Abstract
The first ever insurance reimbursement for an artificial intelligence (AI) system, which expedites triage of acute stroke, occurred in 2020 when the Centers for Medicare and Medicaid Services (CMS) granted approval for a New Technology Add-on Payment (NTAP). Key aspects of the AI system that led to its approval by the CMS included its unique mechanism of action, use of robotic process automation, and clear linkage of the system’s output to clinical outcomes. The specific strategies employed encompass a first-case scenario of proving reimbursable value for improved stroke outcomes using AI. Given the rapid change in utilization of AI technology in stroke care, we describe the economic drivers of stroke AI systems in healthcare, focusing on concepts of reimbursement for value added by AI to the stroke care system. This report reviews (1) the successful approach used by the first NTAP-approved AI system, (2) economic variables in insurance reimbursement for AI, and (3) resultant strategies that may be utilized to facilitate qualification for NTAP reimbursement, which may be adopted by other AI systems used in stroke care.
Keywords: Artificial intelligence; stroke, insurance reimbursement; new technology add-on payment; economics; neuroimaging
Introduction
Reimbursement in healthcare is rapidly evolving to accommodate novel treatments and technologies. 1 Historically, the United States Centers for Medicare and Medicaid Services (CMS) reimbursed hospitals a fixed amount within the Inpatient Prospective Payment System framework for inpatient hospitalizations according to Medicare severity diagnosis related group (MS-DRG, hereafter shortened and referred to as DRG). The DRG is comprised of the average cost per DRG in the prior two years. This cost assessment presents a problem because the system does not immediately recognize the value of nor reimburse new innovative technology used for certain DRGs.
Recently, the CMS has taken a step toward encouraging and financially supporting the adoption of innovative artificial intelligence (AI) technology. 2 In 2020, the CMS awarded the first new technology add-on payment (NTAP) for an AI system for stroke triage. 3 The system accelerates the time-sensitive detection and treatment of patients with stroke. Stroke triage is a critical process that benefits from process improvement due to the time-sensitive loss of neurons that is minimized by the time savings afforded by automated triaging. The clinical benefit of more rapid care is demonstrated in both years of life and tens of millions of annual healthcare dollars saved. 2 The now-approved NTAP for this AI technology provides additional reimbursement to support its use and to bridge the payment gap for the DRG until it recalibrates in 2 to 3 years.1,4,5 Since the core criteria for the NTAP, fulfilled by this first approved NTAP for AI, were previously described in detail, 1 here we expand on the economic drivers and critically analyze how the design and products of that AI technology linked to multiple strategies for NTAP qualification.
In this article we analyze the successful pathway for NTAP approval of the stroke triaging AI system and detail methods that may be used by other AI-based systems to similarly qualify for NTAP-based reimbursement. We describe the economic drivers, opportunities, and threats to qualification for NTAP, and define methods that other healthcare AI software may utilize to prove the impact of their technology, namely through rigorous data collection and cost-analyses that demonstrate substantial improvement in clinical outcomes.
AI in healthcare: Benefits mismatched with cost
The promise of AI in healthcare is gaining widespread acceptance and resources are increasingly being dedicated to rectifying limited or absent financial reimbursement. AI is expected to make substantial contributions to stroke radiology by improving image acquisition, reconstruction and processing, interpretation and pre- and post-test probability triaging, and acceleration of report generation and communication.6–8 By automating and streamlining probability assessments and augmented pre-reads, AI can improve the accuracy of the diagnosis. This improved accuracy correlates to increased diagnostic quality and in turn the level of safety and accelerated time to definitive diagnosis and treatment, and collectively produces superior clinical and financial outcomes. 9 However, in addition to the clinical, scientific, and technical barriers that such technologies must overcome, they must also be financially feasible in terms of the cost of adopting and implementing new technologies. Although the cost may be offset by increased efficiency (e.g. investment in picture archiving and communications systems obviated the need for costly film, file room space, and personnel), this is not always immediately the case. The cost typically falls on hospitals while the financial benefits of successful deployment accrue to payers, though individual radiologists and hospital departments often also benefit in improved reading efficiency and quality.
