Effectiveness of drugs for COVID-19 inpatients in Japanese medical claim data as average treatment effects with inverse probability weighted regression adjustment

BackgroundPrior studies have indicated that drugs against coronavirus disease 2019 (COVID-19) such as antiviral drugs, anti-inflammatory drugs, steroid and antibody cocktails are expected to prevent severe COVID-19outcomes and death. ObjectWe analyzed medical claim data in Japan to assess the effectiveness of drugs againstCOVID-19. MethodWe applied an average treatment effect model with inverse probability weighted regression adjustment, to the Medical Information Analysis Databank managed by National Hospital Organization in Japan. The outcome was death during hospitalization. Subjects were all inpatients, inpatients with oxygen therapy, and inpatients with respiratory ventilators, by three age classes: all ages, 65 years old or older, and younger than 65 years old. Data on physical characteristics, underlying diseases, administered drugs, the proportion of mutated strains, and vaccine coverage were used as explanatory variables for logistic regression. ResultEstimated results indicated that only an antibody cocktails (sotrovimab, casirivimab and imdevimab) raised the probability of saving life, even though these drugs were administered in few cases. On the other hand, other drugs might raise the probability of death. DiscussionResults indicated that only antibody cocktails was effective to save life using an average treatment effect model with inverse probability weighted regression adjustment. No other drugs such as remdesivir, dexamethasone, baricitinib and tocilizumab were found to be effective to save life, even in the pseudo-situation of random assignment.