Background Estimates of fatality rate for patients dying with COVID-19 vary widely Incorporation of the survival time into predictive models increases the accuracy of fatality rate estimates by reducing sampling bias We applied predictive modelling approaches to estimate the current mortality rate and survival time for patients in England with laboratory-confirmed COVID-19 We used these data to develop a model to predict trends in future deaths over time Methods 143,463 and 30,028 cumulative laboratory-confirmed COVID-19 cases and deaths published by Public Health England between 30 January and 14 May 2020 for England were analysed Linear regression analysis was utilised to estimate the mortality rate and survival time for patients in England with laboratory-confirmed COVID-19 A predictive model was established which estimated cumulative deaths until 21 May 2020 Joinpoint trend analysis was performed to identify time periods with significantly different rates in daily deaths Results Fatality rate for patients in England with laboratoryconfirmed COVID-19 was 21 9% (95% confidence interval 21 8% to 22 0%) Survival time for patients who died from SARS-CoV-2 infection was seven days In comparison with reported data, the accuracy of predicted trends for cumulative and daily laboratory-confirmed COVID-19 deaths was >99% and >96%, respectively An estimated 31,420 cumulative laboratory-confirmed COVID-19 deaths were predicted to occur in England by 21 May Predicted daily laboratory-confirmed COVID-19 deaths were significantly different during the following time intervals: 10 5 (6 to 17 March), 111 0 (17 to 27 March), 446 8 (27 March to 4 April), 817 0 (4 to 23 April), 536 3 (23 April to 7 May), and 266 7 (7 to 21 May) daily deaths (P