By Lauren M. Childs
The early weeks and months of 2020 were overshadowed by the rapidly spreading novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that originated in Wuhan, China, in December 2019. By the beginning of April, the coronavirus disease (COVID-19) epidemic had reached nearly every country in the world, with well over a million cases and more than 50,000 deaths [3]. Because the epidemic’s spread is unprecedented in our time, there is much contention regarding the best containment and mitigation strategies. As with most emerging infections, many preventive measures (e.g., vaccines) and curative approaches (e.g., drugs) are not yet available; therefore, countries must rely on non-pharmaceutical interventions—such as individual quarantine—to combat the illness.
Choosing the best intervention strategy is critical during an emerging epidemic, especially when uncertainties surround the disease and health systems are not adequately prepared. Public health officials often initially employ contact tracing, wherein people possibly exposed to the disease become the focus of non-pharmaceutical interventions due to their enhanced risk. These contacts are promptly isolated if they are symptomatic. Often, however, they may appear uninfected when located. Depending on the time since exposure, these individuals might not yet be symptomatic — though they could develop symptoms in the future.
The best way to manage symptom-free contacts is highly disputed. History has relied upon two main strategies: individual quarantine and active symptom monitoring. Similar to isolation, individual quarantine involves the separation of potentially infected individuals from other people; this requires resources to provide necessities, access to secluded spaces, and means of enforcement. In contrast, active symptom monitoring allows individuals to essentially go about their normal routines while checking for symptoms at regular intervals, possibly with daily visits from healthcare workers or more technological-based self-monitoring. Under either intervention, patients are promptly isolated upon detection of symptoms. While individual quarantine is the more effective strategy by definition, it is also costlier and considerably more restrictive.
The success of these strategies relies on fundamental information about the disease’s natural history: most importantly, the timing of when symptoms arise and when individuals are able to transmit the illness. The period between exposure and the onset of symptoms is known as the incubation period, while the period between exposure and infectiousness—i.e., the ability to transmit—is called the latent period (see Figure 1). For many illnesses, symptoms and infectiousness occur at roughly the same time; these two terms are therefore often confused. In order to determine which strategy is most appropriate for mitigating the spread of emerging infections, my colleagues and I built a framework to compare individual quarantine and active symptom monitoring approaches. Our model considers a range of feasibility parameters, including delays in contact tracing, imperfect isolation, and missed contacts [6]. As with many countries during the initial stages of COVID-19, insufficient or nonexistent testing capabilities may also hinder feasibility.
Figure 1. Timing of non-pharmaceutical interventions compared to disease progression. Following the infection event, individuals are noninfectious (green) until the onset of infectiousness, after which they can transmit the disease (orange). They are unobserved prior to and during contact tracing. Once an intervention—individual quarantine or active symptom monitoring—begins, individuals are isolated upon detection of symptoms. The gray area depicts the time before symptom onset. 1a. For diseases where symptom onset occurs after the onset of infectiousness, individuals can transmit prior to isolation (red triangle). 1b. For diseases where symptom onset occurs before infectiousness, no transmission is possible prior to isolation. Figure courtesy of Lauren Childs.
We used a discrete-time stochastic branching model—wherein individuals progress through a susceptible-exposed-infectious-recovered (SEIR) disease process—and focused on the early stages of an epidemic [6, 7]. We utilized data on distributions for the incubation period and serial interval to determine the timing of this progression. Once an individual was symptomatic and thus isolated, we traced a proportion of his/her contacts with a time lag that accounts for the possibility of delays — as would occur during a real epidemic. While infectious, patients generate new infections following a negative binomial distribution with a dispersion parameter that allows for variability in infectiousness [6, 8]. The basic reproductive number \((R_0)\) and the fitted distribution of infectiousness dictate the number of new infections in a given time period. Based on published results for distributions of the incubation period and serial interval, we used particle filtering—i.e., a sequential Monte Carlo algorithm—to fit the maximum duration of infectiousness, time of peak infectiousness, and time offset between the incubation and latent periods.
This framework allowed us to examine the extent to which individual quarantine and active symptom monitoring can mitigate or control an epidemic; we do so by determining the effective reproductive numbers \((R_\rm{eff})\) under each intervention in high and low feasibility settings [6]. For diseases where symptoms emerge prior to or at the onset of infectiousness—like Ebola, pertussis (whooping cough), SARS, and Middle East respiratory syndrome—individual quarantine yields little additional benefit over active symptom monitoring. However, the absolute ability to control a disease, i.e., driving \(R_\rm{eff}<1\), under either scenario is a combination of the basic reproductive number and the feasibility setting. For example, pertussis was uncontrollable with either strategy when \(R_0=5\), though \(R_\rm{eff}\) fell below \(2\) with both non-pharmaceutical interventions. In contrast, short-course illnesses (such as influenza A) or those with extensive pre-symptomatic infectiousness (such as Hepatitis A) showed significantly greater impacts with individual quarantine. In a high feasibility setting where contact tracing and isolation are effective, \(R_\rm{eff}\) fell below \(0.5\) with individual quarantine but remained above \(1\) for active symptom monitoring.
In the ongoing COVID-19 epidemic, a key uncertainty is the extent of transmission that occurs when individuals do not display symptoms. The incidence of pre-symptomatic or asymptomatic cases of COVID-19 inspired considerable debate in the community when an early study reporting pre-symptomatic transmission later came into question [9]. Furthermore, reported serial intervals—the time between symptom onset of an infector and symptom onset of an infectee—have varied considerably. We utilized two published estimates of the serial interval: one short (mean of 4.8 days) and one longer (mean of 7.5 days) [4, 5]. Model results using data from the shorter interval indicate that the mean time of infectiousness is nearly a day prior to symptom onset, suggesting considerable pre-symptomatic transmission [7]. We found the mean time of infectiousness using the longer serial interval to be shortly after symptom onset, although a period of pre-symptomatic transmissibility remains.
