Health in Action The Malaria Atlas Project: Developing Global Maps of Malaria Risk Simon I. Hay*, Robert W. Snow

n allocating public health resources, the guiding principle Ishould be an evidence-based quantifi cation of need. A signifi cant effort to categorize diseases by their global morbidity and mortality impact doi:10.1371/journal.pmed.0030473.g002 has developed during the last decade, epitomized by the Global Burden of MAP Logo Diseases [1] and the Disease Control Priorities projects [2]. But despite these progress has been made [10–21], Endemicity is a measure of the efforts, the evidence base for allocating an evidence-based map of malaria level of malaria challenge in a human resources for malaria control on a transmission intensity for Africa population, and determines the global scale is poor. remains illusive, and there have been average age of fi rst exposure, the rate National reporting on malaria no recent efforts to construct a credible of development of immunity, and continues to be fanciful; Kenya, for evidence-based global malaria map. thus, the expected clinical spectrum example, reported only 135 malaria of disease [25,26]. Therefore, suites deaths in 2002 to the World Health A New Mapping Project of relevant interventions should be Organization [3]. In addition, less The primary goal of the recently tailored to these basic epidemiological than half (22/49) of the malaria- launched Malaria Atlas Project (MAP) foundations [9,27–29]. endemic countries in Africa provided is to develop the science of malaria This is obvious for malaria early information for the most-recent cartography. Our approach will be warning systems, for example, that reporting year, 2003; the rest were fi rst to defi ne the global limits of older [3]. Information on the global contemporary malaria transmission; burden of malaria remains the subject we have initiated this process [12,13], Funding: SIH is funded by a senior research of best guesses rooted in national but will substantially refi ne these layers fellowship from the Wellcome Trust (no. 079091). RWS is a Wellcome Trust principal research fellow reporting systems [3], informed with additional medical intelligence in (no. 079080), and acknowledges the support of estimation based on epidemiological future years. the Kenyan Medical Research Institute. This Health data linked to historical malaria Within these limits, we plan to in Action is published with the permission of the director of the Kenyan Medical Research Institute. distributions [4], or unvalidated then model endemicity using a global This work forms part of the output of the Malaria models of malaria distribution in evidence base of malaria parasite Atlas Project (http://www.map.ox.ac.uk), principally funded by the Wellcome Trust, London, United Africa [5–7]. As a corollary, resource prevalence. This Health in Action Kingdom. allocations for malaria interventions concentrates mostly on how we intend remain driven by perceptions and to achieve this important goal. Once we Competing Interests: The authors have declared that no competing interests exist. politics, rather than an objective have created these global endemicity assessment of need. This status quo is maps, these will then provide a Citation: Hay SI, Snow RW (2006) The Malaria Atlas untenable when global and national baseline to facilitate estimation of Project: Developing global maps of malaria risk. PLoS Med 3(12): e473. doi:10.1371/journal.pmed.0030473 fi nancial resources must be defi ned populations at risk of malaria and to meet needs for new, expensive more-credible predictions of disease Copyright: © 2006 Hay and Snow. This is an open-access article distributed under the terms antimalarial drugs and commodities burden. These maps will also provide of the Creative Commons Attribution License, to prevent infection, and to ensure a platform to help target intervention which permits unrestricted use, distribution, and that these interventions are optimally needs, and may provide a means to reproduction in any medium, provided the original author and source are credited. targeted. measure progress toward national and It has been almost 40 years since the international malaria public health Abbreviations: MAP, Malaria Atlas Project; PR, last global map of malaria endemicity goals at a global scale. parasite rate was constructed [8], and a decade Simon I. Hay and Robert W. Snow are in the Malaria since the need for maps of malaria Why Do We Need Maps of Malaria Public Health and Epidemiology Group, Centre Transmission? for Geographic Medicine, Kenya Medical Research transmission in Africa was fi rst Institute, Nairobi, Kenya. Simon I. Hay is also in advocated [9]. Although substantial Malaria parasite transmission intensity the Spatial Ecology and Epidemiology Group, is spatially heterogeneous [6,22–24]. Department of Zoology, University of Oxford, Oxford, United Kingdom. Robert W. Snow is also at the This heterogeneity has important Centre for Tropical Medicine, John Radcliffe Hospital, The Health in Action section is a forum for individuals implications for risks and age patterns University of Oxford, Oxford, United Kingdom. or organizations to highlight their innovative of progression from malaria infection approaches to a particular health problem. * To whom correspondence should be addressed. to disease, disability, and death [5,25]. E-mail: [email protected]

