Reg's Blog

Medicare Claims, Audits, Denials and AI

AI or Artificial Intelligence has been in the news a lot over the past few months. ChatGPT is the program that I’ve seen the most about. Elon Musk has come forward warning of the advance of AI and its implications for societies. I’ve seen story after story about how AI has the potential to be a “game changer” in medicine and in science advances but also, how it has the potential to produce scary outcomes. Heck, even Joe Rogan is sounding the alarm after a full version of his podcast was done through an AI creation.

As one would suspect, the advances in AI are finding their way into Medicare and Medicaid to adjudicate claims and to detect potential fraud. The first and most prominent use (for AI) is within Medicare Advantage plans. In an analysis published in the health and life sciences publication STAT, the authors found insurers in the Medicare Advantage plans using AI based algorithms to determine post-acute lengths of stay as well as for prior authorizations for certain levels and amounts of care. The purpose is to place a “best practice” construct around certain diagnoses and conditions, reducing variability. Sounds good in theory.

I have been a proponent of the development of clinical algorithms based on certain diagnoses and patient comorbidities. Readers can find some of those algorithms posted on this blog. I also received a U.S. patent for the development of a web based chronic disease management system that involved a highly integrated series of algorithms and pathways to assist patients and physicians with the management of Type 2 diabetes. What I have never been a proponent of is rigidity such that the pathway or the algorithm is the sole determinant of a patient’s care journey and treatment regimen. Every patient is different and some because of the influence of non-medical issues in their life, will require more integrated approaches in their care and treatment plans. For example, where a patient lives (environment, stairs, etc.), who the patient lives with (caregiver?), and what resources the patient has for assistance are all important factors in determining length of stay in a post-acute setting. In other words, some folks need more time, some can advance to discharge sooner.

The government/CMS has been integrating evidence-based algorithms/pathways/protocols into claims reviews and claim adjudication for several years.  InterQual Criteria, a McKesson Health Solutions product has been used by MAC (Medicare Administrative Contractors), QIOs (quality improvement organizations) and Administrative Law Judges for years; two plus decades (https://www.businesswire.com/news/home/20161219005102/en/CMS-to-Continue-Use-of-InterQual-Criteria).  The theory is that highly researched and fine-tuned, evidence-based data tools can provide a proper roadmap for treatment that emphasizes efficiency and reduced variability and negative outcomes.  Code words for “reduce costs”, primarily. I haven’t seen a whole lot of better care, especially in terms of reductions in repeat utilization patters (re-hospitalizations, etc.) among the elderly, especially those with multiple comorbidities.

A rather good report was done on the heels of the STAT article by the Center for Medicare Advocacy.  That report can be downloaded here: AI-Tools-In-Medicare What I noticed as most interesting in the report is the discussion around slippery-slopes and the gaps between what AI does/doesn’t do and what role humans and policy, play.  For example, the Jimmo v. Sebellius case and its implications.  Jimmo’s decision is fundamentally contrary to how AI is being used to determine continued coverage.  Where AI is used to factor when care (and thus coverage) should end under Medicare, Jimmo basically says that coverage is not dependent on improvement or potential for improvement and can continue if the goal is to resist deterioration or is required by the patient’s need for skilled care. 

Coverage does not depend “on the presence or absence of an individual’s
potential for improvement, but rather on the beneficiary’s need for skilled care.” The settlement
re-emphasized what was already provided for by regulation: restoration potential is not the
deciding factor in determining whether skilled care is required. Skilled nursing or therapy
services are coverable when an individualized assessment of the beneficiary’s clinical condition
indicates that the specialized judgment, knowledge, and skills of a nurse or therapist are
necessary to safely and effectively deliver services.  The settlement applies in the skilled
nursing facility, home health, and outpatient physical therapy settings.

As AI use advances within reimbursed health care, the likelihood of a continued disconnect between providers and insurers and ultimately, patients will grove.  We have an aging society that will continue to demand and utilized, more health care resources.  The federal govt. is intent to continue to drive enrollment in Medicare Advantage plans as traditional Medicare Parts A and B continue to have funding challenges and face, default conditions as tax revenues and fees are headed to a condition of inadequacy to fund the outlays.  While evidence-based medicine and the algorithms it can produce have great promise in many regards, reliance on overly broad, one size fits all approaches can cause unintended consequences in terms of overall patient care and quality.  When reducing utilization and thus, saving dollars is the primary goal, a short-sighted impact is likely – the forest for the trees adage applies. A good article to wrap this post is here: https://skillednursingnews.com/2023/03/ai-use-by-medicare-advantage-blamed-for-increased-denial-of-nursing-home-services/

April 17, 2023 - Posted by Reg Hislop III | Health Policy and Economics, Policy and Politics - Federal, Skilled Nursing | AI, Audits, CMS, Compliance, Industry Outlook, Management, Medicaid, Medicare, Medicare Advantage, Money, Payment, RAC, SNF, Trends