Role of AI in Healthcare Utilization Management
Artificial Intelligence (AI) impacts our lives in a multitude of ways, including throughout the healthcare journey. Recently, NORC at the University of Chicago 1 researched the use of AI in insurers’ utilization management (UM) processes on behalf of the consumer representatives of the National Association of Insurance Commissioners (NAIC); the NAIC consumer reps then published a report with the findings and recommendations to protect patients exposed to utilization management that leverages AI.2
Understanding AI in Utilization Management
Stakeholders interviewed included health insurance plans, providers, technologists, regulators, and consumer advocates. All stakeholders recognized the benefits as well as the risks associated with AI use, including the possibility of introducing or perpetuating bias and the need for safeguards that protect access to care and privacy. While recognizing the frenzied evolution of AI as challenging for states (and state insurance commissioners), the report recommends state investment in legislative and regulatory oversight to protect patients, including marginalized populations that may be under-represented in datasets.
Key Research Findings for AI’s Use in UM
NORC conducted a literature review focused on prior authorization (PA) as a form of utilization management, and early efforts by governments to regulate AI. Three types of AI use cases were provided to interviewees as samples: 1) administrative only, 2) decision-making, and 3) learning model.
| Example | Scans Large Datasets | Uses Fixed Inputs to Make Determinations | Evolves Algorithm Based on Data | Decision-making Entity and Authority |
|---|---|---|---|---|
| Administrative Only | X | Human Reviewer | ||
| Decision-Making | X | X | AI can approve coverage/denials reviewed by a human | |
| Learning Model | X | X | X | AI can approve, deny, or escalate to human review using evolving data from prior decisions |
Health plan representatives surveyed were the most optimistic around the benefits of AI in evaluating PA requests, citing the positive impact of AI on delayed responses and inconsistent decisions, as well as removing non-decision-making tasks like information collation from human reviewers. Other stakeholders countered this optimism with concerns about:
- Rigid AI standards
- Lack of transparency into denials (learning model)
- Lack of information around biases included in datasets used.
Opacity of the data considered by the AI system can create or exacerbate historic UM decisions that were inappropriate, including those based primarily on saving money or related to negative social determinants of health, in a way that is difficult to recognize and thus difficult to correct.
Recommendations for AI Governance
Transparency in AI oversight, especially in healthcare, is critical. Key points include:
- Disclosures should be clear, audience-specific, and tailored to the AI use case, avoiding generic approaches.
- Regulators need access to AI implementation and data details to protect consumers.
- Disclosures must address data consent, representativeness, and fairness to ensure accountability and ethical alignment.
Regulatory standards must clearly define responsible stakeholders (e.g., health plans, technology developers) for AI-related UM decisions and hold them accountable.
- Issues leading to consumer harm, such as discrimination or incorrect determinations, must be identified, traced, and corrected.
- Regular audits improve understanding and accountability in AI-based UM.
- Governance structures must address harm to marginalized groups, ensuring datasets represent women, people of color, those with disabilities, and others.
- Policymakers should require health plans to identify and publicly disclose biases and mitigation steps.
Regulators should have oversight authority over mandated robust, accessible appeals processes.
- Human reviewers with authority to overturn AI-driven UM decisions must be embedded to prevent harm.
- Ongoing collaboration among regulators, industry, experts, and advocates is essential for consumer protections and AI advancements in healthcare.
Insights for Stakeholders: Regulation of AI in UM
Regulation of AI use in UM is an important issue that should be addressed across stakeholders, including pharmaceutical manufacturers. Pharma-provided UM resources must evolve to keep pace with AI use. Patient services and field reimbursement teams should identify patient access issues associated with certain patient populations, plans, regions, or providers, and strategies to address these issues should be implemented.
Final Considerations of AI’s Role in UM
AI’s role in healthcare utilization management offers immense opportunities but requires balanced governance to mitigate risks. Transparency, accountability, and collaboration are critical to ensuring that AI benefits all patients while safeguarding their rights. As the healthcare landscape continues to evolve, proactive strategies and informed oversight will be key to unlocking AI’s full potential.
Magnolia Market Access (MMA) leverages extensive expertise in UM claims analysis to help stakeholders address complex access challenges. You can reach us at magnoliamarketaccess.com to discuss your needs.
Magnolia Market Access Authors: Tracy Baroni Allmon
- https://www.norc.org/ ↩︎
- Consumer Rep Report on AI & Health Ins-Consumer Liaison Cmte-11/19/24-FNM; Accessed December 3, 2024. ↩︎