key: cord-0819169-xab2t2j6
authors: Chiu, Weihsueh A.; Fischer, Rebecca; Ndeffo-Mbah, Martial L.
title: State-level impact of social distancing and testing on COVID-19 in the United States
date: 2020-07-07
journal: Res Sq
DOI: 10.21203/rs.3.rs-40364/v1
sha: 2516a85d626e4be6f716874a141109bfb3ab9d85
doc_id: 819169
cord_uid: xab2t2j6

Social distancing measures have been implemented in the United States (US) since March 2020, to mitigate the spread of SARS-CoV-2, the causative agent of COVID-19. However, by mid-May most states began relaxing these measures to support the resumption of economic activity, even as disease incidence continued to increase in many states. To evaluate the impact of relaxing social distancing restrictions on COVID-19 dynamics and control in the US, we developed a transmission dynamic model and calibrated it to US state-level COVID-19 cases and deaths from March to June 20(th), 2020, using Bayesian methods. We used this model to evaluate the impact of reopening, social distancing, testing, contact tracing, and case isolation on the COVID-19 epidemic in each state. We found that using stay-at-home orders, most states were able to curtail their COVID-19 epidemic curve by reducing and achieving an effective reproductive number below 1. But by June 20(th), 2020, only 19 states and the District of Columbia were on track to curtail their epidemic curve with a 75% confidence, at current levels of reopening. Of the remaining 31 states, 24 may have to double their current testing and/or contact tracing rate to curtail their epidemic curve, and seven need to further restrict social contact by 25% in addition to doubling their testing and contact tracing rates. When social distancing restrictions are being eased, greater state-level testing and contact tracing capacity remains paramount for mitigating the risk of large-scale increases in cases and deaths.

The novel coronavirus pandemic (COVID-19) emerged in Wuhan, China in December 2019 1 and has now reached pandemic status, with spread to more than 210 countries and 2 territories, including the United States (US) 1 . The US reported its first imported case of 3 COVID-19 on January 20, 2020, arriving via an international flight from China 2 . Since then, 4 the disease has spread rapidly within the US, with every state reporting confirmed cases 5 within three weeks of the first reported community transmission. As of June 15 th , the US 6 has exceeded 2.1 million cases and 115,000 deaths, heterogeneously distributed across all 7 states 1 . So far, states such as New York and New Jersey have borne the highest burden with 8 more than 379,000 cases and 30,000 deaths and 166,000 cases and 12,000 deaths, 9

respectively, while Montana and Alaska have each reported less than 700 cases and 20 10 deaths each 1 . 11 COVID-19 is caused by a newly described and highly transmissible SARS-like coronavirus 12 (SARS-CoV-2). Severe clinical outcomes have been observed with approximately 20% of 13 symptomatic cases 3, 4 . There is no vaccine and no cure or approved pharmaceutical 14 intervention for this disease, making the fight against the pandemic reliant on non-15 pharmaceutical interventions (NPIs). These NPIs include: case-driven measures such as 16 testing, contact tracing, and isolation 5 ; personal preventive measures such as hand 17 hygiene, cough etiquette, face mask use, eye protection, physical distancing, and surface 18 cleaning, which aim to reduce the risk of transmission during contact with potentially-19 infectious individuals 6 ; and social distancing measures to reduce interpersonal contact in 20 the population. In the US, social distancing measures have included policies and guidelines 21 to close schools and workplaces, cancel and restrict mass gatherings and group events, 22 restrict travel, maintain physical separation from others (e.g. keeping six feet distance), and 1 stay-at-home orders 7 . 2 NPIs and other responses to COVID-19, especially stay-at-home orders, have varied widely 3 across states, leading to spatial and temporal variation in the timing and implementation of 4 mitigation strategies. This variation in policies and response efforts may have contributed 5 to the observed heterogeneity in COVID-19 morbidity and mortality across states 8 . Recent 6 studies suggest that statewide social distancing measures have likely contributed to 7 reducing the spread COVID-19 epidemic in the US 9, 10 . Understanding the extent to which 8 NPIs, such as social distance, testing, contact tracing, and self-quarantine, influence COVID-9

19 transmission in a local context is pivotal for predicting the future course of the epidemic 10 on a state-by-state basis. This in turn will inform how these NPIs should be optimized to 11 mitigate the spread and burden of COVID-19 while awaiting development of 12 pharmaceutical interventions (e.g. therapeutics and vaccines). 13

