Access the full text.
Sign up today, get DeepDyve free for 14 days.
(2011)
Prevention: Estimating Seasonal Influenza- Associated Deaths in the United States: CDC Study Confirms Variability of Flu
Chengjun Sun, Wei Yang, Julien Arino, K. Khan (2011)
Effect of media-induced social distancing on disease transmission in a two patch settingMathematical Biosciences, 230
(2011)
A classical approach for estimating the transmissibility of the 2009 H1N1 pandemic
(2006)
Public Health Agency of Canada: Highlights from the Canadian Pandemic Influenza Plan for the Health Sector
J. Heath (1992)
Mass mediaNature, 359
Katherine Kinnick, D. Krugman, G. Cameron (1996)
Compassion Fatigue: Communication and Burnout toward Social ProblemsJournalism & Mass Communication Quarterly, 73
B. Pourbohloul, A. Ahued, B. Davoudi, R. Meza, L. Meyers, D. Skowronski, Ignacio Villaseñor, F. Galván, P. Cravioto, D. Earn, J. Dushoff, D. Fisman, W. Edmunds, N. Hupert, S. Scarpino, Jesús Trujillo, Miguel Lutzow, J. Morales, A. Contreras, C. Chavez, D. Patrick, R. Brunham (2009)
Initial human transmission dynamics of the pandemic (H1N1) 2009 virus in North AmericaInfluenza and Other Respiratory Viruses, 3
J. Tchuenche, Nothabo Dube, C. Bhunu, Robert Smith, C. Bauch (2011)
The impact of media coverage on the transmission dynamics of human influenzaBMC Public Health, 11
E. Mykhalovskiy, Lorna Weir (2006)
The Global Public Health Intelligence Network and Early Warning Outbreak DetectionCanadian Journal of Public Health, 97
J. Heffernan, L. Wahl (2005)
Monte Carlo estimates of natural variation in HIV infection.Journal of theoretical biology, 236 2
J. Cui, Yonghong Sun, Huaiping Zhu (2007)
The Impact of Media on the Control of Infectious DiseasesJournal of Dynamics and Differential Equations, 20
Rongsong Liu, Jianhong Wu, Huaiping Zhu (2007)
Media/psychological impact on multiple outbreaks of emerging infectious diseasesComputational and Mathematical Methods in Medicine, 8
L. Allen (2008)
An Introduction to Stochastic Epidemic Models
M. Keeling (2000)
Metapopulation moments: coupling, stochasticity and persistence.The Journal of animal ecology, 69 5
F. Brauer (2009)
Mathematical epidemiology is not an oxymoronBMC Public Health, 9
S. Galea (2010)
Macrosocial Determinants of Population Health
Yongfeng Li, Chaoqun Ma, Jing'an Cui (2008)
The Effect of Constant and Mixed Impulsive Vaccination on SIS Epidemic Models Incorporating Media CoverageRocky Mountain Journal of Mathematics, 38
J. Cui, X. Tao, Huaiping Zhu (2008)
An SIS Infection Model Incorporating Media CoverageRocky Mountain Journal of Mathematics, 38
S. Blower, H. Dowlatabadi (1994)
Sensitivity and Uncertainty Analysis of Complex Models of Disease Transmission: an HIV Model, as an ExampleInternational Statistical Review, 62
(2014)
Cite this article as: Collinson and Heffernan: Modelling the effects of media during an influenza epidemic
K. Fanti, E. Vanman, Christopher Henrich, M. Avraamides (2009)
Desensitization to media violence over a short period of time.Aggressive behavior, 35 2
Yi Wang, Jinde Cao, Zhen Jin, Haifeng Zhang, Gui‐Quan Sun (2013)
Impact of media coverage on epidemic spreading in complex networksPhysica a, 392
F. Schweitzer, R. Mach (2008)
The Epidemics of Donations: Logistic Growth and Power-LawsPLoS ONE, 3
K. Smith, R. Rimal, H. Sandberg, J. Storey, Lisa Lagassé, Catherine Maulsby, Elizabeth Rhoades, D. Barnett, S. Omer, J. Links (2012)
Understanding newsworthiness of an emerging pandemic: International newspaper coverage of the H1N1 outbreakInfluenza and Other Respiratory Viruses, 7
D. Mcburney, C. Balaban, D. Christopher, C. Harvey (1997)
Adaptation to Capsaicin Within and Across DaysPhysiology & Behavior, 61
M. Griffiths, Gary Shuckford (1989)
Desensitization to television violence: A new modelNew Ideas in Psychology, 7
citation_title=Highlights from the Canadian Pandemic Influenza Plan for the Health Sector, citation_publication_date= (2006)
Highlights from the Canadian Pandemic Influenza Plan for the Health Sector
J. Heffernan, R. Smith, L. Wahl (2005)
Perspectives on the basic reproductive ratioJournal of The Royal Society Interface, 2
I. Krishnarajah, A. Cook, G. Marion, G. Gibson (2005)
Novel moment closure approximations in stochastic epidemicsBulletin of Mathematical Biology, 67
M. Keeling (2000)
Multiplicative moments and measures of persistence in ecology.Journal of theoretical biology, 205 2
G. Neumann, T. Noda, Y. Kawaoka (2009)
Emergence and pandemic potential of swine-origin H1N1 influenza virusNature, 459
Jianhua Pang, J. Cui (2009)
An SIRS Epidemiological Model with Nonlinear Incidence Rate Incorporating Media Coverage2009 Second International Conference on Information and Computing Science, 3