Hospitals routinely face complex decisions when evaluating if new technology should be purchased. Even if the in-hospital clinical benefit is proven, cost remains an obstacle when the financial benefits of the investment primarily accrue to outside parties, such as the insurance payers. Moreover, estimating the ultimate financial value of the investment can be difficult. If the technology were a standard capital expenditure, the investment decision would be based on a calculated payoff from an estimated demand, and the expected life of the capital equipment, which gives the net present value of the capital expense net of depreciation and including maintenance. With new technology, reticence in purchasing stems from not being able to estimate the demand with confidence, uncertainty over the commitment of the payer (i.e. insurance) to make the reimbursement permanent, and unexpected changes in payment rates, evolution of exclusion rules, and rapidly changing technology resulting in uncertainty over the useful life of the investment. Additionally, the time for the AI to be fully integrated into the hospital’s technology infrastructure and recognized by billing departments can take months, which may supersede the two years that the NTAP covers for reimbursement from its market introduction. In short, multiple factors can slow the use of AI technology, which hinders its adoption.
Policy makers have argued that the mismatch between the parties bearing the costs and reaping the benefits of new technologies is the reason why AI adoption and reimbursement is limited. Based on the DRG regimen, the U.S. system for healthcare reimbursement lacks a mechanism required to incentivize front-line hospitals and their providers to trial and adopt new technologies. The NTAP program, one of several upcoming solutions, can encourage providers and hospitals to adopt new AI technology.
NTAP: A solution to the AI benefit-cost mismatch
The NTAP was introduced in 2001 to provide a temporary payment in addition to the applicable DRG for inpatient services where qualifying new technologies are used and are associated with clinical net benefit. 4 The temporary window for NTAP allows CMS to collect cost information to set future DRG payment rates. Initially, the NTAP amount was no greater than 50% of the price of the new technology or total amount of the case above the existing DRG rate. In 2019 this rate was increased to 65%. 4 The overall premise for this rate changes is that future DRG payments likely will reflect the full-scale adoption of the technology in clinical practice.
Technologies must meet three criteria to be considered for NTAP. The details of each requirement are summarized in the Code of Federal Regulations, 42 CFR 412.87.10,11 Manufacturers and developers of disruptive technologies intending to seek NTAP should appropriately align their application with their Food and Drug Administration (FDA) submission to take full advantage of the time window, within 2–3 years of market introduction, where the NTAP will apply. In brief, the technology must encompass:
-
(1)
Novelty: The new technology is not “substantially similar” to previous or existing technologies.
-
(2)
Minimum cost threshold: The new technology must incur sufficient additional capital or operational expense to warrant additional payment. However, the cost of different AI technologies is often not publicly available. Additionally, individual insurance plans consider different elements in the quality-adjusted life-year metric used to determine minimum cost threshold, and may have their own variable thresholds as well. 12 In the case of stroke triage AI systems, the expense is typically well-defined up front for the individual hospital, as is the quality benefit from the standpoint of superiority (below); hence a threshold for net savings is clearly defined, though for individual hospitals with dynamic case mixes, this calculation may become complicated.
-
(3)
Superiority: The AI system must provide a substantial clinical improvement over existing options. Substantial clinical improvement is a loosely defined term; however, it has been a historically rigorous target where many applications can fall short. The best practice is to provide published evidence of superiority in accepted clinical outcome measures over an accepted current treatment paradigm. For example, recent literature has illustrated how the time savings afforded by AI system in triage stroke on a granular level of seconds to treatment can translate to hours of quality life saved. 13
It is noteworthy that as of fiscal year 2020, technologies that receive an FDA Breakthrough Device Designation are automatically deemed to have met requirements (1) and (3) for NTAP. Additionally, requirements (1) and (2) are similar to the criteria that United States Patent and Trademark Office patent applications must meet. Therefore, technologies that are successfully patented are more likely to also meet two of the three criteria for NTAP reimbursement.
