Some legal and technology experts have grown increasingly concerned by a potential lack of human oversight in artificial intelligence-assisted hospice regulatory enforcement.
The AI-driven efforts to crackdown on fraudulent activity could bring significant impacts for legitimate providers, including risk of revocation and closure.
In the first of a two-part series, Hospice News sits down with legal teams and a technology platform provider to examine the AI evolutions in hospice regulatory oversight. The second story will dig into what the technology utilization trends could mean for the outlook of hospice sustainability and compliance.
AI tools can be helpful in identifying potential fraudulent hospice billing patterns, according to Edo Banach, partner at Foley Hoag LLP. He is also co-chair of the healthcare department at the law firm. A mounting concern is that legitimate providers have little recourse to respond to increasingly AI-enabled audit processes, which can threaten their sustainability, Banach said.
“It’s not only that AI is apparently being used, but there are questions about whether there is appropriate human oversight,” Banach told Hospice News. “I don’t see an appropriate level of human oversight. We can use technology, that’s good and appropriate, but you need human oversight before you accuse someone of committing fraud and before you reject their rebuttal. You can use algorithms to deduce fraud, but it really does require human judgement.
Hospices’ revocation risks in AI unknowns
AI tools are increasingly being used in regulators’ auditing response and revocation processing, according to Banach. Hospices and legal experts alike have questioned the level of human oversight due to the language and nature of the responses. Suspected instances of AI “hallucinations” have been included in regulators’ responses to hospices’ audit appeals, he said.
“We’ve been going through these appeals, and you’re not only seeing what seems like AI on the front end in terms of how hospices are chosen, but you’re also seeing it by CMS contractors processing rebuttals,” Banach said. “We’re seeing responses that clearly contain language that no human being would ever write, and if they saw it, would change it. When we get humans involved, there are times that they turn it around and reverse course.”
Regulators are using AI-driven data analytic tools to hunt for billing “anomalies” such as long lengths of stay or high volumes of repeated recertifications. Medicare auditors have used the data to identify fraudulent patterns. Data analytic tools can scour thousands of billing claims against hospice requirements, but human context is incredibly important in regulatory review, Banach stated.
The U.S. Centers for Medicare & Medicaid Services (CMS) in February proposed the use of a new AI enforcement tool to identify potential fraud, waste and abuse. The Comprehensive Regulations to Uncover Suspicious Healthcare (CRUSH) proposal included consideration of data analytics to track billing claims and enrollment in collaboration with law enforcement. Meanwhile, federal agencies such as the FBI and the U.S. Department of Health and Human Services (HHS) Office of Inspector General (OIG) have also employed AI-driven algorithms to identify fraud.
CMS did not reply to Hospice News’ request for comments related to how the agency is using AI in regulatory oversight processes, including hospice revocations.
Hospices lack clarity around the ways that regulatory agencies are using AI tools in fraud detection, particularly when it comes to new enrollment and revocations. This is according to Howard Young, partner at the law firm Morgan, Lewis & Bockius LLP.
Several recent hospice revocations have occurred in fraud hotbed states with unusually high levels of new enrollment, such as Arizona, California, Nevada and Texas. Agencies such as CMS, HHS-OIG, the FBI and the U.S. Department of Justice have used more data-driven tactics to identify and stop fraudulent operators, Young said. However, it remains unclear as to how data analytic and AI tools are being used to enforce hospice revocation in suspected fraud instances, Young said.
“In terms of public statements from CMS, it’s not abundantly clear exactly where they are using AI tools versus data analytics, because analytics are not necessarily driven by AI,” Young told Hospice News. “But it is clear that they are starting to use AI for various purposes. Crushing fraud [was] a specific purpose of testing AI and machine-learning models against the very large Medicare fee-for-service database, including hospice. The Department of Justice’s National Fraud Detection Center is a multi-agency initiative that’s definitely using AI-assisted reviews and AI-driven data mining to identify potentially abusive claims data.”
Hospices need to have a proactive approach to compliance amid the unknowns of regulatory AI-assisted fraud detection tools, Young indicated. Providers may have hidden revocation risks as they navigate a technology-enabled regulatory landscape, he said.
AI-assisted analytic tools are being used to more efficiently review hospices’ billing and medical claims, according to Steve Shelton, senior director of clinical services at Mosai. Formed in 2025 from the merger of Forcura and Medalogix, the tech platform assists home-based companies with referrals, admissions and clinical management.
Hospices need to understand the unique patterns that may stand out through their own internal compliance and billing review processes, Shelton said. AI algorithms in regulatory oversight practices may not catch the nuances of individualized, person-centered care delivery.
Providers need strong human oversight in their compliance management processes to withstand increased technology use in regulatory scrutiny, he stated.
“Hospices should understand their own organization-wide trends, investigate unusual patterns and ensure the clinical reasons behind them are supported by consistent documentation,” Shelton told Hospice News in an email. “There can be legitimate clinical reasons for differences in utilization, length of stay or other care patterns, so an unusual pattern should be treated as a reason for further review, not as proof of noncompliance. Providers that invest in the appropriate data capabilities and education will be better positioned to identify potential concerns and take proactive measures before an audit.”
Companies featured in this article:
