 Research article
 Open Access
 Open Peer Review
 Published:
Construction of the influenza A virus transmission tree in a collegebased population: cotransmission and interactions between influenza A viruses
BMC Infectious Diseasesvolume 16, Article number: 38 (2016)
Abstract
Background
Coinfection of different influenza A viruses is known to occur but how viruses interact within coinfection remains unknown. An outbreak in a college campus during the 2009 pandemic involved two subtypes of influenza A: persons infected with pandemic A/H1N1; persons infected with seasonal A/H3N2 viruses; and persons infected with both at the same time (coinfection). This provides data to analyse the possible interaction between influenza A viruses within coinfection.
Methods
We extend a statistical inference method designed for outbreaks caused by one virus to that caused by two viruses. The method uses knowledge of which subtype each case is infected with (and whether they were coinfected), contact information and symptom onset date of each case in the influenza outbreak. We then apply it to construct the most likely transmission tree during the outbreak in the college campus.
Results
Analysis of the constructed transmission tree shows that the simultaneous presence of the two influenza viruses increases the infectivity and the transmissibility of A/H1N1 virus but whether it changes the infectivity of A/H3N2 is unclear. The estimation also shows that cotransmission of both subtypes from coinfection is low and therefore coinfection cannot be sustained on its own.
Conclusions
This study suggests that influenza A viruses within coinfected patients can interact in some ways rather than transmit independently, and this can enhance the spread of influenza A virus infection.
Background
Cocirculation of multiple types and subtypes of influenza virus has been commonly observed in human populations [1, 2]. With advanced molecular techniques [3, 4], it is now becoming possible to rapidly detect different subtypes or strains of a disease agent within infected patients. There is accumulating evidence to show that the phenomenon that multiple subtypes of influenza A virus infect an individual simultaneously (i.e. coinfection) is not as rare as we previously thought [5–10]. Hence it is interesting to know how the simultaneous presence of two strains alters, compared to singly infected individuals, the transmissibility of each subtype and of both subtypes together. Some observations of transmission involved with coinfected individuals have been reported [11, 12]. One important observation reported is individuals coinfected with pandemic A/H1N1 and seasonal A/H3N2 influenza viruses within one outbreak at a college campus in Beijing, China during the 2009 pandemic [12]. Liu et al. [12] provides direct evidence that individuals coinfected with different subtypes can transmit viruses separately or simultaneously and provides detailed data for us to quantify the interactions between virus strains. The transmission characteristics can be directly estimated from the transmission tree; however, the difficulty for constructing the transmission tree is that some contacts were missed. Fortunately, recent development in statistical inference [13, 14] allows us to construct the transmission tree of a single influenza A virus from such a partial contact network. In this short report we extend this inference method to construct the transmission tree that includes two influenza A viruses and their coinfection. From the constructed tree we estimate the impact of coinfection on transmission. We use two parameters to characterise the impact: the cotransmission rate and an interaction parameter. Here we define the former as the rate at which two strains simultaneously transmit from doubly infected individuals to susceptible individuals; and the latter as the ratio of transmissibility of each single virus from coinfected individuals to the transmissibility of each single virus from singly infected individuals.
Data
Here we briefly summarize the data collection method and data of Liu et al. [12]. Investigations were conducted on all influenza like illness (ILI) cases identified during the outbreak. Epidemiological, clinical and contact tracing data were collected by interviewing patients and retrieving medical records. Viruses were identified by reversetranscription polymerase chain reaction assays followed by sequence analysis. The heamagglutination inhibition tests were used to detect antibodies to both viruses. The outbreak is reported to have taken place within three buildings (two dormitories and one college clinic). No other cases at the college were reported. Buildings 1 and 2 (with a total membership of 235 and 191 persons, respectively) are next to each other and there is restricted access between the two and to the wider community. Forty five ILI cases were reported from 31 August to 10 September and forty (N = 40) had laboratoryconfirmed influenza A infection. Three different types of infection were reported: 22 infected with pandemic A/H1N1 virus, 12 infected with seasonal A/H3N2 virus and six coinfected with both influenza A viruses. In their sequences no substantial differences were observed between patients with mixed and single infections in either pandemic A/H1N1 or seasonal A/H3N2 virus. The clinical features were similar for patients with different infections and the six coinfected patients showed no more severe symptoms than the singly infected patients. Contacts between infected people are shown in Fig. 2 of Liu et al. [12] but this only extends to the contact network within one dormitory. The ‘index’ case with pandemic influenza A/H1N1 infection was a college student whose symptoms first occurred on 31 August, 2 days after his returning to college. Except for the index case, all patients with A/H1N1 infection had not left the campus during the previous week. In contrast, the source of seasonal A/H3N2 virus infection cannot be determined exactly although available data indicate that A/H3N2 virus might have been prevailing in the college when the pandemic H1N1 virus was introduced. Before the isolation of cases and the initiation of prophylaxis among the campus population (5 September 2009), several patients visited the college clinic and the mixing between students of different dormitories was not frequent in comparison to the mixing between students within each dormitory.
