Trends and prediction in daily incidence and deaths of COVID-19 in the United States: a search-interest based model

Yuan, Xiaoling; Xu, Jie; Hussain, Sabiha; Wang, He; Gao, Nan; Zhang, Lanjing

BACKGROUND AND OBJECTIVES: The coronavirus disease 2019 (COVID-19) infected more than 586,000 patients in the U.S. However, its daily incidence and deaths in the U.S. are poorly understood. Internet search interest was found correlated with COVID-19 daily incidence in China, but not yet applied to the U.S. Therefore, we examined the association of internet search-interest with COVID-19 daily incidence and deaths in the U.S. METHODS: We extracted the COVDI-19 daily incidence and death data in the U.S. from two population-based datasets. The search interest of COVID-19 related terms was obtained using Google Trends. Pearson correlation test and general linear model were used to examine correlations and predict future trends, respectively. RESULTS: There were 555,245 new cases and 22,019 deaths of COVID-19 reported in the U.S. from March 1 to April 12, 2020. The search interest of COVID, “COVID pneumonia,” and “COVID heart” were correlated with COVDI-19 daily incidence with ~12-day of delay (Pearson’s r=0.978, 0.978 and 0.979, respectively) and deaths with 19-day of delay (Pearson’s r=0.963, 0.958 and 0.970, respectively). The COVID-19 daily incidence and deaths appeared to both peak on April 10. The 4-day follow-up with prospectively collected data showed moderate to good accuracies for predicting new cases (Pearson’s r=−0.641 to −0.833) and poor to good accuracies for daily new deaths (Pearson’s r=0.365 to 0.935). CONCLUSIONS: Search terms related to COVID-19 are highly correlated with the trends in COVID-19 daily incidence and deaths in the U.S. The prediction-models based on the search interest trend reached moderate to good accuracies.