Dongmei Xiao, Shigui Ruan
Global Analysis of an Epidemic Model with Nonmonotone Incidence Rate
(2007)
Author ' s personal copy Global analysis of an epidemic model with nonmonotone incidence rate
Background: Mass media is used to inform individuals regarding diseases within a population. The effects of mass media during disease outbreaks have been studied in the mathematical modelling literature, by including ‘media functions’ that affect transmission rates in mathematical epidemiological models. The choice of function to employ, however, varies, and thus, epidemic outcomes that are important to inform public health may be affected. Methods: We present a survey of the disease modelling literature with the effects of mass media. We present a comparison of the functions employed and compare epidemic results parameterized for an influenza outbreak. An agent-based Monte Carlo simulation is created to access variability around key epidemic measurements, and a sensitivity analysis is completed in order to gain insight into which model parameters have the largest influence on epidemic outcomes. Results: Epidemic outcome depends on the media function chosen. Parameters that most influence key epidemic outcomes are different for each media function. Conclusion: Different functions used to represent the effects of media during an epidemic will affect the outcomes of a disease model, including the variability in key epidemic measurements. Thus, media functions may not best represent the effects of media during an epidemic. A new method for modelling the effects of media needs to be considered. Keywords: Mass media, Epidemic, Influenza, Agent-based Monte Carlo simulation Background source of information and are able to incite changes in Influenza causes annual epidemics and occasional pan- behaviour in the public [5]. Individual response to a dis- demics, which have claimed millions of lives throughout ease threat depends on risk perception that is gained history. In the past century four worldwide pandemic largely through information reported by governments to outbreaks of influenza have occurred: 1918, 1957, 1977 mass media: number of infections, hospitalizations and and 2009, [1,2]. According to the Public Health Agency deaths, as provided by the government [6,7]. of Canada, inter-pandemic (or seasonal) influenza affects We have recently seen the use of mass media reports approximately 20, 000 Canadians, with approximately during two infectious disease outbreaks. The first novel 2, 000 to 8, 000 deaths annually [3]. In the USA, it has been infectious disease of the twenty-first century was SARS. It reported that flu-associated deaths can range from 3, 000 haddistinctfeaturessuchasrapid spatialspreadand self- to 49, 000 individuals per year [4]. control, and vast media coverage [6,7] that used to inform Mass media can affect disease transmission during an the public, provide a means of contract tracing, and advise influenza epidemic or pandemic. Attention to health news isolation of exposed individuals. has been increasing in importance during the last few Media coverage was substantial during the H1N1 epi- decades, and thus, media reports can play an important demic in 2009 as well, which may have had an effect on role in defining health issues, since they serve as a major the transmission of influenza by promoting social dis- tancing and self-isolation [8]. The media coverage of this *Correspondence: jmheffer@yorku.ca influenza pandemic was widespread, with an increased Department of Mathematics & Statistics, York University, Toronto, Canada sense of urgency since this influenza strain was related to Modelling Infection and Immunity Lab, Centre for Disease Modelling, York Institute for Health Research, York University, Toronto, Canada © 2014 Collinson and Heffernan; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Collinson and Heffernan BMC Public Health 2014, 14:376 Page 2 of 10 http://www.biomedcentral.com/1471-2458/14/376 −p γ I the 1918 pandemic strain that caused approximately 50 f (I, p ) = e (1) million deaths worldwide [1]. f (I, p ) = (2) Mathematical modelling has been used to study the 1 + p I effect of mass media on epidemics by employing the well- known Susceptible-Exposed-Infectious-Recovered (SEIR) f (I, p ) = 1 − (3) p + I model and various extensions [6-14]. In these studies, mass media has been incorporated using different, but where I is the number of infectious individuals in a pop- qualitatively similar, functions that directly affect disease ulation, γ is the recovery rate, and p , i = 1, 2, 3 is a