Spatial prediction of malaria in the Red River basin, Yunnan, China using geographical information systems and remote sensing.
AbstractThis study aims to identify risk factors for malaria related to landscape, environmental, and socio-economic and human behaviour variables in the Red River basin area, Yunnan, China, to develop a predictive model of malaria spatial distribution, and to utilise these to improve malaria surveillance and control in the basin area. Yunnan is one of the most endemic areas for malaria in China today, particularly in the Red River basin and its border areas. Chloroquine-resistant falciparum malaria is continuing to increase, partly due to immigration and socio-economic development for agriculture in the region. Traditional intensive surveillance systems for malaria are becoming unreliable. The terrain shows considerable variation and some of it is relatively inaccessible. The environment, particularly its land use pattern keeps changing. The present study has used geographical information systems (GIS) and satellite imagery data to identify malaria risk factors related to landscape and environmental data and to develop a predictive model in hope of predicting the risk of malaria transmission and outbreaks and guiding malaria control in the Red River basin area. The work is in two phases: Phase I was a retrospective ecological study. It used basic GIS techniques to analyse routine malaria data, and existing environmental and ecological data in the Red River basin area primarily to identify major determinants of malaria spatial distribution related to landscape and environmental variables. The malaria data were those were reported from villages through health care systems. In view of their limitations of accuracy and coverage, phase II was undertaken. However, The work of the phase I study helped to formulate the specific hypotheses for phase II study. Phase II study was a prospective study. Malaria incidence data were collected in a field survey of the whole study population by our own research team. Malaria incidence data were integrated with altitude data derived from terrain maps and a land use map derived from SPOT 4 imagery data into the GIS. Multilevel Poisson regression modelling were used to model landscape and land use determinants of malaria. The phase I study was carried out in one county of the Red River basin area, Yuanyang County, Yunnan, China. Malaria and population data at the level of administrative village were collected in 131 administrative villages of the county. Terrain maps, the land use map, soil map and administrative boundaries of Yuanyang as well as malaria risk maps were integrated within GIS. The data were analysed by modelling the risk of malaria in the administrative villages and their landscape and environmental variables generated from GIS. The results of the analysis revealed that spatial distribution of malaria was determined by the landscape and land use patterns in the administrative villages. Malaria was negatively correlated with mean altitudes of administrative villages, but more paddy and forest would increase the risk of malaria in the villages. Phase II study was carried out in the Feng Chun Ling Township of Yuanyang County from May to December 1998. The entire population of 24,280 in 5,007 households was included in the study. Around 14% of the study cohort, mostly from the mountains, however, had a history of temporary migration to the lowlands where malaria is highly endemic during the study period. A total of 649 malaria cases (including 3 mixed infections) were identified in the study cohort during the 7-month period. Of the 649 malaria cases, 400 cases were from the population with a history of temporary migration during study period. The overall risk of people with and without temporary migration history were 118.6 and 12.1 per 1000 persons, respectively, during the 7-month study period. The relative risk of the migrated population against non-migrated population were 9.8, suggesting the migrated population had around 10-fold higher risk of malaria than those of non-migrated population. Only 334 indigenous malaria cases out of all malaria cases (649) in the study were used for further analyses and model building. The risk for P. vivax indigenous malaria was 17.8 per 1,000 person years at risk and for P. falciparum was 6.9 per 1,000 person years at risk. Malaria data were integrated with household locations identified by Global Position Systems (GPS), a land use map derived from a SPOT 4 image and terrain maps into GIS. The results of multilevel Poisson regression modelling in phase II study revealed that indigenous malaria is negatively correlated with altitude for P. vivax and P. falciparum. Paddy and forest would increase the risk of malaria, but the effect of forest on malaria reached a plateau at certain level of the forest coverage. The mosquito net is protective for indigenous malaria in the analysis. The protective efficacy for P. vivax is 40% and P. falciparum 29%, respectively. In conclusion, malaria transmission in the study area is primarily determined by the environmental variables particularly altitude, paddy and forest in the Red River basin area. But human behaviour such as temporary migration and the use of mosquito nets play a very important role in determining the malaria spatial pattern in the study area. Malaria control and surveillance should focus on the lower altitude areas and the mobile population in the Red River basin area. The overall temporarily migrated population plus population living below 1,200 metres accounted for 44.2% (10734/24280) of total population, but accounted for 85.7% (559/652) of total malaria cases during the study period, while migrant plus those living under 800 metres accounted for 73.6% (480/652) of cases in only 21.2% (5155/24280) of the population. Future development of land should be aware of the potential for malaria and other vector borne disease risk arising from expansion of paddy field and deforestation, which will provide breeding sites for mosquitoes, particularly in the middle and lower altitude areas. Subsequently, they might result in malaria and other vector borne disease outbreaks or epidemics. The first priority of malaria control strategy in the immediate future is to encourage local residents and 'downhill' migrants to use mosquito nets and to ensure these are regularly treated with insecticides. Chemoprophylaxis and other control measures need to be explored for the temporarily migrating population in the Red River basin area.
Luo, Da-Peng; (2000) Spatial prediction of malaria in the Red River basin, Yunnan, China using geographical information systems and remote sensing. PhD thesis, London School of Hygiene & Tropical Medicine.