eBook: The evolving role of artificial intelligence in payment integrity

Brett Arnold and Anandhi Periynan

Payment integrity is a vital function for healthcare payers as it helps ensure that claims are paid accurately and appropriately. It also helps to reduce fraud, waste, and abuse, which costs the healthcare system billions of dollars every year and can directly harm members. However, executing payment integrity effectively is time and resource intensive, involving complex and dynamic rules, regulations, contracts, and policies. Moreover, payment integrity requires a large amount of data, advanced analytics, and human resources to review and audit claims, and in some cases recover overpayments.

Artificial intelligence (AI) is a technology that offers significant promise for payment integrity as it can help to enhance the value and accuracy of payment integrity programs, as well as reduce the administrative burden and provider abrasion faced by health plans. However, AI is not a solution in and of itself—and should not be used to replace human expertise.

In this eBook, we will explore both the promise and potential pitfalls of AI in payment integrity, how Cotiviti is responsibly deploying AI in our solutions, and how payers can separate fact from fluff when evaluating the AI capabilities of payment integrity partners.

The transformative potential for AI in healthcare is impossible to deny, driving savings that could add up to hundreds of billions of dollars per year across both payers and providers. More important, it could improve outcomes for members by streamlining often tedious processes such as prior authorization and empower care managers to create more tailored care plans for their patients.

In payment integrity specifically, according to McKinsey, 8–30% of health plan administrative costs could be automated with AI, while overall medical costs could be reduced by 0.4–1.7% by improving the effectiveness of claim reviews, edits, and recovery.

Payers that aren’t deploying AI risk falling behind their competitors. Industry analyst Gartner projects that more than 80% of businesses will have deployed generative AI applications or used generative AI application programming interfaces (APIs) by 2026. The bottom line for health plans? Whether they have integrated AI into their operations or not, they should know that some of their competitors already have.

However, payers tell us there are barriers in adopting and implementing AI for their payment integrity needs. These include:

• Lack of resources and expertise. AI requires a significant amount of data, infrastructure, and human resources to develop, deploy, and maintain. Health plans often don’t have the sufficient capacity, capability, or budget to invest in AI, or they may not have the right talent or skills to manage and use AI effectively.
• Lack of trust and understanding. AI is a complex and dynamic technology that can be difficult to comprehend and explain. Payers often have concerns about the reliability, validity, and accountability of AI, especially when it comes to clinical decisions and outcomes.
• Lack of standards and governance. AI is a rapidly evolving and emerging technology that can pose ethical implications. Payers may not have the appropriate policies, procedures, or frameworks to govern the use of AI, or to ensure its compliance with regulations, contracts, and best practices.
• Lack of collaboration and integration. AI is a technology that can create value across the entire healthcare ecosystem involving multiple stakeholders, including payers, providers, and members. An individual health plan likely won’t have effective communication, coordination, or alignment with these stakeholders, or may face resistance or friction from them. Payers also have challenges in integrating AI with their existing systems, processes, and workflows, or in scaling AI across their organization.

Payers that aren’t deploying AI risk falling behind their competitors.

These challenges and barriers can prevent healthcare payers from realizing the full potential and value of AI for their payment integrity programs. Payers need to overcome these challenges and barriers by developing a clear and strategic vision for AI, by building a strong and supportive culture for AI, by aggressively pursuing governance to mitigate the risks of AI, by partnering with trusted and experienced payment integrity vendors, and by learning from the best practices and lessons of other healthcare payers and organizations.

Cotiviti's vision is to enable a high-quality and viable healthcare system, and we embrace the use of AI in support of this vision. We don’t view AI as a product or a solution, but rather a tool to improve the solutions we already have in the payer marketplace and to develop new solutions. This approach is known as augmented intelligence, meaning it supports our human experts’ decisions with relevant information and proprietary insights rather than replacing human judgment or decision-making.

Here are Cotiviti’s four guiding principles when it comes to responsibly deploying AI in our solutions to improve client outcomes.

• AI is a tool, not a solution. AI does not work in a vacuum: it needs data to be trained. It needs prior results generated by humans to learn from. Delivering real value through AI requires not only data and data scientists, but also deep experience and expertise.
• AI should be used to improve results. We don’t use AI for the sake of using AI; we use it to improve business results that matter to our health plan clients and will deliver better results to their members.
• AI must be used responsibly. While AI represents significant opportunity, it also represents risk. Developing and deploying AI capabilities that are safe and trustworthy requires dedicated focus on new governance processes and capabilities. In addition to typical controls for data security and privacy, new policies, procedures, and monitoring are required to manage AI-specific concerns including accuracy, transparency, bias, and accountability. Cotiviti has implemented a strong governance program to define policies for security, transparency, and compliance, and also partners with the Responsible AI Institute. As part of this program, Cotiviti rigorously tests and statistically evaluates new AI models before deploying them into production. We then monitor, evaluate, and iterate on those models to ensure they continue meeting our performance and quality standards.
• AI does not replace human expertise. We strategically limit AI automation to enabling our human specialists to improve performance and client experience. This means that we are not replacing human clinical decision-making or judgment. We use AI to help prepare clinical content, but not to make decisions based on that content.

