Michigan 'flying blind' into virus surge as testing lags, experts say
Correction: A previous version of this story had the incorrect date for when about 9,000 potential COVID-19 specimens were tested by Michigan officials.
Conflicting forecasts of when Michigan will hit its coronavirus peak highlight how the continued lack of testing makes it difficult to predict just when the state will see a "flattening of the curve" or a leveling off on new cases and deaths, experts say.
Without more accurate data, health officials say it makes planning difficult for some medical needs and an end date to the state's social distancing and stay-home policies.
It also creates uncertainty for business and government leaders about when to reopen Michigan's mostly shuttered $527 billion economy.
Some lawmakers, such as U.S. Rep. Tim Walberg, R-Tipton, are wondering if some areas of the state could return to work earlier than others, while Senate Majority Leader Mike Shirkey, R-Clarklake, has formed a bipartisan workgroup that will work with business and medical professions to develop a guide for residents to transition safely back to work.
White House officials on Saturday warned of a peak in deaths for the state by the end of this week. The apex cited in the White House's projection is for the hardest-hit counties of Wayne and Oakland, which account for two-thirds of Michigan's cases and 72% of its deaths.
Gov. Gretchen Whitmer, meanwhile, reiterated Monday she's relying on University of Michigan data in her estimation the virus will hit its maximum impact in cases in late April or early May.
Some statistical models rely on deaths, while others emphasize reported positive cases along with other data and assumptions to project out a peak. But experts say the cases and deaths are both under-reported.
While Michigan is testing thousands more of the sick, there are still many more with mild symptoms who are being treated, sent home and not tested so their numbers aren't showing up in the state's daily case counts, health officials say. Experts also note COVID-19 deaths are being unreported as state resources are being stretched.
Michigan will need more testing to build better models before it can start relaxing any social distancing measures, said Joseph Eisenberg, professor and chairman of the Epidemiology Department at the University of Michigan School of Public Health.
“We need to be able to have a more accurate picture of cases and transmission,” Eisenberg said. “We can only do that through increased testing.
“You are kind of flying blind (without it).”
When cases start to flatten, officials will need to be able to confirm cases are dropping, he said. And robust testing will be needed to ensure when clusters of infections pop up, they are contained immediately, Eisenberg said.
Whitmer acknowledged on Monday the lack of testing had made it difficult to develop the forecasts.
"We’ve never had enough tests to have the kind of robust testing so that we’ve got data that we can feel informs a model that is reliable," the governor said. "... That is not unique to Michigan. That is kind of a countrywide issue that we’ve all got."
How models differ
White House coronavirus response coordinator Dr. Deborah Birx on Saturday projected Wayne and Oakland counties would hit peak caseloads this weekend, citing modeling from the Institute for Health Metrics and Evaluation out of the University of Washington.
Its work is based on COVID-19 deaths because positive cases are so under-reported, said Ali Mokdad, a professor of health metrics sciences at the University of Washington institute.
"Once we project mortality to a point in time, we can back compute the number of people who need to be in the hospital, the number of people who need oxygen, the number who need ICU," he said.
The institute's website predicts as of Wednesday morning that Michigan will peak at 131 new daily deaths on Thursday. On Tuesday, the state announced 118 new deaths.
"There are several models out there, but ours is the only one that does state-to-state comparison (apples to apples)," Mokdad wrote in an email. "We have updated our numbers to account for the new data we have received from all our sources and we are now seeing an earlier peak and a faster drop from locations in Italy and Spain as well as the US."
Michigan officials say the Washington institute is one of at least five sources they are using to predict a later peak.
The others are at Michigan Medicine, the University of Michigan Epimath COVID-19 Modeling group, early modeling from Covid Act Now and assumptions in an academic paper out of Imperial College, said Lynn Sutfin, a spokeswoman for the Michigan Department of Health and Human Services.
When asked at a Monday press conference to explain the difference in the state and White House predictions, Whitmer and Chief Medical Executive Dr. Joneigh Khaldun both said they are using Michigan-specific data.
Infections are slowing in areas of Michigan, particularly Washtenaw and Kent counties, said Dr. Vikas Parekh, an associate chief clinical officer for Michigan Medicine’s adult hospitals and professor of internal medicine.
While he doesn't expect a peak this week, he said it "will be pivotal in understanding the trajectory."
“We are starting to slow, some counties really dramatically," Parekh said. "What the trend this week will really tell us is, are we going to look like China, are we going to look like Italy or we going to look like Washington State?”
Researchers at the UM's Epimath COVID-19 Modeling group are using deaths and positive cases data, along with more than a dozen other measures, such as the time it took for someone to develop symptoms to seek hospital care.