PLoS Medicine | www.plosmedicine.org 2204 December 2006 | Volume 3 | Issue 12 | e473 Box 1. The MAP Web Site balanced risks of iron defi ciency, epidemiological “training” data, malaria disease incidence, and intensity environmental “predictor” data, and a The MAP Web site (http://www. of transmission [33,34]. Furthermore, suite of statistical mapping techniques map.ox.ac.uk) was launched on May 1, optimizing the introduction of to relate the two is considered below. 2006, to further the aims and ambitions diagnostics to rationalize the use of of MAP. The Web site allows users to new, expensive therapies will require An Archive of Parasite Prevalence visualize the current distribution of the better tools to target where this is There are many ways to measure assembled PR data through static maps cost effi cient and where presumptive the abundance of malaria in a in Web browsers, or more interactively treatment remains appropriate. We given location, and they all have through “.kmz” fi les that enable the data anticipate that other interventions are their advantages and disadvantages to be displayed in Google Earth (http:// likely to have health impacts and cost- that have been reviewed elsewhere earth.google.com). We are currently effectiveness balances that may vary [23,24,38]. Regardless of any interested in gathering additional PR under different endemicity conditions, epidemiological preferences, PR data data from the public health community, and we propose to conduct a detailed indisputably constitute the bulk of the and to facilitate communication we systematic review of the evidence. global information available on the have translated the entire Web site into It is often not immediately distribution of malaria endemicity. Spanish and French. apparent when reading guidelines for The PR is the proportion of a sampled MAP is different than previous malaria control that there are many population that is confi rmed positive attempts at mapping malaria, primarily intervention options available, that for malaria parasites, canonically by because it is a global initiative, but also these may need to be appropriately identifying immature “ring stage” because it aims to share data from the combined, and that the optimal mix trophozoites in blood slides [39]. outset. Those supplying useful PR data could depend on the intensity of We have adopted a single and will be provided with the full database malaria transmission in a given area. traditional classifi cation of malaria for their country of interest, provided This would be as true for Africa as it is endemicity based on the PR [40], full permission is granted from the data for other malarious territories of the to standardize our defi nition of risk owners for distribution. In addition, the Old and New Worlds. Global maps of globally. Endemicity is defi ned by entire database will be released in the malaria endemicity should therefore be the PR in the two- to ten-year age public domain after component outputs essential in every step, from selecting cohort (hypoendemic, less than 0.1; have been peer reviewed. We have set a appropriate intervention options mesoendemic, 0.11–0.5; hyperendemic, June 1, 2009, deadline for this release. and identifying requirements and 0.51–0.75), except for the holoendemic A second unique feature of MAP is that budgeting, to planning, implementing, class (greater than 0.75) where the PR it operates with strict inclusion criteria and monitoring at subnational, refers to the one-year age group [40]. for PR data: only random or complete national, and regional scales. This is important because “risk” is a community-sample surveys conducted geographically relative concept: nation post-1985, where parasite species and Large Area Efforts to Map Malaria states in Latin America identify areas of age groups are defi ned and the survey since the 1960s “high” risk that would be classifi ed as involves more than 50 persons to The fuzzy climate-suitability map for low risk in sub-Saharan Africa. minimize sampling error [68]. Extensive stable Plasmodium falciparum malaria To gather global data on PR surveys details of these and additional inclusion transmission was a milestone in the of a suffi cient extent and density rules are provided online in English, mapping of malaria in Africa [10]. to generate endemicity surfaces at Spanish, French, Chinese, and Swahili. It represented the fi rst attempt for moderate spatial resolution requires Thirdly, the MAP project will collect several decades to provide a map of P. combinations of traditional and data on P. falciparum malaria, as well as falciparum transmission at a continental nontraditional search strategies. the often neglected P. vivax parasite. scale, and has been widely used and This process has involved electronic The Web site also allows formal cited by scientists, international searches of formal literature and acknowledgment of those interested agencies, and national malaria control grey literature databases, as well as individuals and institutions who programmes [6,35,36]. using personal contacts with malaria contribute data. We encourage you to However, it has also been widely research scientists and malaria control have a look and send us feedback at misinterpreted, as it represents personnel. Most recently, we have [email protected]. a measure of the likelihood that developed a Web site to guide people stable transmission can occur, in identifying additional data sources rather than ranges of transmission from areas where information is lacking have a rationale only in epidemic- intensity. Furthermore, it has never (Box 1). As of September 10, 2006, prone areas [30,31]. In addition, been formally evaluated against our search has provided 3,036 spatially intermittent presumptive treatment contemporary parasite rate (PR) data independent geopositioned PR surveys of infants is likely to have little impact outside of Kenya [37]. What is required undertaken since January 1985 from on the incidence of clinical malaria for defi ning both disease risks and an aggregate sample of 2,143,979 and anaemia in areas of exceptionally intervention need is a spatial model blood slides in 79 malaria-endemic low transmission [32]. Moreover, that predicts levels of endemicity, countries. The data included 2,728 where one should withhold iron defi ned and validated by empirical survey locations reporting P. falciparum supplementation in young children data and constructed at a global prevalence (Figure 1) and 1,379 demands an understanding of the scale. This approach to assembling locations reporting P. vivax prevalence.

