Sarah is the CTO of Sight Diagnostics, a company transforming healthcare through fast, accurate and less painful blood testing.
The coronavirus pandemic exposed weaknesses in our healthcare system and a lack of agility to react quickly in a crisis. Four months into the pandemic, as governments are working to reopen economies, experts are already warning about the inevitable “second wave.” The collateral damage of the next wave will be dependent on the progress we make over the coming months in testing for the virus, tracking cases and isolating people who are infected to manage the spread of the disease.
The pandemic has rapidly transformed the hype of artificial intelligence (AI) into a new reality for healthcare. Researchers and developers are leveraging the capacity of AI and machine learning to gain a better understanding of the coronavirus and to track the spread. They are attempting to predict where the disease will spread next, why it affects some populations more than others and what measures will help manage and prevent the further spread of infections. The common thread of all these efforts is an enormous amount of data, which is too much for humans to comprehend and analyze on their own but a perfect task for AI and machine learning.
Triage And Diagnostics
Shelter-in-place orders and social distancing have helped flatten the curve in an effort to combat Covid-19, but as the reality sets in that a known cure or vaccine is years away, we must look at the potential for technology to help us get through this pandemic. The coronavirus pandemic has forced hospitals to more quickly adopt AI technologies to help triage and screen patients — and spot those who could develop severe symptoms. Many health systems, such as the Cleveland Clinic, have customized their own chatbots to allow patients to complete a virtual exam to screen for Covid-19. The benefit of a virtual exam is twofold, as it aids physicians in identifying the most critical cases for further testing and keeps people with non-urgent conditions out of the emergency department. Apple released its own Covid-19 screening system, which was “created after consultation with the White House Coronavirus Task Force and public health authorities.”
Researchers are also working to create and verify predictive models that can help hospitals understand which of their Covid-19 patients will get worse and how quickly that will happen. For example, Stanford researchers are using an off-the-shelf AI tool from Epic “to see if it can help identify which hospitalized patients may soon need to be transferred to the ICU.” The next step in preparation is to supplement coordinated triage efforts with more comprehensive diagnostic insights. Polymerise chain reaction (PCR) and antibody testing are the primary ways in which global healthcare systems are testing patients for coronavirus, but both techniques have their limitations. PCR requires up to 48 hours to return a result, and even as companies such as Abbott race to develop faster molecular tests, studies are scrutinizing their accuracy.
The need for reliable, accessible testing to screen for the disease has become increasingly important, and we need to supplement these tests with other data points to improve the triage and diagnostic capabilities. Early research has shown that one of the simplest tests, complete blood count (CBC), which evaluates overall health, can predict the severity and disease progression of Covid-19. When combined with patients’ symptoms and medical history, physicians can begin to get a clearer picture of the disease and better manage its spread. These additional data points enhance the quality of regional reporting. When hospitals report on new cases, they should include a breakdown of gender, sex, age, race and underlying conditions.
Preparing For The Next Wave
As governments around the world begin planning exit strategies to lift shelter-in-place restrictions and restart economies, most plans propose extensive testing, contact tracing and isolation measures, but they overlook the need for a dynamic approach of data collection with a feedback loop. Existing public health reporting measures are reporting metrics with 48-hour delays and often with errors. When media headlines report that cases have spiked in a specific region, it’s not clear whether the increase is due to the daily increases in testing, more testing of symptomatic patients or a true change in the infection rate. Instead, we need to move to scientific surveillance of population samples to properly estimate prevalence. Regions should report both suspected and confirmed hospital cases and note when a patient was admitted, along with the date of the report, to more accurately project the disease’s burden on the regional health system.
We have a small window of opportunity to prepare for the next wave by organizing centralized data collection efforts that keep precise records of each infected patient. We must track demographic information, clinical presentation, disease progress and cause of death to better understand the novel coronavirus. Then we can aggregate the data regionally and model predictions for comorbidities, complications and the true rate of infection spread. AI can help produce sophisticated models to help quickly classify patient risks and suggested treatment pathways, but we first need the organized data available to train an algorithm. If we hope to prevent the second wave of Covid-19 or to address future pandemics, we need to invest now in reliable tracking and benchmarking efforts.