Over the weekend, pandemic models projecting Alabama to be one of the hardest-hit states by COVID-19 shifted to show a massive drop in the number of estimated deaths and the number of ventilators and ICU beds that may be needed as the outbreak peaks in the Yellowhammer State later this month.
That model, which was developed by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington, has received significant attention since it was first published last month and even more since it was cited by members of the White House’s Coronavirus Task Force last week.
The numbers it has projected are credited with decisions that governors in Florida and other Southern states made to issue “stay-at-home” orders after initially being reluctant to do so. Though it’s unclear to what extent, the projections were also used by officials in Alabama as they’ve evaluated their response to COVID-19.
As recently as Saturday, the IHME model suggested Alabama could see the fourth-most fatalities of any state in the U.S. by the end of August — a grim forecast of 5,516 deaths, tens of thousands infected Alabamians and a critical shortage of hospital beds, intensive care units and invasive ventilators.
By Monday morning — two days after Gov. Kay Ivey’s statewide stay-at-home order took effect — the same model was projecting an average of 917 deaths and moved Alabama’s projected peak up to mid-May with the highest demand on hospitals expected around April 18. Instead of needing to find thousands more ventilators for an expected surge of cases, the models now suggest Alabama already has enough.
Update: As of April 9, the IHME has reduced Alabama’s projected number of deaths down again to an average of 634 with a margin of uncertainty between 237 and 1,837.
Dr. Ali Mokdad, a professor of health metrics sciences at IHME who developed its COVID-19 model, said there were a number of factors behind the positive shifts, but the most prominent was the inclusion of better data.
According to Mokdad, the original IHME models were informed by Centers for Disease Control and Prevention (CDC) reports from early COVID-19 cases in the U.S. and the outcomes that were being seen in other countries.
While outbreaks in countries like Spain and Italy have been disastrous, their peaks have come faster than expected.
Speaking to Lagniappe Monday, Mokdad also said using Italy as an indicator of how many Americans might be hospitalized or require intensive care appears to have been an “overestimate” now.
In an April 5 update, researchers were also able to add state-specific data on the pandemic as it’s actually unfolding.
In addition to the general changes in the modeling, Mokdad said new weight given to social distancing efforts in the projections are another reason Alabama’s numbers improved so drastically.
“The reduction in projected deaths is the implementation of these social distancing measures, which usually have an impact about a month down the road,” Mokdad said. “This is the result of the appropriate decisions your governor has made. They’re playing out now in the models because we know they have worked here in other places like Washington and California that took these actions very early on.”
Due to the timing, some were quick to link the improved IHME models directly to Ivey’s April 3 stay-at-home order, but Mokdad said other actions Alabama had taken in the weeks prior had a significant impact as well. He also clarified that IHME’s calculations are based on full adherence to those social distancing measures and don’t take into consideration what they look like in practice.
As Lagniappe has reported, every state, city and county with that type of order provides exceptions for essential activities and essential businesses. Mokdad said Monday Alabama’s decision to close public schools on March 13 — more than two weeks before its first reported death — likely had a significant impact on the spread of COVID-19, but still said further statewide measures were warranted.
“Closing the schools and universities has had a huge impact in Alabama and has saved a lot of lives in the U.S.,” he said. “But a stay-at-home order is still very important. If you close the schools but kids and parents are still going out, you’re only changing where people are getting together.”
Even using the retooled models, though, IHME seems to be projecting numbers of deaths that are significantly higher than what is actually being seen on the ground in Alabama. For instance, the Alabama Department of Public Health reported just 53 deaths as of April 6 while IHME’s model projected anywhere from 56 to 165, with the average being 89.
That means the projection for April 6 was 87 percent higher than the actual number of deaths. April 6 death projections for several other states Lagniappe examined were significantly higher than what actually occurred as well.
There also seems to be inaccuracies with the “total deaths” listed on IHME’s public dashboard, which are supposed to be the actual deaths reported from each state. Between April 4 and April 5, IHME showed Alabama’s total deaths jumping from 57 to 72 even though the state had only reported 53 at the time.
A spokesperson for IHME told Lagniappe that discrepancy in the number of actual deaths was an “error” and would be corrected. They did clarify that the error was with what’s displayed on the dashboard and not with the actual data being used to calculate future projections.
As for the projections trending higher, Mokdad said they are based on what information comes in when it’s reported. Because IHME’s modeling is updated regularly and the number of deaths and new cases in a state can vary significantly from day to day, the projections can shift drastically.
Mokdad said that’s one reason why the projections have such wide margins, though those are expected to shrink.
“There’s a big difference as the reported deaths move up and down and we take the average of those in our statistical models to make the model perform better. What we’re seeing is a function of [Alabama] having fewer reported deaths and fewer deaths earlier on,” Mokdad said. “Unfortunately, we will have more deaths as we move forward, and a week from now our estimates will likely be much stronger and [the scale] will become more narrowed.”
While the shift in the IHME projections for Alabama is good news, Mokdad said the state shouldn’t view it as some kind of all-clear. Pandemic models — especially ones built from live and evolving data — are subject to change.
If the number of confirmed deaths increases, as it is expected to when Alabama reaches the projected peak of its outbreak later this month, he said the projections will likely look worse.
“We shouldn’t be letting our guard down. When something is encouraging, you keep going. You don’t change horses in the middle of the race if you’ve got a good horse,” Mokdad said. “I’m happy for Alabama and what you’ve done with these measures, but it’s time to stick to them so you can reduce the number of deaths and possibly even reopen your businesses earlier than originally expected.”
This page is available to subscribers. Click here to sign in or get access.
It looks like you are opening this page from the Facebook App. This article needs to be opened in the browser.
iOS: Tap the three dots in the top right, then tap on "Open in Safari".
Android: Tap the Settings icon (it looks like three horizontal lines), then tap App Settings, then toggle the "Open links externally" setting to On (it should turn from gray to blue).