PCORI Biweekly COVID-19 Scan: AI-Assisted Radiographic Image Assessment, Dual Point-of-Care Nucleic Acid Assays (October 15-28, 2020)
The Briefing provides an at-a-glance view of some important developments in the information universe surrounding COVID-19. The views presented here are solely those of ECRI Horizon Scanning and have not been vetted by other stakeholders.
As the United States enters this year’s cold and flu season, many people will not only be asking, "Is this a cold or is it the flu?" but "Is this COVID?" The ability to differentiate influenza from COVID-19 is important, not only for individual patients but also for providers, to ensure efficient delivery of care within and across populations. New combined diagnostics are poised to help providers do just that (see Topics to Watch). The flu season is also predicted to challenge artificial intelligence models (see Topics to Watch) trained during the height of the pandemic in the absence of flu.
The good news of lower COVID-19–related death rates, albeit among a late October surge in new cases, is balanced somewhat by concerns about people who have contracted COVID-19 and experience persistent, low-grade symptoms, sometimes for months. Because many of these patients do not require hospitalization and might not have received an official COVID-19 diagnosis, they could be omitted from case numbers. With mounting evidence and many unknowns surrounding the virus’s impact on the nervous system and autoimmune consequences, potential exists for a future burden on the health care system and the health of the American people.
ECRI Horizon Scanning has selected the topics below as those with potential for impact relative to COVID-19 in the United States within the next 12 months. All views presented are preliminary and based on readily available information at the time of writing.
Because these topics are rapidly developing, we cannot guarantee the accuracy of this information after the date listed on this publication. In addition, all views expressed in the commentary section are solely those of ECRI Horizon Scanning and have not been vetted by other stakeholders. Topics are listed in alphabetical order.
Artificial Intelligence–Assisted Radiographic Image Assessment for Determining COVID-19 Prognosis
At a Glance
- Artificial intelligence (AI) algorithms are being investigated to help analyze radiographic images (eg, chest x-ray, computed tomography) of patients with COVID-19, to predict the clinical course.
- Methods to identify and monitor the course of COVID-19 might help health care facilities prioritize care and efficiently allocate resources.
- AI-assisted image assessment modalities in development include AI-driven computational tools to evaluate which patients might need the most extensive treatment (eg, mechanical ventilation) or are more likely to develop pneumonia as well as a machine learning–based model for predicting length of hospital stay.
Methods to identify and monitor patients with COVID-19 who are at risk of developing severe disease are of great importance and might help health care facilities efficiently allocate resources and ensure that the highest-risk patients get more timely access to critically needed care. AI algorithms are being investigated to help analyze radiographic images (eg, chest x-ray, computed tomography), to predict the clinical course of COVID-19.
Examples of AI-assisted image assessment in development include an AI-driven computational tool developed by researchers at Case Western Reserve University that helps health care workers decide which patients need the most extensive treatment (eg, mechanical ventilation). Also, Chinese researchers developed a machine learning-based model for predicting length of hospital stay and an AI-driven model for predicting malignant progression of COVID-19 pneumonia.
AI tools to analyze chest radiographic images might help health care professionals identify which patients are likely to develop severe disease and need intensive interventions. AI assistance might help providers improve triage during coronavirus surges. However, overreliance on AI assistance might lead to inappropriate care if an algorithm misclassifies a patient. Increased COVID-19 image data and technology improvements might help providers develop protocols for triaging patients who have COVID-19.
ECRI internal stakeholders thought early data suggested AI assistance might help health care providers identify which patients were most likely to require intensive treatment and which were not. By helping to improve patient triage, stakeholders also thought AI use could help health care facilities better manage resources while protecting staff and patients both with and without COVID-19. However, stakeholders noted that more data collected from larger patient groups over time are needed to improve the accuracy and usefulness of this technology and some were skeptical that it might ever widely diffuse into clinical practice.
- Area of Potential Impact: Patient outcomes, population health, clinician and/or caregiver safety, health care delivery and process, health care costs
- Category: Systems and management
Dual Point-of-Care Nucleic Acid Assays to Diagnose COVID-19 and Influenza
At a Glance
- Coronavirus and influenza virus infection symptoms can be similar, yet clinicians must be able to differentiate between the 2 respiratory infections to properly manage patients.
- Point-of-care polymerase chain reaction (PCR) assays have been developed to diagnose COVID-19 and influenza within an hour, such as the cobas SARS-CoV-2 & Influenza A/B Nucleic Acid Test and the Xpert Xpress SARS-CoV-2/Flu/RSV.
- The US Food and Drug Administration (FDA) granted Emergency Use Authorization (EUA) to cobas SARS-CoV-2 & Influenza A/B Nucleic Acid Test and Xpert Xpress SARS-CoV-2/Flu/RSV and is in the process of evaluating other multiplexed assays.
The flu season in the Northern Hemisphere usually begins in the fall, peaking between December and February, and tapering around May. Because symptoms for coronavirus and influenza virus infection can be similar, clinicians must be able to differentiate between the two to properly manage patients.