transmission and susceptibility. In general, the chosen parameter used to represent media effects in these func- functions are decreasing functions with respect to the tions. In general, these functions are decreasing functions current number of infected individuals in the population. of I, which represents the fact that, as the number of infec- However, the choice of function seems to be arbitrary. It tions increases in a population, and is reported by mass is possible that the choice of function can change study media, susceptible individuals will practice social distanc- results. For example, public health officials could be inter- ing or control measures, which decreases susceptibility. ested in epidemic measurements such as the peak number Comparing the three functions, we see that, for a given I of infection, peak time, total number of infections and and γ , p and p can be written in terms of p : 1 2 3 end of epidemic, which are all directly related to impor- tant public health resources (i.e. number of hospital bed, − ln p +I antiviral stockpile, vaccination doses). These key mea- 3 c p = (4) surement may vary depending on the chosen media func- γ I tion. Therefore, a sensitivity analysis on these functions is p =.(5) required. p I 3 c In addition to that mentioned above, there is a further drawback of the previous studies which include media These equations demonstrate that, if p is increased, then functions, in that, deterministic systems of ordinary dif- p and p must decrease to have the functions remain the 1 2 ferential equations are employed. Deterministic models same value at a chosen I . can describe the mean behaviour of an epidemic, but In Figure 1, we plot Functions (1)-(3) (dotted, dashed, information surrounding any variability in key epidemic solid lines respectively) for media parameters p , i = 1, 2, measurements cannot be made. A stochastic model is 3 (dotted, dashed, solid lines) with γ = 1/4and I = 300 well suited to this task. Various methods of introducing (top) and I = 1000 (bottom). Note that the functions are stochasticity into disease models exist: [15-19]. Agent- equal when I = 300 (top) and 1000 (bottom). Here, p = Based Monte Carlo (ABMC) simulations, provide a way 10 (left panel), 100 (middle panel), and 1000 (right panel), in which individuals with certain disease characteristics and p1and p2 are determined using Equations (4) and (5). can be tracked in a population. This method also pro- vides flexibility, as changes in biological assumptions can SEIR framework be easily incorporated, which are difficult to include in To compare the effects of different media Functions (1)-(3), other methods. we must choose a standardized model. For the purposes of our study, we have chosen the basic Susceptible-Exposed- In the sections that follow we give an overview of the Infectious-Recovered (SEIR) model with a constant popu- functions used to describe media in the disease modelling lation size N: literature. The functions are then incorporated into a stan- dardized SEIR model, and model results are compared. A S =−f (I, p)βSI stochastic agent-based Monte Carlo (ABMC) simulation E = f (I, p)βSI − σ E is then introduced, and is employed to study the variabil- I = σ E − γ I (6) ity within an epidemic depending on the media function R = γ I chosen. A sensitivity analysis is also completed in order to N = S + E + I + R . determine the importance of certain model parameters on various epidemic outcomes for each media function. Briefly, susceptible individuals (S) are infected by infec- tious individuals (I)withrate β and become exposed Methods (E). Transmission can also be reduced as determined by Media functions f (I, p). Exposed individuals become infectious with rate From the disease modelling literature [6-14] we have iden- σ , and infectious individuals recover with rate γ . The ini- tified three distinct functions employed to present the tial conditions and parameters values for this study can be effects of mass media: found in Table 1. Collinson and Heffernan BMC Public Health 2014, 14:376 Page 3 of 10 http://www.biomedcentral.com/1471-2458/14/376 1 1 1 0.8 0.8 0.8 0.6 0.6 0.6 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 0 500 0 500 0 500 1 1 1 0.8 0.8 0.8 0.6 0.6 0.6 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 0 1000 2000 0 1000 2000 0 1000 2000 Number of Infecteds (I) Figure 1 Media Functions f (I, p) for different values of p. Functions (1), (2) and (3) (dotted, dashed, solid lines respectively) are shown when p = 10 (left), 100 (middle), 1000 (right), γ = 1/4and I = 300 (top), 1000 (bottom). p and p are calculated using equations (4) and (5). 