LVO Stroke - Viz.ai

[AnxietyMonster]

Viz.ai was granted CMS's new added technology funding of 1024$ per patient use. I feel like this is a huge move for them and the field of radiology as well.

Viz.ai Granted Medicare New Technology Add-on Payment

/PRNewswire/ -- In a groundbreaking ruling, CMS has granted Viz.ai the first New Technology Add-on Payment (NTAP) for artificial intelligence software. NTAP,...

www.prnewswire.com

---

Members don't see this ad.

 

[Cognovi]

Viz.ai was granted CMS's new added technology funding of 1024$ per patient use. I feel like this is a huge move for them and the field of radiology as well.

Viz.ai Granted Medicare New Technology Add-on Payment

/PRNewswire/ -- In a groundbreaking ruling, CMS has granted Viz.ai the first New Technology Add-on Payment (NTAP) for artificial intelligence software. NTAP,...

www.prnewswire.com

Can you think of another use case for which this sets the precedence?

 

[maxxor]

Can you think of another use case for which this sets the precedence?

Didn’t CMS approve HeartFlow and then cut the payment a year later?

Mammo CAD also had an add on payment that was recently bundled.

 

[AnxietyMonster]

Alright fair enough, good to know. I thought this would be some major development, partly from the article wording. I should have known better than trust tech-journalism.

Regardless before medical school I remember there being a lot of concern with LVOs and thrombectomy time, hopefully this improves clinical outcomes significantly!

As for other similar applications @Cognovi , I was considering large PEs? Not sure if this would work in clinical practice, I am but an MS3 with a strong interest in radiology. I do greatly appreciate all the insight provided on these forums + the same from Aunt Minnie

 

[Cognovi]

Alright fair enough, good to know. I thought this would be some major development, partly from the article wording. I should have known better than trust tech-journalism.

Regardless before medical school I remember there being a lot of concern with LVOs and thrombectomy time, hopefully this improves clinical outcomes significantly!

As for other similar applications @Cognovi , I was considering large PEs? Not sure if this would work in clinical practice, I am but an MS3 with a strong interest in radiology. I do greatly appreciate all the insight provided on these forums + the same from Aunt Minnie

There are certainly many other emergency diagnoses but none rivals the immediacy of the "time is brain" mantra. Viz.ai had to really torture the data to convince CMS that saving just 40 minutes can improve outcomes. The system boils down to notifying the tertiary hospital's interventionalist directly, rather than waiting for the primary hospital's radiologist to tell the emergency physician and then the stroke neurologist and then the interventionalist. The AI is a necessary triage (but far from perfect at <90% sensitivity and specificity) so that the interventionalists do not become overwhelmed on call essentially having to do prelim reads on CTAs for their entire referral network. But the time savings comes from the workflow of pushing the patient from from OSH CT scanner to the interventionalist's doorstep as soon as possible. It's an app that essentially makes less-well-oiled hospitals follow a care pathway.

 

[Cognovi]

There are certainly many other emergency diagnoses but none rivals the immediacy of the "time is brain" mantra. Viz.ai had to really torture the data to convince CMS that saving just 40 minutes can improve outcomes. The system boils down to notifying the tertiary hospital's interventionalist directly, rather than waiting for the primary hospital's radiologist to tell the emergency physician and then the stroke neurologist and then the interventionalist. The AI is a necessary triage (but far from perfect at <90% sensitivity and specificity) so that the interventionalists do not become overwhelmed on call essentially having to do prelim reads on CTAs for their entire referral network. But the time savings comes from the workflow of pushing the patient from from OSH CT scanner to the interventionalist's doorstep as soon as possible. It's an app that essentially makes less-well-oiled hospitals follow a care pathway.

Since heartflow was brought up, it reminds me that we should also think about the differences and similarities with ST-elevation myocardial infarction, where "time is myocardium." In the status quo hyperacute workflow, noninvasive imaging (coronary CTA) does not play a role. There's no time for people to mess around with radiology at all. Patients go straight from troponin/EKG to cath lab. The interventional cardiologists accept a substantial rate of angio-only (ie, no balloon or stent) due to false positive diagnostic error, distal embolization, or MINOCA.