The team's model is projecting a wider window for positive cases to peak anywhere between mid-April and early May, said Marisa Eisenberg, a UM associate professor of epidemiology and complex systems who isn't related to Joseph Eisenberg.
"We are ... modeling the actual process by which people move through from being not infected to becoming infected but maybe being asymptomatic to then maybe developing symptoms and seeking care," Eisenberg said.
She cautioned that many factors could alter the peak, such as changes to social distancing, and that different parts of the state will likely reach a high point in cases at different times.
Marisa Eisenberg said there have been "Herculean efforts" to get better data over the last several weeks, saying their sources are coming from what other states and countries have experienced with the virus, as well as what local hospitals have documented.
"People are really doing this as well and as fast as they can, but it is tough," she said. "It’s kind of good in a way to have different models though out there.
"You want to see what different assumptions will tell you about what could happen."
An unknown factor that could skew the forecasts are individuals who are infected but don't show symptoms and people who have been diagnosed with mild symptoms but haven't been tested, experts say.
In some early modeling, the lack of knowledge about the prevalence of the virus in asymptomatic people led to an inaccurately high death rate, which caused an "astronomical error" in death projections and falsely informed preparedness measures, said John Ioannidis, professor of medicine, epidemiology and population health at Stanford University in California.
"If you look at the predictions for preparedness for hospital bed need, almost all of them were very off because they assumed the vast majority of people who get infected get very sick," he said.
The true number of Michigan positives still "may be 50 times higher" than what is being reported because of those who are asymptomatic or untested, Ioannidis said. The cases confirmed by the state so far also might be just the “tip of the iceberg," he said.
To get a full picture of how many people actually have the virus, the state and nation would have to develop a comprehensive understanding of how many people have antibodies to the virus in their system, he said.
“I’m pretty confident that the numbers will be very high compared to the numbers that we have now,” Ioannidis said.
Without the underlying data, even the best models are unable to provide a reliable projection for peak cases, deaths or resource needs, he said. At best, those models are “speculative,” Ioannidis said.
Based on the epidemic curve throughout the nation, Ioannidis estimates the state will peak in the next few days.
As of Tuesday afternoon, about 43,500 specimens have been tested statewide, compared with 9,000 on March 25, state officials said.
Sutfin said the state has no way "to capture information about people who visit the hospital (or other health care provider) with symptoms and are not tested."
Whitmer and Khaldun said Monday the state is working to develop ways to track antibodies tests once available.
Oakland County Medical Director Dr. Russell Faust argues the positive case and death totals are undoubtedly underestimates.
“They are not representative of reality,” Faust said. “If you can’t test everybody, you don’t know.”
He also said the models are only "as good as the data you put into them."
As of late last week, he had heard from many emergency room doctors in Oakland County who won’t test patients who come in with obvious COVID-19 symptoms unless they need hospitalization. Faust said when patients are presumed to have coronavirus and aren’t severe, they are sent home and advised to self-isolate and return only if their condition worsens.
Deaths are being undercounted, too, he said, citing conversations with funeral homes in Metro Detroit.
"We know that people are dying that have COVID symptoms, yet their cause of deaths is being reported as something else," Faust said.
Henry Ford Health System has been working on its own predictive modeling for the tricounty area of Wayne, Oakland and Macomb counties, said Dr. Adnan Munkarah, the system's chief clinical officer.
While the five-hospital system has seen a steady increase of day-after-day admissions over the past week, officials also have begun to notice what appears to be a decline in the transmission or contagion rate, he said.
“With any of these models, there are a lot of assumptions you have to make and those assumptions change from one day to another,” Munkarah said.
Numbers that are beginning to slow or level off in other parts of the state also are helpful — short of complete modeling — in understanding the movement of the virus amid mitigation measures, said Bob Riney, chief operating officer of the Detroit-based Henry Ford Health System.
The data seems to show “the penetration of that virus wasn’t in those communities before we put into place social distancing,” Riney said.
Despite the differences in predictions, the state's hospitals are assuming they are going to be dealing with a lot of patients. More reliable data on outbreak projections are important but "all hospitals have already been implementing their surge plans," said Ruthanne Sudderth, a spokeswoman for the Michigan Health and Hospital Association.
"Whether our peak arrives in Michigan in a week, or two weeks or three weeks, it is not going to drastically change the fact that we need more supplies, and we need people to stay home and protect health care workers," Sudderth said.
Despite the unknowns, Khaldun said the prediction models are getting more accurate but remained cautious.
"But again anyone who says they know the specific date where it is going to peak, I just don’t believe that’s true at this point," she said.