Abstract
Background The United States has the highest numbers of confirmed cases of COVID-19, where they took up nearly half in the hot spot states of New York, New Jersey, Connecticut, and California. The workforce in these states was required to work from home except for essential services. It is necessary to evaluate an appropriate date for resumption of business since premature reopening of economy will lead to broader spread of COVID-19, while the opposite situation would cause greater loss of economy.
Methods To consider pre-symptomatic and asymptomatic transmission of COIVD-19, it is crucial to evaluate the unobserved numbers of unidentified infectious individuals but not the observed number of confirmed cases, which reflect the real-time risks of different stage of infectious disease. We proposed an epidemic model in considering the pre-symptomatic transmission and asymptomatic transmission of COVID-19 to evaluate the real-time risk of epidemic for the states of New York, New Jersey and Connecticut, and compared with California state (where it effectively phased reopened on May 8) for assessments of the appropriate Monday for resumption of business.
Results The predicted numbers of unidentified infectious individuals per 100,000 for states of New York, New Jersey and Connecticut which are close to those in California state are 12.23 with 95% CI (10.68, 13.57) on June 1, 25.65 with 95% CI (20.04, 30.43) on June 15, 28.49 with 95% CI (19.10, 38.65) on June 22, respectively, which may cause 11.23%, 15.64% and 17.32% higher than the estimated number of cumulative confirmed cases on July 11, if the second wave of the infection would happen after people return to work.
Conclusions It may be feasible for states of New York, New Jersey and Connecticut to reopen business on June 1 (or even May 18), June 15 and June 22. For the period after resumption of work, if the number of unidentified infectious individuals is still non-zero, the risks for the second wave of the infection would never vanish.
Introduction
The outbreak of novel coronavirus disease (COVID-19) has spread over 200 countries since December 2019[1]. It is unprecedented to have over 7 million cumulative confirmed cases of COVID-19 worldwide[2]. The “battle” against COVID-19 in China has provided experience and likely outcomes of interventions to the ongoing hard-hit areas. As a novel and acute infectious disease, the transmissibility of COVID-19 was unclear at the early stage of epidemic, the Chinese government implemented relatively strict nonpharmaceutical interventions in the hot spot areas, where the public transportations were suspended within and outside of the city in Hubei province since January 23, 2020[3]. All nationwide residents were recommended to stay at home except for essential needs. The holiday season of the Chinese Spring Festival had been prolonged to late February when essential services were recommenced operating gradually outside Hubei province[4]. In April, a comprehensive resumption of business is approaching in China[5].
Epidemic situations in the United States
Meanwhile, the confirmed case of COVID-19 was reported since late January 2020 in the United States[6]. There were over 1000 cumulative confirmed cases on March 13, 2020[7], when the White House declared a national emergency concerning COVID-19 outbreak[8] and issued a “call to action” coronavirus guidelines on March 16, 2020[9]. The United States has become the most severe country of COVID-19 with 366,346, 154,154, 40,468, and 94,743 cumulative confirmed cases in the states of New York, New Jersey, Connecticut, and California until May 24, 2020[10], respectively. Making things worse, New York and California are the two states that contribute the most to the real gross domestic product (GDP) over years in the United States (Figure 1)[11].
The state of New York had the first confirmed case of COVID-19 reported on March 1, 2020, and it proclaimed an executive order on March 16, 2020, including reducing half of the local government workforce, allowing the statewide nonessential workforce to work from home starting on March 17, 2020, and closing all schools starting on March 18, 2020[12]. However, as the rapidly increasing number of additional cases of COVID-19 in the state, the governor announced an aggressive policy of “New York State on PAUSE” on March 20, 2020[13], and required all people in the state to wear masks or face covering in public since April 15, 2020[14]. On May 1, all statewide K-12 schools and college facilities continued to close for the remaining academic year[15]. The guide of “NY Forward Reopening” Plan was available on May 11, 2020 and there were three regions reopened businesses for phase one on May 15, 2020[16]. Until May 24,2020, there had been seven regions reopened for businesses[17].
