Transmission dynamics of the great influenza pandemic of 1918 in Geneva, Switzerland: Assessing the effects of hypothetical interventions

https://doi.org/10.1016/j.jtbi.2005.11.026Get rights and content

Abstract

Recurrent outbreaks of the avian H5N1 influenza virus in Asia represent a constant global pandemic threat. We characterize and evaluate hypothetical public health measures during the 1918 influenza pandemic in the Canton of Geneva, Switzerland. The transmission rate, the recovery rate, the diagnostic rate, the relative infectiousness of asymptomatic cases, and the proportion of clinical cases are estimated through least-squares fitting of the model to epidemic curve data of the cumulative number of hospital notifications. The latent period and the case fatality proportion are taken from published literature. We determine the variance and identifiability of model parameters via a simulation study. Our epidemic model agrees well with the observed epidemic data. We estimate the basic reproductive number for the spring wave R1^=1.49 (95% CI: 1.451.53) and the reproductive number for the fall wave R2^=3.75 (95% CI: 3.573.93). In addition, we estimate the clinical reporting for these two waves to be 59.7% (95% CI: 55.763.7) and 83% (95% CI: 7987). We surmise that the lower reporting in the first wave can be explained by a lack of initial awareness of the epidemic and the relative higher severity of the symptoms experienced during the fall wave. We found that effective isolation measures in hospital clinics at best would only ensure control with probability 0.87 while reducing the transmission rate by >76.5% guarantees stopping an epidemic.

Introduction

Recurrent outbreaks of avian influenza (H5N1) in Asia threaten the human population with the next influenza pandemic as infections have been observed in humans with probable limited human-to-human transmission. Should the new virus subtype get fully adapted for human-to-human transmission, an influenza pandemic could arise (Snacken, 2002, Enserink, 2005) with devastating economic consequences (Meltzer et al., 1999). Therefore, enhancing our understanding of the transmissibility, mechanisms, and key factors under which the influenza virus propagates among populations is critical to devise effective and economic interventions strategies.

The etiological agent of influenza is an RNA virus (orthomyxoviridae family) (Webster et al., 1992) that causes acute upper respiratory tract infection with symptoms including high fever, myalgia, severe malaise, non-productive cough, and sore throats. The duration of the latent period for influenza is about 1.9 days (Mills et al., 2004) followed by an infectious period of about 4 days (Reeve et al., 1980, Moritz et al., 1980). Influenza is transmitted by direct contact (e.g. hand shaking, sweat, etc.), aerosol, and droplets.

Individuals exposed to the influenza virus gain protection or cross-protection. Hence, the influenza virus undergoes continuous evolution in order for annual epidemics to occur. Such changes in the virus composition are known as drifts or shifts. Drifts are the consequence of single-point mutations in the virus antigenic structure while shifts are major gene reassortments which have the potential of generating pandemics (Webster et al., 1992).

The 1918 influenza pandemic known as the “Spanish flu” caused by the influenza virus A(H1/N1) has been the worst in recent history with estimated worldwide mortality ranging from 20 to 100 million deaths (Cunha, 2004). The worldwide 1918 influenza pandemic spread in three waves starting from Midwestern United States in the spring of 1918 (Patterson and Pyle, 1991, Johnson and Mueller, 2002). The deadly second wave began in late August probably in France while the third wave is generally considered as part of normal more scattered winter outbreaks similar to those observed after the 1889/90 pandemic (Patterson and Pyle, 1991). Subsequent flu pandemics are attributed to flu A(H2N2) in 1957 (Asian flu) and A(H3N2) in 1968 (Hong Kong flu).

Underreporting due to the disruptions in the public health system during the 1918/9 pandemic complicate the estimation of attack and death rates in many regions (Patterson and Pyle, 1991), e.g. influenza deaths were not recorded in Russia and the data from developing countries suffers from significant underreporting (Patterson and Pyle, 1991). Nevertheless, in some countries mandatory notifications of flu cases by health care practitioners were implemented during the pandemic. Switzerland provides one of the best databases for the analysis of the 1918/9 influenza pandemic because mandatory notification of flu cases was implemented at the federal level from the beginning of the pandemic (Ammon, 2002). However, there was underreporting from mild cases and cases who were refused admission in overcrowded hospital clinics. In addition, there is good demographic information of the Swiss population at the time of the pandemic.

In this paper, we investigate the 1918/9 influenza pandemic in the Canton of Geneva located in the south western corner of Switzerland and surrounded in its majority by France. We model the transmission dynamics of the spring and fall waves of influenza using an epidemic model that accounts for the known underreporting. Some of the model parameters are estimated via least-squares fitting and the resulting parameter estimates are corroborated via a simulation study. From our fitted model, we estimate the basic reproductive number of the spring and the reproductive number of the fall wave to be 1.49 (95% CI: 1.451.53) and 3.75 (95% CI: 3.573.93), respectively. In addition, we estimate the clinical reporting for these two waves to be 59.7% (95% CI: 55.763.7) and 83% (95% CI: 79–87). We surmise that both the lack of initial awareness of the epidemic and the relative higher severity of the symptoms experienced during the fall wave contributed to the lower reporting rate in the first wave. We found that effective isolation measures in hospital clinics at best would only ensure control with probability 0.87 while reducing the transmission rate by >76.5% guarantees stopping an epidemic.

