Modelling H5N1 transmission in UK poultry
Control of a highly pathogenic H5N1 avian influenza outbreak in the GB poultry flock
Electronic Supplementary Material
James Truscott*1, Tini Garske*1,2, Irina Chis-Ster1, Javier Guitian3, Dirk Pfeiffer3, Lucy Snow4, John Wilesmith2,5, Neil M Ferguson1, Azra C Ghani2+.
1. Department of Infectious Disease Epidemiology, Imperial College London
2. Department of Epidemiology & Population Health, London School of Hygiene & Tropical Medicine
3. Epidemiology Unit, Royal Veterinary College
4. Centre for Epidemiology and Risk Analysis, Veterinary Laboratories Agency
5. Department for the Environment, Food and Rural Affairs
1. Analysis of the Structure of the GB Poultry Flock 3
1.1. Poultry Register Data 3
1.2. Network Data 4
2. Natural History & Epidemiological Parameters 6
2.1. Course of infection in individual birds 6
2.2. Susceptibility to infection and asymptomatic infection 7
2.3. Duration of infection, viral shedding and mortality 7
2.4. Effect of vaccination 9
2.5. Within-farm dynamics 11
2.6. Interventions 15
3. Model Details 16
3.1. Spatial transmission 18
3.2. Network transmission 20
3.3. Fixed network transmission 20
3.4. Model calibration and risk calculation 25
4. Sensitivity Analyses: Scenarios without controls 27
4.1. Incursion scenarios 27
4.2. Proportion of spatial and periodic group transmission 31
4.3. Scaling of infectiousness with number of birds at a premise 33
4.4. Density-dependent spatial transmission 35
5. Additional Results & Sensitivity Analyses: Impact of interventions 36
5.1. Single incursions 36
5.2. Proportion of spatial and periodic group transmission 40
5.3. Density-dependent spatial transmission and size-dependent contact rate 41
5.4. Shortening the infectious period through faster implementation of interventions 42
5.5. Sensitivity to intervention parameters 44
5.6. Sensitivity to vaccine parameters 45
6. References 47
1. Analysis of the Structure of the GB Poultry Flock
1.1. Poultry Register Data
The GB Poultry Register Data (PRD), provided by the Department for the Environment, Food and Agriculture (Defra), contains details of the spatial location, husbandry practices and types of animals kept for poultry premises in Great Britain. It was a legal requirement that all commercial poultry premises keeping 50 or more birds register with Defra by 28th February 2006. This database should therefore be an accurate representation of the commercial poultry population of GB. However, the extent to which this has been achieved is not known. The database also includes 3292 (14%) premises reporting less than 50 birds. These premises are included in our analyses and in model parameterisation (except where explicitly stated otherwise). However it should be noted that it is not known how representative this is of this population of small holdings since registrations of holdings with less than 50 birds were voluntary and it is likely that there are many more small holdings not registered.
The dataset records information on 23,516 premises holding 271 million birds. The spatial location is known for 23,407 (99.5%) premises; in the analyses presented here we restrict to those with known spatial location. There were statistically significant differences between the characteristics of those with and without spatial location data; those without location data were significantly less likely to keep layer chickens (p=0.003), more likely to keep broiler chickens (p=0.01), more likely to have 50 or less birds (p<0.0001) and less likely to be free range (p=0.03) than those with location data. However, as the overall proportion of premises with missing location data is small we do not believe this substantially biases our analyses.
For model parameterisation the species and husbandry purposes were combined into a single set of categories (Table S1). Where more than one category was present at a single premise the premise was assigned to be in the category with the largest number of birds. The data were further stratified according to the party information supplied in the dataset into those belonging to multi-site companies (595 premises from 11 multi-site companies holding 45 million birds) and single-site premises. This structure (primarily relating to broiler and layer chickens) is included in the models.
Table S1: Species/husbandry classification used in the models. Premises are included once in this table if they keep multiple species/husbandry types and are listed in the classification in which they keep the largest number of birds. The number of birds is the total number at the premise.
