In the wake of the December 4, 2024 fatal shooting of UnitedHealthcare’s CEO Brian Thompson, questions were raised about insurance claim denial practices. Several news outlets noted that UnitedHealth Group’s profits have been steadily increasing for several years, though they declined from $22.3 billion to $14.4 billion in 2024. In prior years, UnitedHealth Group’s profit has been growing year over year from $13.8 billion in 2019 to $15.4 billion in 2020, $17.3 billion in 2021, and $20.6 billion in 2022.
What many will not appreciate is that as of December 2024, UnitedHealthcare insured 49 million Americans either directly, as a Medicare Part C managed care organization, or by administering the ERISA employer health plans. UnitedHealth Group also operates a large Pharmacy Benefit Manager (PBM) providing prescription drug coverage. Unlike several of the other major insurance companies, UnitedHealth Group is in the business of owning physician practices, outpatient surgery centers, and even has a stake in the healthcare research and innovation. As the largest stakeholder in American healthcare, UnitedHealth Group is responsible for an increasing amount of healthcare delivery. For years, my colleagues have remarked quietly that UnitedHealth Group might be the most vertically integrated health delivery system in the United States. It is surprising that UnitedHealth Group has only recently been scrutinized by the Department of Justice’s Antitrust Division.
So what now? To fully understand the complexity of UnitedHealthcare and other insurance companies’ claims handling process requires a careful analysis of the legal landscape and other practical considerations, such as the increasing role of artificial intelligence (AI).
I. Current Laws Governing Claims Processing
Because insurance companies are for-profit, they engage in a constant risk-benefit analysis with regard to whether it is worth the time and expense needed to ensure the right outcome in a particular claim. Recent laws have done little to address the profitability calculation inherent in every insurance decision. Although the Affordable Care Act requires that at least 80% of premiums be spent on healthcare and quality improvement and further requires that insurance companies explain annual rate increases greater than 15%, these laws have almost no impact on the processing of an individual claim. At present, there is no specific law which encourages insurance companies to arrive at the correct decision with respect to a claim. In fact, insurance companies are not paid for their outcomes; they are paid for their administrative functions. The more patients that they cover, the higher potential to make money.
Many state laws have attempted to regulate insurance claims processing, but have failed to require much other than timely processing and payment for “clean” claims. Unfortunately, the regulatory impact of these laws has been that insurance companies need to find ways to process claims quickly – but not correctly. These state laws also incentivize insurance companies to find defects in claims, to issue partial or total denial on the first level of review, so that the claims will not be considered “clean” and the insurance company will have additional time for processing.
The goal of these state processing laws was laudable, as they attempted to address the delay suffered by providers whose healthcare services had already been provided to patients and who faced a cash flow problem when staff and overhead were paid, but insurance reimbursement was delayed. Unfortunately, due to the high number of claims to be processed, these laws have had unintended negative consequences. In requiring the claims to be “clean” it puts the onus entirely on the provider submitting the claim to make sure that it meets every single insurance company policy and health plan requirement supporting payment. At the same time, these laws place all the power in the hands of the insurance company in determining whether the claim is “clean” and will be paid upon submission. Over the years of reviewing provider billing and claims data, insurance companies have developed an extensive system of codes to justify denials or partial denials. There are now dozens of denial and partial denial codes that may be cited individually or together to support that a claim is not clean.
Each insurance company also has its own provider payment policies and procedures, some of which are unfair at best, and arbitrary at worst. For example, Anthem Blue Cross Blue Shield recently announced – and later retracted – that it would limit anesthesia provider payments to certain time limits set in advance in their provider policies. The policy drew significant criticism from anesthesiologists, as the payments would be the same regardless of how long the procedure actually took for a specific patient. The policy fails to acknowledge that anesthesiologists do not have complete control over the length of surgery, which is determined by the surgeon, the type of surgery, and the patient’s ability to tolerate the procedure and the anesthesia.