3 c 1 2 Agent-based Monte Carlo simulation by Heffernan and Wahl, [16,22]. Briefly, the ABMC sim- For further comparison of the SEIR model with different ulation moves forward in time following event times: the media functions, we utilize an Agent-Based Monte Carlo next time that an individual changes state within the sys- (ABMC) simulation to capture the variability in the epi- tem. Agents in each of the susceptible, exposed, infectious demic infection curve (that cannot be determined using a and recovered compartments are assigned event times system of deterministic equations like that of System (6)). corresponding to Table 1 when they are introduced into There are various ways of developing an ABMC simula- the simulation, for each event that allows such an individ- tion. We have employed a previous method as developed ual from that compartment to change state. For example, Table 1 Initial conditions and parameter values for model (6), functions (1)-(3) and the ABMC Description Value Range Unit Reference Population S Susceptible 10, 000 EExposed 0 IInfectious 10 R Recovered 0 Parameter R Basic reproductive ratio 1.5 1.3 − 1.7 [20,21] −1 β Contact transmission rate 3.71287e−5 (person-day) Eq. (7) −1 σ Transition rate E to I 1/2 (day) [20] −1 γ Recovery rate 1/4 (day) [20] p Media parameter varies varies [7] ABMC S/β Mean time to transmission S(t)/β Eq. (7) 1/σ Mean exposed time 2 days [20] 1/γ Mean infectious time 4 days [20] 1/p Media parameter varies varies [7] f(I,p) f(I,p) Collinson and Heffernan BMC Public Health 2014, 14:376 Page 4 of 10 http://www.biomedcentral.com/1471-2458/14/376 Table 2 Model and fitted parameters β and p Infection Model β p R i 0 Exposed Susceptible Time to infected −5 no media 3.29 ×10 1.32 −5 (1) 3.33 ×10 0.00365 1.33 Become −5 −5 (2) 3.32 ×10 5.2 ×10 1.33 infectious −5 (3) 3.33 ×10 1000 1.33 Based on 100 days of data generated using System (6) with Media Function (3) Infectious Recovered with σ and γ values listed in Table 1 and p = 1000. Time to infect Recover from Time to recovered infection Figure 2 Schematic of the Agent Based Monte Carlo simulations the means of the lifetime distributions in the ABMC are corresponding to Table 1. simply the reciprocals of the rates used in System (6). However, so that infection event times always depend on the current size of the susceptible population S(t) (similar exposed individuals are assigned a time to become infec- to what is assumed in System (6)), the infection time dis- tious, and infectious individuals are assigned a time to tribution mean is continuously updated. Figure 2, shows recovery and a time to infect a susceptible. The mini- the progression of an individual through the epidemic. mum event time determines the next event to occur in the simulation. The event is then carried out, and the next event is determined. To compare to the SEIR system as Results and discussion described above, we assume exponential distributions for Basic reproductive ratio all parameters. Table 1 lists the parameter values of the The basic reproductive ratio R is defined as the number SEIR model and the mean value of the parameter distri- of newly infectious individuals produced by one infec- butions for the ABMC simulation. Note that, in general, tious individual in a totally susceptible population. For S E I R 10000 400 500 6000 8000 4000 6000 2000 4000 0 0 0 0 500 0 500 0 500 0 500 10000 400 500 6000 8000 4000 6000 2000 4000 0 0 0 0 500 0 500 0 500 0 500 10000 400 500 6000 8000 4000 6000 2000 4000 0 0 0 0 500 0 500 0 500 0 500 Time (days) Figure 3 Epidemic curves using the different values for each media function. Functions (1) (dotted line), (2) (dashed line), and (3) (solid line) are shown when p = 10 (top), 100 (middle), 1000 (bottom). p and p are determined by equations (4) and (5) with γ = 1/4and Ic = 300. For 3 1 2 comparison, the standard SEIR model 6 with no media effect is also shown (dash-dotted line). Population Size Collinson and Heffernan BMC Public Health 2014, 14:376 Page 5 of 10 http://www.biomedcentral.com/1471-2458/14/376 a description of different methods for calculating R see [22]. For System (6), βN R =.