Yale/Harvard public health tool maps local COVID-19 data

This post was originally published on this site

A team of epidemiologists at the Yale School of Public Health and the T.H. Chan Harvard School of Public Health have created a series of models that show in real-time the latest COVID-19 levels county by county throughout the United States, as well as a state-by-state breakdown that feature estimated infection rates, the reproductive number of the virus and the estimated seroprevalence. The tool, covidestim, was created to be a tool for public health officials and the general public alike to make more informed decisions. It is available to everyone for free at the covidestim website.

“In contrast to most other models that attempt to project the behavior of the epidemic in the future, we are focused primarily on generating a better understanding of the current state of the epidemic,” said Professor Ted Cohen, one of covidestim’s creators.

Covidestim is a “nowcasting” model. What is the goal of this model and how does it differ from other models that we have been following?

TC: Our ability to track the progress of the COVID-19 epidemic in real-time is challenged by several features of the disease: 1) not all infections lead to symptoms; 2) not all symptomatic cases are detected, and those that are detected are reported with delay; 3) not all individuals dying from COVID-19 are identified, and those that are identified are reported with delay. Our model uses data about detection completeness and reporting delays of COVID-19 cases and deaths to make real-time estimates of the numbers of new infections occurring in a location on any particular day of the epidemic. We make estimates at both the state- and county-level for the number of infections occurring today (and over the previous course of the epidemic) and the real-time effective reproduction number of COVID-19. In contrast to most other models that attempt to project the behavior of the epidemic in the future, we are focused primarily on generating a better understanding of the current state of the epidemic.

How will this model contribute to the fight against COVID-19?

TC: A rational public health response to the epidemic requires a solid understanding of the current state of the epidemic. Our ability to measure the impact of interventions depends on being able to more closely track the relationship between the timing and intensity of interventions and changes in the rate at which new infections are occurring within specific geographic areas. Our model aims to improve our understanding of the real time fluctuations in the incidence of infection to allow for these insights. We hope that they can also be used by individuals to better understand their current risk level based on local information.

How frequently are these models updated?

TC: The model is updated daily.

Do you see any trends as of early October that are concerning to you as an epidemiologist?

TC: There are several trends that I think are worth keeping our eyes on. First, we continue to observe numerous states in which the incidence of new infections is either stable or increasing from levels that were already concerning. Diving deeper into the county-level estimates reveals that there is marked heterogeneity in terms of infection incidence within these states, and the rate of infections in some hotspot counties are rising suggesting areas for targeted interventions. Using the timeline feature on the county-level maps reveals wave-like behavior of the epidemic, first spiking in the Northeast in March and April, then in the Southern and Western regions in June and July, before beginning to flare in the Midwest in August. We remain far from herd immunity, even in disease hotspots, meaning that renewed vigilance is needed.

Any trends that are hopeful?

TC: The rise and fall of disease in some geographic hotspots like the Northeast region suggest that we are entirely capable of flattening the infection curve with serious commitment to physical distancing and mask use. We can still control this epidemic even before vaccines become available, but this requires clear communication and coordinated action by government.

What was the biggest challenge in creating these models?

TC: The lack of consistent testing and reporting practices have been one of the most difficult challenges we have been grappling with. We also have a relatively small team of researchers working to make this model available; important contributors to this project include, from the T.H. Chan Harvard School of Public Health: Nick Menzies; and from Yale School of Public Health: Melanie Chitwood, Marcus Russi, Kenneth Gunasekera, Joshua Havumaki, Virginia Pitzer, Joshua Warren, and Dan Weinberger.

Are these models available to anyone who wants to use them?

Yes, these models are freely available. We have made our code available, with links through our website and we hope that others use and build on our work. We also provide links on the website which allow users to download files which contain estimates for all states and counties modeled.

How are these models being used by the public and by policymakers?

TC: We don’t quite know yet! Yale has made use of the some of the state-level projections at the beginning of the academic semester and some individuals involved with the CT Reopening Plan have been exploring model use for tracking county-level epidemic behavior. We hope that this new tool can help with intervention planning and evaluation and we will be continuing to invest in model development. Feedback and suggestions would be valuable for us as we try to improve usefulness of covidestim.org

Michael Greenwood

Author: Admin