key: cord-0905939-zukjh1hr
authors: Feng, Zhilan; Glasser, John W.; Hill, Andrew N.
title: On the benefits of flattening the curve: A perspective()
date: 2020-05-27
journal: Math Biosci
DOI: 10.1016/j.mbs.2020.108389
sha: 7ffd042b919518a067b92b9678c1d71a12e548fd
doc_id: 905939
cord_uid: zukjh1hr

The many variations on a graphic illustrating the impact of non-pharmaceutical measures to mitigate pandemic influenza that have appeared in recent news reports about COVID-19 suggest a need to better explain the mechanism by which social distancing reduces the spread of infectious diseases. And some reports understate one benefit of reducing the frequency or proximity of interpersonal encounters, a reduction in the total number of infections. In hopes that understanding will increase compliance, we describe how social distancing a) reduces the peak incidence of infections, b) delays the occurrence of this peak, and c) reduces the total number of infections during epidemics. In view of the extraordinary efforts underway to identify existing medications that are active against SARS-CoV-2 and to develop new antiviral drugs, vaccines and antibody therapies, any of which may have community-level effects, we also describe how pharmaceutical interventions affect transmission.

 Social distancing refers to non-pharmaceutical measures to reduce the frequency or proximity of interpersonal encounters  The impact of these measures on epidemic curves is commonly misrepresented, suggesting a lack of understanding of the underlying mechanisms  As this may affect compliance with recommendations, we describe determinants of the magnitude and timing of peak incidence and the total number of infections  We also describe possible population-level effects of pharmaceutical interventions

The most recent community mitigation guidelines for pandemic influenza (Qualls et al. 2017 ) include a graphic illustrating the goals of non-pharmaceutical interventions, slowing the rate at which new infections occur (incidence), reducing the peak number of infected people (prevalence) and concomitant demands on healthcare facilities and personnel, and decreasing overall infections and deaths. Antimicrobial drugs generally are administered solely for the benefit of individual patients, but may also reduce the magnitude or duration of infectiousness, affecting transmission to others. Similarly, vaccines and monoclonal antibodies may reduce infectiousness as well as susceptibility. Accordingly, we include pharmaceutical interventions in our description of mechanisms underlying efforts to flatten the curve.

Plots of daily numbers of new infections, called epidemic curves, are affected by several factors, including under-reporting. Individual responses to infection by the same pathogen differ, and mild infections are less likely to be reported than those requiring medical attention. Absent widespread testing, few asymptomatic infections would be reported. In the ongoing pandemic, people with few if any symptoms may be infectious , diminishing the effectiveness of control measures based solely on symptomatic infections. Social distancing -increasing inter-personal distances, reducing the number of people attending meetingsinvolves everyone, reducing the risks of being infected if susceptible and of infecting others if infectious.

Assuming that infected people are infectious upon infection, other factors affecting epidemic curves are person-to-person contact rates (i.e., contacts per person per time period) and probabilities of infection on contact between infectious and susceptible people, or infectiousness. The sum of contacts by which susceptible people would be infected, which are products of these rates and probabilities (Hethcote 2000) , estimates the basic reproduction number, or average number of secondary infections per infected person.

The areas beneath epidemic curves represent total numbers of infections. The onset, shape and area beneath these curves all depend on the basic reproduction number, which must be greater than one for epidemics to occur. The smaller this number, the longer the delay to peak incidence, the lower the peak, and fewer the total number of infections (figure 1). Both of the basic reproduction number's constituents may be modifiable, contact rates by social distancing and infectiousness by medications, whereupon they would be termed effective reproduction numbers. Figure 1 . Epidemic curves with varying contact rates. The contact rates decrease by 1 from 10 to 6 per day (left to right), corresponding to reproduction numbers ranging from 2.5 to 1.5. As these numbers decrease, the percent of the population infected decreases from 89% to 58%. Thus, social distancing not only delays and diminishes the peak number of infections, it also reduces the total number of infections.

While drugs are administered to cure infections or mitigate symptoms, they may also affect the magnitude or duration of infectiousness. Clinical trials of remdesivir and other promising antiviral drugs are underway (https://clinicaltrials.gov/ct2/results?cond=COVID-19), but none has yet been proven effective against SARS-CoV-2. This is the mechanism underlying the "treatment as prevention" approach to reducing HIV infection rates. Treatment of infected people with anti-retroviral drugs reduces their viral load, preventing their progression to AIDS. Authorities believe that it also reduces their infectiousness, which would lower the effective reproduction number. This is the rationale for the US national initiative, "Ending the HIV Epidemic: A Plan for America" (https://www.hiv.gov/federal-response/ending-the-hiv-epidemic/overview).

Similarly, unprecedented efforts to develop vaccines (Le et al. 2020) or monoclonal antibodies (e.g., Wang et al. 2020 ) against SARS-CoV-2 also are underway. Should any prove safe and effective, they may reduce infectiousness as well as susceptibility to infection.

But the only way to flatten epidemic curves without also affecting the areas beneath them -as some figures in the media have suggested -is to change the mean interval between being infected and infecting others, or generation time. The pathogenesis of infectious diseases typically is characterized by replication prior to dissemination (Nash et al. 2015) , so that people are not infectious for a while after being infected, called the latent period. Suppose -for sake of argument -that we could modify this period (figure 2). Figure 2. Epidemic curves with varying latency. Intervals from infection to the onset of infectiousness or latent periods increase from 10 to 26 in 4-day increments (left to right), all corresponding to a basic reproduction number of 2.5. Note that the thick red curves are identical in both figures and that, in this one, while the curves differ in shape, the same total percent of the population, 89% is infected.

J o u r n a l P r e -p r o o f Should some infected people not live long enough to cause the average number of secondary infections, the basic reproduction number and generation time would no longer be independent. Be that as it may, the goal is not just to delay and reduce the peak number of infections, possibly below surge capacity as some figures in the media have suggested, it is also to reduce the total number of infections. Social distancing accomplishes all three objectives: a) reducing the peak incidence of infections, b) delaying the occurrence of this peak, and c) reducing the total number of infections during epidemics. rate (contacts per person per time period),  is the probability of infection on contact between a susceptible and an infectious person, and N is the total population size. 

The mathematics of infectious diseases

The COVID-19 vaccine development landscape

Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (SARS-CoV-2)

Pathogenesis of Infectious Disease

Community mitigation guidelines to prevent pandemic influenza -United States

A human monoclonal antibody blocking SARS-CoV-2 infection

In the model that we used to generate these figures, the host population is partitioned into those who are susceptible to infection, S; who have been infected, but are not yet infectious, E; who are infectious, I; who have recovered and are immune, R; and who have died, D.In this system of ordinary differential equations,  is the force or hazard rate of infection,  is the rate of disease progression (its reciprocal is the latent period),  is the rate of recovery (its reciprocal is the infectious period), p is the proportion surviving (thus, disease-induced mortality is 1p), a is the contact J o u r n a l P r e -p r o o f