Utilization management and prior auth: symptoms of illogical system failures.

Utilization management was meant to promote high-value care and serve as the anchoring counter-balance against fraud, waste, and abuse. Instead it has metastasized into a bloated cost center of its own. The U.S. spends over $93.3 billion annually across payers and providers implementing, contesting, and navigating utilization management. Amidst that spend, 75% of prior authorization appeals are overturned, and 1 in 3 physicians report witnessing serious adverse events due to authorization delays, as highlighted by the American Medical Association. This analysis reframes UM as a system logic problem, not just a process or technical flaw.

The country has spent two decades digitizing healthcare, only to discover that integration and automation without alignment does not solve for friction. It uncovers other issues, or worse, moves the friction faster and elsewhere. The pages that follow trace the evolution of UM, build the foundational understanding most commentary skips, and map what needs to change if UM is going to deliver on its original intent. [1,2,3-5]

The infrastructure still does not work

Despite massive investments in EHRs and payer portals, prior authorization remains largely manual. According to the 2024 CAQH Index Report, only 35% of PA transactions were fully electronic, and manual transactions cost $10.92 versus $3.61 when done electronically. That makes PA one of the most expensive manual transactions in healthcare. [6]

Even worse, clinical staff and provider access teams may spend an average of 12 hours per week managing PA, up from 10 hours in 2018. Portals multiply instead of unify, faxes and e-faxes persist, and peer-to-peer calls drag on, sometimes escalating from RN-to-RN, or pharmacist-to-pharmacist, to MD-to-MD. Denial rationales are vague or missing, and workflows vary by payer, geography, employer, and benefit. Payer policy transparency varies alongside them, and to make matters worse, the patient and paying member often has to clear significant hurdles just to reach the specifics of the diagnostic or therapy medical necessity policies that govern their coverage. [3,7]

When misalignment becomes harm

The consequences are not abstract. Denials and delays lead to worse outcomes, especially when patients are stable, the treatment plan is guideline-concordant, and time to therapy is critical.

• Prescription denials for certain medication classes, including β₂-agonists, antiepileptics, and atypical antipsychotics, were associated with 40 to 75% increased odds of acute care visits within 60 days. [3]
• 60% of neurology studies on PA reported care delays, and 25% reported increased disease activity, including breakthrough seizures and multiple sclerosis relapses. [4]
• In cancer care, 69% of patients perceive delays from PA, leading to increased anxiety and reduced trust. [8]
• Systematic reviews across oncology, cardiology, pediatrics, and behavioral health show clear associations between PA and preventable hospitalizations, delayed therapy, and worse disease control. [2,3]

These harms may not be intentional, but they are embedded in the reality of how UM has evolved. UM was originally introduced to reduce unnecessary utilization and rein in skyrocketing healthcare costs. As care delivery evolved and evidence-based medicine matured, UM promised to ensure that interventions were medically necessary, evidence-based, and aligned with the benefit design of each plan. On paper that sounds right. In practice, instead of aligning incentives and workflows, it has added fragmentation. Over time the focus shifted from clinical optimization to cost containment. UM is now the gatekeeper, not the guide. The administrative burden grows, and PA became its most visible symptom.

Foundations first, payers and member benefits

Every UM decision, whether it is about necessity, setting, or duration, passes through a complex operational filter rooted in the patient's specific benefits policy and design. The right clinical decision can still trigger a denial if it does not align with the underlying member's policy architecture. At its core, this is operationalized through a patient-specific overlay that includes:

• Line of Business (LOB). Medicare Advantage, Medicaid (FFS or MCO), Commercial and ASO, and Exchange plans each carry distinct policy constraints and turnaround requirements.
• Network Structure. Whether the plan is HMO, PPO, EPO, or POS determines the referral process, out-of-network allowances, and prior auth necessity.
• Benefit Design. Covered services, exclusions, tiered cost-sharing, utilization thresholds, and the split between medical and pharmacy policy rules.
• Payer-Specific Rules. CMS timelines such as Medicare Advantage's 72-hour urgent PA response, state Medicaid variations, and proprietary commercial plan rules all shape how UM is executed.