Other reimbursement mechanisms for hardware devices are in development and refinement. The Medicare Coverage of Innovative Technology (MCIT) pathway was developed to provide national Medicare coverage for breakthrough devices, certain services related to the devices, and any reasonable and necessary treatments due to complications from the devices. 14 The coverage was projected to last for 4 years from as early as the date of FDA market authorization (which is necessary for use in the US market, but is not tied to reimbursement). However, since introduction of the MCIT in January 2021, CMS delayed its implementation and then discontinued it under the Biden administration in the fall of 2021. After reevaluation of the rule-making process, it was decided there was mismatch between Breakthrough Device Designation and coverage requirements for the Medicare population, and cost to CMS for technologies with yet-to-be-proven clinical or financial benefits. 14 Yet, the NTAP and other programs, such as a future “MCIT replacement,” offer substantial advancements in encouraging the use of disruptive technologies and they require increased focus on companies’ regulatory and reimbursement strategy to allow their qualification for the program. Finally, this process is most relevant for the United States approval process, and globally the steps vary, as discussed later.
Why NTAP and stroke AI?
Novel technology approval for the NTAP is increasing in recent years, and usage is sustained following the NTAP period. 15 Technology to improve treatment of large vessel occlusion (LVO) stroke is similarly increasing in everyday use, yet until recent, they were not eligible for reimbursement. Treatment of LVO strokes by thrombectomy substantially reduces morbidity and mortality, is highly time-dependent, and a cost-effective therapy.2,13,16 Treating patients with LVO strokes requires coordination of clinical resources across a large geographic area. Many patients present with stroke symptoms at community hospitals and urgent care clinics; yet there are few, if any, local facilities capable of effectively performing thrombectomies due to the sophisticated resources required. Thus, technologies to rapidly identify patients and direct them to larger facilities that can provide the appropriate levels of care avoids unnecessary use of costly transfers and leads to better clinical outcomes. Such a solution optimizes the use of healthcare resources, producing superior clinical outcomes for the stroke population and financial outcomes for government and private payers.
As an example, in triage of patients with acute stroke at a mid-size regional hospital, there may be two to four cases per week that require transfer for thrombectomy. For these patients to be eligible for NTAP reimbursement, the hospital must evaluate the aforementioned barriers and costs to determine their breakeven patient volume, and then purchase the AI technology. Then, the NTAP would reduce the breakeven by paying for some costs of acquisition and maintenance in the first 2 years of operation.
The NTAP has the potential to mitigate adoption cost risk in a stroke scenario, but for proven technologies rather than emerging new technology. As with the recent NTAP-approved AI technologies, we note that their base AI system has been available for more than four years. Yet new “add-on” functions, with associated costs of continual development, could far exceed the initial cost of implementation. Hence the CMS and hospitals require data on the effectiveness of the technology. The data will need to show if the new system add-ons make a difference in rates of adoption and clinical outcomes. Analysis of impact is required; particularly as recent studies have raised concern about some NTAP technologies that may not have provided the anticipated clinical benefit. 17
The first approved NTAP for stroke triage AI: The specific strategies employed
A stroke AI system company was granted the first NTAP effective for Medicare discharges occurring between 1 October 2020 and 30 September 2021. 3 It used software that automatically stratifies patients with and without a LVO stroke, then triggers automated alerts and graphical data to the treating physicians intended to accelerate the identification of such patients, reducing the number of phone calls and wait times to assess images and clinical data. To our knowledge, this is the first AI software to be granted this form of additional payment. Hospitals were set to receive incremental payment when the technology was used in connection with hospital discharges occurring between 1 October 2020 and 30 September 2021. 16 To date, several other AI systems are now also granted NTAP status for performing similar functions. In effect, under NTAP, hospitals have a 12-month period when employing this stroke AI technology to be reimbursed up to US$1040 per patient. For the NTAP reimbursement to be approved, specific conditions must be met, as described below.