Section snippets

The 1918 flu pandemic in Geneva, Switzerland

The 1918/9 influenza pandemic affected more than 50% of the population in Geneva, Switzerland (Ammon, 2002). The first wave occurred in July 1918 (“spring wave”), the second deadliest wave in October–November 1918 (“fall wave”), and the third wave was observed at the end of 1918 (“winter wave”). The symptoms presented during the second wave were more severe than for the first wave. Moreover, it seems that individuals infected with the flu were subsequently protected to secondary waves of

Epidemic model

We model the first two waves of the 1918 influenza pandemic in Geneva, Switzerland (Fig. 2) separately using a compartmental epidemic model. The model (Fig. 3) for the transmission dynamics of pandemic influenza classifies individuals as susceptible (Si), exposed (Ei), clinically ill and infectious (Ii), asymptomatic and partially infectious (Ai), hospitalized and reported (Ji), recovered (Ri), and death (Di) where i=1,2 indices the spring and fall waves, respectively. We assume that the birth

Results

Our fitted model for the influenza notifications during the first two waves of the 1918 influenza pandemic in Geneva, Switzerland agrees well with the observed epidemic data (coefficient of determination of 0.99 (Neter and Wasserman, 1974), Fig. 4).

We estimated epidemiological parameters via least-squares fitting of the model to the cumulative number of hospital notifications (Table 1). We estimated the standard deviations of the parameters via a simulation study using the parametric bootstrap (

Discussion

In the context of influenza, mathematical models have been used to study different demographic and epidemiological mechanisms that characterize influenza dynamics such as annual periodicity (e.g. Castillo-Chavez et al., 1989, Boni et al., 2004, Nuño et al., 2005), describe and predict its spread (Elvebaek et al., 1976, Spicer and Lawrence, 1984, Rvachev and Longini, 1985, Flahault et al., 1988, Viboud et al., 2003, Hyman and Laforce, 2003), and evaluate different control strategies that could

Acknowledgements

We thank two reviewers for providing useful comments that helped improve our manuscript. G. Chowell was supported by a Director's Postdoctoral Fellowship from Los Alamos National Laboratory.

References (43)

  • R.M. Anderson et al.

    Infectious Diseases of Humans

    (1991)
  • Boni, M.F., Gog, J.R., Andreasen, V., Christiansen, F.B., 2004. Influenza drift and epidemic size: the race between...
  • F. Brauer et al.

    Mathematical Models in Population Biology and Epidemiology

    (2000)
  • C. Castillo-Chavez et al.

    Epidemiological models with age-structure, proportionate mixing, and cross-immunity

    J. Math. Biol.

    (1989)
  • G. Chowell et al.

    Model parameters and outbreak control for SARS

    Emerg. Infect. Dis.

    (2004)
  • O. Diekmann et al.

    Mathematical Epidemiology of Infectious Diseases: Model Building, Analysis and Interpretation

    (2000)
  • Dubois, J., 2005. E-mail Communication. Office Cantonal de la...
  • B. Efron et al.

    Bootstrap methods for standard errors, confidence intervals, and other measures of statistical accuracy

    Stat. Sci.

    (1986)
  • L.R. Elvebaek et al.

    An influenza simulation model for immunisation studies

    Am. J. Epidemiol.

    (1976)
  • M. Enserink

    Pandemic influenza: global update

    Science

    (2005)
  • N.M. Ferguson et al.

    Strategies for containing an emerging influenza pandemic in southeast Asia

    Nature

    (2005)
  • Cited by (232)

    • Classification of community-acquired outbreaks for the global transmission of COVID-19: Machine learning and statistical model analysis

      2021, Journal of the Formosan Medical Association
      Citation Excerpt :

      In order to have well preparedness, a highly awareness is helpful to control the transmission scale and to reduce the severity of symptoms in the face of second epidemic waves.14–17 In the early epidemic phases, rapidly identifying the clinical, epidemiological and pathogenologic characteristics of pathogens from infective cases and clarifying the travel history, occupation, contact history, cluster gathering (TOCC) for patients are helpful for health authorities to make contingency plan on implementing containment measures in community to prevent the community-acquired infection.16–18 However, when time goes and COVID-19 pandemic still exists very few researches have been conducted to look into why there are uninterrupted and repeated episodes with high case load of community-acquired outbreaks during COVID-19 pandemic.

    View all citing articles on Scopus
    1

    Los Alamos Unclassified Report LA-UR-05-7140.

    View full text