(7) where N = S is the total population size of susceptibles at 120 time zero. Note that the calculation of R is independent of f (I, p). Comparison of media functions Figure 3 plots System (6) using Media Functions (1)-(3) for p = 10, 100, and 1000 (solid line), with and p (dot- 3 1 ted line) and p (dashed line) determined by Equations (4) and (5), and γ = 1/4and I = 300. For comparison, the 0 50 100 150 200 250 300 350 400 c Time (days) standard SEIR model with no media effect is also shown Figure 4 Data and fitted models. The epidemic curves are shown (dash-dotted line). This figure demonstrates that mass resulting from fitting System (6) with Media Functions (1)-(3) (dotted media will affect the epidemic curve. It also demonstrates line, dashed line, solid line) and no media (dash-dotted line) to 100 that the epidemic curve varies depending on the media days of epidemic data (boxes) generated using Media Function (3) function used. with p = 1000 and γ and σ from Table 1. For a further comparison of the media functions, we determine the values of the basic reproductive ratio R and the media constant p of f (I, p) based on epidemic data S E I R 10000 2000 10000 5000 1000 5000 0 0 0 0 0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 10000 2000 10000 5000 1000 5000 0 0 0 0 0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 10000 2000 10000 5000 1000 5000 0 0 0 0 0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 10000 2000 10000 5000 1000 5000 0 0 0 0 0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 Figure 5 Epidemic curves for different media functions. The epidemic curves are shown for Model (6) (dashed line) and the ABMC simulation (gray lines). There are 100 simulations plotted for each. In each panel, we also show the mean of the simulations (sold line). The first row has no media effect and rows (2-4) correspond to Functions (1)-(3). Here, I = 300 and p = 1000, γ = . c 3 Infected Population (I) Population Size Collinson and Heffernan BMC Public Health 2014, 14:376 Page 6 of 10 http://www.biomedcentral.com/1471-2458/14/376 Table 3 Key epidemic measurements for the SEIR model with (a) no media, (b) media function (1), (c) media function (2), and (d) media function (3) Model Peak Magnitude Peak time Epidemic end Total (E+I) (days) (days) (I) (a) 2676.1 21.155 76.6065 9254.8 2623.9 ± 204.09 21.005 ± 19.9 77.1863 ± 15.59 9542.4 ± 255.37 (b) 1061.5 21.7841 162.3647 7537.5 1095.6 ± 46.35 21.001 ± 19.99 168.077 ± 31.24 7578.13 ± 460.95 (c) 858.176 18.544 176.8229 7557.9 902.15 ± 78.85 17.7 ± 15.99 177.07 ± 23.6 7622.2 ± 328.8 (d) 1144.7 23.5169 154.3623 7586.0 1175.9 ± 175.09 22.01 ± 20.99 157.07 ± 28.005 7643.9 ± 229.2 The results of system 6 and the ABMC are listed in the top and bottom row of each section respectively with p = 1000, p and p determined using equations (4) and 3 1 2 (5), and γ = 1/4 and I = 300. 500 simulation runs are used for the ABMC calculations. generated from our models. For an example, Table 2 lists and no media (dash-dotted line). Table 2 and Figure 4 the values of β and p determined through a fitting rou- demonstrate that even though all of the models have a tine (in MATLAB - various routines were used and similar very similar shape and basic reproductive ratio R at the results were obtained) over the first 100 days of an epi- beginning of the epidemic, the resulting epidemic curves demic, with data generated using Media Function (3) with can still vary drastically depending on the media function σ and γ values listed in Table 1 and p = 1000. Figure 4 chosen. Similar observations were made using different plots the resulting epidemic curves of System (6) using numbersofdaysofdatausedtofit β and p, and differ- Media Functions (1)-(3) (dotted, dashed and solid lines) ent Media Functions to generate the data (not shown). Figure 6 LHS-PRCC results for I = 300 for peak magnitude. This sensitivity analysis is done with 1000 bins. The rows correspond to Function (1), Function (2) and Function (3), respectively. The columns of this PRCC figure correspond to p = 10, p = 100 and p = 1000. Here I = 300 and 3 3 3 c γ = . 4 Collinson and Heffernan BMC Public Health 2014, 14:376 Page 7 of 10 http://www.biomedcentral.com/1471-2458/14/376 Figure 7 LHS-PRCC results for I = 1000. This sensitivity analysis is done with 1000 bins. The rows correspond to Function (1), Function (2) and Function (3), respectively. The columns of this PRCC figure correspond to p = 10, p = 100 and p = 1000. 