Predicting COVID-19 in Oregon: it’s all in the numbers

Available in Spanish

photograph of Peter Graven, director of Oregon Health & Science University’s Office of Advanced Analytics
Peter Graven, director of OHSU’s Office of Advanced Analytics

Peter Graven isn’t a fortune-teller, but he tries his best to predict the future.

Graven, director of Oregon Health & Science University’s (OHSU) Office of Advanced Analytics, has helped people in Oregon make sense of the pandemic with  regular forecasts that began in March 2020.

As complex as COVID-19 data collection and interpretation is, Graven works hard to simplify it.

“As a Ph.D., you learn how to make simple things complex, and boy can we,” said Graven, age 44. “But maybe just as importantly, we learn to make complex things simple. A fancy model is only good if people can understand what it’s doing.”

The pandemic has offered Graven an opportunity to do exactly that – to simplify COVID-19 data.

“We are trying to estimate how many people are susceptible the virus and, given how fast we know the virus spreads, how many more people will get infected,” Graven said.

By also including data about how many people are likely to need hospitalization, Graven’s model can help predict how many hospital beds might be available during a COVID-19 surge. It’s the kind of data Oregon Health Authority (OHA) looks at closely.

“Dr. Graven’s projections have proved to be remarkably close to the mark, so we value the preview they give us into where the pandemic is going,” said Paul Cieslak, medical director, communicable diseases and immunizations at OHA. “How many COVID-19 patients are in hospitals tells us how much stress the pandemic is putting on our health care system, and whether we might need to undertake more measures to ‘flatten the curve.’”

Graven has used multiple factors to predict COVID-19 cases and hospital capacity, initially weekly, and now every other week. His methods are similar to how weather forecasters combine factors such as air pressure, moisture levels and wind speed to predict whether the coming week will be sunny or rainy.

Graven specifically includes things such as what COVID-19 variant or subvariant is predominant, how transmissible it is, how well it escapes immunity and how well vaccines are working.

sample page from Peter Graven’s August 5, 2022, forecast

From economics to COVID-19

Graven is the first to admit that his expertise isn’t infectious diseases. His specialty is health economics and using math, data and models to help improve OHSU operations.

“I’m a people person who likes math, being part of a team and helping,” Graven said. “I’ve always thought that numbers can help us connect on things where words aren’t enough.”

Graven was an undergraduate at University of Wisconsin-Madison when he discovered his interest in economics.

“The field seemed to be this nice blend of understanding social and political problems from a mathematical perspective,” Graven said. “And I liked how I was able to communicate things in a numeric way.”

When Oregon recorded its first COVID-19 case Feb. 28, 2020, Graven went to work, knowing his data-oriented mind could help. He combed through hospital data and began researching epidemiological models. As he played around with models, using what scientists knew at the time about the COVID-19 virus, it quickly became apparent that Oregon needed to act fast.

One of Graven’s models during those first few weeks stated, “We’re going run out of hospital beds in Oregon.”

Within days, the state locked down. People stayed away from each other, canceled events and stayed home as much as they could. As a result, Graven’s ominous prediction about hospital beds, thankfully, did not come true.

Forecasting COVID-19

Graven’s model has mostly done well forecasting when surges will peak and how many hospitalizations may be required. For example, his model forecast the surge and timing of the peak of hospitalizations during the Delta wave in summer 2021, as well as for the original Omicron wave in winter 2021 and early 2022.

But that doesn’t mean he’s always right. In February 2021, Graven predicted spikes in cases for four consecutive forecasts, but those spikes never materialized in Oregon or elsewhere. Graven realized his “fear and fatigue” model was off. That model balances how fearful the public is about the spread of COVID-19 with how fatigued people are about wearing masks, not gathering with friends and family and other safety restrictions. And Graven realized he had miscalculated.

The pandemic’s first winter holiday season in 2020, which included a surge, had just passed. Graven initially believed the end of the surge would reduce public fear. He assumed people were tired of safety measures and would venture back into society, take more risks and spread the virus. But, as it turned out, people mostly continued to follow public health measures, longer than Graven anticipated, thereby preventing the spike in cases he had predicted.

Graven tweaked his “fear and fatigue” model, the spike in cases eventually came, and he correctly forecast a surge in hospitalizations for April 2021.

Graven, however, doesn’t take those failures personally. He just commits to doing better the next time.

“I’ve taken a real chance of not trying to hide what my actual prediction is,” Graven said. “If I’m wrong, it’s me. And I will revise it the next week.”

Models can’t predict the future, but they can help us understand what the most likely scenarios are going to be based on what we know. Graven’s approach is to give people information. Rather than saying ‘things are going to get worse,’ his model gives people in Oregon a date to prepare for the next surge, or for when hospitals might be overwhelmed.

When people are able to see a prediction, they can make decisions to protect themselves and others. In this sense, Graven believes he has succeeded.

“The most valuable parts are when I’ve heard my model has given people enough hope that things will get better, and that they can push through a little longer with sacrifices they’ve made for each other.”

Peter Graven and OHSU release a new COVID-19 forecast every other week. The latest forecast was published August 5, 2022. Bookmark this webpage to check future forecasts.