PLoS Medicine | www.plosmedicine.org 2205 December 2006 | Volume 3 | Issue 12 | e473 doi:10.1371/journal.pmed.0030473.g001

Figure 1. Distribution of the n = 3,036 PR Data Points Collected and Geopositioned by September 10, 2006 There are n = 420 PR surveys conducted in the 1985–1989 period, n = 557 in the 1990–1994 period, n = 556 in the 1995–1999 period, and n = 1,503 in the 2000–2006 period.

R In addition, the distribution of Our proclivities are for such (canonically, 0) [38]. These “force the main Anopheles malaria vectors information derived from Earth- of infection” metrics, however, are and the frequency of the inherited observing satellites [44–48] much less frequently measured, [6,22] haemoglobin disorders will constrain because they are globally consistent and while not recorded at suffi cient malaria infection risks and disease measurements, often more frequency to enable mapping, will be outcomes globally. Both will be subject contemporary and often of higher archived by the MAP as companion to a similar intensity of data search spatial resolution than interpolated data to inform modelling. and assembly, which will be described climatologies [47]. Examples of these Moreover, in high-endemicity elsewhere. environmental data [47] have been areas, PR samples are often restricted made available in the public domain, to children, but in areas of low An Archive of Global and will be hosted on a MAP Web endemicity, surveys are usually Environmental Data site when all necessary distribution extended to include all age groups. Malaria is a vector-borne disease and rights have been negotiated. There is PR is therefore confounded by the the culpable anopheline mosquitoes signifi cant potential for improvement interacting factors of the age of the are very sensitive to climate. This in these environmental themes through population sampled, its immune status, has been exploited in many ways to data collected by new generations and the “detectability” of peripheral predict the distribution of malaria of satellite sensors [48]. Continual parasitaemia [23,24]. It is necessary in time [14,30,31,41,42] and space improvement of these data will be part to transform PR into a measure of the [15,20,43]. Mapping the distribution of the ongoing commitment of MAP. force of infection of malaria controlling of malaria requires spatially referenced for these factors. This is because these data (e.g., altitude, temperature, First Steps to Maps of Force interrelated measures are more closely rainfall, and vegetation extent) to be of Infection related to the life-history characteristics matched to the PR-survey positions, As a cross-sectional measure of and dynamics of the Anopheles vector to establish uni and/or multivariate prevalence, the PR is a less direct populations that we will attempt to statistical relationships between malaria measure of malaria transmission than model with environmental data. Our endemicity and the environment. the entomological inoculation rate goal is to generate APfEIR, C, and R These relationships can then be (the number of infective bites per ultimately 0 surfaces from our PR applied to the environmental data, in capita, often expressed annually for data for mapping, and the modelling more or less sophisticated ways (see P. falciparum, hence APfEIR) [22,23], framework within which to perform below), to generate continuous global the vectorial capacity (canonically, these conversions has already been maps [20]. C) or the basic reproductive number developed [23,24].

PLoS Medicine | www.plosmedicine.org 2206 December 2006 | Volume 3 | Issue 12 | e473 Those techniques required to Future Applications at the global level. Considerable standardize PR for age represent an These planned malaria-endemicity effort as part of MAP is required to ongoing challenge, although methods maps will provide the basis for improve our basic maps of malaria by which to achieve this have been increasing the fi delity of morbidity [4], transmission intensity and help identify suggested [11]. The models written to mortality [5,6] and co-infection burden the global distribution of populations perform these conversions will be made estimates [58,59]. These studies lead at risk of malaria. This will involve freely available in the public domain logically to more-accurate commodity assembling the largest-ever collection pending peer review. R is the chosen demand and budget estimation. As of PR data and a signifi cant parallel MAP platform as it is a programming we have argued, these maps may also investment in establishing the required environment that is free to all provide a means to help determine environmental, population, and (http:⁄⁄www.r-project.org). the distribution of intervention malaria vector data. A very considerable research effort Measuring Risk and Managing types and mixes within countries. Companion maps of the global is also required to evaluate those Uncertainty distribution of the main anopheline statistical techniques needed to relate Quantifying the uncertainty in vector species will also be particularly the PR and environmental data prediction has been a neglected area important in helping inform the for extensive map predictions with in the fi eld of malaria mapping and in appropriate modes of control. At the confi dence intervals. This will take time ecology, more generally [49]. Our aim very least, these map suites should and succeed only with the cooperation is to present all risk maps generated augment the objective monitoring of the malaria control community. To through MAP with uncertainty guides, and evaluation of our interventions encourage interaction, MAP will be companion maps that show the spatial in the coming years. There are connected with the philosophy of open variation in predictive accuracy. We also strong arguments for these exercises access, so that all data collected and intend to evaluate the most-accurate being conducted independently of techniques developed can be made procedures for achieving the basic international agencies responsible for available in the public domain rapidly mapping with appropriate robustness the implementation and evaluation of after peer review. measures [11,50–54]. There are interventions [60–62]. Acknowledgments alternative methods available to achieve While we strive to assemble data to mapping with error estimations: Bayesian defi ne an endemicity baseline, the We thank Drs. Carlos Guerra and Andrew Tatem for commenting on the manuscript. [11,21,55], discriminant analyses static maps we generate will represent [52], and logistic-regression [50,52] a “snapshot” of a dynamic malaria References techniques, among others [53,54], and epidemiology. It is important to 1. Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJ (2006) Global and regional burden these will all be systematically tested establish this epidemiological baseline of disease and risk factors, 2001: Systematic as part of the MAP project. The code because history has shown that changes analysis of population health data. Lancet 367: written to implement these techniques 1747–1757. independent of planned interventions 2. Jamison DT, Breman JG, Measham AR, Alleyne will again be written in R and distributed are inevitable [63]. These global G, Claeson M, et al., editors (2006) Disease freely upon acceptance of its products environmental changes will affect the control priorities in developing countries. Washington (D. C.): World Bank Publications. through peer review. populations at risk of malaria. Land-use 1352 p. changes [64], such as deforestation 3. The World Health Organization (2005) The Where People Live [65], may modify vector population