Methods
For a fully traced transmission tree (i.e., the information of the infector v and time of symptom onset t are collected), the infector v(i) for each case i (except the index case) and the duration between symptom onset of case i and symptom onset of its infector v(i): t_{ i }t_{v(i)} should be known. From these it is straightforward to estimate the generation interval distribution and transmissibility of infection. In reality, however, it is rare and difficult to record all the information. Based on the partially known contact tracing data and dates of symptom onset, Hens et al. [14] illustrated an inference method to reconstruct the most likely transmission tree that involved with only one virus. Here we further develop it to a transmission tree during an outbreak that involves with two viruses of similar epidemiological characteristics.
In general, a possible transmission tree can be described by p_{ ij }(v,w,φ;θ), the probability that case j is the infector of case i, given the duration between symptom onset of case i and case j, given the information on the possible infector v and the known contacts w, and given the types φ of infection of both cases. Following Hens et al [14], its total loglikelihood is given by:
The sum runs through all the nonzero p_{ ij }. Here g(∆t θ) denotes the probability density of the generation interval distribution of influenza infection, with θ representing the set of parameters that characterise the probability density distribution. Different distributions such as gamma, lognormal and Weibull can be used to describe the distribution of generation intervals [14]. Here we assume it follows a Weibull distribution:
The distribution has two parameters (i.e., θ = {a, b}): scale parameter a and shape parameter b, such that the mean is T = aΓ(1 + 1/b) and the variance is σ^{2} = a^{2}[Γ(1 + 2/b) Γ(1 + 1/b)^{2}]. Here Γ() is the Gamma function.
Compared with outbreaks that involved one virus and one transmission process [14], this outbreak [12] involved two viruses and five possible transmission processes: from A/H1N1 to A/H1N1; from A/H3N2 to A/H3N2; from coinfection to A/H1N1; from coinfection to A/H3N2 and from coinfection to coinfection. However, the data given in Liu et al. [12] only provide the relevant information for three transmission processes: from A/H1N1 to A/H1N1; from coinfection to A/H1N1; and from coinfection to coinfection. The simple calculations show that the mean generation intervals (and their standard deviations) for the three transmission processes are 1.8 (0.8), 1.2 (0.4), 1.6 (0.6) days, respectively. In this data there is no evidence that they differ. As no data is available for the transmission processes that were involved with A/H3N2 [12], it is difficult to estimate their generation intervals and judge how they differ from those generation intervals involved with pandemic A/H1N1. Nevertheless, a recent systematic review [15] indicates that the generation interval of pandemic A/H1N1 virus was similar to that of the seasonal flu. Further, as observed by Liu et al. [12], the three different types of infections have similar epidemiological characteristics. In view of these, the same generation interval distribution is assumed for the three types of infection.
The probability that case i is infected by case j, p_{ ij }, can be calculated as the probability of observing the duration between the symptom onsets in cases i and j, g(t_{i}t_{j}θ), times the probability of a potentially infectious contact between i and j, π_{ ij }, normalized by the probability of i being infected by any other case k:
The probability of a potentially infectious contact between i and j, π_{ ij }, is based on the contact information (v,w) collected during the outbreak and the types φ of infection of both cases i and j. To distinguish different types of infection and to reflect the fact that there is only limited mixing between students in building 1 and building 2, we define the following,
The variable w defines the extent of contact rates between students in two dormitories in relation to contact rates within dormitories: w = 0 implies that contacts between dormitories are forbidden and w = 1 suggests that contacts between dormitories are the same as that within each dormitory. Student access to the college clinic was not restricted. The original numberings of patients in Fig. 2 of Liu et al. [12] are given in accordance with the order of symptom onset within each building. For the convenience of our analysis, the forty patients have been reindexed in the order of symptom onset as 1, 2, …, 40 (see Fig. 1). If cases i and j form a likely transmission pair (i.e., v(i) = j) and there is only one possible infector, π_{ ij }(v,w,φ) = ψ_{ ij } and p_{ ij }(v,w,φ;θ) =1. If the outbreak investigation reveals that case i is not the index case and does not contact any of the i1 cases that developed symptoms before case i, the probability of a potentially infectious contact is π_{ ij }(v,w,φ) = ψ_{ ij }/η_{i1} Here η_{i1} is the ‘effective’ number of infections that developed symptoms before case i and are of the same type of infection as case i or coinfection, and is calculated as η_{i − 1} = ∑ _{k = 1} ^{i − 1} ψ_{ ik }.
