Considerations for the Adoption of Digital Algorithms and Cardiovascular Decision-Support Tools in Clinical Practice

Abstract

Early detection of cardiovascular disease and implementation of evidence-based treatments can reduce cardiovascular morbidity and mortality. Medical algorithms and decision-making tools provide a compelling option for screening, risk prediction, and treatment management. Such digital tools have the potential to aid both healthcare professionals and patients, providing support to decrease unwarranted diagnostic and treatment variability while guiding personalized care, with the overall objective of improving clinical outcomes. However, incorporating digital tools in healthcare settings is challenging, and evidence the required to support their adoption and understand the limitations can be lacking. A multinational panel of expert cardiologists and emergency physicians across North America, Europe, and Oceania gathered to deliberate on the current landscape of digital tools and medical algorithms, drawing on prior clinical experiences and knowledge of country-specific regulations. In this viewpoint, the evidence to support and guide the adoption of digital tools in cardiovascular clinical practice and the necessary components for successful integration into clinical workflows were discussed. Digital tools must be developed with the needs of the healthcare professionals, other relevant stakeholders (eg, administration personnel), and patients in mind to give them the best chance of widespread adoption. Academia, industry, and regulatory bodies should work together to cultivate and accelerate the implementation of digital tools in healthcare. The considerations discussed here may help decision makers to determine if a digital tool has the components necessary to integrate into the clinical workflow successfully.

INTRODUCTION

Cardiovascular disease (CVD) is the leading cause of death globally.¹ Early detection of CVD and identification of patients at high risk has the potential to significantly reduce mortality and morbidity.² Medical algorithms and other digital tools have the potential to assist healthcare professionals (HCPs) with tasks such as diagnosis, treatment and monitoring, and discharge decisions, which may ease work burdens by supporting clinical judgement.³ Medical algorithms are digitalized, automated, mathematical formulae that gather information from various sources, such as an electronic medical record (EMR), wearable devices, or a laboratory information system, to provide data or a recommendation back to the end user.⁴ The output data of these digital tools are standardized and can be used to promote consistent levels of care within and between institutions.⁵ Algorithms could assist daily HCP decision making for patient care, including in settings with limited resources or periods outside of normal operating hours, or other scenarios where this support may be needed.^6–8

In this viewpoint, we discuss the level of evidence required to support and guide the adoption of digital tools in cardiovascular clinical practice and the necessary components for successful integration.

DISCUSSION

Algorithms to Support Cardiovascular Medicine

Current Use of Digital Tools in Cardiovascular Medicine

There is currently a wide variety of uses for digital tools in cardiovascular medicine. For instance, there are digital tools that can potentially improve the accuracy of implantable cardiac monitors for detecting arrhythmias. In 2021, the US Food and Drug Administration (FDA) approved AccuRhythm AI by Medtronic, an algorithm capable of reducing false atrial fibrillation alerts from implantable cardiac monitors by 74.1% while maintaining true alerts at 99.3%, thereby decreasing the burden of clinical review.⁹ Digital tools may also provide decision support to emergency and acute care physicians in the risk assessment of patients who present with suspected acute coronary syndrome (ACS). The metrics of calculating risk in suspected ACS are increasingly complicated, with the adoption of high-sensitivity cardiac troponin (hs-cTn) assays that require accurate timing and often depend on interpreting change values from repeat testing. Artificial intelligence (AI) models in several studies have been able to facilitate early recognition of ACS using retrospective EMR data; however, these tools require further validation.^10,11 Algorithms capable of accurate risk assessment for suspected ACS could help in the early recognition of high-risk patients requiring treatment and could support decisions to discharge appropriate low-risk patients safely.

Several approaches with digital tools have been introduced to help support personalized care. For instance, the ARTEMIS algorithm, by ACS Pathfinder, was designed to predict the individual probability of myocardial infarction (MI) using hs-cTn measurements.¹² Validated results found that, using a single hs-cTn measurement, the ARTEMIS algorithm allowed direct rule-out of MI with high and similar safety but up to triple efficiency compared with the guideline-recommended strategy.¹² In a retrospective study using a single point-of-care hs-cTn measurement, the ARTEMIS algorithm identified more than twice as many patients as eligible for direct rule-out of MI (35.1%) compared with guideline-recommended European Society of Cardiology (ESC) 0-hour (15.2%) and American College of Cardiology 0-hour (13.8%) pathways.¹³ Cardio Explorer, by Exploris Health, another algorithm based on an AI model, aims to provide personalized predictions of the presence of coronary artery disease (CAD) using a combination of the most significant predictors of CAD. This algorithm has been shown to predict the risk of CAD more effectively than traditional risk scores.^14,15