world malaria report 2005. Geneva: World Health Organization. Available: http:⁄⁄www. Accurate population data are critical dynamics, for example. Population rbm.who.int/wmr2005. Accessed 10 September for the assessment of the effects of growth, urbanization [6], and climate 2006. human population density on malaria 4. Snow RW, Guerra CA, Noor AM, Myint HY, change [66] will additionally affect Hay SI (2005) The global distribution of risk and the attribution of risk to human population dynamics. Other clinical episodes of Plasmodium falciparum populations [56]. These databases are factors such as the progression of the malaria. Nature 434: 214–217. 5. Snow RW, Craig M, Deichmann U, Marsh K also becoming increasingly accessible HIV/AIDS pandemic and changes in (1999) Estimating mortality, morbidity and [47,56,57]. Areas of the world for undernutrition and socioeconomic disability due to malaria among Africa’s non- which we have a particularly poor status [66,67] will infl uence the ability pregnant population. Bull World Health Organ 77: 624–640. understanding of human population of human population to cope with 6. Hay SI, Guerra CA, Tatem AJ, Atkinson PM, distribution will limit the accuracy of malaria infection. These changes Snow RW (2005) Urbanization, malaria MAP and other databases to derive transmission and disease burden in Africa. Nat will be signifi cant over the time span Rev Microbiology 3: 81–90. population at-risk estimates. Countries of international development goals 7. Rowe AK, Rowe SY, Snow RW, Korenromp EL, of specifi c concern are highlighted and targets. MAP aims to develop Armstrong Schellenberg JR, et al. (2006) The burden of malaria mortality among African in the ancillary data section of the plausible scenarios for many of these children in the year 2000. Int J Epidemiol 35: MAP Web site. More contemporary or infl uences and techniques to model 691–704. higher spatial resolution census data their potential impact, as they will be 8. Lysenko AY, Semashko IN (1968) [Geography of malaria: A medico-geographic profi le of supplied to MAP will be forwarded confounders in our ability to evaluate an ancient disease]. In: Lebedew AW, editor. with permission to collaborators critically interventions at scale. Medicinskaja Geografi ja. Moscow: Academy of developing the Gridded Population Sciences. pp. 25–146. 9. Snow RW, Marsh K, le Sueur D (1996) The of the World, version 3 (GPWv3) and Conclusions need for maps of transmission intensity to the Global Rural–Urban Mapping The distribution of populations guide malaria control in Africa. Parasitol Today P. falciparum 12: 455–457. Project (GRUMP) (http:⁄⁄sedac.ciesin. exposed to the risk of and 10. Craig MH, Snow RW, le Sueur D (1999) A columbia.edu). P. vivax malaria is poorly understood climate-based distribution model of malaria

PLoS Medicine | www.plosmedicine.org 2207 December 2006 | Volume 3 | Issue 12 | e473 transmission in sub-Saharan Africa. Parasitol Ochola SA, et al. (2003) Forecasting, warning, 50. Augustin NH, Mugglestone MA, Buckland ST Today 15: 105–111. and detection of malaria epidemics: A case (1996) An autologistic model for the spatial 11. Gemperli A, Vounatsou P, Sogoba N, Smith T study. Lancet 361: 1705–1706. distribution of wildlife. J Appl Ecol 33: 339–347. (2006) Malaria mapping using transmission 32. Schellenberg D, Cisse B, Menedez C (2006) 51. Brooker S, Hay SI, Bundy DAP (2002) Tools models: Application to survey data from Mali. The IPTi Consortium: Research for policy and from ecology: Useful for evaluating infection Am J Trop Med Hyg 163: 289–297. action. Trends Parasitol 22: 296–300. risk models? Trends Parasitol 18: 70–74. 12. Guerra CA, Snow RW, Hay SI (2006) Defi ning 33. English M, Snow RW (2006) Iron and folic acid 52. Rogers DJ (2006) Models for vectors and the global spatial limits of malaria transmission supplementation and malaria risk. Lancet 367: vector-borne diseases. Adv Parasitol 62: 1–35. in 2005. Adv Parasitol 62: 157–179. 90–91. 