Classification / Company / Number (%) of Premises / Number of BirdsChicken Broilers / 10 / 286 (1.2%) / 26.4 million
30 / 119 (0.51%) / 17.7 million
40 / 138 (0.59%) / 17.5 million
50 / 64 (0.27%) / 7.2 million
140 / 207 (0.88%) / 20.6 million
independent / 985 (4.2%) / 60.8 million
small holdings / 264(1.1%) / 31,600
Chicken Layers / 70 / 101 (0.43%) / 5.4 million
80 / 20 (0.09%) / 1.8 million
90 / 14 (0.06%) / 1.7 million
130 / 223 (0.95%) / 12.2 million
independent / 1197 (5.1%) / 19.2 million
small holdings / 5562 (24%) / 0.47 million
Turkeys / 110 / 45 (0.19%) / 3.5 million
independent / 636 (2.7%) / 7.4 million
small holdings / 287 (1.2%) / 57,100
Reared for shooting / independent / 6720 (29%) / 53.9 million
small holdings / 2389 (10.2%) / 0.63 million
Ducks & Geese / independent / 265 (1.1%) / 4.9 million
small holdings / 751 (3.2%) / 73,100
Other / independent / 419 (1.8%) / 9.3 million
small holdings / 2713 (11.6%) / 0.27 million
1.2. Network Data
Network data were obtained from a sample of single-site premises, multi-site premises, poultry slaughterhouses and catching companies through self-completed questionnaires. The information was collected through postal questionnaires completed by personnel at the premises. Given the rapid timescale under which the research was taken, it was not possible to pilot or validate these questionnaires.
In total, 3989 poultry premises, 95 slaughterhouses and 45 catching companies are included, although some of these did not complete questionnaires (for example, slaughterhouses reported as used by poultry premises may not themselves have completed a questionnaire). The data were cross-checked with the Poultry Register Database. A total of 1822 (46%) premises represented in the network data could not be identified in the PRD; of these only 10 had 50 or more birds and hence were legally obliged to register. These 10 premises belonged to a single owner.
Information taken from the slaughterhouse questionnaires was used to derive the distribution of the number of premises served by each slaughterhouse and the distance distribution between poultry premises and slaughterhouse. Ten slaughterhouses were excluded because they did not report any premises. A further 24 slaughterhouses did not complete the questionnaire, leaving a distribution based on 61 slaughterhouses. A similar process was undertaken for catching companies (n=45) and bird supplier premises (n=204). Exponential distributions provided the best fit to these data and were used for model parameterisation.
Premises on average send birds to slaughter 6 times a year i.e. every 2 months. Table S2 shows the statistics obtained for the main recorded contacts in the data. There is some uncertainty over the interpretation of these data as about 85% of premises recorded with zero contacts are actually non-responders. The data presented, limited to those with a response in each category, may therefore over-represent the true frequency of contact.
Summing over these possible modes of contact (for those premises that recorded feed delivery contacts) a single premise has a median of 66 contacts in a year (90% range 6-406) or 1.3 contacts per week. This is positively correlated with the number of birds at the premise (Pearsons correlation coefficient r=0.27, p<0.001). These parameters are used to define the mean node degree in the network model and for the contact frequency for group transmission in the spatial simulation model.
Table S2: The median and 95% range of rate of contact for the main reasons that a premise has contact with others in the poultry industry.
Feed delivery / 469 / 30 (4-204)
Slaughterhouse visit / 2330 / 6 (1-21)
Catching Company / 220 / 4 (1-42.5)
Vaccination / 21 / 6 (1-20)
Cleaning / 171 / 3 (1-24)
Other / 397 / 30 (3-310)
2. Natural History & Epidemiological Parameters
As data on the natural history parameters at the flock level in the case of uncontrolled outbreaks are scarce, we use data on bird-level parameters and a within-farm model to translate these estimates into farm-based parameters. In the following sections we provide justification from the literature for the bird-level parameters and detail the within-farm model used to translate these estimates into farm-based parameters.
2.1. Course of infection in individual birds
Information on the course of avian influenza virus infections within individual birds has been extensively studied using experimental infections, the majority as control arms in vaccination studies. Because of the high doses and the route of infection (mostly through direct inoculation of the virus although a small number of studies report attack rates in contacts of infected birds) it is not possible to directly extrapolate these results to infection in the field. However, in the absence of detailed data on field infections, we have chosen natural history parameters based on these experimental infections.
A second limitation to the review detailed below is the heterogeneity in the strains and types of avian influenza considered. For completeness we present some results from LPAI studies although it should be appreciated that the duration of infection may well be longer and the severity less than for HPAIs. Furthermore, because of the relative paucity of data on H5N1, we also consider other HPAI data. Finally, it has been noted that the pathogenicity of the H5N1 strain of HPAI has changed over time. This change has been documented in detail in ducks. For example, ducks infected with H5N1 in Hong Kong between 1997 and 2002 show either no clinical signs or only very mild disease (Hulse-Post et al., 2005). In contrast, more recent experimental studies in ducks with H5N1 viruses obtained in 2002 from Vietnam and China have killed ducks and other aquatic poultry (Ellis et al., 2004; Sturm-Ramirez et al., 2004; Sturm-Ramirez et al., 2005). These differences between strains within the H5N1 type can also therefore impact on observations of the natural history within individual birds.
