North Dakota developing projection model for COVID-19
GRAND FORKS, N.D. — As Minnesota presents its COVID-19 projection modeling data on Friday, North Dakota is in the process of finalizing its plan with a potential release early next week.
Projection models for the virus have been the subject of much discussion in recent days. During a press conference held Tuesday, April 7, Gov. Doug Burgum indicated that the state was working on forming its data models in response to one of the leading modeling organizations in the nation inputting incorrect data for the state.
North Dakota officials have not yet released their projections for the coronavirus surge’s timing, magnitude or duration. State officials have declined multiple requests by Forum News Service for their predictions, including the number of hospital beds, intensive care beds and ventilators that will be required to deal with the pandemic.
"We’re in the process of finalizing our modeling and surge plan and hope to be able to communicate that early next week," governor spokesman Mike Nowatzki told Forum News Service Thursday, April 9.
As of Friday morning, there have been 278 positive cases of COVID-19 reported in the state, of that number 105 people have recovered. Just over a dozen people are currently hospitalized and seven people have died.
During Thursday’s press conference, Burgum said projection models help to figure out where the peak is going to be and whether there will be enough hospital capacity to meet that peak.
“The model feeds your hospital surge planning, so I don't have a great sense of urgency right now to drive and get out with a model,” he said Thursday.
Forecasts are just one factor to consider in decision making, according to Burgum, noting there is some uncertainty because one of the biggest factors is rate of spread, as it’s difficult to predict human behavior.
“We're going to continue to work on models," said Burgum, noting his team was trying to figure out whether tier-three facilities, such gymnasiums, would be needed. “Our goal is to make sure that we're taking the actions that keep our citizens out of that. That's not where we want to be caring for people in our state so I feel like we're on the right path.”
Even sophisticated models can have flawed outcomes if they don’t have the right data, according to Burgum, who said models also get better with more data.
Burgum said the state had been utilizing the Institute for Health Metrics and Evaluation, a popular projection model out of the University of Washington, but the organization has apparently been inputting incorrect data for North Dakota.
Burgum said the state had been trying to contact IHME to let it know about the inaccuracies. The model had, at one point, said that North Dakota had a day with 17 deaths and another with nine, according to Burgum.
IHME projects North Dakota will hit its peak on April 21 with 672 beds and 138 ICU beds needed. While the state would have enough regular hospital beds, IHME projects the state would be short 52 ICU beds. IMHE also projects 117 ventilators would be needed by that date.
The organization also projects the state would hit a peak of about 15 deaths per day during April 20-22. IHME projects there will be 519 COVID-19 deaths in North Dakota by Aug. 4. Its original data had estimated 22 on April 20, and more than 670 overall.
The organization has adjusted its projections for deaths nationwide, as well. Last week, IHME’s model has dropped its estimates from 93,000 deaths by late summer to 60,400 deaths in the United States.
The North Dakota model being developed is being done, in part, with help from UND.
Unable to disclose details about the model, Ryan Adams, director of the UND School of Electrical Engineering and Computer Science, said the data is trending positively. The state will likely hit its peak in the next week to two weeks and then gradually taper off, Adams said.
“It's not nearly as extreme as some had feared, and certainly not as extreme as what we're seeing in New York and California and other other states,” Adams said. “The fact that we are a rural state is very much an advantage right now.”
Creating models can be tricky, Adams noted. While there are good modelers across the country, North Dakota wanted to do its own to provide a “sanity check.”
“It doesn't help to rely on one group of people in Washington or wherever,” he said. “We wanted to make sure that things were looking consistent and appropriate.”
Incorrect assumptions in areas such as the number of people who may come into contact with an infected person can lead to massive errors in data inputs and outputs, Adams said.
Adams said data shows the state should have enough hospital beds for COVID-19-related cases.
Adams said there are more positive cases than initially anticipated, but noted the primary reason for that is because the state has been more proactive that what the modeling had estimated. For example, the state’s decision to do community testing, regardless whether someone was sick, in the southwest portion of the state did lead to an elevation in the number of positive test cases. However, those numbers don’t seem to be impacting how many are being hospitalized or who need ICU care.
“We have much more capacity that we will need for this virus and it's by a factor of five times more,” Adams said. “So, we have plenty of capacity, which is very, very good, given the circumstance.”