Given the contact information (v,w), the most likely transmission tree can be obtained by finding the values of parameters θ = {a, b} that maximize the total loglikelihood (1). The downhill simplex method [16] was used to locate the maximumlikelihood estimate (MLE), \( \widehat{\boldsymbol{\uptheta}} \), of the parameters. We also used the MCMC method to sample the values of parameters θ and find their medians, 2.5 percentile and 97.5 percentile.
From the probabilities that case i is infected by case j: {p_{1i}, …, p_{ Ni }}, we can sample a transmission tree in which all cases are connected and case 1 is the index case (Fig. 1) and further estimate transmissibility. The case reproduction number is the average number of secondary cases generated by primary cases, which measures the transmissibility of infection. The overall casereproduction number on day t, R_{ t }, can be estimated by summing over all these infectious contacts:
We can further estimate the infection typespecific reproductive numbers when both case i and case j are of the same type of infection:
Here the symbol δ_{ij,TYPE} is defined as: δ_{ij,TYPE} =1 if both cases i and j are of the infection TYPE; δ_{ij,TYPE} = 0 otherwise. To characterise the interaction between two types of virus within coinfected individuals, we further estimate the reproductive numbers of A/H1N1 and A/H3N2 due to coinfected individuals as:
Here ∆ _{ij,TYPE} is defined as: ∆ _{ij,TYPE} =1 if infector case j is coinfected while the infectee case i is singly infected with type TYPE; otherwise ∆ _{ij,TYPE} =0. The strain interactions within coinfections are estimated as:
Here ϕ_{1} measures the effect on infectivity of A/H1N1 within coinfections and ϕ_{2} measures the effect on infectivity of A/H3N2 within coinfections.
Ethics considerations
Data that were used in the analysis of this study were extracted from a previous study [12] and hence did not require Human Resource Ethics committee approval.
Results and discussion
The data shown in Fig. 2 of Liu et al. [12] indicates that the index case is infected with pandemic A/H1N1 virus. Among the six cases that the index case infects, three patients (i.e., case 2, 5, and 15) are coinfected with both A/H1N1 and A/H3N2. It was assumed that seasonal A/H3N2 virus was endemic [12], however, the infectious contacts with A/H3N2 of these three coinfected cases were unknown. Several possible scenarios are possible: cases 2, 5 and 15 were exposed to pandemic A/H1N1 virus when they were still infectious with A/H3N2; cases 2, 5 and 15 were further exposed to seasonal A/H3N2 virus soon after they had acquired pandemic A/H1N1 virus from the index case; the three cases were infected simultaneously with both A/H1N1 and A/H3N2 viruses from the index case who was actually coinfected before contact with cases 2, 5 and 15 but was incorrectly typed as a single infection with pandemic A/H1N1 virus. For simplicity, we use all the contact information for cases within building 1, especially the pathways from the index case to the three coinfections, ignoring the possible pathways for transmitting background endemic seasonal A/H3N2 virus.
The maximum likelihood estimates of the infective contact probabilities p_{ ij }(v,w,φ;θ) are listed in Table 1 from which one sample transmission tree is shown in Fig. 1. The MLEs of the generation interval distribution g(∆tθ) and the time course of the mean case reproductive number R_{ t } are shown in Fig. 2. The generation interval has a mean of 1.72 days and a standard deviation (SD) of 0.74 days. The mean case reproductive number over the whole outbreak is 1.57 with SD = 0.022. Before the isolation of cases and the initiation of prophylaxis among the campus population occurred (5 September 2009), the reproductive number is estimated to be 2.80 with SD = 0.016; after this it declines to 0.341 with SD = 0.047. This clearly shows the effectiveness of isolation and prophylaxis.