2.2. Susceptibility to infection and asymptomatic infection
Experimental studies have demonstrated that chickens are always susceptible to H5N1 infections and there is a near 100% mortality rate. Turkeys appear to be more susceptible to other HPAI and LPAI infections (Capua and Marangon, 2004; Tumpey et al., 2004) and recent experimental data using an H5N1 strain from outbreaks in Turkey in 2005 suggest that this is also the case for that subtype (McNally et al., 2006). Asymptomatic infection has not been reported in chickens or turkeys.
The degree to which ducks are susceptible to infection and onset with clinical signs varies by strain, even within the H5N1 subtype. The most comprehensive study of H5N1 in ducks shows some recent H5N1 viruses retrieved from a variety of locations in South-East Asia can be highly pathogenic and result in clinical signs and death in mallard ducks whilst other only result in asymptomatic infection or mild clinical disease (Hulse-Post et al., 2005). These results are in contrast with earlier studies of H5N1 from the Hong Kong outbreak in 1997 in which all ducks showed few clinical signs (Ellis et al., 2004; Sturm-Ramirez et al., 2004; Sturm-Ramirez et al., 2005).
2.3. Duration of infection, viral shedding and mortality
The data in Table S3 summarise the information obtained from the literature on the duration of infection (days from inoculation to death) as well as any information provided on viral shedding over this period.
For chickens and turkeys, all experimental inoculations result in the death of the birds. For HPAI H5N1 the duration of infection is reported to be between 3 and 5 days following infection. One study of H7N7 HPAI reported that 3 infected contact chickens (those placed in close contact with the inoculated chickens) survived infection (van der Goot et al., 2005) but we assume that this is unlikely for the more pathogenic H5N1. Viral shedding typically occurs rapidly with all studies reporting shedding by 3 days post inoculation (when first tests are typically undertaken). Only one study reported more frequent testing; this showed viral shedding with H7N7 occuring in the buccal cavity from 24 hours post inoculation in chickens and 8 hours post inoculation in turkeys, suggesting a potentially rapid onset of infectiousness in individual birds (Essen et al., 2006).
Table S3: Summary of the duration of infection, mortality rate and viral shedding in experimentally infected birds.
Species / Subtype / Country / Year / Days from inoculation to death / Viral shedding / ReferenceChickens:
HPAI H7N7 / Netherlands/ 2003 / 2 – 5 / - / (van der Goot et al., 2005)
HPAI H5N2 / U.S. / 6 / - / (van der Goot et al., 2003)
HPAI H5N1 / China / 2004 / 2 days / - / (Tian et al., 2005)
HPAI H5N1 / Vietnam / 3-4 / 1/1 at 3 days p.i. / (Webster et al., 2006)
HPAI H7N1 / Italy/1999-2000 / >21 / Buccal cavity from 24 hours p.i.; Cloacal from 3 days p.i. / (Essen et al., 2006)
Turkeys:
HPAI H7N1 / Italy / 1999-2000 / Up to 8 days? / Buccal cavity from 8 hours p.i.; Cloacal from 24 hours p.i. / (Essen et al., 2006)
Ducks:
HPAI H5N1 / China / 2004 / 13/15 died by day 6 / Oropharyngeal and cloacal shedding from 3 days p.i. / (Tian et al., 2005)
HPAI/LPAI H5N1 / Hong Kong / 1997-2003, China / 2004, Vietnam / 2003-2004, Indonesia / 2004, Singapore / 1997 / 6/12 HPAI died, 0/16 LPAI died / Viral shedding from day 7 to 17 / (Hulse-Post et al., 2005)
HPAI H5N1 / Hong Kong / 1997 / 16 hours – 4 days / - / (Shortridge et al., 1998)
HPAI H5N1 / Hong Kong / 1997-2003 / 4 – 6 days / Tracheal and cloacal shedding peaks on day 3; drops from day 6 / (Sturm-Ramirez et al., 2004; Sturm-Ramirez et al., 2005)
HPAI H5N1 / Vietnam / No deaths / Tracheal and cloacal shedding from day 3 p.i. / (Webster et al., 2006)
Geese:
HPAI H5N1 / China / 2004 / All died within 7 days / Oropharyngeal and cloacal shedding from 3 days p.i. / (Tian et al., 2005)
2.4. Effect of vaccination
A range of vaccines have been developed against H5N1 and are in widespread use in parts of South East Asia (notably Vietnam and China (Normile, 2005a; Normile, 2005b)). A growing number of studies have been undertaken on the effectiveness of current vaccines with the majority based on individual birds rather than premises. Table S4 summarises these studies. The current vaccines appear to have high efficacy in protecting individual birds and all also report a reduction in viral shedding.