claims. Incorrectly coded claims can slip through, be denied, or worked around by fraudulent providers, whereas an effective AI model automatically updates NCCI codes and continually monitors for changes. Access to public records also allows the AI model to screen for provider sanctions and repetitive behaviors that may go unnoticed such as frequent prescriptions for expensive medications, or suspicious practitioner addresses.
Insight 04 – AI identifies waste
Waste can be an institutional problem and unless there are specific rules written for the hundreds of thousands of types of potential waste, legacy systems won’t detect them. While some waste is intentional, some reflects overly cautious practitioners or procedures performed without cost considerations.
California case an eye-opener
Cataract surgery is the most common elective surgery among Medicare beneficiaries, with 1.7 million procedures performed annually. It has less than a one percent risk of major cardiac events or death. But when two academic medical centers, Los Angeles County and the University of Southern California evaluated their pre- and post-operative practices, the results were shocking.
Low-risk patients were subject to the following unneccesary pre- and post-op testing:
- 90% pre-op and 24% post-op chest x-rays
- 92% pre-op and 37% post-op lab work done
- 95% pre-op and 29% post-op EKGs
In Washington State, the Choosing Wisely campaign was launched by the medical association to educate healthcare providers and patients about the high cost and potential harm done by unnecessary procedures and tests.
Patterns of waste
When the AI model reports inconsistencies with NCCI codes, the flagged procedures or practitioners may be part of a localized or systemic waste problem. Depending upon the AI model’s predetermined tolerance threshold, it will alert investigators after a certain number of unusual codes are billed by each provider or organization.
Insight 05 – AI streamlines overall operations in organizations
AI can be used across healthcare organizations to streamline operations as each function is related to another. With over 100,000 decisions per second, Mastercard’s AI can process related files quickly and effectively, freeing up staff for other duties. Efficiencies can be potentially felt across the following areas:
- Pharmacy benefit manager
- Utilization management
- Claims processing
- Appeals and grievances
- Transportation
- Behavioral health
- Dental
- Vision
- Sales/marketing
- FWA analytics
- Peer review
- Pre-authorization
- Network management
- Credentialing
- Coordination of benefits
- Subrogation
Predict, analyze, prevent
With overstretched investigators and thousands of transactions a day, payer organizations need to become efficient, effective units that detect fraud and identify waste. By implementing AI, SIUs can become tighter operations that are focused on complex fraud schemes and predictors of upcoming trends.
Streamlining rules, building unique self-learning models that create one-to-one profiles and reporting in real time allow fraud investigators to work with higher returns. Whether payer organizations are set up for prepay or still working within the pay-and-chase paradigm, AI helps to improve efficiency and increase ROI.
Learn more about preventing fraud, waste and abuse in healthcare by downloading our Ebook, Prevent and save: advanced AI for fraud, waste and abuse.
