The headline said it all. “AI should be free … and factored into the cost of doing business.” The words struck fear into the hearts of many AI companies like a man wearing a red cape and devil horns at a Billy Graham revival. Free AI? Seriously? Knowing and respecting the source of the quote, I have to think it was either misstated, misinterpreted, or merely said for shock value.
Michael J. Cannavo.
I grew up with the premise that nothing in life is free and learned that everything has a cost associated with it. You learn that lesson quickly at age 8 when you are mowing lawns using a reel push mower, hand edging and trimming. Shoveling sidewalks on days when school was cancelled provided yet another life lesson. So… free? I think a better way to have said this was either not billed as a separate charge or bundled in with the procedural cost.
Several of the largest reading groups in the country have been billing patients an additional $40 for AI to provide a second opinion on their breast imaging studies. Although many request this, the patient can always elect to opt out of having it read by AI as well. While there is nothing technically wrong with this scenario, it sets up a dividing line between the haves and have nots — those who have the financial resources to have AI complement the radiologist’s interpretation and those who don’t.
Standard of care
If using AI becomes the standard of care down the road then everyone should have access to it, not just those who are willing to pay for it. There are several papers in the National Library of Medicine that reference the term “standard of care.” It’s important to understand that the standard of care is a legal term, not a medical term, something few realize. Basically, it refers to the degree of care a prudent and reasonable person would exercise under the circumstances.
What is considered reasonable and prudent by a radiologist remains to be seen is, as currently no definition or guidelines are in place for this. If a patient has a prior family history of breast cancer, for example, using AI would in almost all cases be considered more than reasonable and prudent. But what if AI isn’t used for whatever reason (except a well-documented patient refusal) or the patient can’t pay for it? Not using AI in this particular case opens up Pandora’s box to a potential lawsuit.
This is especially crucial if the radiologist may have missed a critical finding or could have obtained additional clinical information that AI may have provided. It’s not a question of if this scenario happens but instead when. I’m surprised there hasn’t been a class-action suit on this already.
That brings us back again to who pays. Nearly a dozen AI vendors out of the 600 or so offering U.S. Food and Drug Administration (FDA) – cleared products in the marketplace have had investments made in them over $15M. Most investment brokers and venture capitalists like to see a return of at least a 5 times (5x) their investment and often 10 times (10x)…and that is before the company itself makes any money.
Return on investment
So if AI is offered for “free” how will they get their investment back? One way is if the AI company gets sold to a PACS vendor with very deep pockets. This would allow the investors to be paid back and also put a few bucks in the pockets of those who helped the company grow. Can or will this happen? Probably with a few of the better-known AI companies, but then what about the rest of the marketplace?
It is unfathomable to think that AI will continue to be reimbursed in the hundreds of dollars per study as it is today. As reimbursements will be much less, making up the difference in higher volume is going to be crucial. With 300 million conventional X-rays and 9 million CT exams being done annually in the U.S. alone (growing at a 4-5% compounded annual growth rate) and the balance other imaging studies making the difference that’s a lot of studies.
Is it enough to support the 600+ vendors in the marketplace? No, but realistically 75% of the vendors offering products today will not operating as they currently are in the next two years. That leaves roughly 150 vendors to offer from their AI. How much can they realistically generate revenue-wise per study?
Don’t shoot the messenger, but most I have chatted with seem to agree that conventional x-rays should be priced at under $5 per case and CT’s at less than $20 per case. That still should provide more than enough revenue for companies to sustain growth and development. Some companies with highly specialized algorithms can still charge into the hundreds but those are few and far between.
AI has been spoiled (there is no other word for it) by a government program called the New Technology Add-On Program (NTAP). NTAP provides an additional payment aimed to bridge the gap between the cost of the technology and the existing Medicare Severity Diagnosis-Related Group (MS-DRG) reimbursement rates.
Ideally this would have jump started the AI marketplace but the process to qualify for NTAP dollars and restrictions were so complex that few qualified. This is despite payments that could exceed $1,000 each time the companies’ AI algorithm was used. The actual reimbursement formula is also fairly complex and varies depending on the specific technology, hospital costs, and DRG payment rates. Most times reimbursement was significantly less than the $1,000+ maximum, maybe a quarter of that. That said, we are still talking AI use generating as much (if not more) than 10 times what a radiologist may get for their professional interpretation of the imaging study. Is this right?
