For data junkies
Very interesting visualizations tracking fatalities and healthcare system utlilization and drawdown state by state.
Kindly brought to my attention by @Molly Nixon
Background paper with assumptions
http://www.healthdata.org/sites/defa...2v1-Murray.pdf
Lead paper introduction
http://www.healthdata.org/research-a...ays-and-deaths
Good news is that estimates of US Fatalities throughout summer are less than expected - approx 81,000 fatalities
NOTE
- key assumptions are social distancing measures continue.
- resource rate utilization is variable from state to state. Some states are overwhelmed (NY LA), Some states hold up (WA)
- data is extrapolated from time-series from Wuhan. Is the Wuhan data accurate?
ORIGINAL POST BY MOLLY NIXON
There have been a lot of items circulating with forecasts for COVID-19, but many of them have been hacked together by non-experts (and it turns out this stuff is actually really complicated).
I happen to work at a world class global health research institution here in Seattle (the Institute for Health Metrics and Evaluation) that specializes in estimating health burden (deaths, cases, etc). We've thrown a team of extremely talented experts at COVID modeling for the past few weeks, and just released the first round of models to the public this morning. This initial release covers daily deaths, hospital-bed days, ICU-days and ventilator-days for every US state for the next 4 months.
Importantly - this isn't a static model. We are continuing to incorporate huge amounts of new data and the goal is to update the model EVERY DAY. As the epidemic evolves and different control measures are implemented (or not), our forecasts will change in response.
Our current predictions (from the paper linked below) are:
"Compared to licensed capacity and average annual occupancy rates, excess demand from COVID-19 at the peak of the pandemic in the second week of April is predicted to be 64,175 (95% UI 7,977 to 251,059) total beds and 17,309 (95% UI 2,432 to 57,584) ICU beds. At the peak of the pandemic, ventilator use is predicted to be 19,481 (95% UI 9,767 to 39,674). The date of peak excess demand by state varies from the second week of April through May. We estimate that there will a total of 81,114 deaths (95% UI 38,242 to 162,106) from COVID-19 over the next 4 months in the US. Deaths from COVID-19 are estimated to drop below 10 deaths per day between May 31 and June 6."
The primary mode for accessing the forecasts is a visual tool (linked below) - it's a little slow at the moment, it's getting a lot of traffic. This tool will be updated to reflect our latest model run. The solid lines indicate our estimates, while the shading denotes uncertainty intervals. Please note that the "United States of America" set of figures is missing the horizontal lines that indicate bed availability (which makes sense, because beds in Connecticut are not useful for patients with COVID in New Mexico). If you look at the state-specific forecasts, they include bed availability.
If you want all the nitty gritty of the modeling parameters, an accompanying paper has been submitted to medRxiv