The state of New Jersey had the first positive case of COVID-19 reported on March 4, 2020[18]. The governor of New Jersey recommended the cancellation of statewide public gatherings over 250 individuals since March 12, 2020, suspended visiting state prisons and statewide halfway houses since March 14, 2020, and closed restaurants, bars, movie theaters, gyms, casinos since March 16, 2020[19]. The governor announced an order including the statewide stay at home and closure of statewide non-essential retail industries on March 21, 2020[20]. States park and gold courses were effectively reopened on May 2, 2020[21]. The car gathering provision was permitted resumption on May 12, 2020 and construction and non-essential retail provisions were resumed on May 18, 2020[22]. In-person sales were authorized to reopen at car, motorcycle, and boat dealerships and bike shops by appointment only and with social distancing measure on May 20,2020[23]. Three states of New York, New Jersey and Connecticut signed a multi-state agreement on beaches reopened on May 22, 2020[24].
The state of Connecticut had the first positive case as a Connecticut resident on March 8, 2020[25]. Three states of New York, New Jersey, and Connecticut announced measures to slow the spread of COVID-19 throughout the tri-state area on March 16, 2020[26]. Governor signed an executive order for businesses and residents “stay safe, stay home” on March 20, 2020[27]. Statewide K-12 schools which remained closed for the rest of the academic year were announced on May 5, 2020[28]. The restaurants, offices, retail stores, outdoor museum and zoos as the first phase were authorized to reopen on May 20, 2020[29] and hair salons and barbershops were aligned to reopen in Early June[30].
The state of California had two positive cases of COVID-19 reported on January 26, 2020[31]. One month later, there was the first possible case of local transmission of COIVD-19 in California[32]. The “stay home except for essential needs” order was issued on March 19, 2020, where all individuals were required to stay at home except for the workforce in 16 critical infrastructure sectors[33]. On May 7, 2020, the state public health officer determined the entire state gradually moved to Stage 2 of California’s Pandemic Resilience Roadmap, i.e. reopening the lower-risk workplaces and other spaces[34]. Therefore, California state had phased reopened on May 8.
The timelines of four states (New York, New Jersey, Connecticut and California) of intervention details are shown in Figure 2. According to the capacity of medical facilities in the website of COVID-19.direct[10], there could be 52200 beds and 219,000 hospitals, 25,300 beds and 5,300 hospitals, 8,200 beds and 46 hospitals, and 87,800 beds and 463,000 hospitals in the states of New York, New Jersey, Connecticut, and California, respectively.
On April 13, 2020, there were 7 states joined together to form a multi-state council to combat COVID-19 and to help reopen the economy including the states of New York, New Jersey, and Connecticut[35]. As the epidemic of COVID-19 was getting worse in the United States, the state governors issued a series of interventions to slow the spread of virus, including requiring all individuals to stay at home and reducing the workforce in the industries. The whole social system was slowed down and the employment in total nonfarm was fell over the month change (Figure 3), where the unemployment rates were 14.7% and 13.3% in April and May, 2020[36], and thus, the time to restore the economy in the United States, especially for the states of New York, New Jersey, Connecticut, has been the most significant and consequential decision for the president of the United States and the governors of the states. To recover the economy for people to return to work, the key concern is the risk and severity of the second wave of the infection.
Methods
Data collection
To emphasize the effectiveness of our model for evaluating the real-time risk of epidemic before the second wave of the infection, we focused on the data during the period before the protest in United States[37]. We collected the epidemic data from March 13, 2020 when the national emergency concerning COVID-19 was proclaimed to May 24, 2020 in four states of New York, New Jersey, Connecticut, and California. The data are available on the New York Times[38].
Model
Based on the report of WHO, the transmission of COVID-19 could be caused by the individuals infected with the virus before significant symptoms developed[39], or even the carriers who did not develop symptoms. Pre-symptomatic transmission and asymptomatic transmission would interfere with our judgment on the current magnitude of outbreak of infectious diseases because of the existence of the duration between catching virus and being confirmed by testing. To consider this nature of COIVD-19, it is crucial to evaluate the unobserved number of unidentified infectious individuals but not the observed number of confirmed cases.
To understand the mechanism of COVID-19, we divided a concerned population into four compartments: susceptible (S), unidentified infectious (I), self-healing without being confirmed (H), and confirmed cases (C).
The susceptible (S) individuals, who have no immunity to the disease, are the majority of the population at an early stage of the epidemic.
Unidentified infectious (I) individuals are infectious individuals without being confirmed, which consist of two kinds of individuals: infectors who would have symptoms eventually and asymptomatic carriers who would not develop symptoms. We assumed that I would transfer to H or C eventually.
Self-healing without being confirmed (H) individuals are assumed to be no longer infectious and resistant to COVID-19.
Confirmed cases (C) contain two kinds of individuals: patients in hospital and asymptomatic carriers who are assumed to stay at home, and unable to transmit the disease.