After several weeks of statewide stay-at-home orders, most US states have begun to ease 14 their social distancing requirements 11 , while attempting to increase their testing and 15 contact tracing capacities 12 . Mathematical modeling is a unique tool to help answer these 16 important and timely questions. Models can contribute valuable insight for public health 17 decision-makers by providing an evaluation of the effectiveness of ongoing control 18 strategies along with predictions of the potential impact of various policy scenarios 13 . 19 To address these needs, we developed and validated a data-driven transmission dynamic 20 model to evaluate the impact of social distancing, state-reopening, testing, and contact 21 tracing on the state-level dynamics of COVID-19 infections and mortality in the US. We 22 evaluated the transmissibility of COVID-19 in each state from March, 2020 to early June, 1 2020, to estimate the state-level impact of shelter-in-place and reopening on COVID-19 2 transmission. Finally, we evaluated the degree to which increasing testing efforts (rate of 3 identification of infected cases) and/or contact tracing could curtail the spread of the 4 diseases and enable greater relaxation of social distancing restrictions while preventing a 5 resurgence of infections and deaths. 6

We fit our model to state-level daily cases and deaths data using a Bayesian inference 9 approach (see Online Methods). Model performance assessment for several representative 10 states are shown in Figure 1 

Using the posterior distribution of our model parameters we estimated the effective 10 reproduction number from March 19 th to June 20 th , 2020 and identified the minimum 11 level of transmission achieved in each state (Figure 2A ). We found that for all, except two 12 states (Arkansas and Utah), the minimum value was less than 1 and these values were 13 mainly achieved during the state shelter-in-place ( Figure 2A ). On June 20 th , 2020, 27 states 14 had at least a 0.5 probability that >1. Thus, the model predicts that as states are 15 reopening, a majority of states are at risk of continued increases in the scale of the 16 outbreak and require additional mitigation to contain the spread of the disease. disease transmission in each state relative to its minimum value during state shelter-in-place 7

(median, IQR, and 95% CrI). 8 9 We conducted an analysis of variance to evaluate the contribution of each parameter to the 10 variation in value (Table S1 ). Across states, we found that the largest drivers of 11 variation in are the power parameter for social distancing, , the maximum relative 12 increase in contact after shelter-in-place orders, rmax, and the fraction of contact traced, fC, 13 which together contribute over 60% of variance ( Figure S8 ). This observation is consistent 14 with mobility data alone being insufficient to account for the combined effect of multiple 15 control measures, and suggest that the degree of adoption of non-mobility-related 16 measures, such as enhanced hygiene practices and contact tracing, play a large role in the 1 extent to which a state may reduce disease transmission. 2

For each state, we also estimated the current level of reopening/rebound in disease 3 transmission relative to its lowest transmission rate observed during shelter-in-place 4 ( Figure 2B ). We found that only nine states had a 50% or more rebound in COVID-19 5 transmission by June 20 th , 2020 while eight states had a 25% or less rebound in 6 transmission ( Figure 2B ). 7

Bringing and keeping the effective reproduction number, , below 1 is necessary and 9 sufficient to curtail the spread of an outbreak. We evaluated the probability of keeping 10 <1 for different levels of testing and contact tracing under the June 20 th , 2020 level of 11 state reopening. We found that for 12 states and the District of Columbia have at least 12 0.975 probability of keeping <1, and 22 states have less than 0.025 probability of 13 bringing and keeping <1, under their current level of testing and contact tracing 14 ( Figure S9 ). We found that for most states bringing and keeping <1 may not be 15 possible without increase contact tracing efforts as increasing testing and isolation alone 16 would be sufficient or require extremely high coverage to curtail the epidemic curve with a 17 0.975 probability ( Figure S9 ). 18

To evaluate the impact of scaling up testing and contact tracing on the epidemic dynamics 19

in each state, we assumed a linear "ramp-up" of either testing and/or contact tracing from 20

