Dr. Anthony Chang

“Innovation drives hope, but reimbursement drives adoption.” 

GPT4o when prompted for a quote on financial dimensions of AI in healthcare

Artificial Intelligence is rapidly transforming how healthcare is delivered, documented, billed, and optimized. A topic that I am asked to comment or speak on more often these days is the financial and/or economical aspects of AI deployment in a health system. The so-called “ROI on AI” is a phrase we hear often as AI continues its march to becoming more pervasive in healthcare but accountability needs to be present. To put the following discussion in context, the FDA has cleared over 1,000 AI tools for clinical use with the majority of these tools in the radiological services.

Key financial elements for reimbursement success include: FDA clearance (probably one of the most important elements on this list), published or studied evidence of clinical benefit, economic modeling or real-world return on investment, integration into clinician workflow, application under supervision of a qualified provider, support by payer-specific reimbursement policies, and clear documentation that couples AI action into billing logic and improved outcome.

More specific reimbursement strategies are listed below:

1) Outpatients and AI Services:

• Current Procedural Terminology (CPT) Category III Codes (Emerging Technology Codes): The codes that are now used by providers to bill payers for services that include AI tools include: proprietary laboratory analyses (PLA) codes, category III codes (new technology temporary tracking codes), and permanent category I codes (limited). CPT Category I codes are for established medical procedures and services, while CPT Category III codes are temporary codes (four numbers + T and sometimes called “T codes”) for emerging technologies, services, and procedures and therefore most relevant for AI tools and have potential for the more established Category I codes. Payment for Category III codes, released twice a year, are based on payer policies so may be limited and variable in reimbursement levels. Lastly, CPT Category II codes are supplemental tracking codes used for performance measurement and quality improvement initiatives and has potential relevance for future AI tools in this domain.
• Examples: One example is the IDx-DR AI tool for diabetic retinopathy detection (CPT 92229 for about $50). In addition, the automated analysis of coronary atherosclerotic plaque for coronary artery disease (CAD) using AI can be reimbursed under category III CPT code in the Hospital Outpatient Prospective Payment System (OPPS)(0625T at about $950).
• Healthcare Common Procedure Coding System (HCPCS) C-Codes coupled to Ambulatory Payment Classification (APC). These codes are available under OPPS for FDA-cleared SaaS or per use for temporary extra payment until CMS has enough data to assign a longer term APC. An example is the HeartFlow FFR-CT AI tool (C9606 but linked to APC 5724).
• Clinical Application and Payment Strategy: This payment strategy is good for outpatient or diagnostic services with defined workflows (e.g. imaging or EKG analysis). 

 

2) Inpatients and AI Services:

• New Technology Add-On Payment (NTAP): This is a temporary Centers for Medicare and Medicaid Services (CMS) program that is designed to reimburse hospitals and health systems for using new, high-cost technologies during inpatient care. The AI tool needs to have FDA clearance with clinical benefit with the payment on top of diagnosis related group (DRG) with up to 65% of technical cost.
• Examples: The first AI tool to receive NTAP approval (2020) was Viz.ai for stroke detection with an add-on payment (up to $1,040). In addition, an AI tool for echocardiography in heart failure patients (HFpEF)(Ultromics’ EchoGo Heart Failure AI tool) is also approved for NTAP payment (2023)(XXE2X19 code for about $1,025).
• Clinical Application and Payment Strategy: This payment strategy is good for AI tools that are used during inpatient hospital stays.

There are other potential mechanisms for reimbursement for AI tools that are less straightforward, as they are in transition:

1) Pathway to Payment (PTP): This is a new promising Medicare benefit category called Algorithm Based Healthcare Services (ABHS) for FDA-cleared AI tools for outpatient reimbursement. These AI tools are proposed to receive dedicated ambulatory payment codes (APCs) under Medicare for at least 5 years so innovators have a more favorable timeline for AI service development and implementation as well as real-world cost and utilization data. It is hoped that this program will end the ad hoc basis for AI tools to be reimbursed under CMS. There is hope that the Health Tech Investment Act (S.1399), which is a bipartisan Senate bill introduced in April 2025 to create a new reimbursement pathway under Medicare’s Outpatient Prospective Payment System (OPPS) specifically for ABHS, will bring a more favorable milieu for AI in health innovation (longer structured reimbursement and transition to permanent APC assignment).

 

2) Medicare Coverage of Innovative Technology (MCIT): This was a program for both inpatients and outpatients proposed by CMS to provide immediate Medicare coverage for up to 4 years after FDA market authorization for breakthrough devices in order to reduce the lag between FDA approval and CMS coverage. If an AI tool has FDA breakthrough device designation then it can qualify under this MCIT program. This program was unfortunately revoked in November 2021 by the Biden administration (for insufficient evidence of benefit) and is no longer active, but there are ongoing discussions from MedTech and AI companies to have MCIT 2.0 (Transitional Coverage for Emerging Technologies or TCET) in the near future.

Overall, AI is not reimbursable if the AI tool is administrative (such as automating documentation) but one can use AI in the form of large language models for DRG optimization. In addition, most AI tools that are not FDA-cleared or lack clinical evidence are not likely to be covered by payers. Finally, physician offices have more of a challenge for AI tool reimbursement under physician fee schedule and CPT category I.

In short, the main mechanisms for reimbursement of AI tools currently is relatively fragmented under CPT/Category III for outpatient and NTAP for inpatient structures (the latter more than the former) that are challenging at best. A more structured financial reimbursement strategy for AI tools will be particularly helpful for innovation in this domain and also useful for underserved or rural settings. Future strategies for AI technology reimbursement will include: maturation of AI tools from CPT category III into category I or HCPCS coupled with APC as well as the new PTP/ABHS for outpatients, expansion of NTAP coverage for inpatients, new CMS pathways and payment modifications for coding, AI inclusion in bundled payments, real-world evidence mandates for AI tools for local coverage determinations, and AI tied to quality and equity metrics especially at the population level.

The reimbursement strategies for AI in healthcare will be among the popular topics covered at AIMed25.

AIMed25 is the longest running meeting focused on artificial intelligence in medicine and healthcare for clinicians, healthcare leaders, data scientists, entrepreneurs, and innovators who are shaping the next era of healthcare with artificial intelligence.

Across three action-packed days, experience:

• Breakthrough discussions on Generative AI, Agentic AI, Large Language Models, and iXR

• Specialized tracks on Pediatrics & Neonatology, AI in Medical Education, and AI & Mental Health

• Hot topic breakfast workshops — you choose the conversations

• 20+ clinical subspecialty breakouts

• Scholarship-winning abstract competitions

• The prestigious ABAIM certification course

• Plus, a brand-new Chief AI Officer agenda 

AIMed25 will be held at the Manchester Grand Hyatt in San Diego on November 10-12, 2025. It is more than a conference - it's a community, a movement and a catalyst for change. Secure your ticket today and be part of the future of healthcare.