By Meredith Wilkerson
We often hear from founders: “our product is an AI-enabled solution for solving [issue stemming from out-sized pressures on the US healthcare system].” AI is now a well-recognized buzzword and is seemingly poised to change our path forward in addressing US healthcare needs efficiently and ethically. However, healthtech founders must temper their excitement with an understanding of the speed at which AI-enabled solutions are being adopted into the reimbursable clinical workflow. Of course, I’m referring to regulatory hurdles and other barriers to user adoption.
My focus in this blog is around the regulatory framework and potential user adoption of AI in healthcare. Coincidently, an overarching message from my previous blog (FDA Approval is not the End of the Road) holds true here: the reimbursement strategy must be developed in parallel with regulatory and clinical strategies for a given technology to avoid additional costs, time, and product launch disasters. We at Plains Ventures are seeing startups developing exquisite AI-enabled solutions, and while the FDA has approved 950 AI-enabled products through August 7th, 2024; it appears CMS has only made positive coverage and payment decisions for less than a dozen of them – painfully limiting access for those who can’t afford the cash-pay amount. While we expect that this will improve in the years to come, founders and investors alike would do well to understand that the rate of change may not be as quick as we would like and set their expectations (and capital plans) accordingly.
The relative paucity of reimbursable AI-enabled products in healthcare can be at least partially attributed to the time-intensive requirements for navigating the CMS reimbursement decision process. CMS decisions appear to heavily take into consideration the actual clinical demonstration of the technology’s ability to: 1) influence care decision(s) and show that the improvement in those care decision(s) ends up positively influencing later outcomes; or, 2) provide new diagnostic information or improve diagnostic performance vs. standard of care. There are non-trivial nuances of details here around whether the AI-enabled technology is for out-patient or in-patient settings. Regardless, the main point founders should recognize is this: CMS considers actual clinical performance in their decision-making process. For AI-enabled healthtech, the exact threshold of clinical significance acceptable to CMS is unclear. Thus, many companies spend several years with only an unlisted procedure code for reimbursement. This can hinder widespread adoption, as many healthcare systems are hesitant to adopt technology without a specific reimbursement code (e.g., CPT code) due to the extra reporting necessary to receive reimbursement for an unlisted procedure code. The coveted Category I CPT code is often contingent upon widespread utilization PLUS support from associations/specialty societies to endorse the need for the code. The time (and capital!) it takes to achieve such traction can prove to be a surprise hurdle for which many founders had not planned. If founders haven’t set investor expectations and prior financing terms appropriately, this can manifest as unexpected dilution and missed milestones, which can create extremely painful situations as companies navigate fundraising for the additional need.
Another aspect often overlooked by eager founders developing AI-enabled solutions for healthcare is the actual motivations for payers and providers to adopt these new solutions. These two major stakeholders often have differing motivations for adopting new technology. Founders must have a grasp of how these two groups view value creation of AI-enabled solutions for healthcare. For example, the cardiovascular imaging industry is a prime candidate for AI innovation due to the disproportionate rate of growth in diagnostic radiology workloads. The successful implementation of AI in this care area depends on the ability for workflow integration and collaboration between providers and payors. From the provider perspective, there is the mounting pressure to efficiently minimize a patient’s length of stay while simultaneously improving care – AI solutions that providers are likely to favor should address this factor in a way that can be easily integrated into existing clinical workflows. From the payor perspective, the value of an AI application may differ substantially between healthcare systems. Specifically, in a healthcare system with poor access to specialist radiologists, a radiology AI application may help improve diagnostic accuracy and clinical outcomes; whereas in a healthcare system with good access to specialist radiologists, the same application may only serve to improve the efficiency of the reporting radiologist more than it actually impacts diagnostic accuracy. Ultimately, the actual value of an earlier or more accurate diagnosis can be difficult for providers and payers to quantify and may not be realized for months to years after implementation (read: yet another barrier to adoption that will require founders to persuade their customers to ignore).
Although several AI developers have indeed received CMS reimbursement, a continued lack of standardized reimbursement is likely to significantly inhibit mass progress, stifle innovation, and prevent patients from accessing advanced diagnostic technologies. Health policy experts have noted that some of this lack of clarity to-date is a result of misincentives in reimbursement frameworks – sustained adoption of AI may be especially challenging for fee-for-service payment models because providers and payers must balance a desire for innovation with concerns associated with spending and potential overuse. With time, as value-based payment models (i.e., measuring improvement in quality becomes increasingly important at decreased costs) mature, those adverse incentives should be resolved, and AI-based tools are likely to proliferate across the healthcare landscape. Until these policies mature, though, founders and investors alike would do well to factor this uncertainty into their timeline and capital expectations.
We at Plains Ventures are eager to support healthtech founders who are developing AI-enabled solutions to efficiently and ethically benefit healthcare systems. We are on the lookout for founders with a strong rationale around how their solution can significantly improve the standard of care (e.g., downstream testing/costs which contribute to a patient’s diagnosis and treatment plan) and a strong consideration of today’s regulatory and adoption hurdles specific to AI in healthcare.
To talk to an investment principal and learn more about Plains Ventures, click here.