We introduced Susceptible (S) - Unidentified infectious (I) - Self-healing without being confirmed (H) - Confirmed cases (C) (SIHC) model with four compartments: Susceptible (S), unidentified infectious (I), self-healing without being confirmed (H), confirmed cases (C). We divided I into two part: asymptomatic carriers without being confirmed (I1), infectors who would have symptoms eventually (I2). Similarly, we divided C into two part: confirmed cases who self-quarantine or self-isolate (C1), confirmed patients at hospitals (C2). The transfer diagram of S, I1, I2, H, C1, C2 are given in Figure 1 in the Supplementary.
The dynamic system of numbers of compartments S, I, H, and C at the time t: S(t), I(t), H(t), and C(t) are assumed as: where ρ is transmissibility and θ(t) is the average contact numbers per person at the time t[40], which is assumed to be time-varying; DH is the average length of duration between catching the virus and self-healing without being confirmed. Dc is the average length of duration between catching virus and being confirmed by testing; N is the total number of population.
Let α(t) = ρ θ (t). Generally, θ (t) is controlled by policy interventions, which is a constant at an early stage of the epidemic and decreases along with the implementation of interventions until the control policies are completely carried out, i.e., reaching the lowest level. Based on this point of view, we furthermore assumed α(t) is monotonically decreasing curve[41] with four parameters α0, d, m, and η: where α0 denotes the maximum of α(t) at an early stage of outbreak of infectious disease, d is the time when the control measures start to be effective and the α(t) starts to decline, m represents the duration of decreasing process of α(t), η is the maximum decreasing proportion of α(t) comparing with α0, i.e., the minimum of α(t) is α0 (1 − η). λm is chosen as and ε is fixed to be 0.01. The graph of α(t) is given in Figure 2 in the Supplementary.
By the setting above, we can derive the time-varying reproductive rate Rt[42, 43] by SIHC (susceptible, unidentified infectious, self-healing without being confirmed, and confirmed cases) model:
Let Ω = (α0, η, m, Dc) and Θ = (Ω, d, DH), we introduced randomness to the epidemic model above: Let St, It, Ht, Ct be the observed values of S, I, H, C at the time t with constraint: and At = (It, Ht, Ct) is assumed to be a multi-dimensional Markov process, i.e., for g < t and given Hg: where At(Ag, Θ) is the evolve values of I, H and C at the time t determined by the deterministic dynamic system (1) with the initial value (Sg, Ig, Hg, Cg) and parameters Θ. Poiss(· |v) is the density of multi-dimensional independent Poisson distribution with mean vector v. Similarly, we define It(Ag, Θ), Ct(Ag, Θ), Ht(Ag, Θ), which are the corresponding values at time t determined by the deterministic dynamic system (1) with the initial value (Sg, Ig, Hg, Cg) and parameters Θ.
Let α(t) = α(k), k ≤ t < k + 1, k is an integer. For simplicity, we could assume that Ht (g < t) is not random given the initial value (Sg, Ig, Hg, Cg) and parameters Θ, i.e., Ht|Ag =: Ht(Ag, Θ) a.s., if we know DC.
Notice that the observable data is C1:T, N, where T is the length of time. Assume that H1 = 0 and treat I1:T and H2:T as the missing data, we applied techniques of Markov Chain Monte Carlo (MCMC)[44-47] for approximations of posterior distributions of parameters Ω and unknown numbers of I, H at each time by given priors with values of d, and DH. Once the prior distribution of Ω: π(Ω), was given, we apply the Bayesian method combined with MCMC (Markov Chain Monte Carlo) to simulate the posterior distribution:
For the choice of prior distributions, notice that DC is governed by the mean of incubation period, in terms of studies about the incubation period of COVID-19[48], we choose log-normal distribution with log-mean log (5.1) as the prior distribution of DC. Similarly, we choose log-normal distribution with log-mean log (30) as the prior distribution of m, which implies a priori hypothesis that the control policies are completely carried out one month later, while the priors of other parameters are chosen non-informative or flat priors.
For the fixed parameters d, m, and DH, d is taken the beginning of implementation of important policy interventions, i.e., d = 8 for New York state (“New York State on PAUSE” in Figure 2 in the main text), d = 9 for New Jersey state (“statewide stay at home” in Figure 2 in the main text), d = 8 for Connecticut state (“stay safe, stay home” in Figure 2 in the main text), d = 7 for California state (“stay home except for essential needs” in Figure 2 in the main text). DH is assumed to be 9.5 according to the clinical study of asymptomatic cases[49]. The point estimates of Ω, I1:T and H2:T were presented as the median of the posterior distribution while 95% credible intervals were constructed with 2.5% and 97.5% quantiles.