ROI
AI companies will say they have millions invested in the development of their algorithms and need to show a return on their investment. Radiologists will say they have a minimum of 10 years of advanced educational training to get them where they are now and also must also bear the responsibility for the diagnostic interpretation.
Both sides have valid points. You can’t sue a software company for a missed interpretation though, as the limits of liability in most software agreements clearly exclude that. Like it or not, the bottom line is you can have a radiologist’s interpretation without AI but not AI without a radiologist signing off on the report. Both working together is ideal. That brings us back again to the question, should AI be free?
Ideally, ALL radiographic studies should benefit from the use of AI with very few well-documented exceptions. In the U.S., that amounts to around 400 million procedures (excluding dental which brings the total to almost 700 million with). Worldwide the number increases to around 4 billion, although getting accurate or current data on volume is a challenge.
The key here is AI use has to be an all-or-nothing scenario; it cannot or at least should not be used selectively. Therein lies another conundrum. AI’s acceptance has lagged over the past decade because frankly many radiologists don’t trust AI. While a radiologist can take exception to AI’s findings, many are reticent to do that. Instead, they will document AI’s findings in the final report.
One notable study found that over 70% of low-dose CT lung cancer screenings had incidental findings of which about 10% required clinical follow up. The study did not indicate if these findings were caught by AI or a radiologist during the interpretation process. However, several published studies have shown that AI is exceptional in catching incidental findings. It is important to note that while there is no official mandate for readers to report incidental findings in patients’ radiology reports, the American College of Radiology (ACR) highly recommends it.
Inseparable
In a few short years AI is going to be an inseparable part of the diagnostic interpretation process. Few will deny that. While every personal computer you buy doesn’t have all the apps you need — those you buy separately — it does come bundled with the operating system (OS).
Just as PACS software is considered the OS of radiology, various forms of AI will be part of the OS of PACS as well. Microsoft and Apple get paid every time their OS resides on a machine. Windows and MacOS aren’t free even though it seems like it is. AI vendors will get paid as well for their algorithms being used as part of the PACS software. You will never see it as a line item but rest assured the cost of the AI software is buried in the overall cost. This is how AI should be addressed: bundled in the cost of the PACS or other software.
Most probably the study cost will need to be very slightly increased with the AI cost factored in or radiologists and/or facilities are willing to take a very slight decline in reimbursement. Whatever the cost of AI is, it needs to be bundled into the PACS software cost, not broken out and shown separately. This drives me crazy just like it does when I take my car in for an oil change and have a line item charge for “shop supplies.” Just bury it in the final price please.
Transparency
The cost of AI also needs to be transparent to everyone — patient, facility and provider — with everyone in agreement that the value of AI is worth it. We aren’t quite there yet but are slowly getting there. Speeding up reporting time benefits the radiologist, facility, and patient with some stating that it can reduce dictation time by as much as 50%. I’d buy half of that but it’s still a significant improvement.
The mega groups might pay for this aspect of AI to improve their radiologist’s performance and increase revenue but beyond what they will pay for is a big question mark. Most facilities typically want to know how this benefits them financially as well.
There are a plethora of other benefits of AI like Increasing radiologist’s performance relative to relative value units (RVUs), improved patient care, reduced turnaround time (TAT) and more but showing a hard dollar return on investment (ROI) still remains key.
It comes down to who benefits most often pays the most…… with the bottom line being there is no such thing as a free lunch. Someone, somewhere is gonna hafta pay.
Michael J. Cannavo is known industry-wide as the PACSman. After several decades as an independent PACS consultant, he worked as both a strategic accounts manager and solutions architect with two major PACS vendors. He has now made it back safely from the dark side and is sharing his observations.
His healthcare consulting services for end users include PACS optimization services, system upgrade and proposal reviews, contract reviews, and other areas. The PACSman is also working with imaging and IT vendors to develop market-focused messaging as well as sales training programs. He can be reached at pacsman@ix.netcom.com or by phone at 407-359-0191.
The comments and observations expressed are those of the author and do not necessarily reflect the opinions of AuntMinnie.com.