Delivering real value through AI requires not only data and data scientists, but also deep experience and expertise.

Cotiviti uses AI to enhance our solutions primarily in payment integrity and risk adjustment. Here are representative examples:

• Prepay claim review: We use machine learning to identify prepay policy areas where edits are adjusted or overturned by our health plan clients to revisit policy logic, determining modifications to increase accuracy and reduce provider abrasion. We also use machine learning to rapidly identify circumstances that require a member’s medical record to be required.
• Pre and postpay DRG clinical review: We deploy advanced analytics driven by machine learning to enable health plans to select medical records for review that are most likely to return value back to the plan, rather than casting a wider net that could increase provider abrasion. Evaluation of a client's claims, provider behaviors, and prior results inform the appropriate mix of claim selections to analyze prospectively and retrospectively, increasing value and change rates across both programs.
• Fraud, waste, and abuse (FWA) management and prevention: We use machine learning to analyze prepay claims and other data points to identify suspicious patterns of FWA and prevent inappropriate claims from being paid while still meeting prompt-pay requirements. This also helps flag suspicious providers that warrant immediate investigation, helping special investigations units (SIUs) build ironclad cases to stop bad actors sooner.
• Medical record coding: We leverage natural language processing (NLP) technology to improve the value and quality of DRG validation medical record reviews performed by our expert AAPC- and AHIMA-certified risk adjustment coders. This helps capture relevant information from the record and display it, enabling our trained reviewers to be more thorough and more consistent.
• Risk adjustment suspecting: Our advanced analytical models leverage machine learning to provide our clients with a nuanced examination of their data, predicting which members have the highest probability of missing or incomplete conditions to improve coding accuracy.

Most critically, we are continually evaluating the effectiveness of the results our AI engines deliver to determine how they can be improved and deliver better results back to the health plan.

As payers look to AI to improve their payment integrity programs, they’ll likely encounter a variety of vendors who claim to offer the best and most advanced AI solutions, some of which may provide more spin than fact when it comes to their capabilities and the value they offer. Therefore, health plans need to be ready to ask the right questions, such as:

• How do they market themselves? Be wary of payment integrity partners that describe themselves as “an AI company.” As discussed, AI is a tool, not a solution. What matters is how the vendor actually uses AI to drive tangible value for the client. All vendors should be leveraging AI at this point. In three to five years, no one will be calling themselves “an AI company,” because that will simply be expected. What value will they offer then?
• What is their track record in payment integrity? Before drilling too deep into a vendor’s AI capabilities, find out what payment integrity solutions they have and what results they’ve delivered to their clients. Do they have only prepay solutions, postpay solutions, or both? How much medical cost savings do they typically deliver to their clients? How often are their payment decisions overturned?
• What is the breadth of their data sets? AI needs data to learn from and a significant amount of prior results to be trained. Therefore, a vendor should have access to large and diverse sets of both claims and clinical data. If a vendor does not have a significant payment integrity data set, a health plan’s own data likely will be used to train their inexperienced AI algorithms.
• What consultative support do they offer? Is the vendor’s service model software only, or do they also provide ongoing access to payment integrity and robust clinical expertise? Who is verifying the accuracy of AI-driven recommendations before they are delivered back to the client, and what are their credentials?

By asking these questions, health plans can be more assured their chosen payment integrity vendor will drive actual value for their organization, not simply promise cutting-edge technology that may or may not be effective.

AI is a tool, not a solution. What matters is how the vendor actually uses AI to drive tangible value for the client.

There’s no telling exactly what AI-driven payment integrity solutions will be available five years from now or even one year from now, but one thing is certain: we’re swiftly moving beyond AI being a catchy buzzword or shiny object.

Instead of simply asking their payment integrity partners whether they use AI, it’s imperative for health plans to find out how their use of AI drives value back to the plan in the form of:

• Increased medical cost savings
• Improved administrative efficiency
• Higher coding accuracy
• Improved provider and member experience

By focusing less on the specific technology and more on the human expertise and consultative support enabling that technology, health plans can be more confident they’ve chosen a payment integrity vendor that delivers results that are timely, accurate, and defensible, ultimately helping to lower the cost of healthcare.

About the authors

Brett Arnold Senior Vice President, Product Development  Brett’s role is to understand client needs and lead the process of enhancing Cotiviti’s product portfolio to meet them. This involves engaging with clients, understanding market needs, leading product discovery, developing market strategies, building business cases, managing a product roadmap, defining solutions, and leading market launches.

Anandhi Periyanan Senior Vice President, Research and Development Anandhi aligns with the overall strategic vision of Cotiviti and provides leadership and oversight into new product development, existing platforms, and technology strategy including artificial intelligence for payment integrity solutions. With more than 25 years of industry experience and over a decade in healthcare, she has a background in technology planning and is instrumental in driving change and implementing technology strategy in architecture, innovation, process improvement, and automation, incorporating machine learning in SaaS, and data platforms.