When the U.S. Centers for Medicare and Medicaid Services (CMS) approved a new technology add-on payment (NTAP) for large-vessel occlusion (LVO) stroke triage and notification software in 2020, the decision made a good business case for AI in early detection of stroke.
It meant that the AI technology was seen as a substantial clinical improvement over existing services. It set a precedent, Gregory Nicola, MD, chair of the American College of Radiology (ACR) Commission on Economics wrote in an ACR practice management bulletin at the time. Providers that used Viz LVO, for example, to triage suspected stroke patients could bill Medicare for up to $1,040 per use. The potential drawback was expiration of the add-on payments.
Notably, CMS estimated that overall, additional payments for inpatient cases involving new medical technologies will decrease by $364 million in fiscal year 2024 as a result of NTAP expirations. In addition, the agency changed its NTAP policies starting in fiscal year 2024 to increase transparency and improve efficiency of the NTAP program and application process.
When the U.S. Centers for Medicare and Medicaid Services (CMS) approved a new technology add-on payment (NTAP) for large-vessel occlusion (LVO) stroke triage and notification software in 2020, the decision made a good business case for AI in early detection of stroke.
It meant that the AI technology was seen as a substantial clinical improvement over existing services. It set a precedent, Gregory Nicola, MD, chair of the American College of Radiology (ACR) Commission on Economics wrote in an ACR practice management bulletin at the time. Providers that used Viz LVO, for example, to triage suspected stroke patients could bill Medicare for up to $1,040 per use. The potential drawback was expiration of the add-on payments.
Notably, CMS estimated that overall, additional payments for inpatient cases involving new medical technologies will decrease by $364 million in fiscal year 2024 as a result of NTAP expirations. In addition, the agency changed its NTAP policies starting in fiscal year 2024 to increase transparency and improve efficiency of the NTAP program and application process.
Regardless, the start in 2020 of Medicare payments for AI software to detect LVO stroke on CT scans precipitated a sharp increase in AI utilization. Five five stroke detection algorithms have received clearance from the U.S. Food and Drug Administration (FDA) as computer-aided triage and notification systems for LVO stroke detection by the end of 2020:
- Viz LVO (Viz.ai), February 2018 via the FDA's de novo pathway
- LVO detection algorithm (Aidoc), December 2019
- Cina Head (Avicenna.ai), June 2020
- Rapid LVO (RapidAI), July 2020
- StrokeViewer LVO (NiCo.Lab), November 2020
Other applications have also since received approval, including FastStroke (GE HealthCare), Neuro.ai (TeraRecon), StrokeSens (Circle Neurovascular Imaging), qER (Qure.ai), Rapid NCCT Stroke (RapidAI), and Brainomix 360 Triage LVO (Brainomix).
AI reportedly played a diagnostic role in approximately 38% of the procedures that resulted in mechanical thrombectomy the first year, AuntMinnie.com reported at the year-one mark.
Thrombectomy outcomes
With a few years of stroke AI utilization in the rearview mirror, experts at UTHealth Houston asked the question, "Does implementation of automated LVO detection software for acute stroke triage decrease time to endovascular thrombectomy (EVT) initiation?"
Their study involving four comprehensive stroke centers, which was published September 18, 2023, in JAMA Neurology, examined AI-enabled automated LVO detection from CT angiography (CTA) coupled with immediate secure messaging to the mobile phones of clinicians and radiologists near the time of CT imaging completion. The primary outcome was the effect of AI-enabled LVO detection on door-to-groin (DTG) time. Secondary outcomes included time from hospital arrival to intravenous tissue plasminogen activator (IV tPA) bolus in eligible patients, time from initiation of CT scan to start of EVT, and hospital length of stay.
Exploratory analysis evaluated the impact of AI implementation on 90-day modified Rankin Scale disability outcomes, and the researchers concluded that software implementation was indeed associated with clinically meaningful reductions in EVT treatment times.
