AI in Healthcare Prior Authorization: Present and Future Outlook

Introduction

Artificial Intelligence (AI) is increasingly shaping the healthcare landscape, particularly in the realm of prior authorizations (PAs). With the growing complexity of healthcare systems and the demand for efficient service delivery, AI has become a valuable tool for streamlining PA processes. This guide will explore the current state of AI in prior authorizations, its benefits and challenges, and the recent regulatory updates impacting health plans and medical clinics.

The Current Landscape of Prior Authorizations

Rising Denial Rates in Medicare Advantage Plans

In 2022, Medicare Advantage (MA) plans saw a significant increase in denied prior authorization requests, with a record 3.4 million denials. This marked a rise to 7.4% of all PA requests, compared to an average of 5.7% in previous years. The surge in denials has raised concerns about access to necessary healthcare services, particularly as MA enrollment has grown from 22 million in 2019 to 28 million in 2022.

Many healthcare professionals and policymakers worry that the use of AI in reviewing PA requests may be contributing to these denials, potentially creating barriers to essential care. The low appeal rate of denied requests—just 9.9%—further complicates the situation, suggesting that patients may be unaware of their rights to appeal or may find the process too daunting.

The Role of AI in Prior Authorization

Streamlining the PA Process

AI technology is rapidly transforming the way health insurers manage PAs. By automating many of the manual processes involved in PA requests, AI can significantly reduce the time and effort required from both payers and providers. For example, natural language processing (NLP) can extract key information from submitted documents, while algorithms can assess whether the requested treatment meets the insurer’s medical criteria. This can lead to faster, more consistent decisions and reduced administrative overhead.

The Risks and Challenges

Despite its potential benefits, AI in PA also presents significant risks. AI systems are only as reliable as the data and algorithms they use. If these are flawed, the results can be inaccurate, leading to inappropriate denials of care. A class-action lawsuit against UnitedHealthcare highlighted this issue, with allegations that an AI tool had a 90% error rate, resulting in the denial of medically necessary care for thousands of Medicare beneficiaries.

Moreover, AI can inadvertently perpetuate biases against marginalized communities. Studies have shown that some algorithms may assign lower risk scores to Black patients compared to White patients with similar health profiles, leading to disparities in care. These biases can be especially harmful in PA processes, where swift and accurate decisions are crucial.

Regulatory Oversight and Safeguards

CMS Guidelines and Federal Regulations

In response to growing concerns about the use of AI in healthcare, the Centers for Medicare & Medicaid Services (CMS) has issued new guidelines and regulations. The CMS Interoperability and Prior Authorization Final Rule, finalized in January 2024, aims to streamline the PA process, improve transparency, and ensure that decisions are based on individualized patient circumstances.

The rule mandates that AI and other automated tools can assist in utilization management only if they adhere to specific regulatory criteria, including evidence-based guidelines and the consideration of patient-specific factors. CMS also requires that any adverse medical necessity determinations be reviewed by a qualified physician or healthcare professional, ensuring that human oversight remains a critical component of the PA process.

State-Level Initiatives

At the state level, regulatory oversight of AI in insurance is evolving. Colorado, for instance, has been a pioneer in establishing regulations for the use of algorithms in health insurance. The state’s approach, which includes public listening sessions and input from various stakeholders, could serve as a model for other states and federal agencies.

California has also taken steps to regulate AI in healthcare, requiring licensed physicians to supervise AI tools used in PA and mandating transparency in AI decision-making processes. These state-level initiatives reflect a growing recognition of the need for robust oversight to ensure that AI is used responsibly and equitably in healthcare.

Future Directions and Recommendations

Enhancing AI Accountability

As AI continues to play a larger role in healthcare, there is a pressing need for more comprehensive regulatory oversight. Federal and state agencies should collaborate to establish clear guidelines for the development and use of AI in PA, drawing on lessons from the FDA’s regulation of clinical algorithms. This could include the creation of certification standards and guidelines for AI developers, ensuring that these tools are designed and implemented with patient safety and equity in mind.

Addressing Bias in AI

To mitigate the risks of bias, AI systems should be regularly audited for fairness and accuracy. Developers and healthcare organizations must work together to ensure that AI tools do not perpetuate existing disparities in care. This includes using diverse and representative data sets in the development of algorithms and implementing safeguards to prevent biased decision-making.

Conclusion

The integration of AI into prior authorization processes offers significant potential to improve efficiency and reduce administrative burdens in healthcare. However, these technologies must be implemented with care, transparency, and oversight to ensure that they serve the best interests of patients. By adhering to regulatory guidelines and continuously refining AI tools, health plans and medical clinics can harness the power of AI while safeguarding the quality and equity of patient care.

PCG Software is an AI Healthcare company that has been helping Payers for 30+ years with AI-assisted claims auditing. We have helped several payers with AI-assisted Prior Authorization solutions but have always, and will always state that AI is a tool for finding and quickly analyzing data based on a plan’s or provider’s knowledge of the patient, the provider, services, and the final decision of approval or denial always should remain with the person, not the machine. Apply your rules, auto-approval, and auto-adjudicate per those rules, and consistently review your rules and contracts as a payer, while consistently reviewing your usage of AI as a provider.

If you would like to explore AI solutions for your payer organization or medical practice, contact us today at wschmidt@pcgsoftware.com. Thank you.

Resources: