p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

Available online at www.sciencedirect.com

Public Health journal homepage: www.elsevier.com/puhe

Review Paper

Risk assessment of vector-borne diseases for public health governance L. Sedda a,*, D.W. Morley a,d, M.A.H. Braks b, L. De Simone c, D. Benz a, D.J. Rogers a a

Department of Zoology, University of Oxford, South Parks Road, OX1 3PS Oxford, United Kingdom Centre for Zoonoses and Environmental Microbiology, National Institute for Public Health and the Environment (RIVM), P.O. Box 1, 3720 BA Bilthoven, The Netherlands c Surveillance and Response Support Unit (SRS), European Centre for Disease Prevention and Control (ECDC), €gen 11 A, 171 83 Stockholm, Sweden Tomtebodava b

article info

abstract

Article history:

Objectives: In the context of public health, risk governance (or risk analysis) is a framework

Received 28 March 2014

for the assessment and subsequent management and/or control of the danger posed by an

Received in revised form

identified disease threat. Generic frameworks in which to carry out risk assessment have

26 August 2014

been developed by various agencies. These include monitoring, data collection, statistical

Accepted 28 August 2014

analysis and dissemination. Due to the inherent complexity of disease systems, however,

Available online xxx

the generic approach must be modified for individual, disease-specific risk assessment frameworks.

Keywords:

Study design: The analysis was based on the review of the current risk assessments of

Risk governance

vector-borne diseases adopted by the main Public Health organisations (OIE, WHO, ECDC,

Risk assessment

FAO, CDC etc…).

Public Health governance

Methods: Literature, legislation and statistical assessment of the risk analysis frameworks.

Vector-borne diseases

Results: This review outlines the need for the development of a general public health risk

Statistical analysis

assessment method for vector-borne diseases, in order to guarantee that sufficient information is gathered to apply robust models of risk assessment. Stochastic (especially spatial) methods, often in Bayesian frameworks are now gaining prominence in standard risk assessment procedures because of their ability to assess accurately model uncertainties. Conclusions: Risk assessment needs to be addressed quantitatively wherever possible, and submitted with its quality assessment in order to enable successful public health measures to be adopted. In terms of current practice, often a series of different models and analyses are applied to the same problem, with results and outcomes that are difficult to compare because of the unknown model and data uncertainties. Therefore, the risk assessment areas in need of further research are identified in this article. © 2014 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.

* Corresponding author. Present address: Geography and Environment, University of Southampton, University Road, SO17 1BJ Southampton, United Kingdom. Tel.: þ44 (0) 2380599622; fax: þ44 (0) 2380593295. E-mail address: [email protected] (L. Sedda). d Present address: MRC-PHE Centre for Environment and Health, Imperial College London, St. Mary's Campus, W2 1PG London, United Kingdom. http://dx.doi.org/10.1016/j.puhe.2014.08.018 0033-3506/© 2014 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

2

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

Introduction Public Health (PH) can be defined as the activities engaged in by governments and organizations to prevent disease and to ensure a healthy population.1 This includes the assessment and monitoring of the health of communities and populations at risk, intervention, education, and formulation of public policies, and encompasses all issues that may be detrimental to health.2 In terms of Public Health, risk has been defined by medical epidemiologists and health organizations as ‘the probability of disease developing in an individual in a specified time interval’.3 More generally, the Society for Risk Analysis (SRA)4 defined risk as ‘the potential for realization of unwanted, adverse consequences to human life, health, property, or the environment; estimation of risk is usually based on the expected value of the probability of the event occurring times the consequence of the event given that it has occurred’. In econometrics, risk is quantified by the ‘absolute value of expected losses’.5 Notice that in these latter two definitions the probabilistic concept of risk is conflated with some idea of the (fixed or calculable) economic cost outcome should that risk be realized. PH risk governance is a framework6 that includes risk assessment (or risk estimation), risk management and risk communication. This manuscript focusses on one category of public health challenges, the risk assessment of vector-borne diseases (human illness transmitted by arthropods) which, at least in the tropics, have a huge impact on human health and well-being (for example, WHO estimated there were 627,000 deaths among 207 million cases of malaria in 20127). Emerging vector-borne diseases are of increasing concern in Europe8 and North America (West Nile virus, chikunguya, malaria and dengue for instance) some with autochthonous cases.9 Especially in the case of vector-borne diseases and because of their complexity, the metric used for risk is not uniquely defined. Often risk is synonymous with vector or host suitability,10 the basic reproduction number R0,11 vector capacity,12 vector presence/absence,13 prevalence or incidence of

the disease14 or many others. The same diversity is found in model selection, data resolution, the predictor variables, their transformation, and the input parameters (e.g. models coefficients applied to different areas/events) used to model risk, even for the same disease (e.g.15 for Lyme disease). Hence a deep understanding of the various metrics and models used for estimating risk is of primary importance, especially when previously obtained parameters, values and models are applied outside the domain of the original model fitting15 (e.g. in the case of newly emerging diseases). The goal here is to integrate risk assessment or estimation theory (and models) for vector-borne diseases with the present regulatory, and sometimes mandatory, system of risk governance, as defined by the main health organizations.

Risk governance in international public health organizations and some national agencies Table 1 presents some key contributions to general riskrelated policy over the last fifty years. From the initial basic statements of risk governance guidelines, procedures have developed into clear conceptual frameworks allowing a standard approach to be adopted. Most recently the complementary roles of quantitative and qualitative methods have been directly addressed along with the understanding that an estimation of uncertainty is vital to any risk analysis. The concept of risk governance as a combination of risk assessment, risk management and risk communication (or ‘translating knowledge’16 into advice for informative/legal actions17) originated in the 1960s and 1970s when concerns were raised over environmental carcinogens and food additives.18e22 Subsequently risk governance concepts were formulated by individual scientists,19 and by national and international agencies,20 with a concentration on cancer and harmful chemicals.21,22 In the 1990s, risk assessment became mandatory in Europe for environmental, epidemiological and food health threats (i.e. Council Directive 89/391/ECC, the Decision

Table 1 e Main documents defining risk analysis for public health purposes. Year 1960e1970 1981 1983 1995

Agency

2002 2002 1999e2005

Various US agencies The Society for Risk Analysis (SRA) National Research Council of the US (NRC) Food and Agriculture Organisation of United Nations (FAO) European Commission World Organisation for Animal Health (OIE) World Health Organisation (WHO)

2005 2005

European Commission International Risk Governance Council (IRGC)

2007

European Commission

2009 2012

National Research Council of the US (NRC) European Food Safety Authority

Document Recommendations, Guidelines, standards76 Vision Statement4 Red Book22 Risk analysis and food: the experts' view23 Regulation 178/200277 Terrestrial Animal Health Code41 Health Impact Assessment78; International Health Regulations79; Comparative Quantification of Health Risks80 Regulation 851/200481 White paper on risk governance (towards an integrative approach).43 White paper on the health strategy for years 2008-2013.82 Science and Decisions: Advancing Risk Assessment46 Guidance on risk assessment for animal welfare44

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

No 2119 of 1998/EC, the European Commission Regulation 178/ EC of 2002), was enshrined within the European Union Risk Assessment System and was managed by DG SANCO. The fundamentals of risk governance are similar in the various frameworks of Table 1 and based on the concepts defined by the Red Book and by the CODEX of the FAO/WHO.23 With these documents the division of risk governance into risk assessment, risk management and risk communication becomes the standard form for risk governance for any type of risk. None of the frameworks presented in Table 1 imposes a specific method of risk assessment. Indeed, a unique risk assessment method for Public Health risk governance does not yet exist. The following section outlines the current statistical approaches to risk assessment as a contribution to developing such a method. It concentrates on vector-borne diseases because although these have historically been neglected in many risk assessment frameworks they are likely to become increasingly important in temperate regions in the future.

Risk assessment for public health risk governance To date, risk assessment of vector-borne and other diseases for Public Health applications has depended heavily upon the modellers and what they can and cannot do with the data available. The choice of the model depends on the aetiology of the disease in question. For example, when aetiology is unknown, studies are usually population-based (e.g. smoking/ pollutants related cancers). Usually very large sample sizes are required to remove obvious biases and to reveal what are often very low incremental probabilities of disease risk. In most vector-borne diseases the aetiology is well established. Insect or tick ‘vectors’ pick up infections (i.e. become infected) when they take blood from infectious vertebrate hosts. The vectors in turn become infectious (i.e. capable of transmitting the infection to another vertebrate host) only after a period of incubation which may take several days (insects) or even months (ticks) and is referred to as extrinsic incubation period (EIP). Hence a minimum of three species are involved in vector-borne disease transmission (host, vector and pathogen). In fact such transmission often involves numerous and complex vector and zoonotic (domestic and wild animal) components. In these cases even if risk assessment can be more individual-based, or based on identifiable sub-groups, for example focussed only on those people living where vectors can be found, the number of parameters and variables needed would be greater than is typically used in other sorts of public health models.12,24 More complex systems need more research investment to reach the same level of understanding but vector-borne diseases have been neglected in comparison with other diseases (e.g. cancer), probably because of their lack of importance in the developed world. In 2006, and following an outbreak of the mosquito-borne disease chikungunya (CHIK)25 in Asia, a destination for many tourists and other visitors from Europe, the European Centre for Disease Prevention and Control (ECDC) conducted a CHIK risk assessment for Europe. The risk assessment was based on the knowledge that infected and infectious travellers

3

were returning to Europe, in some cases to places where a known vector of chikungunya, the Asian tiger mosquito Aedes albopictus, had already established (in Europe, this mosquito was first reported in Albania in 1979, and in Italy in 1990). By studying the current numbers of recorded cases in endemic areas as well as imported cases and the presence of local vector species, the ECDC concluded there was a risk of local acquisitions of CHIK virus in Europe, although the magnitude of this risk could not be determined.26 Quantitative models were not applied because of the lack of knowledge of the response of the virus to the environmental conditions of a new geographic area, the preventative measures on the part of both travellers and health professionals, the potential of further spread in southern Europe of Ae. albopictus and finally the competence of local mosquitoes to transmit the virus. In 2009, following a large outbreak of CHIK in Italy in 2007 and after a subsequent large-scale entomological survey, the ECDC published a first risk map for Ae. albopictus in Europe,27 based on a random forest model and multicriteria analysis applied to entomological and environmental datasets. As in all other cases, however, a risk map of a putative vector is not the same as a risk map of the disease it transmits. ‘Anophelism without malaria’ (i.e. the occurrence of the vector Anopheles without the malaria) is a familiar concept to malariologists, and a similar effect is likely to apply to all other vector-borne diseases. Disease pathogens potentially transmitted by many different mosquito species are even less likely to be described by risk maps of just one, or even several of their vector species. For example, the European Union has made several attempts to assess the risk to Europe of Rift Valley fever (the transmission of which in Africa involves several different mosquito species) but, due to the complexity of the disease epidemiology and the difficulties in the transmission parameterization, the risk assessments were predominantly qualitative.28,29 Similar problems arise with diseases of veterinary importance, especially those of high economic impact such as bluetongue (BT), which provoked an extensive legislative response (Directive 75/EC of 2000, EC Regulation 178 of 2002, Commission Decision 393/EC of 2005, Regulation 1266/EC of 2007; Decision, 367/EC, Regulation 394/EC of 2008, Regulation 1304/EC of 2008, Decision 655/EC of 2008, Decision 19/EC of 2009, Regulation 1156/EC of 2009); or foot-and-mouth in the UK in 2001. In both cases control activities against the diseases were eventually informed by predictive risk models each of which subsequently came in for a certain amount of criticism because of the ways in which the models were formulated (including the assumptions on which the models were based), and the values of the models parameters used. These examples show the difficulties in applying modelbased solutions to human and non-human, vector-borne and directly transmitted disease outbreak emergencies when only inadequate or incomplete data are available. This will inevitably be the case with emerging diseases (for example chikungunya, Rift Valley fever and West Nile virus in Europe) but may also apply even to neglected but long-established diseases, such as Lyme disease.8,15 The examples clearly show that, in the face of an imminent public or veterinary health threat, control agencies have