53. Elith J, Graham CH, Anderson RP, Dudik M, 13. Guerra CA, Snow RW, Hay SI (2006) Mapping 34. Sazawal S, Black RE, Ramsan M, Chwaya HM, Ferrier S, et al. (2006) Novel methods improve the global extent of malaria in 2005. Trends Stoltzfus RJ, et al. (2006) Effects of routine prediction of species’ distributions from Parasitol 22: 353–358. prophylactic supplementation with iron occurrence data. Ecography 29: 129–151. 14. Hay SI, Snow RW, Rogers DJ (1998) Predicting and folic acid on admission to hospital and 54. Segurado P, Araujo MB (2004) An evaluation malaria seasons in Kenya using multitemporal mortality in preschool children in a high of methods for modelling species distributions. meteorological satellite sensor data. Trans R malaria transmission setting: Community- J Biogeography 31: 1555–1568. Soc Trop Med Hyg 92: 12–20. based, randomised, placebo-controlled trial. 55. Gemperli A, Vounatsou P (2006) Strategies 15. Hay SI, Omumbo JA, Craig MH, Snow Lancet 367: 133–143. for fi tting large, geostatistical data in MCMC RW (2000) Earth observation, geographic 35. Small J, Goetz SJ, Hay SI (2003) Climatic simulation. Commun Stat Simul Comput 35: information systems and Plasmodium falciparum suitability for malaria transmission in Africa, 331–345. malaria in sub-Saharan Africa. Adv Parasitol 47: 1911–1995. Proc Natl Acad Sci U S A 100: 56. Balk DL, Deichmann U, Yetman G, Pozzi 173–215. 15341–15345. F, Hay SI, et al. (2006) Determining global 16. Kleinschmidt I, Bagayoko M, Clarke GPY, 36. Thomas CJ, Davies G, Dunn CE (2004) population distribution: Methods, applications Craig M, Le Sueur D (2000) A spatial statistical Mixed picture for changes in stable malaria and data. Adv Parasitol 62: 119–156. approach to malaria mapping. Int J Epidemiol distribution with future climate in Africa. 57. Hay SI, Noor AM, Nelson A, Tatem AJ (2005) 29: 355–361. Trends Parasitol 20: 216–220. The accuracy of human population maps for 17. Kleinschmidt I, Omumbo J, Briet O, van de 37. Omumbo JA, Hay SI, Guerra CA, Snow public health application. Trop Med Int Health Giesen N, Sogoba N, et al. (2001) An empirical RW (2004) The relationship between the 10: 1073–1086. malaria distribution map for West Africa. Trop Plasmodium falciparum parasite ratio in 58. Korenromp EL, Williams BG, de Vlas SJ, Gouws Med Int Health 6: 779–786. childhood and climate estimates of malaria E, Gilks CF, et al. (2005) Malaria attributable to 18. Omumbo JA, Hay SI, Goetz SJ, Snow RW, transmission in Kenya. Malaria J 3: 17. the HIV-1 epidemic, sub-Saharan Africa. Emerg Rogers DJ (2002) Updating historical maps of 38. Smith DL, McKenzie FE (2004) Statics and Infect Dis 11: 1410–1419. malaria transmission intensity in East Africa dynamics of malaria infection in Anopheles 59. Mwangi TW, Bethony J, Brooker S (2006) using remote sensing. Photogrammetric Eng mosquitoes. Malaria J 3: 13. Worms and malaria interactions: An Remote Sens 68: 161–166. 39. Gilles HM (1993) The malaria parasites. In: epidemiological viewpoint. Ann Trop Med 19. Omumbo JA, Hay SI, Snow RW, Tatem AJ, Gilles HM, Warrel DA, editors. Bruce-Chwatt’s Parasitol 100: 551–570. Rogers DJ (2005) Modelling malaria risk in essential malariology. 3rd ed. London: Arnold. 60. Murray CJL, Lopez AD, Wibulpolprasert S East Africa at high-spatial resolution. Trop Med pp. 12–34. (2004) Monitoring global health: Time for new Int Health 10: 557–566. 40. Metselaar D, Van Thiel PH (1959) solutions. BMJ 329: 1096–1100. 20. Rogers DJ, Randolph SE, Snow RW, Hay SI Classifi cation of malaria. Trop Geogr Med 11: 61. Attaran A (2005) An immeasurable crisis? A (2002) Satellite imagery in the study and 157–161. criticism of the millennium development goals forecast of malaria. Nature 415: 710–715. 41. Hay SI, Renshaw M, Ochola SA, Noor AM, and why they cannot be measured. PLoS Med 21. Gemperli A, Sogoba N, Fondjo E, Mabaso M, Snow RW (2003) Performance of forecasting, 2: e318. Bagayoko M, et al. (2006) Mapping malaria warning and detection of malaria epidemics 62. Horton R (2005) The Ellison Institute: transmission in West and Central Africa. Trop in the highlands of western Kenya. Trends Monitoring health, challenging WHO. Lancet Med Int Health 11: 1032–1046. Parasitol 19: 394–399. 366: 179–181. 