July 1th -15th, 2020, after which both parameters remain constant. We then predicted the 21 daily number of cases and deaths (Figures 3 and S10). We found that under current levels 1 of reopening and control, at least 26 states would see a continuous increase in cases and 2 deaths ( Figure S10 ). Even with increased testing and contact tracing, some of these states 3 will still experience a short-term increase in cases and deaths (Figures 3 and S10) . For 4 example, Ohio, Texas, and Washington may experience a substantial short-term increase of 5 cases and deaths even if their current testing and contact tracing rate were doubled within 6 the next two weeks ( Figure 3B-D) . Moreover, reported cases may slightly increase during 7 the "ramp-up" period ( Figure 3 ). We also found that in most states additional relaxation of 8 restrictions without simultaneously increasing contact tracing may exacerbate disease 9 dynamics and result in large-scale outbreaks ( Figure S10 ). We next evaluated the maximal degree of rebound in transmission (i.e., level of reopening) 5 permitted while keeping <1 under different testing and contact tracing scenarios 6 ( Figure 4 ). We found that under the current level of testing and contact tracing rate, 36 7 states cannot keep their <1 even with only 25% reopening/rebound in transmission 8 ( Figure 4A ). By doubling the current testing rate, four states (Connecticut, North Dakota, 9

Nebraska, Rhode Island) could keep their <1 even with a 75% level of reopening 10 ( Figure 4B ). By doubling contact tracing, five states (Connecticut, Delaware, Maryland, 11

Pennsylvania, Rhode Island) could remove all mobility restrictions while keeping <1 12 ( Figure 4C ). By doubling both testing rate and contact tracing, 12 states could remove all 13 mobility restrictions while keeping <1 ( Figure 4D ). 14 1 We categorized states by the additional amount of mitigation efforts needed to keep < 8 1 with at least 75% confidence (Figures 5 and S10). We found that under current control 9 efforts, three states (Connecticut, North Dakota, Rhode Island) could continue to curtail 10 their epidemic curve even with an additional 25% reopening ("None" category), and that 11 an additional 16 states and the District of Columbia could curtail their epidemic curve 12 without additional reopening ("Low" category). 13 states could curtail their epidemic curve 13 by doubling their current testing or contact tracing rate ("Moderate" category), while for 11 14 states by doubling both testing and contact tracing are need ("High" category). The 15 remaining seven states (Alaska, Hawaii, Georgia, Florida, Oklahoma, Oregon, Montana) 1 require not only doubling both testing and contact tracing, but also additional social 2 distancing restrictions, in order to curtail their epidemic curve ("Very High" Category). There is a delicate and continuous balance to strike between the use of social distancing 10 measures to mitigate the spread of an emerging and deadly disease such as COVID-19 and 11 the need for re/opening various sectors of activities for the social, economic, mental, and 12 physical well-being of a community. To address this issue, it is imperative to design 13 measurable, data-driven, and flexible milestones for identifying when to make specific 14 transitions with regard to easing or retightening specific social distancing measures. We 15 developed a data-driven SARS-CoV-2 transmission dynamic model not only to make short-16 term predictions on COVID-19 incidence and mortality in the US, but more importantly to 17 evaluate the impact that relaxing social distancing measures and increasing testing and 1 contact tracing would have on the epidemic in each state. 2

We showed that in most states, control strategies implemented during their "shelter-in-3

place" period were sufficient to contain the outbreak, defined as reducing and ultimately 4 maintaining the effective reproductive number below 1 ( <1). However, for the 5 majority of states, our modelling suggests that "reopening" has proceeded too rapidly 6 and/or without adequate testing and contact tracing to prevent a resurgence of the 7 epidemic. Even in states with currently decreasing incidence and mortality, such as New 8

York and New Jersey, additional relaxation of restrictions is likely to "bend the epidemic 9 curve upwards." However, our model predicts that a combination of increased testing, 10 increased contact tracing, and/or scaling back reopening will be sufficient for curtailing the 11 spread of COVID-19. Specifically, doubling of current testing and contact tracing rates 12 would enable the vast majority of states to either maintain or increase the easing of social 13 distancing restrictions in a "safe" manner. Increasing testing and contact tracing rates 14 entails both increasing the number of tests performed per day as well as requiring early 15 identification and isolation of COVID-19. This can be accomplished through active case 16 detection via efficient contact tracing strategies. However, it should also be noted that 17 increased testing and contact tracing will lead to a short-term increase in reported cases 18 because a larger fraction of the infected population is being observed, and that several 19 weeks may pass before these rates begin to show a decline. It is therefore imperative that 20 policymakers and the public recognize that such a surge is actually a sign that testing and 21 tracing efforts are succeeding, and to have the patience to wait several weeks before these 22

