Development of dengue outbreak risk prediction with ecosystem approach in Gombak, Selangor.

Abstract

The dengue infection is influenced by the complex interaction between agent, host, environment and vector. In 2019, the incidence of dengue in Malaysia was 390.4 per 100 000 population with an increment of 61.4% from the previous year. The dengue risk assessment by using classical Stegomyia indexes has been reported to be imprecise in several studies. High larval indices do not necessarily increase the dengue transmission, thus cannot be used to predict dengue outbreak occurrence. The objective of this research is to develop and validate dengue outbreak risk prediction using ecosystem approach in Gombak, Selangor. In phase one study, 40 samples were taken randomly from all dengue locality in Gombak districts. The data collection for phase one study was conducted for six months duration. Information on vector surveillance, both single and outbreak localities were taken, and Unmanned-Aerial-Vehicle (UAV) were deployed to get an aerial photograph. While in phase two study, localities were taken randomly in July 2019 and the photographs were quantified, analysed and stratified accordingly. Eight machine learning methods were used to analyse and predict the outbreak occurrence. The localities have been followed up for six months to monitor the outbreak formation. Twenty-six outbreak and 14 single case localities were analysed in phase one study where three significant risk factors have been associated with outbreak; overall cleanliness with pOR: 16.54 (95% CI: 1.80 – 152.18), water body index with pOR: 14.80 (95% CI: 1.38 – 158.99), and the presence of garden with pOR: 8.87(95% CI: 1.24 – 63.59). The Artificial neural network method shows the highest predicted accuracy (Accuracy: 0.85 with 95% CI = 70.16% - 94.29%), substantial Cohen’s Kappa agreement and better AUC score. The dengue outbreak risk prediction via aerial photography can estimate the risk of dengue outbreak occurrence and valid to be performed as a dengue risk assessment tool.