By employing our framework and parameter estimations, we determined that in high feasibility settings—where most contacts are traced with minimal delay and isolation is nearly perfect—individual quarantine can drive down transmission in more than 95 percent of scenarios [7]. Comparatively, active symptom monitoring reduces \(R_\rm{eff}\) below \(1\) only 12 percent of the time. Low feasibility settings—where about half the contacts are traced, isolation is moderately effective, and longer delays for tracing and treatment exist—cannot sufficiently achieve control with either individual quarantine or active symptom monitoring. Our work assumed current estimates of the basic reproductive number of COVID-19: \(R_0=2.2\) [4, 8]. However, control was impossible even when reducing \(R_0\) to \(1.5\). The inability for disease containment in these cases stems from the fact that many transmission chains are not followed at all, not followed quickly enough, or incompletely stopped to prevent additional spread. Nonetheless, onward transmissions can be blocked for an individual whose contacts are traced, therefore reducing overall burden within the population. Thus, even in situations where containment is inconceivable with individual quarantine and active symptom monitoring, these strategies will help mitigate transmission and can be used in conjunction with other policies.
To more adequately track the burden of individual quarantine in the COVID-19 outbreak, we also considered a fraction of ultimately uninfected but traced contacts who self-quarantine for 14 days—the current recommendation [1, 2]—before returning to normal activities. Thus, the number of individuals under quarantine grows significantly more quickly than the number of cases (see Figure 2) and is likely to rapidly outpace available resources. In these situations, people must exercise more general interventions like mass quarantine, travel restrictions, or social distancing. However, it is important to note that such interventions also depend on our understanding of a disease’s natural history. With COVID-19, which shows evidence of pre-symptomatic transmission, mass quarantine—as occurred on the Diamond Princess cruise ship—forced infected and uninfected individuals to remain together in close quarters and may have led to additional cases.
Figure 2. Simulated number of cumulative infections and individuals currently under quarantine. The daily count of cumulative infections is shown in red and the known infected contacts currently under quarantine appear in blue. Uninfected contacts currently under quarantine that assume a 1:1 ratio of uninfected to infected contacts traced are in dark green, and uninfected contacts currently under quarantine that assume a 9:1 ratio of uninfected to infected contacts traced are in light green. Quarantine is imposed when the cumulative case count reaches 1,000. These results assume a low feasibility setting, a basic reproductive number \(R_0=2.2\), and a shorter mean serial interval of 4.8 days. Figure courtesy of [7].
Contact tracing and the resulting non-pharmaceutical interventions, like individual quarantine and active symptom monitoring, have the potential to be quite effective in combatting certain types of infectious diseases. The extent of each strategy’s efficacy depends on assumptions pertaining to the underlying parameters, such as serial interval, incubation period, and feasibility setting. In some circumstances, particularly when symptoms arise before infectiousness, both approaches may adequately eliminate the disease. In other scenarios, as appears to be the case in the ongoing COVID-19 pandemic, use of these measures alone is insufficient. Nevertheless, individual quarantine and active symptom monitoring do mitigate spread and can complement social distancing and travel restrictions.
References
[1] Centers for Disease Control and Prevention. (2020). Interim U.S. guidance for risk assessment and public health management of persons with potential 2019 novel coronavirus (2019-nCoV) exposure in travel-associated or community settings. Retrieved from https://stacks.cdc.gov/view/cdc/84776.
[2] Chinese Center for Disease Control and Prevention. (2020). COVID-19 prevention and control guidelines. Retrieved from http://www.nhc.gov.cn/jkj/s3577/202003/4856d5b0458141fa9f376853224d41d7/files/4132bf035bc242478a6eaf157eb0d979.pdf.
[3] Dong, E., Du, H., & Gardner, L. (2020). An interactive web-based dashboard to track COVID-19 in real time. Lancet Infect. Dis. To be published.
[4] Li, Q., Guan, X., Wu, P., Wang, X., Zhou, L., Tong, Y., ..., & Feng, Z. (2020). Early transmission dynamics in Wuhan, China, of novel coronavirus — infected pneumonia. New Eng. J. Med., 382(13).
[5] Nishiura, H., Linton, N.M., & Akhmetzhanov, A.R. (2020). Serial interval of novel coronavirus (COVID-19) infections. Int. J. Infect. Dis., 4(93), 284-286.
[6] Peak, C.M., Childs, L.M., Grad, Y.H., & Buckee, C.O. (2017). Comparing nonpharmaceutical interventions for containing emerging epidemics. Proceed. Nat. Acad. Sci., 114(15), 4023-4028.
[7] Peak, C.M., Kahn, R., Grad, Y.H., Childs, L.M., Li, R., Lipsitch, M., & Buckee, C.O. (2020). Modeling the comparative impact of individual quarantine vs. active monitoring of contacts for the mitigation of COVID-19. Lancet Infect. Dis. To be published.
[8] Riou, J., & Althaus, C.L. (2020). Pattern of early human-to-human transmission of Wuhan 2019 novel coronavirus (2019-nCoV), December 2019 to January 2020. Eurosurveill., 25(4).
[9] Rothe, C., Schunk, M., Sothmann, P., Bretzel, G., Froeschl, G., Wallrauch, C., ..., & Hoelscher, M. (2020). Transmission of 2019-nCoV infection from an asymptomatic contact in Germany. New Eng. J. Med., 382, 970-971.
Lauren M. Childs is an assistant professor of mathematics at Virginia Tech, where she develops and analyzes mathematical and computational models that examine biologically-motivated questions. Understanding the pathogenesis and spread of infectious diseases, such as malaria and dengue, is a main focus of her work.