Moreover, since π(Ω, I1:T, H2:T |C1:T, N, d, m, DH. H1) has no close form, we apply Gibbs sampler imbedded Metropolis-Hastings steps for updating parameters. Let be the k-th sample, given initial , the details of updating are given in Algorithm S1 and Algorithm S2 in Supplementary.
Results
To assess the risk and severity of the second wave of the infection after the resumption of business, it is important to consider the unobserved numbers of unidentified infectious individuals in the community. Our approach is based on publicly available data to simulate the possible outcomes of an outbreak of COVID-19 at different business reopening dates and to provide recommendations as to when it is reasonable safe for people back to work.
The estimated results for four states are given in Figures 3 in the Supplementary. The estimated cumulative numbers of confirmed cases on July 11 are 395638 with 95% CI (391461, 399830) for New York state, 179220 with 95% CI (173988, 185356) for New Jersey state, 52720 with 95% CI (49306, 56948) for Connecticut state, and 213263 with 95% CI (192476, 232771) for California state, which estimated proportions of cumulative infected individuals without being confirmed (including I and H) are 20.11% with 95% CI (19.66%, 20.65%), 32.57% with 95% CI (31.77%, 34.48%), 34.72% with 95% CI (33.03%, 37.86), and 36.85% with 95% CI (35.33%, 38.62%), respectively. Normally, higher testing rate would lead to smaller proportion of cumulative infected individuals without being confirmed. The estimated proportions were consistent with the reported totally test per 100,000 of four states (9312, 7434, 6321, and 4396 for states of New York, New Jersey, Connecticut, and California, respectively)[50].
Focusing on states of New York, New Jersey, and Connecticut, the predicted numbers of daily new infected individuals and new confirmed cases, are given in Figure 4 and the details of infectious diseases maps for states of New York, New Jersey, and Connecticut are given in Figures 4-6 in the Supplementary and the daily trajectories of COVID-19 in combined three states were shown in an additional graphics interchange format (gif) file 1. Note that there are gaps between the numbers of daily new confirmed cases and new infected individuals per day in Figure 4, which is caused by the pre-symptomatic transmission and asymptomatic transmission of COVID-19. We called this phenomenon as lagging effect of daily new confirmed cases.
For considering the potential risk of infectious disease, we present the estimated numbers of unidentified infectious individuals per 100,000 for states of New York, New Jersey, Connecticut, and California, which are given in Figure 5. Based on Figure 5, the estimated timing of the peak of unidentified infectious individuals in the states of New York, New Jersey, and Connecticut had passed. Compared with California state, where it effectively reopened the economy on May 8, the numbers of unidentified infectious individuals in the states of New York, New Jersey and Connecticut was higher and showed different downward patterns.
We considered the possible timing for resuming business in the different Mondays. The predicted numbers of unidentified infectious individuals per 100000 for states of New York, New Jersey, Connecticut, and California on May 11, May 18, May 25, June 1, June 8, June 15, June 22 and June 29 are given in Table 1. The dates that the predicted numbers of unidentified infectious individuals per 100000 for states of New York, New Jersey and Connecticut close to those in California state are May 18, June 15 and June 22, respectively.
To realize the potential risk of resuming business on different Mondays, we simulated the possible outcomes that the second wave of the infection would happen after people return to work. There could be maximum contact degree once all individuals back to work, and then the stringent interventions were implemented to curb the wave of infectious disease a week later. In doing so, for the starting days of June 1, June 8, June 15, June 22 and June 29, we modified the estimated by estimated to last 7 days and specifying to its minimum value afterward, i.e., representing the most intensity of control policies for states of New York, New Jersey, and Connecticut accordingly. The simulated results are given in Figure 6.
For New York, the predicted numbers of cumulative confirmed cases per 100000 on July 11 for the simulated outbreak of epidemic when people would return to work on June 1, June 8, June 15, June 22 and June 29 are 11.23%, 7.65%, 5.02%, 3.10%, and 1.68% higher than the estimated number of cumulative confirmed cases per 100000 on July 11 if no resumption of business.