Elements of the first approved AI system are applicable to other developers within the evolving AI landscape. All AI developers aiming to pursue CMS payment opportunities should be familiar with the details of the CMS’ review and approval process. This process, encompassing the three criteria above, is broadly applicable to most AI systems, and involves preparation of a comprehensive application. The steps of and the components for this application detail: the technology itself, how it meets all three NTAP criteria, its FDA approval and clearance status, the procedure, DRG codes associated with the technology, charge per use and volume of cases, and reporting of adverse events. 18 Importantly, the CMS clearly describes the application process and is approachable to applicants who have demonstrated their AI will reduce cost. Beyond these minimum criteria, strategies can be used to construct the application. Below, are two elements specific to the approved AI stroke system that other AI developers must pay attention to:
First, CMS initially stated concerns as to whether the use of AI, an algorithm or software, which is intangible, may be considered a unique mechanism of action. After additional discussions, CMS agreed that the AI software does not use the same or a similar mechanism of action to achieve a therapeutic outcome when compared to existing treatments because there are currently no FDA approved or cleared technologies that use computer-assisted triage and notification to rapidly detect a LVO and shorten time to physician notification. The final determination from the CMS describes the comments received and additional analyses which addressed these concerns, with a specific focus on the mechanism of action.
Second, the substantial clinical improvement determination from the CMS centered on the evidence showing the AI software reduced time to notification and treatment. CMS appeared to rely heavily on a substantial literature base and medical society recommendations where it is unambiguously understood that reducing the time to treatment of LVO strokes leads to better outcomes. Although the company provided data on performance of their AI, as well as early results from a study showing improved outcomes, the crux of the argument appeared to be that increased time efficiency is known to correlate with improved clinical outcomes. The key takeaway for AI developers is that evidence must be presented of efficiencies gained that must be: (1) demonstrated through rigorous data collection, and (2) are widely understood to improve clinical outcomes.
Therefore, it can be concluded that a major element of the NTAP approval in this case was for a form of robotic process automation. 19 In this case, robotic process automation involved packaging parts of imaging interpretation, image transfer and physician notification steps into an automated process that is triggered by the algorithmic interpretation of an imaging study which accelerates decision making. Accordingly, the sensitivity and specificity of the imaging interpretation is less important than the process acceleration. Additionally, robotic process automation advised systems may prove to reduce operator variability, which will affect process quality. 20
Strategies for future stroke AI platforms
Stroke technology companies that aim to use AI to improve stroke care should evaluate the opportunity provided by NTAP to assess how their clinical and regulatory strategy can best support an application. We recommend that distinct benefits of using AI technology are framed in one of two ways. First, as improving outcomes and overall cost of care. Second, in cases where cost of care may increase, this has to be offset by substantial increases in operational efficiency. 21 In the second case, the resource-based relative value scale may require adjustment, since the effect of the AI offsets the decrease in effort. In addition to the three criteria specified by CMS in the NTAP regulations (i.e. novelty, cost, and clinical superiority), to accelerate the effective implementation of their technologies we suggest that companies demonstrate to healthcare providers that their AI produces outputs that are:
-
(1)
Fiscally quantifiable: The technology must address a process or intervention that, when implemented, results in quantifiable cost benefits. To the extent practical, the company should estimate the amount of financial benefit that accrues to each relevant stakeholder (e.g. payer, performing provider, and downstream provider or other beneficiary). 18 The technology should identify a breakeven usage volume and then illustrate how the NTAP substantially reduces this volume, or the initial investment, to allow adoption of the technology.
-
(2)
Explainable: A human physician should be able to understand the output classification and the basis for the software program’s weighing of classifiers to make the clinical recommendation. In the first approved AI system, the output data includes imaging that can be assessed visually by the physician, with the blocked artery forming the basis for making a clinical decision. As a counter example, another machine learning algorithm may suggest initiation of antibiotics in a patient in the intensive care unit without identifying the current clinical parameters that support the recommendation. Such a recommendation is less likely to be accepted by the physician as they would be accountable for a decision made without clinical insight.
-
(3)
Actionable: A therapeutic decision is predicated on the output data. In the first approved AI software case, the software output demonstrates an LVO that can be treated. Level 1 A evidence supports the use of thrombectomy to treat patients with LVO stroke, and the recommendation can prompt therapy. 22 In contrast, a piece of software that automatically outlines the skull size may be interesting, but presently would not influence any therapeutic decision; this type of software would unlikely find a ready market.
