The above map uses data from the CDC's tracking of current hospital capacity and current and historical case numbers. Use the arrows to step back or forward by a single day, or to jump to the earliest reported data from roughly mid-April. The heatmap shows how occupied hospital ICUs are in each state, without regard to how many beds are represented. To see how many beds a state had available and occupied at a particular date, you can mouse over it and the data will appear in the legend.

Before July 30th, the CDC used a different method for reporting hospital capacity. Between April 15th and July 7th, I have interpolated this estimate against the first canonical data available from July 30th. Between July 7th and July 29th, there is no usable data from the CDC on hospital capacity. Because the CDC reports case data daily, but not hospital data, there are some dates for which the case graph updates but not the capacity graph or the map.

If you choose "Scaled," the heatmap applies a scaling factor based on the rolling average of new cases. That is, if the case trend is positive (a rising number of COVID-19 cases) then low ICU availability becomes even more of a critical issue. On the other hand, if daily new cases are declining, then what appears to be low bed availability might reflect a spike in cases that is now abating. You can change the impact that the scale has by dragging the slider.

I have chosen 65% ICU capacity as a good baseline for what counts as being fairly safe; if you think this is too aggressive or insufficiently so, you can choose a different value and the heatmap will automatically update.

If you choose "Change/Time," the map will instead track capacity changes compared to the previous dataset (generally one day earlier, although in some cases the CDC provides more than one drop per day). This is probably less useful of a visualization, but it is available if it helps.