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

4

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

little time to worry about the characteristics of disease models. Instead, they must be seen to be doing something to combat the diseases, whether or not this is informed by quantitative models. Sometimes this involves excessive application of the precautionary principle, for example the localised and repeated use of insecticide which although it diminishes disease risk may not eliminate it entirely but certainly does create other environmental and health risks.30 The greater the impact of an event, or the longer its duration, the more likely an inappropriate control (either too much or too little; the precautionary approach encourages the former) will be applied. Policy must therefore encourage the development of robust models in advance of any new threat and, ideally, these models should be tested during simulation exercises,15 as was done in 2004 in the aftermath of the 2001 foot-and-mouth outbreak in the UK and following no fewer than three independent reports into the outbreak31 and one Department of Health longitudinal study on the health and social impacts of it.32 But what should these robust models look like, and how may they be developed? The final section explores some of the options.

Models for public health risk governance The distinction made in the previous section, between models for diseases without or with known aetiologies, highlights the dichotomy between statistical and biological approaches to disease modelling. Statistical models look for statistical associations between a dependent variable (e.g. the number or percentage of disease cases) and a suite of independent variables that may be either correlated with risk, indicators of risk or causal drivers of risk (the distinction between these three different types of independent variables is not obvious at the start of many studies but tends to become clearer over time, as was the case with lung cancer). The statistical model of risk therefore seeks robust statistical descriptions of levels of risk based on the measured levels of the independent variables. Various kinds of regression and odds ratio models can be employed here. In contrast to statistical models, biological models consist of components, each of which captures one or other important biological feature of the know infection pathway. For example, a model for the extrinsic incubation period (defined above) will incorporate temperature as a causal variable determining the duration of the EIP. During the EIP, however, the insect may die either from abiotic (temperature, rainfall) or biotic (predation) effects each of which may be incorporated into full biological models of disease transmission. Characteristic of many biological systems are thresholds (pathogens will not develop at all below a threshold temperature), break-points (mortalities may be low at low densities and then increase above a certain level of abundance of competitors, parasites or predators of the vectors) and nonlinearities (insect mortality rates increase at increasing rates at the upper end of their thermal tolerance ranges). The ideal biological model must therefore be able to describe all aspects of each and every threshold, break-point and non-linearity at all points of the transmission pathway and this is often

beyond the limits of the present knowledge and capabilities. Instead, relatively simplistic but still biologically - based models can go a long way towards describing the overall characteristics of diseases, if not their finer points. Classic examples of simple but powerful biological models for vectorborne diseases involve calculation of the key epidemiological variable, R0, the disease's basic reproductive number, or the number of secondary cases arising in an entirely susceptible population from one case of the disease at the present time. This index can be used to predict if a vector-borne disease can establish after introduction in a certain area and under certain climatic and environmental conditions.33 It has several advantages over statistical methods, such as the clear biological and epidemiological interpretation of its values, and the flexibility for simulating different scenarios. However, even the relatively simplified R0 equation requires estimates of many different parameters and the knowledge of the relevant biological processes involved (for example, the R0 equation for insect vectors is quite different from that for ticks). In addition, theoretical studies show that R0 is related to disease prevalence at equilibrium (i.e. the former might therefore be used to predict the latter). But it is a moot point as to how often diseases are in fact in equilibrium, and therefore how often the R0 approach is applicable; vector-borne diseases are often seasonal and many emerging and re-emerging vector-borne diseases of interest are clearly not in equilibrium. An alternative approach for vector-borne diseases is therefore to fall back on the sorts of statistical methods applied to diseases of unknown aetiology. This would appear to be a backward step (why ignore all the information on causal pathways in the transmission of vector-borne diseases?) but it may be the only approach to take when the key aspects of transmission was unclear, all of which must be understood for the biological approach to succeed. For example, West Nile virus in the USA infects at least 100 species of mammals and birds and is potentially transmitted by up to 70 or more different mosquito species. In such a context, statistical models ignore many of the details of biological transmission and look for direct correlations between environmental variables and vector or disease outcomes.10 This approach clearly ‘short-circuits’ the often tortuous biological pathways involved in transmission, and depends for its success on close, predictable and unchanging links between predictor and predicted variables. That this approach is often successful is due entirely to the sensitivity of many vectorborne diseases to climate (most of the parameters of the R0 equation for vector-borne diseases are vector-related and hence sensitive to climate and environment). Risk assessments here are based on modelled correlations between the vector's distribution and or abundance, or past disease presence/incidence, and sets of environmental variables. In the first case (based on vector distribution/abundance), analyses are excellent predictors of risk to individuals. It is, however, difficult to translate them into measures of human disease incidence or prevalence, both population-level measures of risk. In the second (based on past incidence/prevalence) the risk assessment is clearly based on the past exposure of the local human population to disease risk, something that is affected not just by the absolute risk in each area, but by the amount of human exposure to that risk, dependent on both

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

Organisation CDC ECDC NaTHNac NHS CDC

ProMed HealthMap

BioCaster

EpiSPIDER

WHO

FAO

OIE Google

Description Travel health. Travel health. Travel health. Travel health. Arbonet is a site for displaying vector-borne disease maps based solely on recorded cases of each disease (i.e. not on models). Maps are updated continuously as new surveillance data are acquired. An internet reporting system aimed at providing early warning of global threats to human health An online cartographic resource maintained by the Computational Epidemiology Group of the Children's Hospital, Boston, U.S.A. The site offers real-time surveillance of emerging health threats around the world. It uses an algorithm that monitors eleven sources of news and reports, among them ProMED-mail, Google News, WHO, OIE and ECDC. Real-time surveillance website that tracks over 1700 RSS feeds and plots emerging health events in Google Maps. Identifies potential outbreaks by screening news reports and Twitter feeds (only English texts) with a naive Bayesian classification service, then refining the output with web-based text miners. FluNet website receives influenza virological data from national laboratories and WHO regional databases. EMPRES-i is a global animal disease information system based on a collection of official (e.g. governmental) and unofficial (e.g. personal contacts, NGOs) sources and displayed on Google Maps WAHID provide access to the data held by OIE on past, present and immediate animal disease notification. Google Flu trends uses regional search engine results relating to influenza to detect possible outbreaks.

Website

Reference

http://wwwnc.cdc.gov/travel/ http://www.istm.org/eurotravnet/main.html http://www.nathnac.org/travel/ http://www.travax.nhs.uk/ http://www.cdc.gov/ncidod/dvbid/westnile/USGS_frame.html

83

http://www.promedmail.org/.

87

http://healthmap.org/en/.

88

http://born.nii.ac.jp/_dev/.

89

http://www.epispider.org/.

90

http://www.who.int/influenza/gisrs_laboratory/flunet/en/.

91

http://empres-i.fao.org/eipws3g/

92

http://www.oie.int/wahis_2/public/wahid.php/Wahidhome/Home

93

http://www.google.org/flutrends/

94

84 62 85 86

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

5

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

Table 2 e Surveillance and risk communication websites with a focus on vector-borne diseases.

6

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

human population density and behaviour. The resulting risk assessment, therefore involves the often unacknowledged component of human behaviour, which is sometimes more important than human population in determining disease incidence or prevalence.34 Commercial trade and human travel add additional levels of complexity. Examples here include the connectivity/ mobility analysis of humans (e.g. air travel,35 and goods24) in which risk is often calculated from a crude or probabilistic combination of suitable habitat in the arrival area for the pathogen or vector (brought in by tourism or trade) and the rate at which such importation occurs. As a result of the application of the above approaches, risk can be expressed as incidence, prevalence, occurrence, cumulative cases (as for the West Nile Virus map used by CDC) or, alternatively, as other metrics that generally represent a portion of the disease transmission process, for example the probability of pathogen establishment according to the probability of introduction,24,35 the probability of contact between host and vectors but not humans36 or between host capacities (a rate of vector infection from the host species), vector-host feeding preferences and, finally, human biting rates per vector.37 Risk metrics are hence based on the information available (see the attempt of38 to produce a uniform scheme) and are rarely expressed as a probability (apart from the few examples given here), which is the canonical form for risk as defined for public health (see the introduction section). The statistical and mechanistic approaches to risk modelling and prediction are many and varied, including discriminant analysis, generalized linear models (e.g. logistic regression, autoregressive regression, Poisson models etc…), generalized additive models, mixed models, tree based classifications, fractals, spectral density, wavelets, agent based, cellular automata, maximum entropy, multicriteria analysis, community models among others.12,24,36,39,40 Examples of applications of these models can be found in the quantitative risk assessment as defined by the World Organisation for Animal Health (OIE) (Terrestrial Animal Health Code41); WHO (numerical approaches to the quantification of risk42); International Risk Governance Council (IRGC)43; EFSA44; International Organisation for Standardization (ISO 31000)45; and the National Research Council of the United States.46 It is well established that different models often give different outputs (e.g. the dengue models compared in47), hence a procedure for their validation and comparison is needed. A further complication, which is often ignored in vector and disease risk mapping, is that of spatial dependency of the disease, which arises for two main reasons. Firstly, simple geographic proximity means that epidemiological outcomes often depend on disease events occurring in nearby locations (e.g. pathogen transmission may occur between areas that are close to each other, or infected vectors might fly, or infected humans move between them). Secondly the environmental conditions that favour the occurrence of a vector or disease in one area are likely to be similar in nearby areas and so promote the occurrence of the same disease in those areas, too. Whatever the cause of spatial dependency is, it may be used to improve the predictions of vector-borne disease models to render them truly spatial. Spatial dependency is included in models in many different ways, from the simple auto-logistic

regression approach of considering the occurrence of the species only in the most adjacent grid squares to the more complex approach of using the techniques of kriging and cokriging (directly or combined with deterministic models). The unwary biologist or modeller should be warned that kriging and cokriging often applies only to predicting errors from a modelled ‘mean’ surface, the latter imagined to be constant or trending in a simple way (e.g. dependent on environmental variables).48 The biological interest is therefore really in the trend surface rather than in the kriging results.49 Kriging thus contributes an important and usually neglected component to risk maps; the prediction of errors indicating the variability of the modelled result at different points in space (and time, with spatio-temporal models), and hence the statistical significance of the trend. The size of such errors generally depends upon the amount and variability of the data available; errors are minimized when there is a sufficient amount of local data, all with similar mean values (which is not always the case for vector-borne diseases such as malaria).50 Risk maps are experienced still51,52 based simply on the unmodelled distribution of diseases, pathogens or conditions, or real-time data from online warning systems and media sources (Table 2), or from informal exchanges between public health professionals (‘Event-based’ surveillance).53 Despite increasing use of disease mapping, for which often quality cannot be evaluated, it is urgent to define a policy for disease risk mapping assessment (within the risk governance framework) in order to obtain products of similar statistical quality. Uncertainty is a common feature of risk assessment of meteorological events and earthquakes, but it is usually not applied in vector-borne disease risk mapping. In the past, uncertainty was obtained from sensitivity analyses by simply varying (within fixed ranges) the values chosen for the parameters and variables. Rarely do sensitivity analyses consider variation of more than one parameter/variable at a time. Interactions between variables may be investigated using hyper-cube sampling of variables, but gives no real indication of the likely variability of each of the parameters in the field.54 Today, the most successful models use a Bayesian approach to predicting both the level of risk and the uncertainty of that prediction.55 This approach is usually applied when knowledge of the process is available and converted into distributional forms, but is also used for hypothesis testing and uncertainty measurement. These new Bayesian approaches are ideal for the relatively small datasets often found in epidemiology, for which the frequentist approaches are unable to include prior knowledge56 although they are robust for small sample size. Frequentist studies are based on the hypothetical question ‘Given ten thousand versions of reality, on how many occasions would we observe an outcome as extreme as that shown by the data?’ If the occasions are very few, then the result is likely to be ‘significant’, i.e. it is unlikely to be a member of the ten thousand versions of reality generated by the hypothesis being tested. Unfortunately, however, there is only one version of reality and so the above question remains a hypothetical one. The one version of reality, that the authors do have, generates the data that they record. These data may have been produced by a variety of