To measure the interactions between two influenza A viruses, we first estimated the values of infection typespecific reproductive numbers before the initiation of isolation and prophylaxis occurred. Because most transmissions that involve A/H1N1 and coinfections are known, the estimates of the infection typespecific reproductive numbers are stable at R _{ e } ^{1} = 1.44 and R _{ e } ^{C} = 0.67, and the reproductive number of A/H1N1 due to coinfection is R _{e,C} ^{1} = 2.01. In contrast, transmissions that involve A/H3N2 were missing and so estimates of their reproductive numbers show some uncertainty: R _{ e } ^{2} has a mean 0.48 and 95 % confidence interval of [0.45, 0.51], and the reproductive number of A/H3N2 due to coinfection is R _{e,C} ^{2} = 1.60 [1.58, 1.62]. Although the estimate of R _{ e } ^{1} for A/H1N1 is in agreement with the usual estimates [15, 17], that of R _{ e } ^{2} for A/H3N2 is much smaller than other estimates which range from 1.19 to 1.37 [17]. The likely reason for this difference lies in the fact that pandemic A/H1N1 is a novel virus while A/H3N2 is an endemic seasonal virus in the study region [12] so some preexisting immunity against A/H3N2 due to the previous infections and/or vaccination reduces its susceptibility and hence its spread. In the presence of coinfection, infectivity of A/H1N1 increases moderately with ϕ_{1} = 1.4. However, the increased infectivity of A/H3N2 in the presence of coinfection is large: ϕ_{2} = 3.28 [95 % CI: 3.18—3.38] and so coinfection is estimated to contribute greatly to the outbreak. This is not inconsistent with Liu et al. [12]’s observation that “no patients had detectable hemagglutination inhibiting antibodies against pandemic H1N1 virus in their acutephase samples”. Although the presence of coinfection can increase the infectivity of each component influenza virus, coinfection itself cannot spread successfully because the cotransmissibility measured by R _{ e } ^{C} is estimated to be 0.67, which is less than one.
The above results are obtained under the assumption of w = 25 %, which is used to reflect the observation “although there were activities outside the dormitory, the members of the 2 buildings did not interact with each other to any significant extent” (p1361, second column, Liu et al. [12]). If different values of w are chosen, the estimation of transmissibility involved with A/H3N2 virus changes although the results of the overall transmissibility and those involved with A/H1N1 remain nearly the same. With a low value of w (i.e. the mixing between two dormitories become more restrictive), the reproductive number of A/H3N2 due to coinfection (R _{e,C} ^{2} ) decreases while R _{ e } ^{2} increases; this consequently reduces the value of ϕ_{2}. For example, the estimate of ϕ_{2} becomes 3.00 [2.92, 3.13] when w =0.1. This is the outcome because without the index case of A/H3N2 infection, the only possible infectors of A/H3N2 cases 11, 12, 13, 14 in building 2 are coinfection cases 2 and 5, which is independent of the limited mixing rate between two dormitories (w). In the situation without limitation in mixing rates between two dormitories (i.e., w = 1.0), the estimate of ϕ_{2} is 4.57 [4.42, 4.70]. This result implies that in the absence of any unidentified index cases of A/H3N2 infection, coinfection can enhance the transmissibility of each component virus.
However, all the cases reported in building 2 were infections with A/H3N2 virus alone. In view of the restrictive mixing between two dormitories, it is likely that there is an unknown index case of A/H3N2 infection within building 2, although it was not reported. An analysis listed in the Additional file 1 shows that such a hidden index case could change the above conclusion: the reproductive number of A/H3N2 due to coinfection (R _{e,C} ^{2} ) might not exceed the reproductive number by its own (R _{ e } ^{2} ) and therefore coinfection could not enhance the infectivity of A/H3N2. Jombart et al. [18] have developed a Bayesian method to reconstruct disease outbreaks by combining epidemiological and genomic data. This may allow for the tests of whether coinfection cases in building 1 are the infectors of A/H3N2 cases in building 2 or whether there is a hidden index case of A/H3N2 infection in building 2. Unfortunately, the representative sequences deposited in GenBank by Liu et al. [12] were not complete genome sequences and were not marked with the relevant symptom onset information. Hence they cannot help to distinguish and/or find the potential sources of A/H3N2 infection in the transmission tree. Conditional on the available information, the evidence about how coinfection alters the infectivity of A/H3N2 virus is lacking. Nevertheless, the conclusion about enhanced infectivity of A/H1N1 within coinfection is not affected.