successes are reflected as declining rates of reported cases. 23

Like all modeling studies, our study has several limitations due to modelling assumptions 1 and the quality of available data. The initiation of social distancing measures, such as stay-2 at-home orders in the US, for mitigating the spread of COVID-19 has occurred concurrently 3 with increased promotion and application of other NPIs such as hygiene practices 4 (e.g. hand hygiene, surface cleaning, cough etiquette, and wearing of face mask). These 5 hygiene practices coupled with the avoidance of physical contact whenever possible 6 (keeping six feet apart) could impact the spread of COVID-19 by reducing both the risk of 7 exposure and the risk of transmission of SARS-CoV-2 from infected patients 14, 15 . Though 8 our model explicitly accounts for the differential contribution of social distancing (mobility 9 reduction) versus hygiene practices and physical distancing to reducing COVID-19 10 transmission, we assume that the impact of hygiene practices and physical distancing was a 11 function of social distancing (mobility reduction). While cell phone mobility data may 12 continue to accurately reflect the contact rates, the impact of enhanced hygiene practices is 13 more difficult to measure independently. As several states are easing their social distancing 14 requirements, especially their stay-at-home orders, compliance with hygiene practices 15 would become even more important for reducing individuals' risk of getting or 16 transmitting the pathogen. However, keeping a high population-level adherence to these 17 measures is required to mitigate the spread of the COVID-19 epidemic in a city, state, or 18 nation 16 . As states are reopening various aspects of their economy, data on compliance 19 with enhance hygiene practices and physical distancing are needed to improve the 20 estimation of these measures' population-level impact on reducing disease transmission. 21

Additionally, consistent with previous COVID-19 modeling studies 17-19 , our model uses a 22

simple functional form to model increases in testing rate from early March to June, 2020. 23

This testing rate was estimated through model fitting to daily reported case and mortality 1 data. Particularly in states that have seen a substantial increase in testing capability and 2 efforts during the month of May, our simple time varying assumption may underestimate 3 the current level of testing and contact tracing. However, it should be noted that increased 4 testing capacity does not necessarily lead to increased rate of testing if individuals are 5 unaware, unwilling, or unable to be tested 20 . Having contact tracing and date of symptoms 6 onset data would enable us to compute a better estimate the current testing and contact 7 tracing rate in each state. Our also model assumes that all individuals who test positive to 8 COVID-19 are effectively isolated for the rest for their infectious period and no longer 9 contribute to disease transmission. Though voluntary compliance to COVID-19 self-10 quarantine recommendations may be high across the US, it is likely not 100%. Therefore, 11

the assumption of effective isolation of all identified cases may cause our model to slightly 12 overestimate the impact of increase testing rate on disease dynamics. However, we 13 anticipate that this assumption would only have a marginal impact on the qualitative 14 nature of our results. Finally, our model does not explicitly account for age-stratified risk 15 of disease transmission and mortality. This age-stratification is important for designing and 16 evaluating social distancing and testing strategies that are targeted towards the elderly 17 population which are at higher risk of COVID-19-induced hospitalization and death 21 . As 18 reopening the economy becomes an imperative for states across the US, age-or risk-19 targeted interventions may be a valuable tool to mitigate the burden of the pandemic. 20

Future modeling studies could investigate the effectiveness of age-or risk-targeted non-21 pharmaceutical and potential pharmaceutical (vaccine or therapeutic) interventions for 22

controlling the spread and burden of COVID-19. 23

In sum, we use a data-driven mathematical modeling approach to study the impacts of 1 social distancing, testing, and contact tracing on the transmission dynamics of SARS-CoV-2. 2 Our findings emphasize the importance for public health authorities not only to monitor 3 the case and mortality dynamics of SARS-CoV-2 in their state, but also to understand the 4 impact of their existing social distancing measures on SARS-CoV-2 transmission and 5 evaluate the effectiveness of their testing and contact tracing programs for promptly 6 identifying and isolating new cases of COVID-19. As reported case rates are increasing 7 widely across US states because social distancing restrictions have been eased to allow 8 more economic activity to resume, we find that most states need to either significantly 9 scale back reopening or enhance their capacity and scale of testing, case isolation, and 10 contact tracing programs in order to prevent large-scale increases in COVID-19 cases and 11 deaths. 12