-
(4)
Impactful: The decisions supported by the software needs to have a demonstrable clinical impact in practice and in the peer reviewed literature. In the case of the first approved AI software, the proposed benefit is acceleration of time to treatment, which is a powerful predictor of ischemic stroke outcomes. 23 CMS agreed that the automated identification of these strokes and acceleration of physician notification leads to potentially substantial clinical benefit. In contrast, software that identifies imaging features of a tumor that indicates a probable genetic mutation for which there is no meaningful change in therapies offered would not be a clinically actionable insight. On the other hand, tumor cell mutations that differentiate effective treatment strategies, including radiosensitivity or chemotherapeutic sensitivity would all be impactful versions of such a class of software.
-
(5)
Globally Applicable. Finally, technology incorporating successful safety and use testing in both the United States and internationally will be best positioned for eligibility and similar reimbursement outside of the US. Since stroke is the leading cause of disability both in the US and globally, such AI algorithms may actually be more impactful in other markets than the US. 24 In countries where the software is already used, a full understanding of the digital health and economic environment priorities is necessary to build the evidence that public payers require. Some countries do not have clear pathways for technology reimbursement, while others do. An example of the latter falls under the Digital Health Act (Digitale–Versorgung–Gesetz or DVG) in Germany, in which technology may be included in the reimbursement catalog by satisfying safety, privacy, and efficacy benchmarks. 25 Overall, reimbursement schemes are country-specific, without unified criteria for AI-based medical software.
Ultimately, it is the quantification of impact of such software that allows developers to address benchmarks for balancing costs and outcomes.
Importance of payment durability
While the initial NTAP payment for the adoption of stroke triage system is an encouraging first example, providers and new technology developers must continue to validate the clinical benefits of the new tools. Similarly, the CMS and other payers should analyze and openly share clinical and financial outcomes data that result from the use of these tools. Collectively this effort should establish which programs have sustained clinical benefits for patients and sustained value for the healthcare system, where value is defined as outcomes divided by costs for a relevant class of patients. 26 These validation studies may be in the form of traditional medical research as well as using observational data to look at the impact in healthcare payments for a disease class. The ultimate goal is for providers and technology developers to collect sufficient data on reduced stroke rehabilitation payments or stroke related deaths, in the case of stroke AI software, to justify DRGs reimbursements by the CMS and insurance plans in geographies where there is widespread use.
NTAP financial benchmarks and limitations
Several limitations exist for NTAP in stroke AI and for other AI-based technologies. First, for stroke specifically, if the AI triage functions ideally, the patient will receive a successful thrombectomy treatment that translates into lower expense for the hospital than the DRG payment, which makes the case ineligible for NTAP. Yet the benefits for the healthcare system are immense: an annual 10% increase in a good reperfusion rate for all thrombectomy-treated patients in the United States is estimated to save $21 million and $37 million for the healthcare system and society, respectively.2,13 This savings is partially driven by prevention of time-delays in stroke triage, which avoids accumulation of disability-adjusted life-years. For patients who are not triaged successfully or whose thrombectomy is unsuccessful, they have worse outcomes, and both the patient and healthcare system may endure more costs, yet qualification for the NTAP is likely. Second, the NTAP will assist in covering expenses but may still leave hospitals financially short. Third, the new AI technology from time of validation to widespread hospital installation and integration may take years, which may not provide sufficient time to demonstrate benefit over the limited time the AI is eligible for the NTAP (e.g. two or up to 3 years). Additionally, the economic analysis for NTAP and subscription-based services may favor high-volume centers, which have sufficient patient volume and can purchase AI add-ons that qualify for the NTAP. Conversely, the NTAP may be advantageous to some smaller and less resourced centers because their identification of the LVO stroke adds value to the process and enables eligibility for NTAP reimbursement, which offsets their inability to perform high-margin LVO thrombectomy to offset AI software costs. Another limitation of NTAP is that it only applies to technology or services deployed in a hospital setting. Many of the approved or emerging AI and machine learning clinical tools are meant for use in an outpatient setting or unregulated spaces. For example, the first fully automated AI diagnostic system approved by the FDA in any medical field is a system that can screen for referable diabetic retinopathy in a primary care setting. 27 Likewise, there is currently no NTAP counterpart for outpatient based services to encourage adoption of innovative technologies. Finally, if the pace of AI continues to reach record rates of healthcare integration, potentially overwhelming the CMS NTAP process, there may be changes to the fundamental eligibility for an NTAP qualified case.
