One (Real) Solution To Mitigate Medicaid Fraud and Waste: Artificial Intelligence

Editors: Patti Boozang and Amanda Eisenberg 
Authors: Jared Augenstein, Anne O’Hagen Karl, Randi Seigel and Kaylee O’Connor

Click here to subscribe to receive more content like this!
Help shape The 80 Million: What should we write about next?

tl;dr

• States have new opportunities to use artificial intelligence (AI) to run better Medicaid programs, including improving quality and efficiency, and reducing fraud and waste.

• Three areas states should consider for AI use include: (1) eligibility and enrollment; (2) program oversight and integrity; and (3) network adequacy and access monitoring.

• Investing in AI leveraging the federal match rates for Medicaid administrative activities is essential to states accessing and deploying AI technology.

The 80 Million Impact

A new, primary talking point among Republicans seeking to rationalize proposed deep cuts to federal Medicaid funding is that the nation needs “to address fraud and waste in Medicaid.” We don’t know of a single state Medicaid director who would disagree with that statement. State Medicaid agencies serve dual roles of regulator and payor — chief among those roles is stewardship of federal and state Medicaid funding to ensure that the lion’s share of resources goes to health care coverage and access for Medicaid enrollees. Slashing federal Medicaid funding isn’t a solution to fraud and waste in Medicaid, but AI is.

Three priority areas in which Medicaid programs and the Trump Administration should consider investing in and deploying AI technology include:

• Eligibility and enrollment

• Program oversight and integrity

• Network adequacy and access monitoring

Medicaid application and eligibility

Medicaid application and eligibility verification is complicated, time consuming, and riddled with red tape for people and Medicaid agencies. Applications for coverage are tedious and confusing. Detailed questions about household sources of income and composition are based on arcane and cumbersome definitions. There are myriad additional questions about employment, potential disability, pregnancy, and other health coverage an applicant may have. Indeed, the Healthcare.gov “single streamlined application” for health coverage, including Medicaid and CHIP, is a dozen pages of instructions and questions. Many applicants (at least 12%) wait 45 days or more for states to determine Medicaid eligibility based on their applications. And while states must ensure that people can submit applications and renewal forms online and by phone, paper-based processes remain the norm in many states.

AI tools can help by translating Medicaid applications and renewal forms into more digestible, culturally, and linguistically appropriate formats. The technology can even “pre-populate” in real time with available (and accurate) data sources that can be mined for applicant eligibility information. It would be game changing for people applying for and renewing coverage while reducing Medicaid program paperwork, caseworker burden and administrative costs. AI tools can also be deployed to help states automate review of applications and renewal forms to improve speed and accuracy of eligibility determinations and lower the resource burden for Medicaid program staff.

Program oversight and integrity

AI is known for its ability to quickly identify patterns and recognize irregularities within large sets of data. AI capability can be crucial for weeding out fraud and abuse in Medicaid claims payment systems by identifying providers upcoding or billing for services that weren’t deliver. As always, human oversight is critical to ensure that patterns potentially signaling fraud by an AI system are, in fact, fraudulent.

Today, most Medicaid programs are run through Medicaid managed care plans. States spend significant resources monitoring managed care plans. This includes fraud and abuse oversight. But some of the most resource intensive oversight activities in Medicaid agencies relate to ensuring that plans are complying with their contract terms and meeting important quality and care delivery metrics. Too often, states are awash in plan data and lack the capability and bandwidth to analyze that information fully. Here, AI can pitch in to assess data and identify underperforming plans.

AI can also help states ensure that program oversight is responsive to enrollee and provider complaints regarding operational or care delivery issues in Medicaid. For example, enrollees and primary care providers might complain about access issues for people who need mental health outpatient therapy. States can us AI to efficiently integrate and mine data from diverse data sources (e.g., claims systems, complaint databases, population health management systems and provider network data) and assess those complaints, including whether they are targeted geographically or for certain populations.

Network adequacy and access monitoring

Another powerful AI use case is looking at provider networks to identify and intervene where there are potential provider access issues. Medicaid agencies receive reports on grievances and appeals from providers. Similarly, the agencies receive complaints from members that certain doctors have refused to treat them due to their insurance. They’re also seeing claims data that show how frequently a provider is accused of denying access to Medicaid patients.

AI is good at synthesizing information and finding patterns, especially across diverse types of data sets, whether it's a grievance letter or a structured spreadsheet. Medicaid agencies can ask AI to find out if a provider really is available to see patients and question why they’re an outlier. Again, it requires a human touch, but an AI algorithm could help provide insights and ensure a state’s provider directory is up to date and accurate.

The Bottom Line

The integration of AI into Medicaid program oversight, eligibility operations, population health management and quality improvement holds tremendous promise to improve Medicaid program efficiency and quality, making it work better for the 80 million people enrolled in Medicaid. Deploying AI in Medicaid will require technical assistance and technology investment at the state program level. And as our colleague Kevin McAvey points out in his recent 80 Million blog, The Stealth Cut to Federal Funding That Would Undermine Medicaid Efficiency, Integrity, and Innovation, Congressional proposals to cut the federal Medicaid spending by reducing Medicaid administrative matching rates for things like Medicaid IT systems would be antithetical to AI investment. AI is a specific and concrete solution that can be applied to reducing Medicaid fraud and waste, and one that states, and their federal Medicaid partners, can work together to fund and deploy.

Help shape The 80 Million: What should we write about next?