New Report Urges Guardrails and Oversight of AI Use by Health Insurance Plans

DENVER – Today, a nationwide cohort of consumer representatives released a report pushing the National Association of Insurance Commissioners (NAIC) to monitor and regulate the increasing use of artificial intelligence (AI) by health insurance plans. The report highlights Colorado’s efforts to prevent discrimination in the use of AI, but the report also underscores concerns about how rapid development of AI tools without regulatory guardrails may lead to consumer harm, including discrimination, breaches of privacy, and incorrect denials of coverage and care. The report focuses on the use of AI in prior authorization processes in health insurance when a patient and their provider must get approval from the health insurance company in advance for a course of treatment or procedure.

In a letter to the NAIC, the advocates say, “The speed of technological advances in AI is far outpacing the changes in state and federal health insurance regulation and oversight needed to protect consumers…. unregulated use of AI could also exacerbate existing bias and discrimination, particularly for marginalized and disenfranchised communities who already experience disparate health care access challenges.”

 Key findings from the research include the following:

●  AI is regularly used by health insurance plans to decide approvals. There is increasing evidence this can increase bias and discrimination and that insurers may prioritize denying care to save money when care should be approved.

●  Proper guardrails should be in place to ensure that patient coverage, access, privacy, and safety are protected

●  The report highlights states like Colorado that are at the forefront of ensuring health insurance company accountability in the use of AI. However, despite Colorado being a lead state, the report underscores the challenge of keeping up with the proliferation of the use of AI and the importance for all states to act quickly.

“Colorado is a leader in regulating the use of AI in health insurance, but even as a leader, the development of AI is progressing so rapidly, it is difficult for the state to keep up in designing our regulations,” said Adam Fox, deputy director of the Colorado Consumer Health Initiative and a designated Consumer Representative to the NAIC. “This report further underscores the need for Colorado and other states to move as quickly as possible to ensure AI does not lead to inappropriate denials of coverage and care or have discriminatory impacts on consumers. Insurers must be held accountable for the governance, monitoring, and ultimately the impact of the use of AI.”

Colorado passed legislation, SB21-169, that is designed to ensure use of algorithms and predictive models, including AI, in all types of insurance does not lead to discrimination. Colorado’s Division of Insurance has been working to develop the regulations under the legislation to hold insurers accountable in their use of AI and is currently developing regulations focused on health insurance. The report, however, highlights the need for all states to move quickly to ensure regulatory guardrails and consumer protections are in place as AI continues to develop rapidly.

The report from the NAIC consumer representatives emphasized that:

 1. Meaningful transparency is critical

●  It must be clear, to both regulators and consumers, when AI is being used by health insurance plans for the purposes of Utilization Management (UM – prior authorization processes are a prominent form of UM) and what role the AI plays in making determinations about coverage for care.

●  Transparency must extend to disclosures about the data used to develop, train, and test the AI tools (with an emphasis on consent for use and representativeness of the population), and the extent to which any AI tool can begin to train itself.

●  Existing nondiscrimination and data regulation laws should be assessed for their applicability to AI in UM.

●  Regulatory standards must identify which parties are accountable (e.g., health plans, technology developers, etc.) when AI tools are used in UM decisions that lead to consumer harm, including discrimination, breaches of privacy, and incorrect adverse determinations (coverage/care denials).

●  Regular audits, conducted on behalf of state regulatory agencies by parties with specialization in testing AI technologies, can be an effective way to both understand the ways AI is used in making UM decisions and hold the plans accountable for its use.

●  AI tools intended for UM decisions should be built on standards of care that aim to achieve the highest level of quality, and penalties for non-compliance need to be significant enough to have influence.

●  Governance structures that measure and prevent harm to historically marginalized and disenfranchised populations must be required.

3. Regulators should ensure that health insurance companies require oversight by humans with the appropriate clinical training regarding decisions that impact patient outcomes

●  Robust and accessible appeals processes for coverage denials need to be established and considered a guaranteed right for all health insurance consumers.

●  Human oversight must be embedded into UM when AI is used and those reviewers must have the authority and ability to overturn decisions made by the AI without undue consequences.

●  AI regulation must be considered an evolving practice that relies on collaboration between regulators, technical experts, industry stakeholders, consumers, and consumer advocates.

This national report stresses that all states need to understand the urgency of this issue and develop the necessary guardrails rapidly to protect consumers. While the report holds Colorado out as an example, there is more work Colorado must do, and quickly, to ensure the use of AI in health insurance does not harm or discriminate against consumers.