For the COVID-19 HotSpotting Score, investigators use a combination of indicators to help predict upticks of cases and potential hospital admissions – with up to six weeks' lead time.
In a study published this week in BMJ Open, Kaiser Permanente researchers put forth a method to predict upcoming COVID-19 surges up to six weeks in advance.
By examining electronic health record data from Kaiser Permanente in Northern California, the team was able to zero in on ten indicators that, they say, can help effectively forecast an upcoming surge when combined.
"This current COVID-19 surge has shown us how challenging it is to have a reliable, long-range forecast of COVID’s impact on hospitals," said Dr. Vincent Liu, lead author on the study, in an email to Healthcare IT News.
"By knitting together diverse streams of health system data, we can identify the earliest signals of renewed COVID activity impacting our patients and contextualize our findings against other prediction tools," said Liu, who is a research scientist with the Kaiser Permanente Division of Research, as well as being a practicing intensivist with the Permanente Medical Group and regional director of hospital advanced analytics for Kaiser Permanente in Northern California.
WHY IT MATTERS
Based on 35 million data elements, the investigators ultimately incorporated 10 indicators into "the COVID-19 HotSpotting Score," or CHOTS.
They identified four major indicators:
- Patient calls that activated regional "cough and cold" protocols.
- Patient-initiated "influenza-like illness" email communications.
- New positive COVID-19 tests.
- COVID-19 hospital census numbers.
The also noted another six minor ones:
- Patient calls that activated regional COVID-19 protocols.
- Respiratory infection visits (routine).
- Respiratory infection visits (urgent care).
- COVID-19 visits (routine).
- COVID-19 visits (urgent care).
- Respiratory viral testing.
Although many of the individual indicators signaled an upcoming surge within one to three weeks, the combined CHOTS significantly increased the lead time to as far as six weeks prior to a surge, said the Kaiser team.
"Over the course of 2020, COVID-19 surprised us at nearly every turn, making longer-term predictions of its impact on our patients, health system, and communities extremely challenging," said Liu in a statement.
"At the same time, shorter-term predictions – looking only one to three weeks out – left little time to respond adequately,” he said.
After CHOTS went live in June 2020, the team evaluated it against actual COVID-19 hospital activity through the end of the year.
"The correlation of the regional CHOTS with hospital census was very strong, peaking with a 28- to 35-day lead time, but with continued correlation when tested out to six weeks," said Kaiser representatives in a press release.
Researchers note that public health officials and individual health systems could use the forecasting information to help prepare for increased patient numbers – and know when relief is on the way.
They also flag a few of the study's limitations, such as the fact that the tool's generalizability could vary across settings and geographies.
In addition, they mention that they developed CHOTS during a time of "great uncertainty" following the first wave of COVID-19 in California.
"As a result of the extreme urgency to prepare our health system, we depended on clinical judgment and heuristics, in addition to prior health-system influenza patterns, to develop our score," they wrote in the study.
"With the luxury of time, more advanced machine learning or statistical techniques may have produced different calculations," they added.
THE LARGER TREND
Given the immense strain on resources that COVID-19 has continued to put on hospitals, multiple teams of researchers have tried to develop predictive tools that can help health systems prepare for the different factors affecting demand – such as length of hospitalization, respiratory failure likelihood, clinical severity and patient outcomes.
More broadly speaking, chief information officers have spoken to the importance of integrated supply chains, which could respond to fluctuations in need.
ON THE RECORD
"We use machine learning and artificial intelligence every day in our research group to develop predictive models to improve patient care. We applied these tools when we were developing CHOTS, but didn’t find that they improved the tool’s value," said coauthor Patricia Kipnis, principal statistician at Kaiser Permanente's division of research, in a statement.
"Our research group focuses on pairing the right algorithm with the right use case, and, in this case, a simpler tool showed excellent performance and could be readily implemented and shared," she added.