Thread by @joshuasweitz

Today, we are releasing an update on GA-level forecasts: a Metapopulation AGe-structured Epidemiological (MAGE) model for COVID-19 in Georgia, USA.

The central projections can be seen in the figures with a full report available here:

weitzgroup.github.io/MAGEmodel_covi…

MAGE is an age-structured COVID19 epidemic model extending a SEIR framework to include hospitalization, demography, and commuting information for all of GA’s 159 counties.

MAGE development was led by @BeckettStephen w/key contributions from @marian_dm12 and @friendly_cities.

We initialized the model by fitting to data for Georgia on March 28, 2020 (including existing county-level heterogeneity); and projected forward in time under a scenario with a 50% reduction in COVID-19 transmission rates due to social-distancing interventions.

Fits to hospitalizations and deaths hew close to reports from @GaDPH, but the model’s estimate of the number of cases is far above reports. Despite model uncertainty, we interpret this as a fundamental problem of ascertainment.

Hence, whereas there are approximately 25,000 reported cases via the @GADPH, we anticipate that the total prevalence is likely to be in the range of 125,000-250,000 (i.e., serological testing is urgently needed).

dph.georgia.gov/covid-19-daily…

We use the same ascertainment level to convert ‘model’ cases into ‘reported’ cases. Even with uncertainty in ascertainment, we estimate that the vast majority (on the order of 95%+) of GA residents remain susceptible.

In addition, we project that 20,000+ new cases and 1,000+ new deaths by the end of June are plausible, even with continued efforts to reduce contacts via social distancing.

Instead, more aggressive efforts that complement social distancing with testing/tracking could lead to sustained downturns in cases and fatalities.

Given that the vast majority of us are immunologically naive, returning to business as usual is not a viable option. Instead, a combination of social distancing, testing (PCR and serology), contact tracing, and other measures will be needed to reduce COVID19 spread.

As we were preparing our model for release, we learned that @drake_lab reached similar findings (albeit independently, and using different modeling approaches):

2019-coronavirus-tracker.com/index.html

These independent reports take fundamentally different routes to similar conclusions.

In summary: COVID19 is a long-term challenge. The impacts are heterogeneous across the state and policy makers should take steps now to scale-up interventions (including testing/tracing) to reduce transmission to enable safer economic re-engagement.

weitzgroup.github.io/MAGEmodel_covi…