"In this cluster randomized trial including 243 patients treated with thrombectomy over a one-year period, implementation of automated large vessel occlusion detection software led to a statistically significant reduction of 11 minutes in time to thrombectomy initiation," wrote Juan Carlos Martinez-Gutierrez, MD, and colleagues from McGovern Medical School at UTHealth Houston.
Further, a Johns Hopkins Hospital-Intermountain Medical Center joint analysis of the benefits of successful thrombectomy treatment, published in 2022, estimated that an annual 10% increase in a good reperfusion rate for all thrombectomy-treated patients in the U.S. would save approximately $21 million and $37 million for the healthcare system and society, respectively. They said the savings would be driven partially by the prevention of time delays in stroke triage, which avoids accumulation of disability-adjusted life-years.
Stroke care in rural areas
AI applications in stroke identification may be especially useful for neurovascular specialists, radiologists, and stroke and trauma teams that oversee critical care in rural areas where stroke mortality rates are higher. The American Academy of Neurology has endorsed telestroke for delivering acute stroke consultation, proper reimbursement for on-call consultations, and use of advanced technological services.
In an example of combining resources to improve stroke care in the rural communities in southern Missouri, for example, Texas County Memorial Hospital (TCMH) received a grant from Rural Citizens Access to Telehealth to support the implementation of RapidAI's software in the hospital’s radiology department. TCMH is classified as a Level III stroke facility. The closest Level I (comprehensive) stroke center is 90 miles away.
Noncontrast CT
Several vendors have also developed AI software for detecting and triaging cases of suspected LVO stroke on noncontrast CT (NCCT) exams, including RapidAI and Brainomix. A multireader, multicase study published in September 2023 in Frontiers in Neurology by radiologists and neurosurgeons at the Mayo Clinic in Rochester, MN, found that decision support using e-ASPECTS software (Brainomix) significantly improved the accuracy of the Alberta Stroke Program Early CT Score (ASPECTS). This score is a scale used to gauge the extent of ischemic change in the middle cerebral artery territory of acute ischemic stroke (AIS) that can be quantified on an NCCT scan of the brain.
The study's authors, who include David Kallmes, MD, from the department of radiology at Mayo Clinic Rochester, noted that guidelines in both the U.S. and Europe recommend using ASPECTS alongside clinical and other imaging criteria to guide patient selection for reperfusion therapy treatment in AIS; a drawback, they said, being inconsistent scoring of ASPECTS, even by trained raters.
"This study demonstrates the ability of an FDA-cleared artificial intelligence (AI) decision support tool (e-ASPECTS) to improve the performance of US physicians when deriving the ASPECTS," according to the authors. The study called attention to the impact of AI software on physician performance in ASPECTS scoring. Agreement with a reference standard (expert consensus read with reference to follow-up imaging) was evaluated with and without software support.
While the study was supported by Brainomix and had several limitations, consistent improvement was seen in both neurology and neuroradiology-qualified doctors, after a subgroup analysis based on the clinical training of the readers.
Stroke systems of care
Demand for AI software may also be sparked by stroke systems of care (SSOC), which have been developed by states using tiered models and various combinations of legislation, protocols, and additional supports, according to the U.S. Centers for Disease Control and Prevention (CDC). The CDC also funds the Coverdale Stroke Program and has resources devoted to stroke policy and stroke systems of care interventions.
Signs of the times, however, may be what comes of a recent Massachusetts law requiring development of a new triage framework for stroke care and emergency medical services. The Berkshire Eagle reported that the decision "marks a departure from existing state law .... As it crafts new guidelines, [the Department of Public Health] DPH can consider transporting patients across state lines, determine plans for triaging and transporting suspected stroke patients and establish other criteria to gauge 'which level of care is the most appropriate destination ...' "
Furthermore, February 7, 2024, marks one year since the American Heart Association issued new recommendations for structuring stroke programs. These recommendations included AI-supported neuroimaging on the list of ideal foundational requirements for all types of stroke centers.