Forward-Looking Serial Intervals Correctly Link Epidemic Growth To
Total Page：16
File Type：pdf, Size：1020Kb
Forward-looking serial intervals correctly link epidemic growth to reproduction numbers Sang Woo Parka,1 , Kaiyuan Sunb , David Champredonc, Michael Lid , Benjamin M. Bolkerd,e,f, David J. D. Earne,f , Joshua S. Weitzg,h , Bryan T. Grenfella,b,i, and Jonathan Dushoffd,e,f aDepartment of Ecology and Evolutionary Biology, Princeton University, Princeton, NJ 08544; bDivision of International Epidemiology and Population Studies, Fogarty International Center, NIH, Bethesda, MD 20892; cDepartment of Pathology and Laboratory Medicine, University of Western Ontario, London, ON N6A 3K7, Canada; dDepartment of Biology, McMaster University, Hamilton, ON L8S 4L8, Canada; eDepartment of Mathematics and Statistics, McMaster University, Hamilton, ON L8S 4L8, Canada; fM. G. DeGroote Institute for Infectious Disease Research, McMaster University, Hamilton, ON L8S 4L8, Canada; gSchool of Biological Sciences, Georgia Institute of Technology, Atlanta, GA 30332; hSchool of Physics, Georgia Institute of Technology, Atlanta, GA 30332; and iPrinceton School of Public and International Affairs, Princeton University, Princeton, NJ 08544 Edited by Nils Chr. Stenseth, University of Oslo, Oslo, Norway, and approved December 4, 2020 (received for review June 4, 2020) The reproduction number R and the growth rate r are critical epi- Since generation intervals measure time between infection demiological quantities. They are linked by generation intervals, events, which can be difficult to observe in practice, genera- the time between infection and onward transmission. Because tion intervals are often replaced with serial intervals. The serial generation intervals are difficult to observe, epidemiologists often interval is defined as the time between when an infector and substitute serial intervals, the time between symptom onset in an infectee develop symptoms (7). While generation and serial successive links in a transmission chain. Recent studies suggest intervals both measure the time scale of disease transmission, that such substitution biases estimates of R based on r. Here we they measure fundamentally different quantities. In particu- explore how these intervals vary over the course of an epidemic, lar, previous studies have noted that, in many contexts, serial and the implications for R estimation. Forward-looking serial intervals are expected to have larger variances than generation intervals, measuring time forward from symptom onset of an infec- intervals but have the same mean in many contexts (7–10). Serial tor, correctly describe the renewal process of symptomatic cases intervals can, in some cases, even take negative values in the pres- and therefore reliably link R with r. In contrast, backward-looking ence of presymptomatic transmission (11), whereas generation intervals, which measure time backward, and intrinsic intervals, intervals must be positive. POPULATION BIOLOGY which neglect population-level dynamics, give incorrect R esti- Although these distributions were clearly and distinctly mates. Forward-looking intervals are affected both by epidemic defined over a decade ago (7), the need for a better concep- dynamics and by censoring, changing in complex ways over the tual and theoretical framework for understanding their differ- course of an epidemic. We present a heuristic method for address- ences is becoming clearer as the COVID-19 pandemic unfolds. ing biases that arise from neglecting changes in serial intervals. Researchers continue to base inferences about COVID-19 on We apply the method to early (21 January to February 8, 2020) both generation and serial intervals without clearly distinguishing serial interval-based estimates of R for the COVID-19 outbreak between them (e.g., refs. 11–15), and, in some cases, explic- in China outside Hubei province; using improperly defined serial itly conflate the definitions of the two intervals (e.g., refs. 16 intervals in this context biases estimates of initial R by up to a fac- and 17). This confusion is apparent even in standard software tor of 2.6. This study demonstrates the importance of early contact tracing efforts and provides a framework for reassessing genera- tion intervals, serial intervals, and R estimates for COVID-19. Significance generation interval j serial interval j reproduction number j The generation and serial interval distributions are key, but infectious disease modeling different, quantities in outbreak analyses. Recent studies sug- gest that the two distributions give different estimates of R he reproduction number R is one of the most important the reproduction number as inferred from the observed r R characteristics of an emerging epidemic, such as the cur- growth rate . Here, we show that estimating based on T r rent pandemic of COVID-19 (1). The reproduction number is and the serial interval distribution, when defined from the r defined as the average number of secondary infections caused by correct reference cohort, gives the same estimate as using a primary infection. The value in a fully susceptible