Mapping the likelihood of introduction and spread of HPAI Virus H5N1 in Indonesia using multicriteria decision modelling

Will de Glanville, Kim Stevens, Solenne Costard, Raphaëlle Métras, Dirk Pfeiffer

The spatial distribution of disease risk and its visual presentation through risk maps can assist in the design of targeted animal disease surveillance and control strategies. This approach is particularly useful in situations in which empirical data are not readily available (Clements et al 2006), or when data are only available on some aspects of the epidemiology of a multi-factorial disease (Tachiiri et al 2006). In such circumstances data on known risk factors can be used to determine those areas in which a specific disease is most likely to occur using knowledge driven models, such as multicriteria decision modelling (MCDM) (Pfeiffer et al 2008). MCDM is an example of a static knowledge-driven modelling approach that can be used to produce qualitative or quantitative estimates of risk ‘based on existing or hypothesized understanding of the causal relationships leading to disease occurrence’ (Pfeiffer et al., 2008). Knowledge of the risk factors associated with the occurrence of a disease and their interrelationships are used to drive the model. The objective of this study was to use a multicriteria decision modelling (MCDM) approach to provide a qualitative estimate of the spatial distribution of the risk of spread of highly pathogenic avian influenza virus (HPAIV) subtype H5N1 in Indonesia. MCDM involves the following sequence of analytical steps (Pfeiffer et al 2008): 1. Defining the objective(s) 2. Defining the factors 3. Defining the relationship between each factor and the risk 4. Sourcing digital maps of the factors and constraints 5. Standardising the maps so that they can be compared 6. Defining the relative importance of each factor in relation to the objective 7. Combining all factors and constraints to produce a final weighted estimate of risk for each location in the study area 8. Sensitivity analysis It is important that the user of the outputs of these models is aware of the assumptions made in defining and quantifying the model inputs and any potential sources of information bias when interpreting the results of such analyses. This is especially important with knowledge-based models. This report details the methods used to produce risk maps illustrating the risk of spread of HPAIV in Indonesia. It is intended that the risk map be used to assist disease control and surveillance at the country level in Indonesia, whilst taking into account the limitations of the MCDM methodology.