Our overall approach is as follows: 1) develop a mathematical model (an SEIR-type 14 compartmental model) that incorporates social distancing data, case identification via 15 testing, isolation of detected cases, and contact tracing; 2) assess the model's predictive 16 performance by training (calibrating) it to reported cases and mortality data from March 17 19 th to April 30 th , 2020 and validating its predictions against data from May 1 st to June 20 th , 18 2020; and 3) use the model, trained on data through June 20 th , 2020, to predict future 19 incidence and mortality. The final stage of our approach predicts future events under a set 20 of scenarios that include increased case detection though expanded testing rate, contact 21 tracing, and relaxation or increase of measures to promote social distancing. All model 22 fitting is performed in a Bayesian framework in order to incorporate available prior 1 information and address multivariate uncertainty in model parameters. 2

Our model is illustrated in Figure 1 , with parameters and prior distributions listed in Table  4 1. We modified the standard SEIR model to address testing and contact tracing. In our 5 model formulation I class also includes infectious pre-symptomatic individuals. With 6 respect to testing, separate compartments were added for untested, "freely roaming" 7

infected individuals ( ), tested/isolated cases , fatalities . In balancing considerations 8 of model fidelity and parameter identifiability, we made the reasonably conservative 9 assumptions that all tested cases are effectively isolated (through self-quarantine or 10 hospitalization) and thus unavailable for transmission, and that all COVID-related deaths 11 are identified/tested. 12

With respect to contact tracing, the additional compartment represents unexposed 13 contacts, who undergo a period of isolation during which they are not susceptible before 14 returning to ; while and represent contacts who were exposed. Again, the reasonably 15

conservative assumption was made that all exposed contacts undergo testing, with an 16 accelerated testing rate compared to the general population. We assume a closed 17 population of constant size for each state. 18

The ordinary differential equations governing our model are as follows: 19

The testing rates and , the fatality rate , and the recovery rate of traced contacts are 9 each composites of several underlying parameters. The testing rate defined as 10

where

,0 is the current testing coverage (fraction of infected individuals tested), 12 is the test sensitivity (true positive rate), and is rate of testing for those 13 tested, with a typical time-to-test equal to 1/ . The time-dependence term models the 14 "ramp-up" of testing using a logistic function with a growth rate of 1/ days −1 , where is 15 the time where 50% of the current testing rate is achieved. Similarly, for testing of traced 16 contacts, the same definition is used with the assumption that all identified contacts are 17 tested, ,0 = 1 and at a faster assumed testing rate , : 18

Because all contacts are assumed to be tested, the rate at which they enter the 1 "recovered" compartment is simply the rate of false negative test results: 2

The fatality rate is adjusted to maintain consistency with the assumption that all COVID-19 4 deaths are identified, assuming a constant infected fatality rate ( ). Specifically, we first 5 calculated the fraction of infected that are tested and positive 6

Where is the fraction of contact identified through contact tracing. 8

Then the case fatality rate ( ) = / ( ). Because the = /( + ), this 9 implies 10

The model is "seeded" cases on February 29, 2020. Because in the early stages of 12 the outbreak, there may be multiple "imported" cases, we only fit to data from March 19, 13 2020 onwards, one week after the U.S. travel ban was put in place 22 . 14 Our model is fit to daily case and death data (cumulative data are not used for fitting 15 because of autocorrelation). To adequately fit the case and mortality data, we accounted for 16 two lag times. First, a lag is assumed between leaving the compartment and public 17

reporting of a positive test result, accounting for the time it takes to seek a test, obtaining 18 testing, and have the result reported. No lag is assumed for tests from contact tracing. 19

Second, a lag time is assumed between entering the fatally ill compartment and 20 publically reported deaths. Additionally, we use a negative binomial likelihood in order to 1 account for the substantial day-to-day variation in reporting results. The corresponding 2 equations are as follows: The term ( ) is 0 before , linear between and , and constant at a value of 14 after that, and made continuous by approximating the Heaviside function by a logistic 15 function. The reopening time is defined as days after , and the maximum relative 16 increase in contacts happens days after that. 17

We selected the functional form above for ( ) because it was found to be able to represent 18 a wide variety of social distancing data, including cell phone mobility data from Unacast 24 19 and Google 25 , as well as restaurant booking data from OpenTable 26 . We used these 20 different mobility sources to derive state-specific prior distributions because different 21 social distancing datasets had different values for , , , , , and ( Figure S2 ). 22