population— and the generation interval distribution. We apply our frame- work to COVID-19 serial interval data from China, outside the “basic” reproduction number R0—allows us to predict the extent to which an infection will spread in the population, and Hubei province (January 21 to February 8, 2020), revealing sys- the amount of intervention necessary to eliminate it in simple tematic biases in prior inference methods. Our study provides cases (2). Since the reproduction number represents an aver- the theoretical basis for practical changes to the principled R age (2, 3), it fails to capture heterogeneity among individuals or use of serial interval distributions in estimating during across space. The reproduction number also fails to provide any epidemics. information about the time scale of disease transmission. Author contributions: S.W.P., K.S., D.C., M.L., B.M.B., D.J.D.E., J.S.W., B.T.G., and J.D. Estimating the reproduction number R is often challenging. designed research; S.W.P., K.S., B.M.B., D.J.D.E., J.S.W., and J.D. performed research; Direct estimates based on observed infections will typically be S.W.P. analyzed data; and S.W.P., K.S., D.C., M.L., B.M.B., D.J.D.E., J.S.W., B.T.G., and J.D. biased down when some infections cannot be observed. A com- wrote the paper.y mon method of estimating R near the beginning of an epidemic The authors declare no competing interest.y is based on the population-level exponential growth rate r, which This article is a PNAS Direct Submission.y can often be estimated robustly from case reports (4, 5). The This open access article is distributed under Creative Commons Attribution License 4.0 growth rate r and the reproduction number R are linked by (CC BY).y the generation interval distribution (6), where the generation 1 To whom correspondence may be addressed. Email: [email protected] interval is defined as the time between when an individual (infec- This article contains supporting information online at https://www.pnas.org/lookup/suppl/ tor) is infected and when that individual infects another person doi:10.1073/pnas.2011548118/-/DCSupplemental.y (infectee) (7). Published December 23, 2020. PNAS 2021 Vol. 118 No. 2 e2011548118 https://doi.org/10.1073/pnas.2011548118 j 1 of 12 Downloaded by guest on September 26, 2021 for estimating R, such as EpiEstim, in which the serial interval an infector and an infectee (e.g., generation and serial intervals). We define distribution is used to infer time-dependent R (18). These stud- the delay as the time difference between the primary event and the sec- ies are examples of many—indeed, it is a common practice to use ondary event. In some cases, the primary event always occurs before the the serial and generation intervals interchangeably. secondary event (e.g., the time from infection to onset of symptoms in a One source of confusion arises from an apparent discrepancy single individual, or the generation interval between two individuals). In other cases, the delay can sometimes be negative (e.g., the time from onset between the generation interval and serial interval viewpoints. of symptoms to onset of infectiousness in a single individual, or the serial While the epidemic is growing exponentially, the spread of infec- interval between two individuals). tion can be characterized as a renewal process based on previous At the individual level, we can define the time distribution between incidence of infection, the associated generation interval distri- a primary and a secondary event that we expect to observe for a single bution, and the average infectiousness of an infected individual. infected individual by averaging across individual characteristics—we refer It is well established that this renewal formulation allows us to to this distribution as the intrinsic distribution. For example, the intrin- link the exponential growth rate of an epidemic r with its repro- sic incubation period distribution describes the expected time distribution duction number R using the generation interval distribution (6). from infection to symptom onset of an infected individual. Likewise, the However, the serial interval distribution also describes a renewal intrinsic generation interval distribution describes the expected time distri- bution of infectious contacts made by an infected individual. However, the process—in this case, the creation of a new symptomatic case intrinsic time distributions are not always equivalent to the corresponding based on a symptomatic case in the previous generation. Since realized time distributions at the population level (i.e., the distribution of both renewal processes, based on either generation or serial time between actual primary and secondary events that occur during an interval distributions, describe the same underlying exponen- epidemic; Fig. 1). For example, an infectious contact results in infection tially growing system, both should provide the same correct link only if the contacted individual is susceptible (and has not already been between the reproduction number R and the epidemic growth infected)—this is one mechanism that causes realized generation intervals rate r.