The Collaboration for the Diagnosis and Evaluation of Acute Coronary Syndrome (CoDE-ACS) and Heart Failure (CoDE-HF) decision support tools, by the University of Edinburgh, are further examples of validated machine learning models developed to assist the diagnosis of MI and acute heart failure, respectively.^16,17 Both tools combine clinical variables associated with heart failure, such as age, heart rate, blood pressure, and renal function, with measurements of hs-cTn (CoDE-ACS) and N-terminal pro B-type natriuretic peptide (CoDE-HF) to determine the probability of CVD. The models discriminate well for MI [CoDE-ACS, area under the curve (AUC) 0.953] and heart failure (CoDE-HF, previous heart failure AUC 0.846, no previous heart failure AUC 0.925). CoDE-ACS provides low and high probabilities of MI to create a pathway that is optimal for patient flow, according to local clinical priorities.¹⁶ For instance, in settings where emergency department capacity is limited, a lower CoDE-ACS score to identify those with a high probability of MI could be applied to reduce the proportion of patients who require observation and serial testing within the department. CoDE-HF can also be embedded within the clinical workflow as part of the triage pathway in the emergency department to facilitate more efficient assessment.¹⁷ The machine-learning algorithm myocardial-ischemic-injury-index (MI³) has also been developed to provide an individualized, objective assessment of the likelihood of MI using hs-cTn values and the age and sex of the patient.¹⁸ MI³ was able to discriminate MI with an AUC of 0.963. Given that fewer factors are used in the MI³ model, it has the potential to avoid inconsistent assessment of symptoms or patient history.¹⁸ These tools have an advantage over prior approaches that used fixed hs-cTn thresholds or risk scores alone by providing flexible, personalized decision support. However, knowledge of the extent to which the implementation of these decision support tools improves healthcare resource utilization and/or patients’ outcomes is currently limited, and further studies are needed to assess their impact on real-world practice.

Current Challenges With the Adoption of Digital Tools

Despite the potential benefits, adopting digital tools in a healthcare setting is challenging. The level of evidence required for their adoption can vary based on the function and operation,¹⁹ and there are limited data currently available to support the impact of medical algorithms on hard outcomes for efficacy and safety.²⁰ Limitations to the use of digital tools must also be considered. If tools are standalone and cannot effectively integrate into clinical workflows and existing systems such as EMRs, it is difficult for a tool’s recommendations to be documented and tracked; thus, HCPs will be less likely to use them. Before implementation, digital tools must demonstrate clinical value and align with regulatory guidelines, such as those applied by the FDA or the European Medicines Agency [including Medical Device Regulation (MDR) guidelines]. They must also allow customization to meet the needs of the community being served and to ensure that any recommendations generated are appropriate for the resources available within the setting in which they are being used. This customization is particularly important for rural and remote areas with limited resources. HCPs may also hesitate to adopt a digital tool that increases the work burden, even if it might improve patient care.

Considerations for the Adoption of Digital Tools

To develop the considerations for the adoption of digital tools, a multinational panel of expert cardiologists and emergency physicians across North America, Europe, and Oceania gathered to deliberate on the current landscape of digital tools and medical algorithms, drawing on prior clinical experiences and knowledge of country-specific regulations. A premeeting survey highlighted 4 general topics for discussion: (i) development, (ii) assessment, (iii) evidence, and (iv) adoption of digital tools, with the questions outlined in Table 1. A face-to-face meeting was held to discuss and debate the feedback. A hypothetical case study was also considered to highlight the challenges that would impede the implementation, timelines, and scale of adoption of CVD digital tools in an emergency department (Table 1).

TABLE 1.