With respect to the reduction in transmission probability , we assumed that during the 1 "shelter-in-place" phase, hygiene-based mitigation paralleled this decline with an 2 effectiveness power , and that this mitigation continued through re-opening. 3

Finally, we define an overall "reopening" parameter that measures the "rebound" in 4 disease transmission ⋅ relative to its minimum, defined to be 0 during shelter-in-place 5 Finally, for (c), we additionally evaluated changes in reopening → + for values of 16 +25% or -25%, for a total of 12 scenarios (4 different levels of testing and tracing, and 3 17 different levels of reopening). We then ran the SEIR model forward in time until August 31, 18 2020. For all three intervention parameters , , and , we assumed a "ramp-up" period 19 of 2 weeks from July 1-July 14, 2020. To summarize the relative urgency of mitigation in 20 each state, we categorized states based on which scenarios resulted in the IQR of ( ) 

Johns Hopkins Coronavirus Resource Center

First Case of 2019 Novel Coronavirus in the United States

Clinical course and mortality risk of severe COVID-19. The 6

Characteristics of and Important Lessons from the 8

COVID-19) Outbreak in China: Summary of a Report of 9 72314 Cases from the Chinese Center for Disease Control and Prevention

Effectiveness of isolation, testing, contact tracing and physical 12 distancing on reducing transmission of SARS-CoV-2 in different settings

Physical distancing, face masks, and eye protection to prevent person-15 to-person transmission of SARS-CoV-2 and COVID-19: a systematic review and meta-16 analysis

What is social distancing and how can it slow the spread of COVID-19? | Hub

Geographic Differences in COVID-19 Cases, Deaths, and Incidence -21

United States

Social distancing to slow the U.S. COVID-19 epidemic: an 24 interrupted time-series analysis

Social Distancing Has Merely Stabilized COVID-19 in the US

Lifting Social Distancing Measures in America: State Actions & Metrics | The Henry J

Kaiser Family Foundation

Mathematical 35 models in the evaluation of health programmes

Respiratory virus shedding in exhaled breath and efficacy of face 1 masks

Face masks for the 3 public during the covid-19 crisis

To mask or not to mask: Modeling the potential for face mask 5 use by the general public to curtail the COVID-19 pandemic

The impact of changes in diagnostic testing practices on estimates 8 of COVID-19 transmission in the United States

Projection of COVID-19 Cases and Deaths 11 in the US as Individual States Re-open

Outbreak dynamics of COVID-19 14 in China and the United States

states see COVID-19 testing supply improvements, but challenges abound -The 17

Predictions of COVID-19 dynamics in the UK: short-term 22 forecasting and analysis of potential exit strategies

The COVID Tracking Project | The COVID Tracking Project

State of the Industry | OpenTable

Bayesian statistical inference for SBML-coded systems 36 biology models

Inference from Iterative Simulation Using Multiple 1 Sequences

RStudio | Open source & professional software for data science teams -RStudio

The R Project for Statistical Computing

The Incubation Period of Coronavirus Disease

From Publicly Reported Confirmed Cases: Estimation and Application

Temporal dynamics in viral shedding and transmissibility of COVID-19

Social contacts and mixing patterns relevant to the spread of 14 infectious diseases

Estimates of the severity of coronavirus disease 2019: a model-based 16 analysis

Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-18 Infected Pneumonia

Pattern of early human-to-human transmission of Wuhan 20 2019 novel coronavirus (2019-nCoV)

Reconciling early-outbreak estimates of the basic reproductive 23 number and its uncertainty: a new framework and applications to the novel 24 coronavirus (2019-nCoV) outbreak

Presymptomatic SARS-CoV-2 Infections and Transmission in a 29 Skilled Nursing Facility

Interpreting a covid-19 test result

Epidemiology, clinical course, and outcomes of critically ill 35 adults with COVID-19 in New York City: a prospective cohort study

Early epidemiological analysis of the coronavirus disease 2 2019 outbreak based on crowdsourced data: a population-level observational study

Case-Fatality Rate and Characteristics of Patients 5 Dying in Relation to COVID-19 in Italy

Map: How states are reopening after coronavirus shutdown -Washington Post