Dengue risk index as an early warning
2013 (English)Manuscript (preprint) (Other academic)
Introduction: A dengue early warning forewarns stakeholders and promotes timely prevention. Besides accuracy and timeliness, an effective early warning system must be comprised of a structure that allows clear and comprehensible communications to stakeholders, and facilitates planning of actions that corroborate with risks. To aid such communication and planning efforts, this study established a risk-stratified forecast strategy which relies on uniformly interpreted risk indices derived from forecasted dengue cases.
Methodologies & Findings: We adopted the Poisson forecasting model developed by Hii et al. (2012) as model-1 and established a model-2 that considered only temperature and rainfall. We validate and compared the models for their forecast precision and sensitivity to diagnose outbreak and non-outbreak. Models were trained using data from 2001-2010. Forecast precision for the period 2011-2012 was analyzed using six cross-validations of 16-weeks forecast and root mean square errors. Operating Characteristic curve was used to analyze sensitivity of models. Forecasts were then translated into dengue risk indices according to estimated alert and epidemic thresholds.
Results showed that model-1 and model-2 explained about 84% and 70% of variance in dengue distribution, respectively. Average RMSE was 28 for model-1 and 33 for model-2 during cross-validations. ROC area was 0.96 (CI=0.93-0.98) for model-1 and 0.92 (CI=0.88-0.96) for model-2 in 2004-2010. The two models were able to forecast outbreak about 90% accuracy with around 10% false positive in 2011-2012. Monthly and seasonal calendar risk index and weekly time series risk index were established using color scheme to represent risk levels.
Significance: Translation of a forecast to dengue risk index permits rapid and clear interpretation of forecast; thus enhances the effectiveness of an early warning. Further studies on feasibility of developing an automated forecast-control-calibration-system using different forecasting methods to allow parallel forecast for comparison and monitoring will enhance sustainability of forecast precision.
Place, publisher, year, edition, pages
Public Health, Global Health, Social Medicine and Epidemiology
Research subject Public health
IdentifiersURN: urn:nbn:se:umu:diva-68014OAI: oai:DiVA.org:umu-68014DiVA: diva2:615568