The original method of reconstructing the transmission tree by Hens et al. [14] relies on three assumptions: all cases are observed; all of them except the index case are infected by another observed case; and the generation interval distribution remains unchanged. To apply to such outbreaks as studied here involving two viruses, some assumptions have been strengthened. There must be two index cases that were singly infected with different viruses or one index case that was coinfected with the two viruses. There are more than one transmission processes. In our study example, there are five transmission processes: A/H1N1 to A/H1N1; A/H3N2 to A/H3N2; coinfection to A/H1N1; coinfection to A/H3N2 and coinfection to coinfection. Unfortunately the available data does not provide direct information for all the transmission processes, and it is not directly possible to assess the heterogeneity in generation intervals among transmission processes. In view of the similar estimates for generation intervals of different transmission processes [15], the generation interval distribution has been assumed to be unchanged over different types of infection as well as over the course of the outbreak.
Another aspect is how coinfection is generated. It could be due to an infection with one virus becoming a coinfection, or it could be a consequence of a cotransmission. And it is possible that going from A/H1N1 to coinfection is easier than going from A/H3N2 to coinfection, or vice versa. Though this is an interesting issue [19] the limited information that we can collect from Liu et al. [12] cannot allow us to detect the order in which the two viruses are acquired by an coinfected individual and therefore no way to investigate the effect of the order in which the two viruses are acquired.
Conclusion
Reconstructing transmission trees provides useful information about generation interval and transmission rate of infectious diseases, which are important for designing containment strategies. In this study, the method of reconstructing the plausible transmission tree from the incomplete data of an outbreak caused by one virus (Hens et al. [14]) has been extended to the outbreak caused by two influenza A viruses. Our estimates of epidemiological characteristics such as the generation interval and the transmission rate of influenza A virus are well within the ranges estimated by others [15, 17]. Our estimation shows that although coinfection with A/H1N1 and A/H3N2 viruses cannot be sustained by cotransmission, it enhances the single transmission of both viruses. However, the concluded enhancement of A/H3N2 virus infectivity within coinfection should be taken with caution owing to the unknown infection source of seasonal H3N2 virus.
Crossimmunity, which characterises the interaction between different viruses when one virus reinfects individuals recovered from previous infection with another virus, has been well recognized and measured. Due to their relative rareness, coinfection and interactions between viruses within coinfections have not yet attracted the attention they deserve. To our knowledge this is the first analysis that estimates the interactions between influenza A viruses within coinfection. Theoretical modelling illustrates their potential role in generating the recurrent epidemics and alternation of the dominant virus in seasonal influenza [20, 21]. Surely this urges more empirical studies to investigate this important issue of influenza and other infectious diseases caused by multiple strains.
Abbreviations
 ILI:

influenza like illness
 MLE:

maximum likely estimate
 SD:

standard deviation
References
 1.
Denoeud L, Turbelin C, Ansart S, Valleron AJ, Flahault A, Carrat F. Predicting pneumonia and influenza mortality from morbidity data. PLoS ONE. 2007;2(5):e464. doi:10.1371/journal.pone.0000464.
 2.
Finkelman BS, Viboud C, Koelle K, Ferrari MJ, Bharti N, Grenfell BT. Global patterns in seasonal activity of influenza A/H3N2, A/H1N1, and B from 1997 to 2005: viral coexistence and latitudinal gradients. PLoS ONE. 2007;12:e1296. doi:10.1371/journal.pone.0001296.
 3.
Pajak B, Stefanska I, Lepek K, Donevski S, Romanowska M, Szeliga M, et al. Rapid differentiation of mixed influenza A/H1N1 virus infections with seasonal and pandemic variants by multitemperature singlestranded conformational polymorphism analysis. J Clin Microbiol. 2011;49:2216–21.
 4.
Nie S, Roth RB, Stiles J, Mikhlina A, Lu X, Tang YW, et al. Evaluation of Alere I influenza A & B for rapid detection of influenza A and B. J Clin Microbiol. 2014;52:3339–44.
 5.
Sonoguchi T, Naito H, Hara M, Takeuchi Y, Fukumi H. Crosssubtype protection in humans during sequential, overlapping, and or concurrent epidemics caused by H3N2 and H1N1 influenzaviruses. J Infect Dis. 1985;151:81–8.
 6.
