Researchers at Oakland, California-based Kaiser Permanente have developed a tool to help health systems predict Covid-19 case surges up to six weeks in advance.
Covid-19 cases are ticking up in the U.S., with a fourfold increase in new cases per day over the past month. This tool can help providers manage and mitigate upcoming surges of the deadly disease, the researchers said in a study describing how they developed the score published in BMJ Open.
“Covid-19 has been a dynamic threat so that no individual wave or surge looks precisely like another which affects the performance of prediction and forecasting models,” said Dr. Vincent Liu, study author and regional director of hospital advanced analytics for Kaiser Permanente in Northern California, in an email. “This variability is a byproduct of differences in the biology of the disease, the policies to combat viral spread, and in individual and community social distancing practices.”
To create their prediction tool, dubbed the Covid Hotspotting Score, Liu and his colleagues examined millions of health system data points to identify the most informative leading indicators of Covid-19 surges on hospitalizations.
“By knitting together these diverse streams of data into a single hotspotting score, we were able to extend the lead time for identifying emerging Covid-19 activity up to six weeks,” Liu said.
The researchers conducted the study at an integrated health system in Northern California that includes 21 hospitals and 4.5 million members. They incorporated 10 variables into the score, including four major indicators, like new positive Covid-19 tests, and six minor indicators, such as respiratory viral testing.
The indicators, passively collected from the system’s EHR, strongly correlated with forthcoming Covid-19-specific hospital capacity — up to six weeks before three successive Covid-19 waves.
The Covid Hotspotting Score has helped Kaiser Permanente prepare its leaders, clinicians and hospitals for surges, and the algorithm is now available to other health systems so they can use it with their own data, Liu said.
A simpler version of the score is also available for facilities that may not have the same depth of data that Kaiser has, including patient-initiated calls and emails, he added.
But the tool does have its limitations, researchers said in the study. These include the fact that the score was developed using data from a single health system in a single U.S. region, thus resulting in a lack of indicators of local Covid-19 activity that still likely impacts hospitalization rates.
“Additional external validation of this tool will also be important to assure its generalisability beyond our system,” they wrote.
Photo: loops7, Getty Images