In Japan, the differences in characteristics, severity, and mortality of novel coronavirus disease (COVID-19) patients between the first and second surges of infections have not been fully understood. This study is a retrospective cohort study of COVID-19 patients confirmed between February 1 and August 31, 2020 in Osaka Prefecture, Japan. Publicly available information on patients was collected from the website of Osaka Prefecture. Patients were divided into two groups according to the date of the positive laboratory test result: the first surge (February 1 to May 22) and the second surge (May 23 to August 31). Patients' characteristics were compared between the two groups. A multivariable Cox proportional-hazards model was applied to compare severity and mortality between the two groups, where sex, age group at the onset date, city of residence, and days to test positive were adjusted. Hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated. A total of 8,541 patients included 1,780 and 6,761 patients in the first and second surges, respectively. Age at the onset date was younger in the second surge (p < 0.001), and median of days from the onset date to the positive test date shortened from 7 to 6 days (p < 0.001). The multivariable Cox proportional-hazards model revealed that both severity and mortality were lower in the second surge than in the first surge (severity: HR: 0.51 [0.39-0.67]; mortality: HR: 0.37 [0.25-0.56]). In conclusion, severity and mortality were lower in the second surge than in the first surge among COVID-19 patients in Osaka Prefecture, Japan.