2.3. Data collection and analysis

The data were collected using the survey function of REDCap. Query and data cleaning was conducted to address apparent errors such as duplicate record entry, missing values and outliers. Site monitoring was not conducted.

Continuous variables were described using median and interquartile range and categorical variables using frequencies. For the analysis of prognosis, the Kaplan-Meier estimate and Cox proportional hazards model were used. The date of the first dose of favipiravir was used as the baseline date, and a period up to the date of entry of clinical outcome was handled as the follow-up period. Only the clinical outcome recorded as died in hospital was defined as death. Age, sex, severity of illness, and underlying diseases (diabetes, cardiovascular diseases, chronic lung diseases, and immunosuppression) were considered to possibly affect death, and each of these factors was selected as explanatory variables of the Cox proportional hazards model. No correlations suspected of multicollinearity were noted between each variable. The proportional hazards assumption was confirmed using Schoenfeld's partial residuals of the independent variables included in the model. Age was handled as a continuous variable, and other variables were handled as categorical variables. We used R (version 4.1.3 patched) for all statistical analyses. A two-sided p < 0.05 was considered statistically significant.

Interim reports of the study have previously been released on the Japanese Association of Infectious Diseases website (https://www.kansensho.or.jp/modules/en/) in May and September 2020, and April and November 2021.