Discussion Topics for the Increased Adoption of Digital Tools in Healthcare

Topic	Discussion Question
Development	How can academia and industry better share data for research and commercial use to cultivate and accelerate the adoption of digital tools in healthcare?
Assessment	What are the expected benefits or key metrics to be improved by integrating digital algorithms and decision-support tools into clinical practice?
Assessment	What are the potential risks and limitations of implementing digital algorithms and decision-support tools?
Assessment	With respect to the clinical/hospital ecosystem, what are the relevant interconnected and downstream impacts to expect from digital algorithms and decision-support tools?
Evidence	Which type and level of evidence is needed to convince stakeholders (eg, clinicians, administrators, and information systems) and key decision-makers to adopt digital algorithms and decision-support tools? Which outcomes are most compelling?
Evidence	What have you learned from prior experiences with digital algorithms and decision-support tools in your clinical practice?
Evidence	Are FDA/CE-certified clinical decision support tools preferred over nonregulated ones for adoption?
Adoption	What are the main challenges or barriers hindering the adoption of digital algorithms and decision-support tools?
Adoption	Which factors influence front-line staff acceptance and willingness to adopt digital algorithms and decision-support tools?
Adoption	How is funding for digital algorithms and decision-support tools handled if insurers do not reimburse them through fee-for-service and they would need to be paid out of the hospital budget? If the hospital pays, which budget supports this expense, and who controls that budget?

CE indicates Conformité Européene; FDA, Food and Drug Administration.

Type and Level of Evidence Needed to Facilitate the Adoption of Digital Tools

Evidence to facilitate the adoption of digital tools should provide a comprehensive scientific foundation and reasoning for their use and demonstrate their clinical value with measures that are relevant to key stakeholders and decision makers. Figure 1 summarizes the key considerations for the adoption of digital tools. The perspectives of HCPs, administrators, regulators, information system technicians, and patients need to be considered before the implementation of digital tools.

Figure 1.

Key considerations for the adoption of digital tools. GDPR indicates General Data Protection Regulation.

Healthcare Professionals

For many HCPs, evidence that implementing digital tools results in improved outcomes in line with the clinical gold standard and a more favorable benefit-to-risk ratio is of primary importance. Data must demonstrate the accuracy of diagnostic and prognostic tools and an improvement in the implementation of evidence-based therapies and processes. A direct clinical and/or cost-effectiveness improvement must also occur, such as a reduction in unnecessary tests or hospitalizations, or a reduction in the length of hospital stay. As an example from CVD care, a study by Attia et al²¹ confirmed that using an AI-enabled electrocardiogram (ECG) could identify silent atrial fibrillation in patients during normal sinus rhythm with an overall accuracy of 79.4%. It is important to note that, although the study demonstrated predictive performance, additional validations from prospective trials are needed before such AI-enabled ECG tools can be adopted in practice. Importantly, these trials need to demonstrate clinical value in terms of improving patient outcomes.

Some trials have investigated the use of AI in clinical practice and yielded positive results, such as AiCure, which demonstrated a 50% increase in treatment adherence.²² However, although such studies are a key source of evidence, many studies involve only a single center, with limited reporting of patient characteristics; this can raise concerns about the generalizability of the observed results.²³ Robust large-scale clustered trials should be used to help promote inclusive enrollment, reduce contamination between the intervention and control arms and provide a more representative assessment of the effect of adopting digital tools on clinical outcomes. However, such large-scale, potentially costly, outcome trials may not be required for algorithms designed to enhance the use of existing guideline-directed medical therapy. Rather, demonstration of increased utilization of evidence-based therapies could be sufficient to assess the clinical value of such algorithms. However, there should be evidence that the algorithm accurately reflects the guidelines and improves guideline adherence.

The design of digital tools should allow for the model to be updated and retrained to prevent overpredicting or underpredicting risk when the standard of care changes, such as with new, improved guidelines, evidence, or medications.²⁴ Similarly, cardiovascular predictive models trained on a particular therapeutic strategy might be less accurate with the increased adoption of preventive therapies. In a study from the United Kingdom, the inclusion of a downward secular trend of CVD incidence into a risk prediction model substantially reduced risk predictions of CVD (35.7% risk reduction for females and 33.2% for males).²⁵ The risk of overfitting is significant when using machine learning to develop a model. Therefore, proper validation in appropriate studies is key to guarantee that the model is accurate.

Iterative assessments of evolving risk must also be considered. For instance, the criteria for recommending the cessation of further tests and discharging a patient at low risk of acute MI should differ from the criteria for a patient with a high likelihood of acute MI. An example in practice was observed by Ashburn et al,²⁶ where caution was required from HCPs as the European Society of Cardiology 0/1-hour hs-cTnT algorithm could not safely predict 30-day cardiac death or MI in patients with known CAD. Different HCPs may also prioritize different features of a digital tool based on their individual preferences and requirements. Emergency physicians may wish to optimize the negative predictive value of a digital tool to exclude the need for further evaluation for acute MI. In contrast, a cardiologist may want to optimize the positive predictive value when considering invasive treatment options in this same population. With sufficiently large datasets and an adaptable model, AI may have the potential to cover both scenarios.