Conclusion
In 2020, an AI system for expediting triage of time-sensitive stroke was the first technology based on AI for which CMS awarded the NTAP. The acknowledgement by the CMS of the clinical benefit of this AI software required a unique mechanism of action and linking of efficiency to clinical outcomes. We have provided additional examples of specific strategies that may be employed by other clinical applications of AI to create impactful solutions that qualify for NTAP reimbursement. Finally, we believe that the NTAP supports an accelerating and exciting rate of change in stroke technology by mitigating the risks of adoption. To take advantage of this opportunity, technology developers, providers, and payers need to focus on the clinical impact and value added they bring to the healthcare system
Footnotes
Declaration of conflicting interests: The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding: The author(s) received no financial support for the research, authorship, and/or publication of this article.
ORCID iD
Nick M. Murray https://orcid.org/0000-0003-3861-0958
References
- 1.Chen MM, Golding LP, Nicola GN. Who will pay for AI? Radiol Artif Intell 2021; 3: e210030. DOI: 10.1148/ryai.2021210030. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kunz WG, Hunink MG, Almekhlafi MA, et al. Public health and cost consequences of time delays to thrombectomy for acute ischemic stroke. Neurology 2020; 95: e2465–e2475. DOI: 10.1212/WNL.0000000000010867. [DOI] [PubMed] [Google Scholar]
- 3.Centers for Medicare & Medicaid Services (CMS) HHS . Medicare program; hospital inpatient prospective payment systems for acute care hospitals and the long-term care hospital prospective payment system and final policy changes and fiscal year 2021 rates; quality reporting and medicare and medicaid promoting interoperability programs requirements for eligible hospitals and critical access hospitals. In: Service CfMM, (ed.). CMS-1735-F. Washington, DC: Federal Register; 2020, pp.58432–59107. [Google Scholar]
- 4.Clyde AT, Bockstedt L, Farkas JA, et al. Experience with medicare’s new technology add-on payment program. Health Aff (Millwood) 2008; 27: 1632–1641. DOI: 10.1377/hlthaff.27.6.1632. [DOI] [PubMed] [Google Scholar]
- 5.Manz CR, Bekelman JE, Doshi JA. The changing characteristics of technologies covered by medicare’s new technology add-on payment program. JAMA Netw Open 2020; 3: e2012569. DOI: 10.1001/jamanetworkopen.2020.12569. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Yu KH, Beam AL, Kohane IS. Artificial intelligence in healthcare. Nat Biomed Eng 2018; 2: 719–731. DOI: 10.1038/s41551-018-0305-z. [DOI] [PubMed] [Google Scholar]
- 7.Murray NM, Unberath M, Hager GD, et al. Artificial intelligence to diagnose ischemic stroke and identify large vessel occlusions: a systematic review. J Neurointerv Surg 2020; 12: 156–164. DOI: 10.1136/neurintsurg-2019-015135. [DOI] [PubMed] [Google Scholar]
- 8.Algra PR, Morozov S, Ranschaert ER. Artificial Intelligence in Medical Imaging: Opportunities, Applications and Risks. 1st ed. London, UK: Springer International Publishing: Imprint: Springer, 2019, p. 1. online resource (XV, 373 pages 104 illustrations, 381 illustrations in color. [Google Scholar]
- 9.Maddox TM, Rumsfeld JS, Payne PRO. Questions for artificial intelligence in health care. JAMA 2019; 321: 31–32. DOI: 10.1001/jama.2018.18932. [DOI] [PubMed] [Google Scholar]
- 10.Administration SS . Payment to Hospitals for Inpatient Hospital Services. Section 1886(d)(5)(K). Woodlawn, MD: SSA, 2020. [Google Scholar]