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

different mechanisms which they can test in the Bayesian modelling approach to answer the question; ‘Given this particular dataset, is hypothesis/model A more likely than hypothesis/model B?’ Not only do Bayesian statistics distinguishes between alternative hypotheses, it also includes estimation of the errors of the different parameters involved, allowing the output models not only to predict mean risk, but also the errors and uncertainties of those mean risks.34,57 From the practical point of view, the main differences between the Bayesian and the more traditional frequentist approach are the use of prior and posterior distributions. The latter is the main output of the Bayesian approach and expresses as a probability, the weighted average of the prior (a priori knowledge) and the data (through its likelihood) via the Bayesian theorem.58 In addition to statistical and biological models, a third model category exists in Public Health applications, based on expert opinions, or other experience, and referred to as participatory models.59 An example is the vaccination-advice map for yellow fever60 based on expert opinion that decided specific thresholds for certain environmental variables thought to be key to transmission (for example, elevation, land cover etc …).61 This map categorizes risk into four classes: endemic, transitional, low risk and no risk.61 Vaccination is ‘recommended’ for those areas that are classified as endemic and transitional, and also for low risk areas if a prolonged stay is expected. In brief, this semi-quantitative risk assessment is the result of a complex process involving data acquisition, data analysis and data discussion in which modellers, national experts and epidemiologists are involved.60 The role of the health practitioner using such a map is to merge the information given by the guidelines with individual characteristics of the person to be exposed to the disease risk (age, gender, medical history, vaccination history) and trip information (destination, departure date, duration of the travel, purpose of travel).62 Whilst the final decision here (to vaccinate or not) is based on a subjective weighing of the evidence by the medical practitioner or health agency involved, this decision is likely to be better informed if based upon more accurate, quantitative risk maps. The unplanned surveillance and control in Sudan, for areas that are recognized as having endemic transmission for yellow fever, may have contribute to the recent uncontrolled outbreaks in Sudan, which required a mass-vaccination treatment of the population.63 In summary, while risk governance is a framework that is very homogeneous between international organisations and scientists,1 the risk assessment component is not standardised, and often not defined at all. For example, in the OIE terrestrial code a list of statistics necessary for risk assessment is given without mention of their uncertainty. EFSA describes the full risk assessment framework, which includes uncertainty, but is not flexible enough to be adapted for general use. Other agencies require semi-quantitative approaches with a strong stakeholder component (IRGC and WHO). The seminal work of1 adapted the IRGC risk governance framework to the case of vector-borne diseases for which they provide generic, but not methodological, guidelines for a preassessment and appraisal of the risk. Apart from the useful indication on the data requested for public health utility, the authors did not however, explore the issues related to the

7

existing variability in the data, methods and outputs used for risk estimation. Risk assessment for vector-borne diseases was only defined in its general aspects, and a framework which guarantees high quality disease mapping and comparable analyses, was missing.

Conclusions Risk governance is the main component of PH measures (prevention, management, surveillance and communication) and is scientifically and legally defined. Although the same basic framework is followed, each risk assessment tends to be uniquely determined by the details of the particular disease under study, the data available for it and the region and time period over which those data were collected. In a recent review on the impact of drought on vector-borne diseases for the UK,64 the authors underlined the undefined risk associated with climate change and human adaptation due to the lack of relevant data.65 In addition, the competence of existing or newly introduced insect vectors in Europe to transmit diseases that are currently confined mostly to the tropics36,11,66 are unknown as they ways in which they are transported.24,35,37 At present time, when different models and analyses are applied to the same problem, they often produce results that are difficult to compare quantitatively because of unknown model and data uncertainties. Hence, new dynamic and flexible approaches are required, determined by the Public Health framework and the inherent complexity and unique characteristics of each system being investigated. Anthropological, zoological or entomological (or a combination of these) disease risk assessment67 needs to be addressed quantitatively, wherever possible, and submitted with a quality assessment in order to evaluate tolerability and adaptability of the risk1; and to enable successful public health measures to be adopted,68 since simple interventions are often impossible (for many vector-borne diseases there is no available treatment or vaccine69). Successful application of control measures in the future depends on the reliability of the uncertainty estimations.40 Even if uncertainty sometimes cannot be reduced,70 it is important for policy-makers to have an accurate estimate of it to support action if the risk estimates warrant it, in order to avoid over- and under-estimations of risk.49 In conclusion the following recommendations have been made for the future: 1) A standard risk assessment within a risk governance framework for vector-borne diseases as proposed by1 should be produced and accepted, that includes experiences and lessons from the various agencies involved in risk assessment (e.g. OIE, FAO, EFSA, WHO, SANCO) and from OneHealth actions71; 2) A comparative review should be carried out of Public Health model selection decisions and model evaluations (as done by72 for species' distributions); 3) Quantitative approaches should be used in participatory analyses leading to risk assessment73,74;

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

8

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

4) In conformity with the definition of risk in public health, risk should be expressed as a probability; 5) All models used for making public health decisions should include estimates of model uncertainty75; 6) The spatial and temporal validity of models (i.e. up to what distances and times are the model error levels acceptable) should always be assessed; 7) Finally, data should routinely be gathered that can contribute to Public Health preparedness for future outbreaks of vector-borne and other diseases. preparedness for which quality and availability of the data need to be assessed for future emergencies.

Author statements Ethical approval Not required. This research is a review and does not include human subjects.

