Ordinal Logistic Regression Models Application to Convalescent plasma therapy in COVID-19.
COVID-19, a potentially serious disease with high transmissibility and global distribution, has changed the world dynamics. Drugs, alternative treatments and vaccines were tested, one of the treatments evaluated was convalescent plasma. The New York University (NYU) - Grossman School of Medicine, monitored hospitalized patients who received such treatment in 6 countries simultaneously: Belgium, Brazil, Spain, USA, India and Netherlands. Data from several parallel studies were uploaded to a central repository. Some patient characteristics have already been analyzed, such as: sex, origin, age, comorbidities and treatment outcomes. To evaluate the performance of the convalescent plasma treatment, the ordinal logistic regression model will be used. The regression model is promising for the analysis, since the response variable of the model has an ordinal scale very well defined by the WHO. Different ordinal regression models will be tested including proportional odds, partial proportional odds, continuous-ratio and stereotype models. We expect to verify which of the regression models best describes the studied database, identify whether the performance of the treatment depends more on the characteristics of the donors or the recipients, and identify which variables are the most important when choosing this treatment.