Improving malaria risk assessment in Blantyre district, Malawi by optimizing Anopheles surveillance using open-source and real-time data

疟疾是马拉维的一个重要公共卫生问题，仅2017年就报告了430多万例病例。这是一种通过按蚊在人与人之间传播的疾病。疟疾可以通过治疗病例、将按蚊从一个地区清除（例如使用杀虫剂）以及在当地人口和蚊子之间建立屏障（例如使用蚊帐）来预防。在过去十年中，对蚊虫控制和医疗保健的大量投资有助于减少疟疾病例。虽然这是成功的，但资源还不足以将疟疾发病率降低到每年400万例以下。由于我们不可能治疗所有人，也不可能在所有地方使用蚊虫控制方法，因此需要以最有效和最有效的方式分配资源。这需要更好地了解按蚊和疟疾疾病的动态。要了解马拉维的疟疾动态，监测蚊子是必不可少的。不幸的是，蚊子监测是一项昂贵且耗时的活动，需要经验丰富且训练有素的人员。提高蚊子监测效率和效果的一个重要方法是针对特定地区：不是从所有地区收集蚊子，而是只将监测重点放在高风险地区。这些目标区域可以使用过去（几年前和几天前）的数据来确定，这些数据显示了有助于预测未来的趋势。在疾病暴发方面已经进行了更大规模的这样做，预防活动集中在高风险省份或地区。虽然有价值，但在这些大范围内实施控制活动是不可行的。在此期间，我想在更精细和更实用的规模上使用类似的技术。将开发一种有针对性的蚊虫监测工具，1)确定高风险地区，2)在这些高风险地区中选择应进行蚊虫监测的地点，以进行准确的风险评估。首先，我将确定与按蚊增加和随后疟疾病例增加有关的变量（如温度、降雨、土地覆盖和人口密度）。将分析蚊子的历史数据、疟疾疾病数据和捕捉环境变量的卫星数据，以确定趋势。其次，与按蚊增加相关的变量和值将用于开发一个预测模型，以确定高风险地区。每日卫星数据将包括在模型中（实时数据），以便进行最新预测。该模型将与一个强大的蚊子抽样框架联系起来，该框架将确定这些高风险地区中应该进行监测的特定地点。将通过将该工具与目前的蚊子监测活动进行比较，在现场验证该工具。最后，预测模型将被改编成一个用户友好的界面，它需要有限的培训，并提供明确的方向。我将与马拉维卫生部密切合作，在病媒控制规划中实施这一决策支持工具。该奖学金的目的是帮助确定资源较少的高风险地区，弥合病媒控制规划与先进定量方法之间的差距。虽然统计模型和其他先进的统计方法可以提高病媒控制规划的功效和效率，但由于其复杂性，它们很少被使用。我将开发一种易于使用的工具，以确定应该进行蚊子监测的特定区域。它将提供一种经济上可行的抽样策略，改善有限资源的分配。将调查其他地理区域和其他蚊媒疾病（如登革热和寨卡病毒）监测方法的未来发展情况。

Malaria is an important public health problem in Malawi, where in 2017 alone more than 4.3 million cases were reported. It is a disease transmitted between people by a mosquito group called Anopheles. Malaria can be prevented by treating cases, removing Anopheles mosquitoes from an area (e.g. using insecticides) and creating a barrier between the local population and the mosquitoes (e.g. using bed nets). Significant investment in mosquito control and medical care has helped reduce malaria cases in the last decade. Although this has been successful, resources have not been enough to decrease malaria incidence below the annual 4 million cases. As we cannot treat everyone nor use mosquito control methods everywhere, resources need to be allocated in the most efficient and effective way. This requires a better understanding of both the Anopheles mosquitoes and malaria disease dynamics. To understand the malaria dynamic in Malawi, mosquito surveillance is essential. Unfortunately, mosquito surveillance is a costly and time-consuming activity that requires highly experienced and well-trained people. An important way in which mosquito surveillance can become more efficient and effective is by targeting specific areas: instead of collecting mosquitoes from all area, surveillance is focussed on high risk areas only. These targeted areas can be identified using data from the past (years ago and days ago), which show trends that help predict the future. This has already been done on a larger scale for disease outbreaks, where prevention activities are focussed on high risk provinces or regions. Although valuable, it is unfeasible to implement control activities throughout these large areas. During this fellowship I want to use similar technology at a much finer and more pragmatic scale. A targeted mosquito surveillance tool will be developed that 1) identifies high-risk areas and 2) selects sites within these high-risk areas where mosquito surveillance should take place for accurate risk assessment. Firstly, I will identify variables (such as temperature, rainfall, land cover and human population density) associated with an increase in Anopheles mosquitoes and subsequent increase in malaria cases. Historical mosquito data, malaria disease data and satellite data that captures environmental variables, will be analysed to identify trends. Secondly, the variables and values associated with an increase in Anopheles mosquitoes will be used to develop a predictive model that identifies high risk areas. Daily satellite data will be included in the model (real-time data) for up-to-date predictions. This model will be linked to a robust mosquito sampling framework that identifies specific sites within these high-risk areas where surveillance should take place. This tool will be validated in the field by comparing it to current mosquito surveillance activities. Finally, the predictive model will be adapted into a user-friendly interface that it requires limited training and provides clear direction. I will work closely with the Malawian Ministry of Health to implement this decision supporting tool in the vector control program. The aim of this fellowship is to help identify high-risk areas with less resources and bridge the gap between vector control programs and advanced quantitative methods. While statistical models and other advanced statistical approaches can improve efficacy and efficiency of vector control programs, they are rarely used due to their complexity. I will develop an easy-to-use tool that identifies defined areas where mosquito surveillance should take place. It will provide an economically feasible sampling strategy that improves the allocation of the limited resources available. Future development of the surveillance approach to other geographic areas and other mosquito-borne diseases (e.g. dengue and zika) will be investigated.