Anthem Blue Cross Blue Shield – which covers 45 million Americans – admitted to announcing the policy without first auditing any claims to see if anesthesia billing was even a problem. To be clear, this payment cap would not have been tied to medical necessity, quality of care, or patient outcomes. Although Anthem has reversed its decision to implement this policy, there is nothing to stop Anthem or any other insurance company from imposing this artificial cap on payments to anesthesia times either now or in the future.
II. AI in Health Claims Processing
Due to the significant volume of claims received by large insurance companies, they have been experimenting with artificial intelligence (AI) to process claims on a large scale using sophisticated algorithms. Although health regulators have been quick to address AI tracking of health data linked to specific individuals, they have been slow to adopt policies which would regulate the use of AI to determine claims denials.
At present, insurance companies are using AI to analyze large datasets, automate decision making, and identify patterns in billing data that tend to support a finding of fraud. For example, insurance companies may use AI to review single provider’s bills over time to look for anomalies. A single provider’s bills contain sufficient information to show that the provider may be billing from 10 different physical locations on the same day, even though one person cannot be in 10 places at the same time. Often times, these anomalies are never investigated or remedied, even when they appear on the face of the claims.
Aside from data mining, it remains an open question whether insurance companies currently utilize AI to make claims determinations, to initiate investigations of providers with questionable claims, or to substantiate findings of fraud. Ensuring the correct outcome on claims should be a top priority for insurance companies, but in the case of the health plans they administer for other companies, they have no financial downside risk when a different company is paying for the care. If insurance companies are going to be harnessing the power of AI in claims processing, then AI should be regulated to mitigate the significant human impact of denied claims.
III. Suggestions for a Path Forward
With the relative lack of understanding of the “black box” that is insurance claims processing, there are a number of safeguards and regulatory practices to consider implementing.
A first major recommendation is reexamining state laws to address and remove vague language regarding “clean” claims and instead focusing on correct outcomes. Insurance companies receive a great deal of information from providers and should use that information to determine whether what is being submitted is feasible, anomalous, or fraudulent, on the one hand, or is being submitted in good faith, on the other hand. AI could be used to aid in the process of correct claims determination, but only if there are safeguards in place to identify outlier situations requiring more careful consideration by a human.
Regardless of whether the claim is clean, where a provider’s bills contain information to show that the provider treats the same patients over time, delivers care at reasonable prices, and bases care on what is best for the patient, that provider should be rewarded with higher and faster insurance reimbursements, even if the provider is not sophisticated enough to submit a perfectly clean claim. Currently, only Medicare is implementing with a Quality Payment Program for providers.
Second, insurance payment policies should be subject to regulatory oversight to prevent arbitrary and capricious claims processing decisions. As a recent example, when setting up the No Surprises Act Independent Dispute Resolution Process (IDR), regulatory defined the requirements for pre-IDR negotiations, IDR administrative fees, and other requirements for both payors and providers. Unfortunately, the IDR process once again requires providers to go without payment during the IDR process and requires providers to submit separate claims for review – with separate administrative fees – if the claims occurred more than 30 business days apart or involve different health plans. At present, there is no way to cite IDR outcomes, which leads to potentially conflicting decisions from one IDR to the next. This process also places the timing of payment in the hands of insurance companies who may delay payment even after losing the IDR. There is much to be learned from the implementation of this process, and from the legal challenges to it.
Ideally, payment systems should incentivize necessary medical services and access to care, support the quality of care, and enable efficient use of resources. Policies that discriminate in arbitrary ways should be avoided, and changes to payment policies should be supported by sound reasoning. For example, several insurance companies provide discriminatory payments for advanced practice nurses performing the same role as physicians in primary care and anesthesiology. Discriminatory provider payments were made illegal in the insurance, Medicare, and ERISA context, but federal authorities have not been enforcing this law. Discriminatory payment policies stand to harm insurance companies and patients in the long term. Processing a lower payment for healthcare providers – who are practicing within the full scope of their license under applicable state law – simply lowers the supply of willing providers, particularly those serving in rural areas where there are already significant primary care and anesthesia provider shortages.
In closing, a complex problem necessitates complex solutions. These proposals are properly termed “suggestions” as they should be evaluated for their relative success or failure before being implemented permanently.