3 3 3 Therefore, it can be concluded that key epidemic mea- Sensitivity analysis surements – peak number of infections, time of peak, end Within public health settings, a goal is to identify key of epidemic, and total number of infections – will also vary characteristics of an epidemic that drive the infection depending on the Media Function chosen. with the hope of determining public health measures that System (6) is useful in describing the mean behaviour can be implemented so that control or eradication of the of an epidemic, but it is unable to provide estimates of pathogen can be achieved. By performing sensitivity anal- variation of important public health measures. A stochas- ysis on System (6) with Equations (1)-(3), parameters that tic simulation lends itself well to demonstrating variation most affect epidemic outcomes for each media function within an epidemic. Here, we employ an Agent-Based can be identified, informing policy makers so that appro- Monte Carlo (ABMC) simulation. Figure 5 shows 100 sim- priate public health measures can be put into place. To ulation runs of the ABMC simulation when no media is conduct the sensitivity analysis, we use Latin Hypercube involved (top), and when Media Functions (1)-(3) are used Sampling (LHS) and partial rank correlation coefficients (second to bottom rows). Each simulation run (gray line) (PRCC) [23]. is shown, as well the mean of the simulation runs (solid We first conduct the sensitivity analysis with I , p c 3 line) and solution of the System (6) with the correspond- and γ constant to directly compare the media functions. ing media functions (dashed line). Note that the mean Figures 6 and 7 show the PRCC values determined for (solid line) and solution to System (6) are in agreement. peak magnitude when I = 300 (Figure 6) and 1000 This figure demonstrates that variability in the key epi- (Figure 7) for Functions (1)-(3) (top to bottom) when p = demic measurement occurs. Table 3 lists the mean and 10, 100 and 1000 (left, middle and right columns respec- standard error of each epidemic measurement when no tively). These figures demonstrate that for each set of media is considered (section a), and Media Functions (1)- parameters the PRCC value is similar. This is true for each (3) are included in the simulation (Sections b-d). In each key epidemic measurement - the PRCC values are similar section we list the simulation mean and standard error between different values of p and I . However the PRCC 3 c (top row), and the solution of System (6). values between key measurements are different i.e. the Collinson and Heffernan BMC Public Health 2014, 14:376 Page 8 of 10 http://www.biomedcentral.com/1471-2458/14/376 Figure 8 PRCC for system (6). Results are shown for peak magnitude, peak time, epidemic end time, and total number of infections (left to right) when no mass media function is considered, and functions (1)-(3) are used (top to bottom). Here, p = 100 and p and p are determined by 3 1 2 Equations(4) and(5). PRCC values are different if comparing peak magnitude to This also explains the very different epidemic curves pro- epidemic end time (not shown). duced by System (6) when different media functions are For further sensitivity analysis, all parameters are employed, notwithstanding the similarities in the media allowed to vary. Figure 8 shows the result of the sensi- functions when plotted at a specific level of media. tivity analysis for all four outcomes key to public health: peak number of infections, peak time, epidemic end time, Conclusion and total number of infections. The results demonstrate Technology and media play an increasing role in daily life. that the SEIR model is more (or less) sensitive to certain Mass media that is transmitted via technological media parameters depending on what Mass Media Function (1)- can therefore be used to inform the public during pan- (3) is used. For example, changes in β have a large effect on demics and epidemics. An understanding of the effects peak magnitude when Function (1) is used, but it has lit- of media during an epidemic or pandemic threat can aid tletonoeffectwhenFunctions (2-3)represent theeffects in the development of public health policy. Of particu- of media. Variable p also has a large effect on peak mag- lar interest to public health are the effects of media on nitude when media is represented by Function (1). It has key epidemic measurements - peak magnitude of infec- a similar effect when Function (2) is used, but it has no tion, time of peak, end of epidemic, and total number of effect on the system utilizing Function (3). infections. The sensitivity analysis indicates that models that Mathematical modelling has been used to study the include mass media influence will greatly depend on effect of media on epidemics by employing functions in different parameters, depending on the media function the transmission terms of mathematical models [6-14]. chosen. This makes it very difficult for policy makers to A survey of the literature identified three functions that determine an effective public health intervention strategy. have been utilized to represent media in disease modelling Collinson and Heffernan BMC Public Health 2014, 14:376 Page 9 of 10 http://www.biomedcentral.com/1471-2458/14/376 [6-14]. We have conducted