Data from: Nets versus spraying: a spatial modelling approach reveals indoor residual spraying targets Anopheles mosquito habitats better than mosquito nets in Tanzania

AbstractThe global implementation of malaria interventions has averted hundreds of millions of clinical malaria cases in the last decade. This study assesses predicted Anopheles mosquito distributions across the United Republic of Tanzania before large-scale insecticide-treated net (ITN) rollouts and indoor residual spraying (IRS) initiatives to determine whether mosquito net usage by children under the age of five and IRS are targeted to areas where historical evidence indicates mosquitoes thrive. Demographic and Health Surveys data from 2011-2012 and 2015-2016 include detailed measurements of mosquito net and IRS use across Tanzania. Anopheline data are far less intensively collected, but we constructed a Maxent-built baseline mosquito habitat suitability (MHS) map (AUC=0.872) with Tanzanian Anopheles occurrence records from 1999-2003. This MHS model was tested against independently-observed georeferenced Plasmodium falciparum cases from the Malaria Atlas Project, with ~87% of cases from 1999-2003 (n=107) and ~84% of cases from 1985-2012 (n=919) occurring in areas of high predicted suitability for mosquitoes. We compared the validated MHS with subsequent malaria interventions using mixed effects logistic regression. Specifically, we assessed whether Anopheles habitat suitability related to the frequency that ≥1 child in a household reportedly slept under a mosquito net when that intervention later became widely available, and whether IRS was reportedly applied to dwellings over a one-year period. There was no evidence that mosquito net use the night before the survey related to MHS from 2011-2012 and marginally significant evidence (p<0.05) from 2015-2016 (β=1.466, 95% C.I.=0.848-2.103, marginal R2=0.020, respectively). However, the likelihood of IRS treatments rose relatively strongly in the 12 months prior to both surveys (β=13.466, 95% C.I.=10.488-16.456, marginal R2=0.144, and β=6.817, 95% C.I.=5.439-8.303, marginal R2=0.136, respectively). IRS treatments have therefore been targeted more effectively than mosquito nets toward areas where anopheline habitat suitability was previously found to be high., Usage notesMaxent_dataThis zipped file contains the ASCII files representing the environmental factors used to make the Maxent model and map.Anopheles_c2001This CSV file contains the combined Anopheles mosquito coordinates, where mosquitoes were collected in Tanzania between 1999-2003, for the building of the Maxent model. These data, combined with the ASCII environmental files, will reproduce the Maxent model and map when used as input in the Maxent software.摘要 全球疟疾干预措施在过去十年间已累计避免了数亿例临床疟疾病例。本研究旨在评估坦桑尼亚联合共和国在大规模推广杀虫剂处理蚊帐（insecticide-treated net, ITN）与实施室内残留喷洒（indoor residual spraying, IRS）项目前的按蚊（Anopheles mosquito）预测分布情况，以明确5岁以下儿童使用蚊帐及IRS措施是否精准覆盖了历史证据显示的蚊虫滋生区域。 本研究使用了2011-2012年及2015-2016年的人口与健康调查（Demographic and Health Surveys, DHS）数据，该数据包含坦桑尼亚全国范围内蚊帐及IRS使用情况的详细统计信息。按蚊采样数据相对匮乏，但本研究借助1999-2003年坦桑尼亚按蚊出现记录，构建了基于Maxent模型的基线蚊虫生境适宜性（mosquito habitat suitability, MHS）地图（曲线下面积（Area Under the Curve, AUC）=0.872）。 本研究通过疟疾地图集项目（Malaria Atlas Project, MAP）提供的独立观测地理参考恶性疟原虫（Plasmodium falciparum）病例数据对该MHS模型进行验证：1999-2003年的病例中约87%（n=107）、1985-2012年的病例中约84%（n=919）均分布在模型预测的蚊虫高适宜生境区域内。 本研究采用混合效应逻辑回归分析，将验证后的MHS模型与后续实施的疟疾干预措施进行关联分析。具体而言，我们评估了两点：一是当蚊帐干预措施大规模推广后，家庭中至少1名儿童报告睡前使用蚊帐的频率是否与按蚊生境适宜性相关；二是一年内报告对住宅实施IRS的情况是否与按蚊生境适宜性相关。 2011-2012年的调查数据未显示调查前一晚的蚊帐使用情况与MHS存在关联；而2015-2016年的数据则显示出边际显著性关联（p<0.05，β=1.466，95%置信区间（Confidence Interval, CI）=0.848-2.103，边际决定系数（marginal R²）=0.020）。但两次调查前12个月内的IRS实施概率均呈现显著上升趋势（两次调查对应的参数分别为β=13.466，95%置信区间（Confidence Interval, CI）=10.488-16.456，边际决定系数（marginal R²）=0.144；以及β=6.817，95%置信区间（Confidence Interval, CI）=5.439-8.303，边际决定系数（marginal R²）=0.136）。因此，相较于蚊帐干预措施，IRS措施更精准地覆盖了此前被判定为按蚊生境高适宜性的区域。 使用说明 Maxent_data：该压缩文件包含用于构建Maxent模型及地图的环境因子ASCII文件。 Anopheles_c2001：该CSV文件包含1999-2003年间在坦桑尼亚采集的按蚊合并坐标数据，用于构建Maxent模型。将该数据与ASCII环境文件结合后，作为Maxent软件的输入即可复现本研究的Maxent模型及地图。