It is also critical to understand that UM is intentionally focused on medical necessity qualification and functionally separated from claims payment. That said, if authorization is required as a condition for payment and is not received, then those services or treatments will not be paid for by the health plan's coverage and member benefits. Failing to account for this operational matrix results in denial churn, unnecessary appeals, provider abrasion, and downstream care delays. It is not just about what care is appropriate. It is about whether that care is authorized under the terms of that patient's policy, in that plan, at that moment.

Later in this space I reference a "270/271 Eligibility and Benefits" transaction, which is the real-time transaction that surfaces this patient-level benefit detail before care is rendered. For those wanting to learn more about payers, their products, and their structures, consider visiting AHIP or AMCP.

What utilization management really is

UM is a clinical and administrative umbrella for activities that assess the appropriateness, cost, and timing of healthcare services. Most commentary centers on prior authorization, but PA and ePA (electronic prior auth) are just one piece. UM spans medical and pharmacy benefits and coverage, and its complexity is most easily described using a four-dimensional model. That model captures when UM occurs (temporal), what it targets (functional), how it is operationalized (operational), and the infrastructure that enables it (technical). This is not just an academic framework. It is a practical tool for providers, developers, and the MedTech and life sciences industry to understand how the UM ecosystem comes together, so as to navigate future hurdles in care and innovation.

The temporal dimension governs when intervention happens along the care continuum, shaping access, cost, clinical decision-making, and the level of friction imposed on guidance as a contingency of coverage. Note that notification is not ubiquitous. It is one example of the reporting burdens required in cases like post-acute transitions for some lines of business and some payer regions. A benefit silo problem may also remain, where different UM processes for the same intervention can be intentionally disparate depending on whether it is billed under the medical or pharmacy benefit, as well as site of care. That leaves the potential for a fragmented landscape with little coherence or predictability.

The functional dimension defines the core intent of UM around the service or therapy in question, whether ensuring medical necessity, optimizing site-of-service, managing inpatient length-of-stay, or sequencing treatments through step therapy. It reflects the specific levers payers use to influence clinical and economic outcomes. The operational dimension layers in the mechanics of implementation, from criteria application and appeals to benefit integration and policy overlays. It is where payer rules, benefit design, and provider workflows intersect, with high impact on compliance, denials, and member experience. I cannot overemphasize the communication demand this creates, because health plans are held to regulatory standards that vary across lines of business, some state regulated and others federally regulated, with significant nuance. This remains a significant area of opportunity.

The technical dimension has been the biggest focus of the industry, with the aim of advancing the infrastructure and capabilities that enable or hinder UM. That includes fax and e-fax, portals, platforms, APIs, HL7 Da Vinci Project FHIR approaches such as CARIN Prior Auth and CARIN RTPBC, and data semantic interoperability enablers like RxNorm, LOINC, SNOMED, CPT, and HCPCS to support complex data orchestration. This is also the space for EDI exchanges, clearinghouse interfaces, and the integration of vendors and third-party administrators (TPAs) involved in various aspects of medical or pharmacy benefits, including PBMs.

Does UM only need a technical tune-up

To be clear, PA can work when used appropriately, and despite pushes from many, it is not going away anytime soon. Even the Centers for Medicare and Medicaid Services will be introducing PA in the near future, where many will be left scratching their heads as it exposes the variability in Medicare Administrative Contractor (MAC) Local Coverage Determination (LCD) policies. In Medicare Part D, PAs for certain drugs reduced utilization by 26.8% and lowered spending by $96 per beneficiary-year, with no reported adverse effect on outcomes. But instead of targeting high-variability, low-evidence use cases, PA is often applied where it is least needed, like radiation oncology, which accounts for only 3 to 4% of cancer costs yet is required in more than 50% of cases. [1,4,3]

This reflects a deeper issue. Payment rules, coverage policies, and clinical workflows are built on outdated ontologies.

