Inside the Model that Predicts Surges in COVID-19 Cases

During the COVID-19 crisis, it’s been important for health systems to learn as soon as patients with COVID-19-related symptoms present in the ambulatory setting. There’s just one problem: current electronic health record (EHR) technology does not allow for providers to easily understand and interpret information about patients in the telehealth, physician office and urgent care setting across multiple EHRs. Without this capability, health systems are deprived of key knowledge about potential cases or surges.

However, limitations in the EHR haven’t stopped leading healthcare providers from predicting COVID-19 cases that may require more complex care.

Innovative technology is helping leading health systems and hospitals predict COVID-19 cases and respond to an uptick.

Health systems have been working with integrated data that leverages clinical data to create predictive models about demand, thereby projecting the health systems’ caseload. With up to a week’s head start, operational leaders can plan for the proper staffing, equipment and supplies across their hospitals.

Here’s how intelligence is keeping providers ahead of the curve in their response to COVID-19:

• By aggregating symptom data in the ambulatory setting and via telehealth. When patients think they have a fever or other possible COVID-19 symptoms, they are likely to contact a primary care physician or urgent care center. The encounter could occur via a virtual visit or an in-person consult. Either way, health systems may get the first indication that a patient is developing a potentially severe COVID-19 infection before the patient requires an acute care setting. This is where providers’ capability to see and aggregate the symptoms in an ambulatory setting becomes imperative, as symptoms may occur about a week prior to hospitalization.
• By turning aggregated data into valuable, actionable information. When patients present in an ambulatory setting or via a virtual visit, Premier® technology can flag suspected or confirmed COVID-19 patient cases directly in the EHR, at the point of care. Natural language processing and machine learning enable this to occur. Natural language processing reads, interprets and contextualizes free text within the medical record, and extrapolates and interprets the patient information to predict surge. By scanning free text in physicians’ notes for terms such as “trouble breathing” and “loss of taste,” the technology quickly identifies patients who are presenting with symptoms and signs associated with COVID-19. Later, these findings may be associated with a patient’s future COVID-19 test results. The data enables the modeling of disease progression, which helps determine the patients who are the most likely to eventually be hospitalized. This enables hospitals to foresee potential future patient volumes.
• By using aggregated data to prepare for a surge. With the foresight presented by data from ambulatory settings, hospitals have time to allocate the beds, staff and supplies to handle the incoming cases. Likewise, if the data projects a decline in cases, providers can begin scheduling elective surgeries and ramping up non-emergent services.

The power to predict an uptick in cases is essential to managing operations in a pandemic.

With COVID-19 remaining in the foreground for the foreseeable future, providers need technology that helps them predict waves, manage clinical care, project needed supplies and model a hospital’s COVID-19 census.

Premier has created a COVID-19 early warning system, leveraging syndromic surveillance, to enable providers with the critical insights they need to safely care for their populations. See this technology in action.

2021 update: Premier's syndromic surveillance solution, described below, was recognized by FastCompany as an honorable mention for its 2021 World Changing Ideas Award! Check out the full list of finalists here.