Bristol Bay Borough shows up on multiple COVID-19 data-modeling sites as one of the nation’s hottest hotbeds of infection.
On Georgia Tech’s popular COVID-19 risk modeling site, Bristol Bay Borough shows up as the only county in the country where the risk of someone at a gathering of 10 people having COVID-19 is as high as 99%.
On the New York Times’ interactive data page on July 15, Bristol Bay Borough represented the highest cases per capita, both overall and in recent cases. That’s 6,818 cases per 100,000 people. The borough also holds the record for recent cases, at 3,708 infections per 100,000 residents. And therein lies the problem.
“They don’t take into account the full influx that happens for an eight-week period in Naknek,” said Mary Swain, director and CEO of the Camai Community Health Center in Naknek, one of the borough’s three communities. “It’s great to put in the population when our population is only 700-850 depending on what population site you look at. However, if you account for the 10- to 13,000 that potentially come in during the season, for those eight weeks, that modeling could look a lot different, and it is not taken into account in the majority of the modelings that I have seen.”
Ask a local health professional about the apparent huge outbreak in Bristol Bay, and they’ll tell you what most local residents would say as well: nearly all of the cases have been in what the state calls “seafood industry” workers, or people running the salmon processing plants in the area. Those outside the plants have been industry-affiliated–fishermen or workers at fishing lodges.
The bottom line, according to Swain,“There has not been one resident case in Bristol Bay Borough.”
And, Swain said, as of Friday, only one of the borough’s 62 positive cases is still in the borough. That’s because the 61 other cases have either recovered from the virus and tested negative, or have physically left.
So why do the national models continue to show such enormous risk and sky-high per-capita cases in the borough?
Jared Parrish, a senior epidemiologist with the State of Alaska, sees Alaska as a fairly consistent anomaly.
“The old adage of a model is, ‘All models are wrong, but some are useful,’ and what we have continually seen here in Alaska is that many of the national models don’t translate very well,” he said.
Parrish said he and his colleagues have talked about developing Alaska-specific models that will better account for the wide variations in seasonal population, and other unique features of Alaska that make nation-wide models fail to accurately represent the state.
Joshua Weitz, a biological sciences professor and risk modeler at Georgia Tech, said the Bristol Bay Borough was an example of the model’s limitations.
“The kind of extreme variation that you’re talking about a sort of a 13-fold difference, something from a thousand to 13,000, that’s a tremendous difference,” he said.
Weitz said the problem is essentially that the models didn’t account for the massive variability of Bristol Bay Borough’s seasonal population, or the rigid COVID-19 mitigation measures at processing plants that have kept the virus out of the local population.
“We make an assumption that the risk is homogeneous, so people need to be aware that models come with assumptions,” he added.
That assumption — that the borough’s population is currently at it’s off-season year-round population count of at most 850 people–is what puts it at the top of the charts for national reporting.
Weitz added that the Georgia Tech risk modeler does update, so in theory, Bristol Bay Borough won’t look like such a risky, infectious place in the future. Since the beginning of this week, the risk in the borough has dropped from above 99% to 99% flat.