PCR assays have been developed that can diagnose both COVID-19 and influenza. These include cobas SARS-CoV-2 & Influenza A/B Nucleic Acid Test (Roche Molecular Systems, Inc) and Xpert Xpress SARS-CoV-2/Flu/RSV (Cepheid, Inc). These assays are intended for use in point-of-care settings, including temporary screening facilities, physician office laboratories, urgent care, and long-term nursing facilities.
Although we found no data specific to the performance of these dual tests, the single virus Xpert Xpress assay has shown that it could perform at least 90% as well as real-time PCR. However, some COVID-19 point-of-care tests have had concordance only perform 70% (or less) as well as real-time PCR. In August 2020, Cochrane had published a summary reviewing the accuracy of point-of-care antigen and molecular-based tests for diagnosis of COVID-19.
A patient's nasal or nasopharyngeal sample is loaded into a single-use cartridge or tube that contains the reagents needed for nucleic acid extraction, amplification, and detection of coronavirus and influenza virus. The cartridge or tube is placed in an analyzer that controls the reaction and reports whether viral nucleic acids have been detected. The manufacturers claim that their assays require a small sample volume and minimal hands-on time to yield accurate results in less than an hour. The FDA has granted EUA to cobas SARS-CoV-2 & Influenza A/B Nucleic Acid Test and Xpert Xpress SARS-CoV-2/Flu/RSV and is in the process of evaluating other multiplexed assays.
Patients with COVID-19 and influenza have similar symptoms in the early stages of disease. Dual COVID-19 and flu tests might allow health centers to accurately determine the cause of patient symptoms.
Initial comments from ECRI internal stakeholders suggest that accurate and rapid tests to distinguish between both viruses will be critical for taking appropriate actions to safeguard both individual and population health. A dual test might decrease testing time and costs, compared with using separate tests. As point-of-care tests, these will be particularly helpful for triaging and managing patients on the same day they are tested as well as limiting disease spread. Differentiating from respiratory syncytial virus (RSV) is also important for managing pediatric patients.
Stakeholders stated concern that health-center adoption of these tests might be limited because of the growing number of available diagnostics and pandemic fatigue. Also, patients living in remote areas might not have access to health centers that offer dual tests.
- Area of Potential Impact: Health care delivery and process, health care costs, health care disparities, patient outcomes, population health
- Category: Screening and diagnostics
Horizon scanning is a systematic process that serves as an early warning system to inform decision makers about possible future opportunities and threats. Health care horizon scanning identifies technologies, innovations, and trends with potential to cause future shifts or disruptions—positive or negative—in areas such as access to care, care delivery processes, care setting, costs of care, current treatment models or paradigms, health disparities, health care infrastructure, public health, and patient health outcomes
The PCORI Health Care Horizon Scanning System (HCHSS) conducts horizon scanning to better inform its patient-centered outcomes research investments. Initially, PCORI defined the HCHSS project scope to focus on interventions with high potential for disruption in the United States in 5 priority areas: Alzheimer’s disease and other dementias, cancer, cardiovascular diseases, mental and behavioral health conditions, and rare diseases. In addition, the system captures high-level disruptive trends across all clinical areas, which may lead PCORI to expand the project scope to include other priority areas in the future.
In early 2020, the COVID-19 pandemic created a fast-moving, widespread public health crisis. In May 2020, PCORI expanded its HCHSS to elucidate the landscape of potentially impactful applications for COVID-19. The HCHSS COVID-19 supplement scans for, identifies, monitors, and reports on emerging and available COVID-19-related treatments, diagnostics, preventive measures, management strategies, and systems changes with potential for high impact to patient outcomes—for individuals and populations—in the United States in the next 12 months.
The HCHSS COVID-19 supplement produces 3 main outputs:
- Biweekly COVID-19 Scans (eg, this document) provide ECRI Horizon Scanning with a vehicle to inform PCORI and the public in a timely manner of important topics of interest identified during ongoing scanning and topic identification or through the ECRI stakeholder survey process.
- Status Reports (quarterly) briefly list and describe all COVID-19-related topics identified, monitored, and recently archived.
- High Impact Reports (every 4 months) highlight those topics that ECRI internal stakeholders (eg, physicians, nurses, allied health professionals, public health professionals, first responders, health systems experts, clinical engineers, researchers, business and finance professionals, and information technology professionals) have identified as having potential for high impact relative to COVID-19 in the United States.
Commentary in this COVID-19 Scan reflects preliminary views of ECRI Horizon Scanning and internal ECRI stakeholders.
The information contained in this document has not been vetted by other stakeholders.
We welcome your comments on this Scan. Send them by email to [email protected].
Posted: November 6, 2020
What's Happening at PCORI?
The Patient-Centered Outcomes Research Institute sends weekly emails about opportunities to apply for funding, newly funded research studies and engagement projects, results of our funded research, webinars, and other new information posted on our site.