• Payers enforce coverage based on claims logic rooted in administrative data ontologies.
• Clinicians use evidence-based guidelines and literature to guide decisions, none of which are defined for alignment to administrative data ontologies.
• Providers operate in workflows divorced from reimbursement triggers.
• Patients are left navigating opacity, including delays, substitutions, or denials they cannot predict or influence.

We have built solutions that optimize steps without questioning whether the steps themselves should exist.

Evidence does not move fast enough

Some of this misalignment is due to lagging evidence integration. Consider the pattern.

• It takes 9 years on average for clinical research to make it into guidelines. [9]
• Payer policy integration of published evidence often lags at least 1 to 3 years behind that evidence.
• Even after guideline updates, clinical practice adoption can take another 6 to 14 years, with a median of 14 years from guideline recommendation to 90% uptake. [10,11]
• Meanwhile, about half of guidelines are outdated within 5.8 years. [12]

So even when new therapies emerge or consensus evolves, coverage policy and clinical behavior often remain out of sync for over a decade. Utilization management then activates blunt enforcement tools against outdated foundations.

A final rule with real teeth and real blind spots

In February 2024, the Centers for Medicare and Medicaid Services finalized the Interoperability and Prior Authorization Final Rule (CMS-0057-F), which signals a significant step toward process standardization. Entities impacted by CMS-0057 must meet HL7 FHIR-based API requirements by January 2027. In the meantime, CMS has exercised enforcement discretion of the X12 278 transaction for HIPAA-covered entities that perform PA workflows electronically using HL7 FHIR. Use of the X12 278 is anticipated to decrease as the industry approaches the regulatory deadlines. [13]

There is a major gap. The rule excludes pharmacy benefits under Parts B and D, due to differing NCPDP data standards for PA, although payers can still share this information by API. Of note, pharmacy is estimated to account for 35% of PAs and 57% of projected savings.

That drug gap is now its own rulemaking. In April 2026, CMS issued the 2026 Interoperability Standards and Prior Authorization for Drugs proposed rule (CMS-0062-P), which extends the electronic prior authorization requirements, the shortened decision timelines, and the transparency obligations of CMS-0057-F to cover drugs. Comments closed June 15, 2026. [15]

What might actually fix this

Large language models and AI-based tools, including OpenEvidence, OpenAI's ChatGPT, Anthropic's Claude, Atropos Health, Abridge, Hippocratic AI, and many others, have proven valuable across many use cases. Yet they have not been pointed at the one UM problem they are most well suited to solve. Rather than using these tools to automate coverage appeals after denials occur, the bigger value would be to compress the evidence-to-coverage-to-practice lag before misalignment creates friction.

• Informing and supporting more timely payer policy updates and provider-facing decision support.
• Synthesizing clinical trials and real-world evidence in more pragmatic, near real-time fashion.
• Mapping new evidence or guidelines to billing codes, NDCs, and coverage frameworks for actuarial analysis and forecasting review.

Used correctly, these tools could help achieve what UM originally set out to do, which is to optimize care, reduce waste, and build trust between payer and provider on what value-oriented, evidence-backed care truly can be.

Where this goes next

If UM and prior authorization are symptoms and not the root cause, then real transformation lies in how we structure and align value for the system beneath them. Three questions carry the analysis forward from here.

The first is the operating layer itself. A UM TPA vendor ecosystem sits between payers and providers, and the entities that actually pull the operational levers are frequently not the ones named on the insurance card. Understanding who administers, delegates, and adjudicates is the prerequisite to changing any of it.

The second is the standards question. HL7 FHIR, X12 278, NCPDP, and CARIN are genuine steps forward, and each deserves credit for the interoperability it unlocks. Each also carries real limits in utility that shape what true interoperability can and cannot deliver. Treating the standards as a finished answer rather than a foundation is how organizations end up digitizing the same dysfunction they meant to retire.