Evidence that a digital tool could ease HCPs’ work burdens may be compelling for its adoption. In the United States, the prevalence of burnout symptoms has been reported to be near or exceeding 50% in both training and practicing physicians.²⁷ Alert fatigue, where overwhelming alerts desensitize a user, is also an important consideration for HCPs with recognized patient safety concerns in intensive care settings.²⁸ Therefore, the design of these tools should prioritize streamlined workflows with minimal user actions and documentation time, as well as support the reduction of alert fatigue.²⁷ The ability to seamlessly integrate the tools into the current workflows while avoiding redundant input of data into the EMR will likely encourage adoption by HCPs. Tools need to be intuitive and designed with the end user in mind. They should demonstrate that they can reliably support HCPs while allowing them to maintain control of the treatment pathway and clinical decision-making. As such, careful attention should be given to the user interface design and experience.

Administrators

For administrators, proof that medical algorithms deliver a quantifiable return on investment may be convincing evidence for their adoption. AI tools have been shown to potentially save costs in treatment and diagnosis compared with traditional methods.²⁹ As an example, implementing AI diagnostic algorithms alongside confirmatory high-specificity polymerase chain reaction for colorectal cancer in the United States has been estimated to save $400 million compared with next-generation sequencing alone.³⁰ A decision support tool that improves the identification of patients at low risk of ACS who could be safely discharged while enabling admission of patients at higher risk could be beneficial in terms of both cost-effectiveness and patient safety. Tools that can demonstrate significant economic value compared to the current standard may be more likely to be adopted.

Reimbursement schemes, such as those seen in remote patient monitoring,³¹ would also be desirable for administrators. Initiatives for pay-by-performance and reimbursement linked to guideline adherence and quality would be expected to encourage the wider implementation of digital tools in clinical practice. Intuitive tools that require minimal training for use would be beneficial. Even without a direct financial cost, internal stakeholder resources required for integration and maintenance may be redirected from other projects, representing an opportunity cost. For this purpose, a digital tool that could demonstrate guideline and/or protocol adherence to standard operating procedures could be beneficial to promote regular audits.

To encourage adoption, digital tools must ensure better resource stewardship, efficiency, and quality metrics. This could include accelerated processes or a reduction in unnecessary testing. It will also be essential to demonstrate that users frequently comply with the tool recommendations and accept their use. Indeed, when HCPs lack confidence in the accuracy of medical algorithms, it can lead to unnecessary additional tests, which decreases efficiency in clinical practice.³²

Regulators

For regulators to support the adoption of medical algorithms, evidence that the tool conforms with region-specific regulations is also required; regulations vary by region, and heterogeneity exists concerning standard operating procedures for evaluation. For example, the FDA’s traditional medical device guidelines in the United States were not designed for adaptive AI and machine learning technologies.³³ Predetermined change control plans that outline anticipated updates to a device or software must be used to gain approval for AI or machine learning tools that undergo postmarket modifications. The Medicines and Healthcare Products Regulatory Agency in the United Kingdom is investigating how to update current regulations for continuous learning AI medical tools and ensure performance is maintained.³⁴ In Europe, medical AI tools must align with the AI Act and MDR guidelines. If the tool operates with a degree of autonomy or continuous learning, it may not fully meet the MDR’s current requirements.³⁵

Data privacy compliance and easy integration into existing data privacy systems are required for digital tools. Appropriate safeguards should be in place to maintain privacy and patient agency to reduce the risk of data breaches.³⁶ Hospitals may incur significant costs to ensure they comply with the General Data Protection Regulation in Europe, and fines may be administered if General Data Protection Regulation violations occur. Data privacy is of increased importance with digital tools that utilize large datasets. A 2018 study found that an algorithm could reidentify 85.6% of adults and 69.8% of children in a physical activity cohort study, even though protected patient health information identifiers had been removed.³⁷

