On Ocober 17, the Senate released a 54-page report on escalating denials in Medicare Advantage plans. (Note - Large file; 140 mb).
The report also documents the concentration of Medicare Advantage providers, despite the existence of "hundreds of MA plans" - with three master plans covering 60% of MA patients.
###
See subscription coverage at STAT PLUS.
See open access coverage at Healthcare Dive:
Insight
The report includes a focus on high denial rates for post-acute-care claims (e.g. hospital discharge to Part A nursing home stay (rehab).) An AHRQ report to Congress last month highlighting post-discharge care as a cause of high variance in the cost of sepsis patients. Note, however, that in many cases post-acute care may be critically valuable for the care of such patients. See an October 21 news item about Blue Shield automating prior auth.
AI Corner
Here's an auto-generated Chat GPT 4o article from the Senate report. (Since the Senate PDF is 140mb, I converted it to a text file (.txt) before uploading to Chat GPT).
Senate Report Sheds Light on
Medicare Advantage Claims Denials and Automation Concerns
A recent Senate investigation has revealed alarming findings regarding the use of prior authorization and artificial intelligence (AI) technologies in Medicare Advantage (MA) plans. The report highlights that major insurers—UnitedHealthcare, Humana, and CVS—have used prior authorization to significantly limit access to post-acute care, a critical component of recovery for many seniors. These denials have raised serious concerns about the impact of automation on patient care and access to medically necessary services.
The Senate's Permanent Subcommittee on Investigations (PSI) examined over 280,000 documents from the three largest Medicare Advantage providers, which together cover nearly 60% of all MA enrollees. The investigation found that these companies denied prior authorization requests for post-acute care at rates far higher than for other types of care. For example, in 2022, UnitedHealthcare and CVS denied post-acute care requests at rates three times higher than their overall denial rates, while Humana’s denial rate was a staggering 16 times higher.
Automation and AI: Faster Decisions, More Denials
One of the most troubling findings is the role that automation and predictive technologies play in these denials. The investigation revealed that UnitedHealthcare's implementation of a machine-assisted prior authorization system led to a sharp increase in denial rates. Testing of the technology showed faster processing times but also more adverse decisions—indicating that AI systems flagged more cases for denial, often based on cost-saving measures rather than medical necessity.
The automation of prior authorizations has been framed as a tool to reduce clinician workload and speed up the decision-making process. [*] However, internal documents show that UnitedHealthcare was aware that the use of automation would lead to more denials, and the company approved the use of these technologies despite these concerns. For instance, UnitedHealthcare’s denial rate for skilled nursing facility care increased by nine times between 2019 and 2022.
[*] Press release about Safeforce.com and Blue Shield ramping up automation - here.
CVS and Humana: Similar Strategies, Growing Denial Rates
CVS, which saw a 57.5% increase in prior authorization requests for post-acute care between 2019 and 2022, also developed algorithms that focused on predicting which requests had a high probability of being denied. The company’s internal data showed that the financial savings from denials vastly outweighed savings from approvals, suggesting a strategic emphasis on denying care to protect profits. In one year, CVS saved over $660 million by denying inpatient facility requests.
Humana, on the other hand, saw a 54% rise in denials for long-term acute care hospital requests after instituting training sessions that focused on how to deny care and successfully uphold those denials during appeals. Although the extent of AI’s role in these decisions remains unclear, the company’s internal policies indicated that third-party contractors could use AI systems to assist in the denial process.
Recommendations for Reform
The Senate subcommittee’s report calls for stricter oversight and more transparency in how MA insurers use prior authorization and automation. Among the key recommendations are:
- Enhanced Data Collection: The Centers for Medicare & Medicaid Services (CMS) should require insurers to break down prior authorization data by service category, allowing for targeted audits where denial rates are unusually high.
- Audits of Denials: CMS should conduct targeted audits if data shows increases in adverse determination rates, particularly for post-acute care.
- Regulating AI and Automation: CMS should expand regulations to ensure that predictive technologies do not unduly influence human reviewers, who may face pressure to follow algorithmic recommendations without fully considering the medical needs of the patient.
The Human Cost
While automation is intended to streamline healthcare processes, this report underscores the risks of relying too heavily on AI-driven decisions, particularly when the stakes involve access to vital post-acute care services. The findings of the Senate investigation suggest that these practices may prioritize cost-saving measures over patient care, leaving vulnerable seniors at risk of being denied the treatments they need during critical recovery periods.
The report serves as a wake-up call for policymakers, highlighting the need for regulatory updates that safeguard patient care from the overreach of automation in healthcare.
This story reflects the growing tension between efficiency and ethics in the administration of healthcare services, particularly as AI technologies play an increasingly central role in decision-making processes.
##
See Oct 23 article in Pro Publica about Evicore.