The third is incentive design. Performance-based pathways like gold carding, where providers with high approval rates earn exemption from prior auth altogether, point toward a smarter model. The early data is promising, but these models depend on smarter logic, not just looser guardrails. UM reform is not about bypassing controls. It is about building systems that learn, trust, and adapt. UM was built to ensure the right care at the right time. Too often it now achieves the opposite.

References

1. Howell S, Yin PT, Robinson JC. Quantifying the economic burden of drug utilization management on payers, manufacturers, physicians, and patients. Health Aff (Millwood). 2021;40(8):1206-1214. doi:10.1377/hlthaff.2021.00036
2. Murphy J, Beauchamp N, Sun KJ, et al. Adverse effects of health plan prior authorization on clinical effectiveness and patient outcomes, a systematic review. Am J Med. 2025;138(2):132-43. doi:10.1016/j.amjmed.2024.07.010
3. Chen WC, Carpenter C, Sidiqi B, et al. Integrating prior authorization into clinical workflows for care access and practitioner experience. JAMA Netw Open. 2025;8(1):e250011. doi:10.1001/jamanetworkopen.2025.0011
4. Gotlieb E, Joseph B, Blank L, Jetté N. Barriers and consequences of prior authorization for neurologic medications. JAMA Neurol. 2025;82(4):427-35. doi:10.1001/jamaneurol.2024.5912
5. American Medical Association. One in three doctors has seen prior auth lead to serious adverse event. Chicago (IL), AMA, 2023 Mar. Available from: ama-assn.org
6. CAQH. 2024 CAQH Index Report, measuring progress in adoption of electronic business transactions. Washington (DC), Council for Affordable Quality Healthcare, 2024 Jan. Available from: caqh.org
7. 2024 AMA prior authorization physician survey. AMA. Available from: ama-assn.org
8. Chino F, Baez A, Elkins IB, et al. The patient experience of prior authorization for cancer care. JAMA Netw Open. 2023;6(10):e2315526. doi:10.1001/jamanetworkopen.2023.38182
9. Borchert F, Wullenweber P, Oeser A, et al. High-precision information retrieval for rapid clinical guideline updates. NPJ Digit Med. 2025;8(1):14. doi:10.1038/s41746-024-01078-2
10. Khan S, Chambers D, Neta G. Revisiting time to translation, implementation of evidence-based practices in cancer control. Cancer Causes Control. 2021;32(5):479-86. doi:10.1007/s10552-021-01401-1
11. Putera M, Roark R, Lopes RD, Udayakumar K, Peterson ED, Califf RM, Shah BR. Translation of acute coronary syndrome therapies, from evidence to routine clinical practice. Am Heart J. 2015 Feb;169(2):266-73. doi:10.1016/j.ahj.2014.09.015
12. Shekelle PG, Ortiz E, Rhodes S, et al. Validity of the Agency for Healthcare Research and Quality clinical practice guidelines, how quickly do guidelines become outdated? JAMA. 2001;286(12):1461-7. doi:10.1001/jama.286.12.1461
13. Centers for Medicare and Medicaid Services. CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F). Baltimore (MD), CMS, 2024 Jan. Available from: cms.gov
14. Kannarkat JT, Warring W, Brennan T. Advancing interoperability and prior authorization reform. JAMA Health Forum. 2024;5(6):e241193. doi:10.1001/jamahealthforum.2024.1193
15. Centers for Medicare and Medicaid Services. 2026 CMS Interoperability Standards and Prior Authorization for Drugs Proposed Rule (CMS-0062-P). Baltimore (MD), CMS, 2026 Apr. Available from: cms.gov

All views, analyses, and frameworks presented here reflect independent professional judgment informed by more than two decades of experience across payer strategy, clinical transformation, and health system operations. They do not represent the views or positions of any current or former employer or affiliated organization.