Information Systems

To facilitate adoption, evidence that any digital tool can interface and function with, or can be integrated into, an electronic platform such as an EMR may need to be demonstrated, with minimal licensing or adjustments required. They should also be suitably assessed for robustness before implementation, ensuring minimal to no software bugs when live. Interfacing with EMRs would be expected to provide a wealth of patient data for a digital tool to assess and predict risk. For example, Kwon et al³⁸ developed an AI algorithm trained on EMR data to predict cardiac arrest using ECGs. The tool could predict cardiac arrest within 24 hours of hospital admission (AUC 0.913 and 0.948, internal and external validation, respectively) using both 12-lead and single-lead ECG. Such results are only possible through a well-integrated toolset with ready access to EMR data. Beyond this, tools must be scalable and adaptable for use in settings such as emergency departments or clinics. This may be difficult for traditional predictive modeling techniques, as each outcome to be predicted requires the creation of a custom dataset with specific variables.³⁹ However, this challenge must be addressed when implementing a digital tool, as algorithms tied to a certain dataset may not apply to healthcare settings that lack access to these data.⁴⁰

Patient Perspective

Alongside key decision makers, it is necessary to consider the patient perspective when adopting digital tools. Such tools, which could be tailored to the individual, can potentially improve the patient experience by allowing them to be more engaged in their healthcare.⁴¹ Indeed, studies have demonstrated that personalization leads to better patient engagement and outcomes.⁴² A previous study also found that patients welcomed AI use in the care process; treatment recommendations from physicians using AI led to a higher intention to follow the recommendation than recommendations from AI on its own.⁴³ Informative tools that present the logic behind recommendations may benefit patients and HCPs and allow for shared decision-making. Patient surveys to assess treatment satisfaction before and after implementation of the tool could help ensure patient needs are met.⁴⁴ Digital tools and devices that offer disease monitoring and information sharing directly with a patient’s healthcare provider may also encourage their adoption. A review of patient perspectives of digital health tools noted that patient empowerment, self-management, and personalization were key facilitators to their use.⁴⁵ Collaboration with patients during the design, study, and implementation process of digital tools should be considered whenever appropriate.

Key Metrics to Demonstrate Value

The benefits of implementing a digital tool for cardiovascular care should be demonstrated through key metrics. These metrics might include improved quality, effectiveness, and safety outcomes, increased adherence to treatment guidelines, provider and patient satisfaction, timeliness of care (including length of stay in emergency departments or chest pain clinics), adequate use of high-cost diagnostics and therapies, and admission rates. These metrics should be monitored before and after integration to illustrate a tool’s benefits. However, metrics should be tool-specific, relevant to the intended use, and strong enough to demonstrate value. For instance, the frequency of out-of-hospital visits could be considered a less informative metric, as it can depend on the healthcare system and local practices. In contrast, a reduction in mortality rate would be more compelling. Tracking patient-focused outcome metrics, such as patient function and quality of life, may also provide evidence of value stretching beyond the clinical impact, demonstrating the longer-lasting benefits of digital tools.

Challenges or Barriers to Adoption

Despite the potential benefits of digital tools in CVD clinical outcomes, challenges such as alert fatigue and a high work burden can hinder their adoption. Key common challenges and barriers are summarized in Figure 2. High rates of alerts, with a large number being irrelevant or low priority, can desensitize HCPs. Indeed, studies in hospitals in the United States have shown high override rates of alerts of 77%–90%,⁴⁶ most commonly including drug interaction alerts.⁴⁷ Alerts are helpful when they provide additional context, such as the priority of the alert, decision support for further action, and the date of the last alert, without being too numerous. A study by Rayo et al⁴⁸ demonstrated that decision support using annotated visualizations rather than pop-up alerts resulted in significantly fewer inappropriate diagnostic imaging tests and was preferred for all patient scenarios.

Figure 2.

Key challenges and barriers to digital tool adoption. IT indicates information technology.

Implementing digital tools in healthcare may present data access challenges. Large costs to access EMR data from data aggregators, engagement from EMR vendors, and a lack of uniformity in the EMR data may hamper data access.⁴⁹ Large observational studies, collaboration among diagnostics, pharma, and device sectors, and the use of federated databases and data harmonization for healthcare improvement are needed. These approaches could ensure patient diversity in data sets to reduce avoidable bias and improve the generalizability of tools to patient groups traditionally underrepresented in research.

Implementation of Digital Tools

Once a digital tool has been determined to meet the necessary requirements for use, its implementation might proceed. Facilitating implementation would benefit from prospective clinical trials that evaluate the most effective ways in which a novel tool might be deployed and the context in which benefits from its use might be expected. Beyond this, developing and maintaining digital tools for decision support systems may be costly, time-consuming, and complex.⁵⁰ A local committee focusing on digital tool deployment should be convened to aid implementation. The committee may report to a medical executive committee or senior physician governing body that might approve digital tool interventions and oversee accountability and data access. Practices that could assist digital tool governance and implementation are summarized in Figure 3.