Funding L.S. and D.J.R. are supported by the European Union grant HEALTH-F3-2010-261504 EDENext; this paper is catalogued by the EDEN Steering Committee as EDENext255 (http://www. edenext.eu). The contents are the sole responsibility of the authors and do not necessarily reflect the views of the European Commission.

Competing interests None declared. The funders had no role in the study design; in the collection, analysis and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication.

references

1. Schmidt K, Dressel KM, Niedrig M, Mertens M, Schule SA, Groschup MH. Public health and vector-borne diseases e a new concept for risk governance. Zoonoses Public Health 2013;60:528e38. 2. Fleming ML, Parker E. Introduction to public health. Elsevier Health Sciences APAC; 2009. 3. Dicker R, Coronado F, Koo D, Gibson Parrish R. Principles of epidemiology in public health practice; 2006:512. 4. Thompson KM, Deisler Jr PF, Schwing RC. Interdisciplinary vision: the first 25 years of the Society for Risk Analysis (SRA), 1980e2005. Risk Anal 2005;25:1333e86. € rk KDC, Regula G, Hernandez J, Knopf L, Fuchs K, 5. Sta Morris RS, Davies P. Concepts for risk-based surveillance in the field of veterinary medicine and veterinary public health: review of current approaches. BMC Health Serv Res 2006;6:20. 6. Gostin LO, Fund MM. Public health law and ethics: a reader. University of California Press; 2010. 7. WHO. World malaria report 2013. World Health Organization; 2014. 8. Medlock J, Jameson L. Ecological approaches to informing public health policy and risk assessments on emerging vector-borne zoonoses. Emerg Health Threats J 2010;3:e1.

9. Nasci RS. Movement of Chikungynya virus into the Western Hemisphere. Emerg Infect Dis 2014;20:1394e5. 10. Brownstein JS, Rosen H, Purdy D, Miller JR, Merlino M, Mostashari F, Fish D. Spatial analysis of West Nile virus: rapid risk assessment of an introduced vector-borne zoonosis. Vector Borne Zoonotic Dis 2002;2:157e64. 11. Robin CA, Lo Iacono G, Gubbins S, Wood JL, Newton JR. The accuracy of the National Equine Database in relation to vector-borne disease risk modelling of horses in Great Britain. Equine Vet J 2013;45:302e8. 12. Zavaleta JO, Rossignol PA. Community-level analysis of risk of vector-borne disease. Trans R Soc Trop Med Hyg 2004;98: 610e8. 13. Romi R, Boccolini D, Menegon M, Rezza G. Probable autochthonous introduced malaria cases in Italy in 2009-2011 and the risk of local vector-borne transmission. Euro Surveill; 2012::17. 14. Winters AM, Eisen RJ, Delorey MJ, Fischer M, Nasci RS, Zielinski-Gutierrez E, Moore CG, Pape WJ, Eisen L. Spatial risk assessments based on vector-borne disease epidemiologic data: importance of scale for West Nile virus disease in Colorado. Am J Trop Med Hyg 2010;82:945e53. 15. Lorenz A, Dhingra R, Chang HH, Bisanzio D, Liu Y, Remais JV. Inter-model comparison of the landscape determinants of vector-borne disease: implications for epidemiological and entomological risk modeling. PloS one 2014;9:e103163. 16. Matthews GW, Markiewicz M, Beitsch LM. Legal frameworks supporting public health department accreditation: lessons learned from 10 states. J Public Health Manag Pract 2012;18:E8e16. 17. Bennett P. Risk communication and public health. OUP Oxford; 2010. 18. Committee on Risk Assessment of Hazardous Air Pollutants, Board on Environmental Studies and Toxicology, Commission on Life Sciences. Science and judgment in risk assessment. National Academies Press; 1994. 19. Rowe WD. An anatomy of risk. New York; London: Wiley; 1977. 20. EPA. Health risk and economic impact assessments of suspected carcinogens: interim procedures & guidelines. In: Agency USEP, editor. Federal register; 1976. p. 21402e5. 21. OTA. Assessment of technologies for determining cancer risks from the environment. Washington, D.C.: U.S. Office of Technology Assessment; 1981 June. 22. National Research Council N. Risk assessment in the federal government: managing the process. The National Academies Press; 1983. 23. Joint Consultation FAO and WHO Expert FW. Application of risk analysis to food standards issues. Geneva: WHO WHO: FNU/FOS/ 95.31995. 24. Thomas SM, Tjaden NB, van den Bos S, Beierkuhnlein C. Implementing cargo movement into climate based risk assessment of vector-borne diseases. Int J Environ Res Public Health 2014;11:3360e74. 25. Burt FJ, Rolph MS, Rulli NE, Mahalingam S, Heise MT. Chikungunya: a re-emerging virus. Lancet 2012;379:662e71. 26. Depoortere E, Coulombier D. Chikungunya risk assessment for Europe: recommendations for action. Euro Surveill 2006;11. E060511 2. 27. ECDC. Development of Aedes albopictus risk maps; 2009. 28. EFSA Panel on Animal Health and Welfare. Scientific opinion on rift valley fever. EFSA J 2013;11:1e48. 29. EFSA Panel on Animal Health and Welfare. The risk of rift valley fever incursion and its persistence within the community. EFSA J; 2005::1e128. 30. Matthews G, Zaim M, Yadav RS, Soares A, Hii J, Ameneshewa B, Mnzava A, Dash AP, Ejov M, Tan SH, van den Berg H. Status of legislation and regulatory control of public health pesticides in countries endemic with or at risk of

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

31.

32. 33.

34.

35.

36.

37.

38.

39. 40. 41. 42.

43. 44. 45.

46. 47. 48. 49. 50.

