Quantitative analysis of the impact of various urban socioeconomic indicators on search-engine-based estimation of COVID-19 prevalence

Numerous studies have proposed search engine-based estimation of COVID-19 prevalence during the COVID-19 pandemic; however, their estimation models do not consider the impact of various urban socioeconomic indicators (USIs). This study quantitatively analysed the impact of various USIs on search engine-based estimation of COVID-19 prevalence using 15 USIs (including total population, gross regional product (GRP), and population density) from 369 cities in China. The results suggested that 13 USIs affected either the correlation (SC-corr) or time lag (SC-lag) between search engine query volume and new COVID-19 cases (p <0.05). Total population and GRP impacted SC-corr considerably, with their correlation coefficients r for SC-corr being 0.65 and 0.59, respectively. Total population, GRP per capita, and proportion of the population with a high school diploma or higher had simultaneous positive impacts on SC-corr and SC-lag (p <0.05); these three indicators explained 37–50% of the total variation in SC-corr and SC-lag. Estimations for different urban agglomerations revealed that the goodness of fit, R2, for search engine-based estimation was more than 0.6 only when total urban population, GRP per capita, and proportion of the population with a high school diploma or higher exceeded 11.08 million, 120,700, and 38.13%, respectively. A greater urban size indicated higher accuracy of search engine-based estimation of COVID-19 prevalence. Therefore, the accuracy and time lag for search engine-based estimation of infectious disease prevalence can be improved only when the total urban population, GRP per capita, and proportion of the population with a high school diploma or higher are greater than the aforementioned thresholds.