Why is it different for stroke?

 

[scoopdaboop]

So can anyone tell me where the money goes to? Hospital system which will then disperse to radiologists?

 

[Cognovi]

So can anyone tell me where the money goes to? Hospital system which will then disperse to radiologists?

My guess is that the add-on fee will go wherever the technical component of the scan's fee goes, ie, whoever owns the scanner and employs the technologists, as they are likely the ones paying the Viz.ai subscription fee. That's usually the hospital, especially as these are emergency department patients and not some other outpatient imaging center. Then the dispersement comes down to whatever general financial arrangement the hospital has with the radiology group, like whether the radiologists are employed or contracted. It'll filter down eventually and create the same incentives as if the radiologists were getting the money. Just look at how all radiology practices choose to add on CAD to mammography even though many radiologists don't find it useful and it takes up time to review. It's for collecting money from the payor (until 2017, CAD was an add-on code; now it's ubiquitous and bundled in). In the Viz.ai situation, it's a win-win for the primary hospital's radiologists: you no longer have to rush to read stroke code CTAs and you (indirectly) get paid for it. If I understand correctly, the reimbursement is not going to the tertiary hospital's interventionalist who now has to do a wet read on their cell phone app to determine whether the AI was right or not and then call the access nurse to coordinate transfer. Sounds like a raw deal unless there is some money flow where primary stroke center (the OSH where the radiologist is) pays to affiliate with a comprehensive stroke center (where the interventionalist is) to send patients.

 

[Little Green Mensch]

If we had the ability to detect large vessel stroke with high-sensitivity point-of-care biomarkers and 12 scalp electrodes, maybe we'd fire up the biplanes more often. Would still need a dry head CT to make sure that it wasn't a hemorrhage. And CT perfusion still for everyone beyond 6 hours.

Biomarkers for CVA would be very helpful, though, would potentially save sorting through a lot of negative CTAs for "stroke."

 

[AnxietyMonster]

My guess is that the add-on fee will go wherever the technical component of the scan's fee goes, ie, whoever owns the scanner and employs the technologists, as they are likely the ones paying the Viz.ai subscription fee. That's usually the hospital, especially as these are emergency department patients and not some other outpatient imaging center. Then the dispersement comes down to whatever general financial arrangement the hospital has with the radiology group, like whether the radiologists are employed or contracted. It'll filter down eventually and create the same incentives as if the radiologists were getting the money. Just look at how all radiology practices choose to add on CAD to mammography even though many radiologists don't find it useful and it takes up time to review. It's for collecting money from the payor (until 2017, CAD was an add-on code; now it's ubiquitous and bundled in). In the Viz.ai situation, it's a win-win for the primary hospital's radiologists: you no longer have to rush to read stroke code CTAs and you (indirectly) get paid for it. If I understand correctly, the reimbursement is not going to the tertiary hospital's interventionalist who now has to do a wet read on their cell phone app to determine whether the AI was right or not and then call the access nurse to coordinate transfer. Sounds like a raw deal unless there is some money flow where primary stroke center (the OSH where the radiologist is) pays to affiliate with a comprehensive stroke center (where the interventionalist is) to send patients.

Similar to CAD, I wonder what unintended consequences this will have. I suspect you are right that there might be a decreased rush on stroke code CTAs. Therefore with the <90% sensitivity and specificity, will there be a not-insignificant # of patients who actually do WORSE?

The law of unintended consequences. Hopefully the good outweighs the bad.

 

[Cognovi]

If we had the ability to detect large vessel stroke with high-sensitivity point-of-care biomarkers and 12 scalp electrodes, maybe we'd fire up the biplanes more often. Would still need a dry head CT to make sure that it wasn't a hemorrhage. And CT perfusion still for everyone beyond 6 hours.

Biomarkers for CVA would be very helpful, though, would potentially save sorting through a lot of negative CTAs for "stroke."

Exactly: imaging is required to select patients with stroke who would benefit most from thrombectomy. We know from the multiple earlier thrombectomy trials that taking less well-selected patients makes for a negative study.