51. 52. 53.

major vector-borne diseases. Environ Health Perspect 2011;119:1517e22. Donaldson A, Lee R, Ward N, Willkinson K. Foot and mouth e five years on: the legacy of the 2001 foot and mouth disease crisis for farming and the British countryside. Newcastle Upon Tyne: University of Newcastle Upon Tyne; 2006. Peck DF. Foot and mouth outbreak: lessons for mental health services. Adv Psychiatric Treat 2005;11:270e6. Hartemink N, Vanwambeke SO, Heesterbeek H, Rogers D, Morley D, Pesson B, Davies C, Mahamdallie S, Ready P. Integrated mapping of establishment risk for emerging vector-borne infections: a case study of canine leishmaniasis in southwest France. PloS one 2011;6:e20817. € hning D, Lesaffre E, Viel J-F, Lawson A, Biggeri A, Bo Bertollini R. Disease mapping and risk assessment for public health. John Wiley & Sons; 1999. Huang Z, Das A, Qiu Y, Tatem AJ. Web-based GIS: the vectorborne disease airline importation risk (VBD-AIR) tool. Int J Health Geogr 2012;11:33. Tran A, Ippoliti C, Balenghien T, Conte A, Gely M, Calistri P, Goffredo M, Baldet T, Chevalier V. A geographical information system-based multicriteria evaluation to map areas at risk for Rift Valley fever vector-borne transmission in Italy. Transbound Emerg Dis 2013;2(60 Suppl.):14e23. Bosch J, Munoz MJ, Martinez M, de la Torre A, Estrada-Pena A. Vector-borne pathogen spread through ticks on migratory birds: a probabilistic spatial risk model for South-Western Europe. Transbound Emerg Dis 2013;60:403e15. Hay SI, Battle KE, Pigott DM, Smith DL, Moyes CL, Bhatt S, Brownstein JS, Collier N, Myers MF, George DB, Gething PW. Global mapping of infectious disease. Philos Trans R Soc Lond B Biol Sci 2013;368:20120250. Meltzer MI. The potential use of fractals in epidemiology. Prev Veterinary Med 1991;11:255e60. Lawson AB. Statistical methods in spatial epidemiology. Wiley; 2006. World Organization for Animal Health O. Terrestrial animal health code 2012. Renouf Publishing Company Limited; 2012. Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. Comparative quantification of health risks: Global and regional burden of disease attributable to selected major risk factors. Geneva: World Health Organization; 2004. Renn O, Walker KD. Global risk governance: concept and practice using the IRGC framework. Springer London, Limited; 2008. EFSA Panel on Animal Health and Welfare. Guidance on risk assessment for animal welfare. EFSA J 2012;10:1e30. Hopkin P. Fundamentals of risk management: understanding, evaluating and implementing effective risk management. Kogan Page; 2012. National Research Council N. Science and decisions: advancing risk assessment. The National Academies Press; 2009. Rogers DJ, Suk JE, Semenza JC. Using global maps to predict the risk of dengue in Europe. Acta trop 2014;129:1e14. Rogers DJ, Sedda L. Statistical models for spatially explicit biological data. Parasitology 2012;139:1852e69. Gelfand AE, Diggle P, Guttorp P, Fuentes M. Handbook of spatial statistics. Taylor & Francis; 2010. Clayton D, Kaldor J. Empirical Bayes estimates of agestandardized relative risks for use in disease mapping. Biometrics; 1987::671e81. Maantay JA. Geospatial analysis of environmental health. Springer London, Limited; 2011. Kitron U. Risk maps: transmission and burden of vectorborne diseases. Parasitol Today 2000;16:324e5. Keller M, Blench M, Tolentino H, Freifeld CC, Mandl KD, Mawudeku A, Eysenbach G, Brownstein JS. Use of unstructured event-based reports for global infectious disease surveillance. Emerg Infect Dis 2009;15:689e95.

9

54. Rothman KJ, Greenland S, Lash TL. Modern epidemiology. Wolters Kluwer Health/Lippincott Williams & Wilkins; 2008. 55. Elliott P, Best N. Spatial epidemiology: methods and applications. Oxford University Press; 2001. 56. Ghosh M, Natarajan K, Waller LA, Kim D. Hierarchical Bayes GLMs for the analysis of spatial data: an application to disease mapping. J Stat Plan Inference 1999;75:305e18. 57. Elliott P, Wartenberg D. Spatial epidemiology: current approaches and future challenges. Environ Health Perspect 2004;112:998e1006. 58. Pelosse P, Kribs-Zaleta CM, Ginoux M, Rabinovich JE, Gourbiere S, Menu F. Influence of vectors' risk-spreading strategies and environmental stochasticity on the epidemiology and evolution of vector-borne diseases: the example of Chagas' disease. PLoS one; 2013::8. 59. Fuller D, Troyo A, Alimi T, Beier J. Participatory risk mapping of malaria vector exposure in northern South America using environmental and population data. Appl Geogr 2014;48:1e7. 60. Hill DR. Mapping the risk of yellow fever infection. Curr Infect Dis Reports 2012;14:246e55. 61. Jentes ES, Poumerol G, Gershman MD, Hill DR, Lemarchand J, Lewis RF, Staples JE, Tomori O, Wilder-Smith A, Monath TP. The revised global yellow fever risk map and recommendations for vaccination, 2010: consensus of the Informal WHO working group on geographic risk for yellow fever. Lancet Infect Dis 2011;11:622e32. 62. Field VK, Ford LMM, Hill DR. Health information for overseas travel: prevention of illness in travellers from the UK [Rev. ed.]. London: National Travel Health Network and Centre; 2010. 63. Markoff L. Yellow fever outbreak in Suda. N Engl J Med 2013;368:689e771. 64. Brown L, Medlock J, Murray V. Impact of drought on vectorborne diseasesehow does one manage the risk? Public health 2014;128:29e37. 65. Braks M, van Ginkel R, Wint W, Sedda L, Sprong H. Climate change and public health policy: translating the science. Int J Environ Res Public Health 2014;11:13e29. 66. Eisen L, Wong D, Shelus V, Eisen RJ. What is the risk for exposure to vector-borne pathogens in United States national parks? J Med Entomology 2013;50:221e30. 67. Eisen RJ, Eisen L. Spatial modeling of human risk of exposure to vector-borne pathogens based on epidemiological versus arthropod vector data. J Med Entomol 2008;45:181e92. 68. Pohjola MV, Leino O, Kollanus V, Tuomisto JT, Gunnlaugsdottir H, Holm F, Kalogeras N, Luteijn JM, Magnusson SH, Odekerken G, Tijhuis MJ, Ueland O, White BC, Verhagen H. State of the art in benefit-risk analysis: environmental health. Food Chem Toxicol 2012;50:40e55. 69. Setbon M, Raude J. Population response to the risk of vectorborne diseases: lessons learned from socio-behavioural research during large-scale outbreaks. Emerg Health Threats J 2009;2:e6. 70. Renn O, Klinke A, van Asselt M. Coping with complexity, uncertainty and ambiguity in risk governance: a synthesis. Ambio 2011;40:231e46. 71. Palatnik-de-Sousa CB, Day MJ. One health: the global challenge of epidemic and endemic leishmaniasis. Parasites Vectors 2011;4:197. 72. Elith J, Leathwick JR. Species distribution models: ecological explanation and prediction across space and time. Annu Rev Ecol Evol Syst 2009;40:677e97. 73. Stirling AC, Scoones I. From risk assessment to knowledge mapping: science, precaution, and participation in disease ecology. Ecol Soc; 2009::14. 74. Hongoh V, Hoen AG, Aenishaenslin C, Waaub JP, Belanger D, Michel P. Spatially explicit multi-criteria decision analysis for managing vector-borne diseases. Int J Health Geogr 2011;10:70.

