Spatial models of avian influenza Approaches used & refinement of the scope Empirical studies based on H5N1 distribution data Overall aproaches Scale: extent and resolution Focus & analytical methods Main results Gaps Marius Gilbert Université Libre de Bruxelles, Belgium mgilbert@ulb.ac.be
Search of litterature: Limitedto peer-reviewedjournals; ISI web of Science (1987-present) and CAB abstracts (197» - present) contents Search terms «Avianinfluenza»AND«Spatial»,«Risk»,«Cluster», «Map», «Distribution» Search made in April 2010; Cross checking in the result references for potentially interesting complementary papers
70 papers covering a fairly wide range of disciplines Classified in 6 categories: Empirical studies analysing AI distribution data (n = 36) Studies on wild bird migratory patterns in relation to existing AI cases, or as a way of assessing introduction risk (n = 15); Studies on poultry production systems and trade patterns in relation to AI spread or introduction risk (n = 5) Mathematical modelling (n = 6) Molecular ecology papers with a spatial component (n = 5) Review papers (n = 3)
70 papers covering a fairly wide range of disciplines Classified in 6 categories: Empirical studies analysing AI distribution data (n = 36) Studies on wild bird migratory patterns in relation to existing AI cases, or as a way of assessing introduction risk (n = 15); Studies on poultry production systems and trade patterns in relation to AI spread or introduction risk (n = 5) Mathematical modelling (n = 6) Molecular ecology papers with a spatial component (n = 5) Review papers (n = 3)
Empirical studies analysing AI distribution data (n = 36) Focus on HPAI H5N1 (n = 32) I. Descriptive studies (n = 5): no hypothesis is formally tested through quantitative analysis; II. Space-time statistics (n = 6): analyses of the space-time distribution of cases without accounting for external variables III. Risk-factor geospatial analyses (n = 21): analyses of HPAI H5N1 distribution data in relation to predictors or risk factors;
Empirical studies analysing AI distribution data (n = 36) Focus on HPAI H5N1 (n = 32) I. Descriptive studies (n = 5): no hypothesis is formally tested through quantitative analysis; II. Space-time statistics (n = 6): analyses of the space-time distribution of cases without accounting for external variables III. Risk-factor geospatial analyses (n = 21): analyses of HPAI H5N1 distribution data in relation to predictors or risk factors;
I. Descriptive studies Global / continental scale: Kilpatrick et al (2006): wild bird and trade flow in relation to HPAI H5N1 introductions Park & Glass (2007): Discuss broad-scale seasonality of HPAI H5N1 in relation to seasonality of human influenza Sengupta et al. (2007): broad-scale distribution of HPAI H5N1 cases in relation to continental scale ecological regions Country: Cecchi et al (2008): Distribution of HPAI H5N1 in Nigeria; Tiensin et al. (2007): Distribution of HPAI H5N1 in Thailand.