For New Jersey, the predicted numbers of cumulative confirmed cases per 100,000 on July 11 for the simulated outbreak of epidemic when people would return to work on June 1, June 8, June 15, June 22 and June 29 are 31.05%, 22.51%, 15.64%, 10.13%, and 5.73% higher than the estimated number of cumulative confirmed cases per 100000 on July 11 if no resumption of business.
For Connecticut, the predicted numbers of cumulative confirmed cases per 100000 on July 11 for the simulated outbreak of epidemic when people would return to work on June 1, June 8, June 15, June 22 and June 29 are 46.44%, 35.27%, 25.63%, 17.32%, and 10.27% higher than the estimated number of cumulative confirmed cases per 100000 on July 11 if no resumption of business.
Discussion
To understand the spread of COVID-19 for the states of New York, New Jersey, Connecticut, we applied an epidemic model considering pre-symptomatic and asymptomatic transmission of COVID-19 to estimate the numbers of daily new infected individuals and new confirmed case for these states. Normally, the numbers of new infected cases mirrored the actual degree of transmission, while the numbers of new confirmed cases, often lagged the real situation by the existence of the duration between catching virus and being confirmed by testing. According to Figure 4, such a lagging effect could not be neglected at an early stage of epidemic, would tend to disappear over time.
For the control of epidemic, it was reasonable to alleviate the lagging effect in the process of epidemic, which could be achieved by shortening the Dc in SIHC model. Notice that Dc was also affected by the testing rate, except the length of incubation period. The estimated of all four states were smaller than the reported average length of incubation period: 5.1[48] (2.37 with 95% CI (2.32, 2.44) for New York, 4.53 with 95% CI (4.31, 4.89) for New Jersey, 4.94 with 95% CI (4.55, 5.46) for Connecticut, and 4.99 with 95% CI (4.61, 5.37) for California), it implied the policy enhancing the testing rate may be implemented in all four states, which was effective to reduce the number of unidentified infectious individuals.
The decision for the resumption of business is not only a public health issue, but also economic issue. What we are focusing on is the epidemiological feasibility of returning to work at an early date, there are still many other factors needed to be paid attention[51]. The higher numbers of unidentified infectious individuals were obtained, the higher risks for the resumption of business would have been. According to Figure 6, we concluded that there could be risks for the resumption of businesses on June 1 for the states of New York, New Jersey and Connecticut. If governors of states delayed one week, or more, for the resumption of businesses, that would significantly scale down the simulated magnitude of the second wave of the infection. But the benefit of delaying decision would be smaller and smaller when the current control policies continue to be effective.
Conclusion
Due to the fact that California state had been in process of reopening economy on May 8, compared with it, it may be feasible for states of New York, New Jersey and Connecticut to reopen business on June 1 (or even May 18), June 15 and June 22, which may cause 11.23%, 15.64% and 17.32% higher than the estimated number of cumulative confirmed cases per 100,000 on July 11, if the second wave of the infection would happen after people return to work. Meanwhile, for the period after resumption of work, if the number of unidentified infectious individuals is still non-zero, the risks for the second wave of the infection would never vanish, which underscore the necessity of maintaining epidemic prevention measures, such as, wearing a mask and keeping a social distance.
Data Availability
To emphasize the effectiveness of our model for evaluating the real-time risk of epidemic before the second wave of the infection, we focused on the data during the period before the protest in United States. We collected the epidemic data from March 13, 2020 when the national emergency concerning COVID-19 was proclaimed to May 24, 2020 in four states of New York, New Jersey, Connecticut, and California. The data are available on the New York Times.
https://www.mndaily.com/article/2020/06/pf-the-george-floyd-protests-a-visual-timeline
https://www.nytimes.com/interactive/2020/us/coronavirus-us-cases.html
Declarations
Ethics approval and consent to participate
Not applicable
Consent for publication
Not applicable
Availability of data and materials
The data that support the findings of this study are available from the New York Times
Competing interests
The authors declare that they have no competing interests
Funding
Not applicable
Authors’ contributions
TT, XW and HZ developed the idea and research. TT, JT and YJ wrote the first draft of the manuscript and all other authors discussed results and edited the manuscript. TT, JT and YJ collected and validated epidemiological data and census data. All authors read and approved the final manuscript
Additional files
Additional graphics interchange format (gif) file 1: A COVID-19 distribution map of three states of New York, New Jersey, and Connecticut. The daily cumulative confirmed cases from March 8 to June 1, 2020 were presented.
Acknowledgements
We would like to thank all individuals who are collecting epidemiological data of the COVID-19 outbreak around the world.