a comparison of these func- Inclusion of mass media reports and media waning in tions to determine the effects of media function on key a model of multiple seasons with seasonal forcing is an epidemic measurement and variability within these mea- interesting direction for study. surements. We first demonstrated that by including mass Competing interests media in System (6) the peak magnitude of the epidemic The authors declare that they have no competing interests. and the total number of infections would decrease. We also determined that the time to peak and the end of Authors’ contributions the epidemic would also occur earlier. However, we also MSC performed the literature review. JMH and MSC developed the models, conducted the analysis and wrote the paper. Both authors read and approved demonstrated that, although the functions are similar in the final manuscript. shape and magnitude, the resulting epidemic curve can be quite different (Figure 5). Therefore, the key epidemic Acknowledgements measurements that are used to inform public health policy The authors would like to thank Nicholas Geard, Amy Greer, James McCaw, will be different. Furthermore, we demonstrated that vari- Seyed Moghadas, Jianhong Wu and Huaiping Zhu for many helpful discussions. The authors would also like to thank the reviewers for their ability in the key epidemic measurements also depends on comments and suggestions that have added value to this study The work was the media function chosen (Figure 5 and Table 3). funded by an Ontario Early Researcher Award, Mathematics of Information A sensitivity analysis on System (6) with the differ- Technology and Complex Systems (MITACS), and the Natural Science and Engineering Research Council of Canada (NSERC). Simulations were ent media functions was also conducted. Obtained from conducted on Sharcnet. this analysis was the insight that some parameters are important for some outcomes and not for others. We can Received: 4 September 2013 Accepted: 4 April 2014 conclude that due to the different fixed functions result- Published: 17 April 2014 ing in very different epidemic behaviours, there is no clear References control strategy present. Also, as a reult of the differ- 1. Brauer F: Mathematical epidemiology is not an oxymoron. BMC Pub ent behaviours from the different media functions, we are Health 2009, 9(Suppl1):S2. 2. Neumann G, Noda T, Kawaoka Y: Emergence and pandemic potential unsure as to which is the best function to use to model of swine-origin H1N1 influenza virus. Nature 2009, 459:931–939. mass media. This suggests that a function representing 3. Public Health Agency of Canada: Highlights from the Canadian Pandemic media may not be the best course for modelling the effects Influenza Plan for the Health Sector. Canada: Public Health Agency of Canada; 2006. of media during an epidemic. We suggest that perhaps, a 4. Centers for Disease Control and Prevention: Estimating Seasonal Influenza- separate model compartment representing media reports, Associated Deaths in the United States: CDC Study Confirms Variability of Flu. such as those incorporated into surveillance data [24] Atlanta, GA, USA: Centers for Disease Control and Prevention; 2011. 5. Viswanath K, Ramanadhan S, Kontos EZ: Mass media. In Macrosocial could be used. Development of such a model is a course Determinants of Population Health. Chap.13 Edited by Galea S, Ann Arbor. for future work. MI, USA: Springer; 2007:275–294. Mass media reports can affect social behaviour, that 6. Liu R, Wu J, Zhu H: Media/psychological impact on multiple ultimately, affects transmission of disease. However, an outbreaks of emerging infectious diseases. Comput Math Methods Med 2007, 8:153–164. individual’s response to a media stimulus will wane over 7. Tchuenche JM, Dube N, Bhunu CP, Smith RJ, Bauch CT: The impact of time [13,25-29]. Models that employ a media function media coverage on the transmission dynamics of human influenza. such as those studied here are difficult to modify to BMC Pub Health 2011, 11(Suppl1):S5. 8. Pang J, Cui J-a: An SIRS epidemiological model with nonlinear involve a waning response to a media stimulus over the incidence rate incorporating media coverage. In Second International age of an epidemic. This can be incorporated easily into a Conference on Information and Computing Science. USA: IEEE; model whereby media is represented as a model compart- 2009:116–119. 9. Cui J-a, Sun Y, Zhu H: The impact of media on the control of infectious ment. A study of the effects of ‘waning media’ is a course diseases. J Dyn Differential Equations 2008, 20:31–53. for future work. 10. Cui J-a, Tao X, Zhu H: An SIS infection model incorporating media The current study has employed a simple SEIR model coverage. Rocky Mt J Math 2008, 38:1323–1334. 