22. Hay SI, Rogers DJ, Toomer JF, Snow RW 42. Hay SI, Snow RW, Rogers DJ (1998) From 63. Hay SI, Guerra CA, Tatem AJ, Noor AM, (2000) Annual Plasmodium falciparum predicting mosquito habitat to malaria seasons Snow RW (2004) The global distribution and entomological inoculation rates (EIR) using remotely sensed data: Practice, problems population at risk of malaria: Past, present, and across Africa: literature survey, Internet access and perspectives. Parasitol Today 14: 306–313. future. Lancet Infect Dis 4: 327–336. and review. Trans R Soc Trop Med Hyg 94: 43. Hay SI, Packer MJ, Rogers DJ (1997) The 64. Feddema JJ, Oleson KW, Bonan GB, Mearns 113–127. impact of remote sensing on the study and LO, Buja LE, et al. (2005) The importance 23. Smith DL, Dushoff J, Snow RW, Hay SI (2005) control of invertebrate intermediate host of land-cover change in simulating future The entomological inoculation rate and and vectors for disease. Int J Remote Sens 18: climates. Science 310: 1674–1678. Plasmodium falciparum infection in African 2899–2930. 65. Guerra CA, Snow RW, Hay SI (2006) A global children. Nature 438: 492–495. 44. Hay SI (2000) An overview of remote sensing assessment of closed forests, deforestation and 24. Smith DL, McKenzie FE, Snow RW, Hay and geodesy for epidemiology and public malaria risk. Ann Trop Med Parasitol 100: SI (2007) Revisiting the basic reproductive health application. Adv Parasitol 47: 1–35. 189–204. number for malaria and its implications for 45. Hay SI, Lennon JJ (1999) Deriving 66. Hay SI, Tatem AJ, Guerra CA, Snow RW malaria control. PLoS Biol. In press. meteorological variables across Africa for the (2006) Infectious diseases: Preparing for the 25. Snow RW, Omumbo JA, Lowe B, Molyneux study and control of vector-borne disease: future. Report T8.2: Population at malaria CS, Obiero JO, et al. (1997) Relation between A comparison of remote sensing and spatial risk in Africa: 2005, 2015 and 2030. London: severe malaria morbidity in children and level interpolation of climate. Trop Med Int Health Department of Trade and Industry. Available: of Plasmodium falciparum transmission in Africa. 4: 58–71. http:⁄⁄www.foresight.gov.uk/ Lancet 349: 1650–1654. 46. Hay SI, Tucker CJ, Rogers DJ, Packer MJ Previous_Projects/Detection_and_ 26. Snow RW, Marsh K (2002) The consequences (1996) Remotely sensed surrogates of Identifi cation_of_Infectious_Diseases/ of reducing transmission of Plasmodium meteorological data for the study of the Reports_and_Publications/Final_Reports/T/ falciparum in Africa. Adv Parasitol 52: 235–264. distribution and abundance of arthropod T8_2.pdf. Accessed 10 September 2006. 27. Greenwood BM, Bojang K, Whitty CJM, Targett vectors of disease. Ann Trop Med Parasitol 67. Snow RW, Hay SI, Marsh K (2006) Infectious GAT (2005) Malaria. Lancet 365: 1487–1498. 90: 1–19. diseases: Preparing for the future. Report 28. Shiff C (2002) Integrated approach to malaria 47. Hay SI, Tatem AJ, Graham AJ, Goetz SJ, Rogers T5.8: Malaria in Africa: Sources, risks, control. Clin Microbiology Rev 15: 278–293. DJ (2006) Global environmental data for drivers and disease burden: 2005–2030. 29. The World Health Organization (2006) mapping infectious disease distribution. Adv London: Department of Trade and Industry. Malaria vector control and personal protection: Parasitol 62: 37–77. http:⁄⁄www.foresight.gov.uk/ Report of a WHO study group. WHO technical 48. Tatem AJ, Goetz SJ, Hay SI (2004) Terra and Previous_Projects/Detection_and_ report series, No. 936. Geneva: World Health Aqua: New data for epidemiology and public Identifi cation_of_Infectious_Diseases/ Organization. 62 p. health. Int J Appl Earth Obs Geoinform 6: Reports_and_Publications/Final_Reports/T/ 30. Hay SI, Rogers DJ, Shanks GD, Myers MF, Snow 33–46. t5_8.pdf. Accessed 10 September 2006. RW (2001) Malaria early warning in Kenya. 49. Clark JS (2003) Uncertainty in ecological 68. Jovani R, Tella JL (2006) Parasite prevalence Trends Parasitol 17: 95–99. inference and forecasting. Ecology 84: and sample size: Misconceptions and solutions. 31. Hay SI, Were EC, Renshaw M, Noor AM, 1349–1350. Trends Parasitol 22: 214–218.

PLoS Medicine | www.plosmedicine.org 2208 December 2006 | Volume 3 | Issue 12 | e473