PERFORMANCE OF GOOGLE TRENDS FOR EARLY DETECTION OF DENGUE INFECTION EPIDEMICS IN JAKARTA AND YOGYAKARTA

Background: Early detection of disease outbreak is among the most critical role of the sub-national authorities as mandated by the health decentralization policy. Given the continuous growth of Internet penetration and dependencies of the society on the digital ecosystem, it is essential to investigate the potential innovations to improve the existing surveillance system using digital epidemiology. Several studies, including in Indonesia, have assessed the roles of Google Trends (GT) to improve dengue surveillance systems. However, they were mostly located in specific areas or national level only. No reports are available to compare the performance of GT for early detection of dengue outbreak among high burdened provinces. Aims: This study aimed to examine the correlation between GT data on dengue-related query terms with the official dengue surveillance reports in Jakarta and Yogyakarta Province. Methods: Relative Search Volume of GT data for dengue were collected from the area of Jakarta and Yogyakarta between 2012 to 2016. Those data were compared with the official dengue reports from the Indonesian Ministry of Health using Pearson correlation and Time-lag correlation, performed with Stata version 13. Results: GT data are positively correlated with the routine surveillance report in Jakarta (r = 0.723, p-value= 0.000) and Yogyakarta Province (r = 0.715, p-value= 0.000). In Jakarta, search term of DBD demonstrated a very strong correlation for lag-1 (r =0.828, p-value= 0.000). This finding indicates that GT data could possibly detect the dengue outbreak a month earlier, especially in Jakarta. Hence, GT data can be used to monitor disease dynamics and improve the public awareness of a potential outbreak in near-real-time. Conclusion: GT data were positively correlated with the routine surveillance report in Jakarta and Yogyakarta Province. Early warning system utilizing GT data is potentially more accurate in Jakarta than in Yogyakarta. We assume that it is related with the larger population as well as the Internet use activities that drives the higher volume of Google search on dengue in Jakarta compared to Yogyakarta. Further studies involving other digital data sources, for example, Twitter, online news, and administrative data from the national health insurance are essential to strengthen the current surveillance system with the new digital epidemiology approach. Keywords: dengue, digital epidemiology, Google Trends, Indonesia