Empirical studies analysing AI distribution data (n = 36) Focus on HPAI H5N1 (n = 32) I. Descriptive studies (n = 5): no hypothesis is formally tested through quantitative analysis; II. Space-time statistics (n = 6): analyses of the spacetime distribution of cases without accounting for external variables III. Risk-factor geospatial analyses (n = 21): analyses of HPAI H5N1 distribution data in relation to predictors or risk factors;
II. Space-time statistics (n = 6) Identification of clusters in space and/or time; Locations and/or timing of clusters are discussed in relation to an external factor, though this association is not formally tested. Studies at continental scales (1), and country-scale on Romania (2), China (1), Vietnam (1) and Thailand (1) Methods included cluster identification techniques (3), geostatistical analysis (2) or some custom techniques (1)
II. Space-time statistics: continental scale Si et al. (2009): Identification of space-time clusters in the distribution of HPAI H5N1 recorded outbreaks. Overlaid over broad-scale - Waders - flyways: «To deduce the role of wild birds in H5N1 spread over long distance, bird migration patterns were compared with the disease trajectory. All regional and local clusters identified along flyways that follow the migration routes were considered potentially related to wild birds
II. Space-time statistics: continental scale: Si et al (2009) No alternative hypothesis is proposed; The hypothesis is not formally tested; Yet, they conclude: «In conclusion, the spread of the H5N1 virus, as quantified by the spacetime clusters, was found to be associated with the timing, location and direction of continental bird migration, suggesting that wild birds spread H5N1 over long distances. Disease clusters were also detected at sites that are known overwintering areas, and at times when these areas were frequented by migratory birds, suggesting that wild birds are involved in short distance H5N1 spread as well
II. Space-time statistics: country-level papers (4) In Romania, Ward et al. (2008) used space-time statistics to identify different patterns interpreted as different phases in the epidemic (introduction vs. local spread). Farnsworth et al. (2009) presented a re-analysis of the same data, though with more elaborated analyses techniques. Minh et al. (2009) analyse the space-time distribution of HPAI H5N1 in Vietnam and describing changing patterns in the disease spatial patterns that are discussed in relation to interventions. Souris et al. (2010) used space-time scanning to identify potential introduction points in Thailand, and attempt to relate them to external variables. They find no departure from randomness. They used a method that may well have produced this result (i.e. circular reasonning), and make rather strange intepretations. Oyana et al. (2009) analyse distribution of HPAI H5N1 in China, and provide a few descriptive outputs, though quite inconclusive.
Empirical studies analysing AI distribution data (n = 36) Focus on HPAI H5N1 (n = 32) I. Descriptive studies (n = 5): no hypothesis is formally tested through quantitative analysis; II. Space-time statistics (n = 6): analyses of the space-time distribution of cases without accounting for external variables III. Risk-factor geospatial analyses (n = 21): analyses of HPAI H5N1 distribution data in relation to predictors or risk factors;
III. Risk-factor geospatial analyses (n = 21) 6 studies testing the effect of one or two factors (e.g. temperature, wild birds) Rivas et al (2010): Role of roads on HPAI distribution in Nigeria Ottaviani (In press) and Reperant (2010): role of temperate isoline on the distribution of H5N1 in Europe Ward et al. (2009): Role of wild birds on HPAI introduction in Romania Kuo et al. (2009): Analysis of space-time distribution of human cases over Asia and used trade as one of the predictors. Li et al. (2004): Analysed the distribution of H5N1 outbreaks in China using a transmission model that uses the distance to railways as one predictor 15 studies where the presence of H5N1 is tested against a set of risk factors of predictors, sometime with mapped outputs.
Authors Extent Resolution Method Sub-national Biswas et al. (2009) Bangladesh Farm Log. Reg Henning et al (2008) Vietnam AdmL3 Log. Reg National Yoon et al. (2005) South Korea Farm Regression Fang et al (2008) China AdmL3 Log. Reg Gilbert et al (2006) Thailand AdmL3 Log. Reg Loth et al (In press) Bangladesh AdmL3 Log. Reg Paul et al (2010) Thailand AdmL3 Log. Reg Pfeiffer et al (2007) Vietnam AdmL3 Log. Reg Tiensin et al (2009) Thailand AdmL3 Log. Reg Ward et al 2008 Romania Village Log. Reg Regional / continental Gilbert et al (2008) Mekong countries AdmL3 Log. Reg Adhikari et al. (2009) SouthAsia Not. Spec. (OIE & local) ENN Williams et al (2008) West Africa 0.01 degrees (OIE) ENN Williams & Peterson (2009) Northeaster Africa 0.01 degrees (OIE) ENN Global Hogerwerf et al (In press) Global Country & Provinces Log. Reg / Mvar