- 11.Regulations CoF . Title 42 - Public Health, Chapter Iv - Centers for Medicare & Medicaid Services, Department of Health and Human Services, Subchapter B - Medicare Program, Part 412 - Prospective Payment Systems for Inpatient Hospital Services, Subpart F - Payments for Outlier Cases, Special Treatment Payment for New Technology, and Payment Adjustment for Certain Replaced Devices, Subjgroup - Additional Special Payment for Certain New Technology, Section 412.87 - Additional Payment for New Medical Services And Technologies: General Provisions. Hague, Netherlands: International criminal court, 2010. [Google Scholar]
- 12.Danzon PM, Drummond MF, Towse A, et al. Objectives, budgets, thresholds, and opportunity costs-a health economics approach: an ispor special task force report [4]. Value in Health 2018; 21: 140–145. [DOI] [PubMed] [Google Scholar]
- 13.Aronsson M, Persson J, Blomstrand C, et al. Cost-effectiveness of endovascular thrombectomy in patients with acute ischemic stroke. Neurology 2016; 86: 1053–1059. DOI: 10.1212/WNL.0000000000002439. [DOI] [PubMed] [Google Scholar]
- 14.Services CfMM . Medicare Program; Medicare Coverage of Innovative Technology and Definition of Reasonable and Necessary. Woodlawn, MD: CMS; 2021. [Google Scholar]
- 15.Morey JR, Katana M, Li L, et al. Adoption and trends in the medicare new technology add-on payment program. J Gen Intern Med 2021; 36: 2174–2176. DOI: 10.1007/s11606-020-05930-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hassan AE. New technology add-on payment (NTAP) for Viz LVO: a win for stroke care. J Neurointerv Surg 2021; 13: 406–408. DOI: 10.1136/neurintsurg-2020-016897. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Judson TJ, Dhruva SS, Redberg RF. Evaluation of technologies approved for supplemental payments in the United States. BMJ 2019; 365: l2190. DOI: 10.1136/bmj.l2190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Centers for Medicare & Medicaid Services (CMS) . New Medical Services and New Technologies. Woodlawn, MD: CMS, 2021, https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/newtech [Google Scholar]
- 19.van der Aalst WMP, Bichler M., Heinzl A. Robotic process automation. Business & Information Systems Engineering 2018; 60: 269–272. [Google Scholar]
- 20.Murrell K. Operator variability and its industrial consequences. International Journal of Production Research 1961; 1: 39–55. [Google Scholar]
- 21.Letourneau-Guillon L, Camirand D, Guilbert F, et al. Artificial intelligence applications for workflow, process optimization and predictive analytics. Neuroimaging Clin N Am 2020; 30: e1–e15. DOI: 10.1016/j.nic.2020.08.008. [DOI] [PubMed] [Google Scholar]
- 22.Powers WJ, Rabinstein AA, Ackerson T, et al. 2018 Guidelines for the early management of patients with acute ischemic stroke: a guideline for healthcare professionals from the american heart association/american stroke association. Stroke 2018; 49: e46–e110. DOI: 10.1161/STR.0000000000000158. [DOI] [PubMed] [Google Scholar]
- 23.Goyal M, Jadhav AP, Bonafe A, et al. Analysis of workflow and time to treatment and the effects on outcome in endovascular treatment of acute ischemic stroke: results from the swift prime randomized controlled trial. Radiology 2016; 279: 888–897. DOI: 10.1148/radiol.2016160204. [DOI] [PubMed] [Google Scholar]
- 24.Collaborators GBDN. Global, regional, and national burden of neurological disorders, 1990-2016: a systematic analysis for the global burden of disease study 2016. Lancet Neurol 2019; 18: 459–480. DOI: 10.1016/S1474-4422(18)30499-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Gerke S, Stern AD, Minssen T. Germany’s digital health reforms in the COVID-19 era: lessons and opportunities for other countries. NPJ Digit Med 2020; 3: 94. DOI: 10.1038/s41746-020-0306-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Bindra P. The Core Elements of Value in Healthcare. Chicago, IL: Health Administration Press, 2018. [Google Scholar]
- 27.Abramoff MD, Lavin PT, Birch M, et al. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. NPJ Digit Med 2018; 1: 39. DOI: 10.1038/s41746-018-0040-6. [DOI] [PMC free article] [PubMed] [Google Scholar]