Figure 3.

Governance and implementation of digital tools.

CONCLUSIONS

In summary, medical algorithms and decision support tools can potentially improve clinical outcomes and provide greater accuracy and consistency of decision-making in CVD treatment. While digital tools could provide a more efficient workflow and improved patient outcomes, several considerations should be addressed before they can be implemented in the clinical setting. Digital tools must be developed with the needs of the HCPs, other relevant stakeholders (eg, administration personnel), and patients in mind to give them the best chance of widespread adoption. Academia, industry, and regulatory bodies should work together to cultivate and accelerate the implementation of digital tools in healthcare. The considerations discussed here may help decision makers to determine if a digital tool has the components necessary to integrate into the clinical workflow successfully.

ACKNOWLEDGMENTS

Editorial support for this manuscript, under the direction of the authors, was provided by Joe Ward, Sean Littlewood, and Fraser Harris of Springer Health+, Springer Healthcare Ltd, UK, and was funded by Roche Diagnostics International AG (Rotkreuz, Switzerland).

DISCLOSURES

This work was supported by Roche Diagnostics GmbH (Penzberg, Germany). C.W.B. is a paid speaker for CE Symmetry, Roche Diagnostics, and Octapharma, an investigator for Abbott Laboratories, Monaghan Medical, Roche Diagnostics, and Sanofi, an advisory board participant for AstraZeneca, Pfizer, Roche Diagnostics, SERB Pharmaceuticals, and Salix Pharmaceuticals, a consultant for Abbott Laboratories, Terumo, Monaghan Medical, Pfizer, Roche Diagnostics, and UltraSight, and an advisor to Lucia Health Guidelines, Quai MD, and Vera Health. C.M.G. has received consulting fees from Roche. E.G. received honoraria for lectures from AstraZeneca, Bayer Vital, Boehringer Ingelheim, Daiichi Sankyo, Lilly Eli Deutschland, Novartis, and Roche Diagnostics. He serves as a consultant for Boehringer Ingelheim, BRAHMS Thermo Fisher Scientific, Roche Diagnostics, and has received research funding from Bayer Vital, BRAHMS Thermo Fisher Scientific, Daiichi Sankyo, and Roche Diagnostics. D.A.M. is a member of the TIMI (Thrombolysis in Myocardial Infarction) Study Group, which has received institutional research grant support through Brigham and Women’s Hospital from Abbott, Abiomed, Amgen, Anthos Therapeutics, AstraZeneca, Bayer HealthCare Pharmaceuticals, Inc., Daiichi Sankyo, Eisai, 4TEEN4, Intarcia, MedImmune, Merck, Novartis, Pfizer, Regeneron Pharmaceuticals, Roche, Siemens Healthcare Diagnostics, and Zora Biosciences. D.A.M. reports consulting fees from Abbott Laboratories, Merck & Co, Novartis, Regeneron, and Roche Diagnostics. J.L.J. reports equity holdings in Fibrosys, Imbria Pharma, and Jana Care, current/recent grant support from Abbott, Applied Therapeutics, AstraZeneca, BMS, Novartis Pharmaceuticals, consulting fees from Abbott Diagnostics, Beckman Coulter, Jana Care, Janssen, Novartis, Prevencio, Quidel, and Roche Diagnostics and serves on clinical endpoint committees/data safety monitoring boards for Abbott, AbbVie, Amgen, CVRx, Medtronic, Pfizer, and Roche Diagnostics. C.P. has received research grants, consulting fees and honoraria from Roche Diagnostics, and consulting fees and honoraria from Abbott. H.-P.B.-L.R. has received unrestricted research grants from Roche Diagnostics and consulting fees and/or lecture fees from AstraZeneca, Boehringer-Ingelheim, Novartis, Roche Diagnostics and Vifor, and payment for expert testimony by Novartis. L.B.D. has received grant support from RCE Technologies and Roche Diagnostics; serves on clinical endpoint committees for Abbott, Pfizer, QuidelOrtho, and Tosoh, and has received consulting fees from Abbott, QuidelOrtho, Roche Diagnostics, and Siemens. M.T. has received research and education grants from Abbott, Alere, Beckman Coulter, QuidelOrtho, Radiometer, Roche, and Siemens Healthineers. He has received consultant, speaker, and/or advisory board payments from Abbott, Alere, Beckman Coulter, QuidelOrtho, Radiometer, Roche, Siemens Healthineers, and Upstream Medical Technologies. All other authors declare that they have no conflict of interest.