(Bloomberg Opinion) -- The Trump administration's Covid-19 task force prepared the country for the possibility that 100,000 to 240,000 Americans could die from the virus at a Tuesday evening press conference. Those shocking figures aren't a guess or worst-case scenario but come from a group of  widely respected scientists whose models all assume that serious social- distancing measures are in place. They paint a grim picture, but it’s important to remember they are still only projections, and can change based on actions taken now.

According to task force experts Dr. Deborah Birx and Dr. Anthony Fauci, it's not too late to reduce mortality and ease pressure on hospitals — if Americans commit to keeping apart. That view is based in part on a separate model developed by the University of Washington, which was used as a reference point at Tuesday’s briefing. It is unique in that it provides evolving, state-by-state projections that more closely reflect the data on the ground on any given day, including the most recent death rates and distancing information. This model, which assumes much more rigorous lockdown measures will be broadly implemented soon, predicts a mean death toll of around 94,000. But it also carries a warning: Its latest projection, while lower than the others, has climbed 10,000 since its initial release last week. 

Only robust action will stop that number from continuing to climb. If the administration is willing to put this model on slides at a press conference, it should be willing to listen to what it says. That means actively pushing for a more consistent and restrictive national shutdown instead of the current patchwork effort. The numbers demand it. 

The University of Washington model comes from the school’s Institute for Health Metrics and Evaluation, a group that spent years analyzing the burden of diseases around the world. Instead of creating a model that projects how many people are likely to become infected and then applying expected death rates, the group works backward from death data. According to Dr. Ali Mokdad, a professor of health metrics sciences at the Institute, that decision was a result of America’s lagging testing data. He feels that U.S. mortality data is a more reliable place from which to model the future. From there, the institute reverse-engineers projections of potential state peak demand for ICU beds and ventilators that can help pull resources to needy states. 

The swing factor for any coronavirus model is the extent to which social-distancing measures slow infection rates. The IHME model not only assumes that holdout states will broadly enact measures like stay-at-home orders, travel bans, and business closures in a fairly short time period, but that these measures will be instituted effectively through May.  This hopeful assumption contributes to an early projection that may underestimate mortality because states and the Trump administration still aren’t doing what they need to.

President Donald Trump’s decision Sunday to extend social-distancing guidelines to April 30 was a start. But optional “guidelines” aren’t  strong enough, especially when accompanied by mixed messages. Florida Governor Ron DeSantis, for example, has refused to enact a broad stay-in-place order in a high-risk state. When asked if the governor should do so Tuesday, Trump said DeSantis “knows what he's doing" and signaled that he'd largely defer to local leaders.

States and the administration won’t be able to hide from the fact that they haven’t taken appropriate steps. Wide variation in the level and enforcement of distancing measures will eventually show up in the data, in Florida and beyond. 

"We're modeling mortality," Dr. Moktad said. "And death is a function of social distancing. So, in a way, we're capturing how much people are adhering to this message." 

The early answer appears to be “not well enough.” Early positive data from states that took rapid action, such as California and Washington, is an exception so far. Rising mortality projections will lead to strain on hospitals. If unchecked, that will lead to shortages that could lead to more deaths. Even in its possibly optimistic early state, the model projects that the U.S. will be short by more than 18,000 ICU beds at peak demand. 

IHME’s projection isn't perfect. It's attempting to model dozens of changing situations, a difficult task made more onerous by limited data. More generally, no model should be used in isolation. The best decisions come from multiple estimates. However, the administration is telling us that this particular model is informing its decision making. Its explicit message is that more forceful measures are needed. 

This column does not necessarily reflect the opinion of Bloomberg LP and its owners.

Max Nisen is a Bloomberg Opinion columnist covering biotech, pharma and health care. He previously wrote about management and corporate strategy for Quartz and Business Insider.

©2020 Bloomberg L.P.