Authors Poultry Anthr W. Birds EcoClim Land/Crop Water Topo SocEco Sub-national Biswas et al. (2009) x x x Henning et al (2008) x x x x x x National Yoon et al. (2005) x Fang et al (2008) x x x x Gilbert et al (2006) x x x x x x Loth et al (In press) x x x Paul et al (2010) x x x x x Pfeiffer et al (2007) x x x x x x Tiensin et al (2009) x x x x Ward et al (2008) x x x Regional / continental Gilbert et al (2008) x x x x Adhikari et al. (2009) x x x Williams et al (2008) x x Williams & Peterson (2009) x x Global Hogerwerf et al (In press) x x x
Risk Factors Significance / Sign Country Poultry / production system/ other domestic hosts / practices Chicken and ducks in different shelters - - Bangladesh Feeding with slaughter remnants of purchased chickens + Bangladesh Buffalo density - - - Vietnam Poultry flock density ++/- - Vietnam Farm type is layer +++ SouthKorea Number of chicken houses +++ SouthKorea Free-grazing duck density/ domestic duck density +++, +,+,+++,+++ Thailand, Vietnam, World Native chickens density +++, +/-, - - - Thailand Cock density +++, +/-, ++ Thailand Density of farm ducks + Thailand Chicken density +, +/- Vietnam, World Quail flocks in subdistrict + Thailand Duck by chicken density + Global Number of commercial units ++ Bangladesh Slaughterhouses in Subdistrict ++ Thailand Anthropogenic variables (H. pop density, distance to road, road density) Min distance to highway - - -, + China, Thailand Human population density +++, +, ++/--, ++ Thailand, Bangladesh, Vietnam Roads per subdistrict +,+,++ Bangladesh, Thailand, Romania Distance to the nearest city -, - Thailand Agricultural population density + Global
Risk Factors Significance / Sign Country Wild Birds Contact with resident birds + Bangladesh Eco-climatic variables May-Oct NDVI ++, - Vietnam Annual precipitation - - - China Land use and crops Sweet potato yield - - - Vietnam Cropping intensity +, +++ Thailand, Vietnam % area used for aquaculture + Vietnam % land used for rice + Vietnam Water Access to water + Bangladesh Min distance to lake*min distance to wetland - - - China Number of rivers and streams ++ Romania Topography Elevation - - -/+++, -/+,-/+,- - / + + Thailand, Vietnam Socio-Economic % Housholds with electricity - - Vietnam Farm managed by owner & family - - - SouthKorea Purchasing power per capita - Global
Poultry variables Effect of chickens depends on scale and production systems Effect of ducks is very clear in Thailand & Vietnam, less so in other published papers, but see Gilbert et al. (submitted) A number of other variables are found significant, but their effect is not easily generalisable
Anthropogenic variables There is good agreement between all models indicating that areas with high human population density, road density, low distance to cities and highways are at higher risk. None of the studies has tested very specifically some more detailed risk factor that may explain this pattern. It is cited as the effect of trade, but this has not been formally quantified (but see Magalhaes et al. 2010) Wild birds Empirical studies statistically testing the effect of wild birds are virtually non-existing; Several studies looking at the distribution of wild bird migratory data in relation to location of H5N1 outbreaks, but none looking at the influence it may have on the risk.
Eco-climatic variables Varying effects and lack of straightforward causal interpretation; Landuse Percentage of rice cropping, cropping intensity stands out in a number of studies (see also V. Martin talk on China). Also reports of effect of aquaculture. The effect can be thought through the influence on animal densities (e.g. free-ranging ducks), or through the associated irrigation network favouring water-borne transmission Water Proximity to water, percentage of water came out significant in three very different studies (Bangladesh, China, Romani). Given the possible effect of irrigation, worth exploring further.
Topography Points at floodplains. Socio-economic Not explored in details, negative associations with indicators of economic development
Gaps and recommendations Geographical 6 5 Thailand Vietnam Quantitative analyses of the effect of trade; Papers 4 3 China Quantitative analyses of the effect of wild birds; Accounting for differences in production systems Analyses of a reduced set of risk factors than can be obtained and compared across regions; 2 1 0 Egypt Indonesia 0 100 200 Human cases Include test of water-borne transmission in all studies