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

10

p u b l i c h e a l t h x x x ( 2 0 1 4 ) 1 e1 0

75. Reiter P, Thomas CJ, Atkinson PM, Hay SI, Randolph SE, Rogers DJ, Shanks GD, Snow RW, Spielman A. Global warming and malaria: a call for accuracy. Lancet Infect Dis 2004;4:323e4. 76. Suter Ii GW, Norton SB, Barnthouse LW. The evolution of frameworks for ecological risk assessment from the red book ancestor. human and ecological risk assessment. An Int J 2003;9:1349e60. 77. European Commission. Regulation (EC) No 178/2002 of the European Parliament and of the Council of 28 January 2002 laying down the general principles and requirements of food law, establishing the European Food Safety Authority and laying down procedures in matters of food safety. Official J Eur Communities 2002;45:1e24. 78. Mindell JS, Boltong A, Forde I. A review of health impact assessment frameworks. Public Health 2008;122:1177e87. 79. World Health Organization W. International health regulations (2005). Nonserial Publication: World Health Organization; 2006. 80. Ezzati M, Lopez AD, Rodgers AA, Murray CJL. Comparative quantification of health risks: global and regional burden of disease attributable to selected major risk factors. Organisation mondiale ; 2004. de la sante 81. European Commission. Regulation (EC) No. 851/2004 of the European Parliament and of the Council of 21 April 2004 establishing a European centre for disease prevention and control. Official J Eur Union 2004;47:1e11. 82. European Commission. White paper: together for health: a strategic approach for the EU 2008-2013 [white paper]; 2007 [in press]. 83. Brunette GW, Kozarsky PE, Magill AJ, Shlim DR, Whatley AD, editors. CDC health information for international travel: the yellow book 2012. Oxford: Oxford University Press; 2012. 84. Parola P. Eurotravnet. Trop Med Int Health 2011;16:51. 85. Team HW. Travax-travel health-HPS; 2006.

86. Lindsey NP, Brown JA, Kightlinger L, Rosenberg L, Fischer M. State health department perceived utility of and satisfaction with ArboNET, the U.S. National Arboviral Surveillance System. Public Health Rep 2012;127:383e90. 87. Zeldenrust ME, Rahamat-Langendoen JC, Postma MJ, van Vliet JA. The value of ProMED-mail for the Early Warning Committee in the Netherlands: more specific approach recommended. Euro Surveill; 2008::13. 88. Freifeld CC, Mandl KD, Reis BY, Brownstein JS. HealthMap: global infectious disease monitoring through automated classification and visualization of Internet media reports. J Am Med Inform Assoc JAMIA 2008;15:150e7. 89. Collier N, Doan S, Kawazoe A, Goodwin RM, Conway M, Tateno Y, Ngo QH, Dien D, Kawtrakul A, Takeuchi K, Shigematsu M, Taniguchi K. BioCaster: detecting public health rumors with a web-based text mining system. Bioinformatics 2008;24:2940e1. 90. Tolentino H, Kamadjeu R, Fontelo P, Liu F, Matters M, Pollack M, Madoff L. Scanning the emerging infectious diseases horizon e visualizing ProMED emails using EpiSPIDER. Adv Dis Surveillance; 2007::2. 91. Flahault A, Dias-Ferrao V, Chaberty P, Esteves K, Valleron AJ, Lavanchy D. FluNet as a tool for global monitoring of influenza on the Web. JAMA J Am Med Assoc 1998;280:1330e2. 92. Welte VR, Vargas Teran M. Emergency prevention system (EMPRES) for transboundary animal and plant pests and diseases. The EMPRES-livestock: an FAO initiative. Ann N Y Acad Sci 2004;1026:19e31. 93. Jebara KB, Caceres P, Berlingieri F, Weber-Vintzel L. Ten years' work on the World Organisation for animal health (OIE) worldwide animal disease notification system. Prev veterinary Med 2012;107:149e59. 94. Pelat C, Turbelin C, Bar-Hen A, Flahault A, Valleron A. More diseases tracked by using Google TRENDS. Emerg Infect Dis 2009;15:1327e8.

Please cite this article in press as: Sedda L, et al., Risk assessment of vector-borne diseases for public health governance, Public Health (2014), http://dx.doi.org/10.1016/j.puhe.2014.08.018

Risk assessment of vector-borne diseases for public health governance.

In the context of public health, risk governance (or risk analysis) is a framework for the assessment and subsequent management and/or control of the ...
410KB Sizes 0 Downloads 7 Views