11. Li Y, Ma C, Cui J: The effect of constant and mixed impulsive that includes three parameters. This model implicitly vaccination on SIS epidemic models incorporating media coverage. assumes that individuals mix at random in the population, Rocky Mot J Math 2008, 38:1437–1455. the age and sex of individuals is unimportant, and that 12. Schweitzer F, Mach R: The epidemics of donations: Logistic growth and power laws. PLoS ONE 2008, 3:e1458. the population size remains constant over the epidemic. 13. Sun C, Yang W, Arino J, Khan K: Effect of media-induced social These assumptions are not true reflections of reality. An distancing on disease transmission in a two patch setting. Math interesting direction for future work is to consider the Biosci 2011, 230:87–95. 14. Xiao D, Ruan S: Global analysis of an epidemic model with effects of mass media reports on individual decision mak- nonmonotone incidence rate. Math Biosci 2007, 208:419–429. ing strategies, and study how mass media can impact the 15. Allen LJS: An Introduction to Stochastic Epidemic Models. In contact structure of a population network [30]. Mathematical Epidemiology. Lecture Notes in Mathematics. Volume 1945. Edited by Brauer F, van den Driessche P, Wu J. Lubbock, TX, USA: Springer; ThesimpleSEIRmodel employed in this studywas 2008:81–130. used to study one season of influenza. Typically, seasonal 16. Heffernan JM, Wahl LM: Monte Carlo estimates of natural variation in forcing is used to model multiple seasons of influenza. HIV infection. J Theor Biol 2005, 236:137–153. Collinson and Heffernan BMC Public Health 2014, 14:376 Page 10 of 10 http://www.biomedcentral.com/1471-2458/14/376 17. Keeling MJ: Metapopulation moments: coupling, stochasticity and persistence. J Anim Ecol 2000, 69:725–736. 18. Keeling MJ: Multiplicative moments and measures of persistence in ecology. J Theor Biol 2000, 205:269–281. 19. Krishnarajah I, Cook A, Marion G, Gibson G: Novel moment closure approximations in stochastic epidemics. Bull Math Biol 2005, 67:855–873. 20. Pourbohloul B, Ahued A, Davoudi B, Meza R, Meyers LA, Skowronski DM, Villasenor I, Galvan F, Cravioto P, Earn DJD, Dushoff J, Fisman D, Edmunds WJ, Huper N, Scarpino SV, Trujillo J, Lutzow M, Morales J, Contreras A, Chavez C, Patrick DM, Brunham RC: Initial human transmission dynamics of the pandemic (H1N1) 2009 virus in North America. Influenza Other Respi Viruses 2009, 3:215–222. 21. Mostaço-Guidolin LC, Pizzi NJ, Moghadas SM: A classical approach for estimating the transmissibility of the 2009 H1N1 pandemic. Can Appl Math Q 2011, 19:185–194. 22. Heffernan JM, Smith RJ, Wahl LM: Perspectives on the basic reproductive ratio. J R Soc Interface 2005, 2:281–293. 23. Blower SM, Dowlatabadi H: Sensitivity and uncertainty analysis of complex models of disease transmission: An HIV model, as an example. Int Stat Rev 1994, 62:229–43. 24. Mykhalovskiy E, Weir L: The global public health intelligence network and early warning outbreak detection. Can J Public Health 2006, 97(1):42–44. 25. Fanti KA, Vanman E, Henrich CC, Avraamides MN: Desensitization to media violence over a short period of time. Aggress Behav 2009, 35:179–187. 26. Griffiths MD, Shuckford GLJ: Desensitization to television violence: a new model. New Ideas Psychol 1989, 7:85–89. 27. Kinnick KN, Krugman DM, Cameron GT: Compassion fatigue: Communication and burnout towards social problems. JMass Commun Q 1996, 73:687–707. 28. McBurney DH, Balaban CD, Christopher DE, Harvey C: Adaptation to capsaicin within and across days. Physiol Behav 1997, 61:181–190. 29. Smith KC, Rimal RN, Sandberg H, Storey JD, Lagasse L, Maulsby C, Rhoades E, Barnett DJ, Omer SB, Links JM: Understanding newsworthiness of an emerging pandemic: International newspaper coverage of the H1N1 outbreak. Influenza Other Respi Viruses 2012, 7:847–853. 30. Wang Y, Cao J, Jin Z, Zhang H, Sun G-Q: Impact of media coverage on epidemic spreading in complex networks. Physica A: Stat Mech Appl 2013, 392(23):5824–5835. doi:10.1186/1471-2458-14-376 Cite this article as: Collinson and Heffernan: Modelling the effects of media during an influenza epidemic. BMC Public Health 2014 14:376. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color ﬁgure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit
BMC Public Health – Springer Journals
Published: Dec 1, 2014
Keywords: public health; medicine/public health, general; epidemiology; environmental health; biostatistics; vaccine
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.