Ghedin E, Fitch A, Boyne A, Griesemer S, DePasse J, Bera J, et al. Mixed infection and the genesis of influenza virus diversity. J Virol. 2009;83:8832–42.
 7.
Perez DR, Sorrell E, Angel M, Ye J, Hickman D, Pena L et al. Fitness of pandemic H1N1 and seasonal influenza A viruses during coinfection. PLoS Current. 2009; doi:10.1371/currents.RRN1011.
 8.
Peacey M, Hall RJ, Sonnberg S. Pandemic (H1N1) 2009 and seasonal influenza A(H1N1) coinfection, New Zealand, 2009. Emerg Infect Dis. 2010;16:1618–20.
 9.
Lee N, Chan PKS, Lam W, Szeto CC, Hui DS. Coinfection with pandemic H1N1 and seasonal H3N2 influenza viruses. Ann Intern Med. 2010;152:618–9.
 10.
Myers UA, Kasper MA, Yasuda CY, Savuth C, Spiro DJ, Hallpin R, et al. Dual infection of novel influenza viruses A/H1N1 and A/H3N2 in a cluster of Cambodian patients. Am J Trop Mede Hyg. 2011;85:961–3.
 11.
Toda S, Okamoto R, Nishida T, Nakao T, Yoshikawa M, Suzuki E, et al. Isolation of influenza A/H3 and B viruses from an influenza patient: confirmation of coinfection by two influenza viruses. Jpn J Infect Dis. 2006;59:142–3.
 12.
Liu W, Li ZD, Tang F, Wei MT, Tong YG, Zhang L, et al. Mixed infections of pandemic H1N1 and seasonal H3N2 viruses in 1 outbreak. Clin Infect Dis. 2010;50:1359–65.
 13.
Wallinga J, Teunis P. Different epidemic curves for severe acute respiratory syndrome reveal similar impacts of control measures. Am J Epidemiol. 2004;160(6):509–16.
 14.
Hens N, Calatayud L, Kurkela S, Tamme T, Wallinga J. Robust reconstruction and analysis of outbreak data: influenza A(H1N1)v transmission in a schoolbased population. Am J Epidemiol. 2012;176:196–203.
 15.
Boëlle PY, Ansart S, Cori A, Valleron AJ. Transmission parameters of the A/H1N1 (2009) influenza virus pandemic: a review. Influenza Other Respi Viruses. 2011;5:306–16.
 16.
Nelder JA, Mead R. A simplex method for function minimization. Comput J. 1965;7:308–13.
 17.
Biggerstaff M, Cauchemez S, Reed C, Gambhrir M, Fineli L. Estimates of the reproduction number for seasonal, pandemic, and zoonotic influenza: a systematic review of the literature. BMC Infect Dis. 2014;14:480.
 18.
Jombart T, Cori A, Didelot X, Cauchemez S, Fraser C, Ferguson N. Bayesian Reconstruction of Disease Outbreaks by Combining Epidemiologic and Genomic Data. PLoS Comput Biol. 2014;10(1):e1003457. doi:10.1371/journal.pcbi.1003457.
 19.
Davis BM, Aiello AE, Dawid S, Rohani P, Shrestha S, Foxman B. Influenza and communityacquired pneumonia interactions: the impact of order and time of infection on population patterns. Am J Epidemiol. 2012;175:363–7.
 20.
Zhang XS, Cao KF. The impact of coinfections and their simultaneous transmission on antigenic diversity and epidemic cycling of infectious diseases. BioMed Res Intern. 2014; Article ID 375862. http://dx.doi.org/10.1155/2014/375862.
 21.
Zhang XS. Strain interactions as a mechanism for dominant strain alternation and incidence oscillation in infectious diseases: seasonal influenza as a case study. PLoS ONE. 2015;10(11):e0142170. doi:10.1371/journalpone.0142170.
Acknowledgements
This research was funded by the Public Health England. The authors gratefully acknowledge the assistance of Tom Nichols and also express gratitude to Drs Niel Hens and Michael Worobey for their constructive and helpful comments.
Author information
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Conceived and designed the study: XSZ and DD; Performed the study: XSZ; wrote the paper: XSZ and DD. Both authors read and approved the final manuscript.
Additional file
Additional file 1:
Effect of an (unidentified) index case of A/H3N2 infection in building 2. (DOCX 186 kb)
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. 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.
About this article
Received
Accepted
Published
DOI
Keywords
 Coinfection
 Cotransmission
 Inference
 Influenza
 Outbreak
 Strain interactions