Results:

The cohort included 3,082 patients from 680 healthcare facilities from across the United States, of whom 61% were men, 23% were Black, 37% were Hispanics, and 69% were <70 years old. A total of 561 (18.2%) had low titers of anti-SARS-CoV-2 IgG, 2,006 (65%) had medium titers, and 515 (16.7%) had high titers. Clinical characteristics and therapies administered were similar between the three groups. Death within 30 days post-plasma transfusion occurred in 830 (26.9%) patients; 29.6% in the low titer group, 27.4% in the medium titer group, and 22.3% in the high titer group. Patients in the high titer group compared to the low titer group had a relative risk of 0.75 (95% confidence interval [CI], 0.61-0.93). The relative risk in the fully adjusted model was 0.82 (95% CI, 0.67-1.00). In subgroup analyses, the association differed according to whether the patients were mechanically ventilated; with lower risk of death among those who received convalescent plasma but were not on mechanical ventilation and those treated within 3 days of diagnosis. There was no effect in patients already mechanically ventilated. Using machine learning, the most important variables associated with risk of death at 30 days were age, receipt of mechanical ventilation, and anti-SARS-CoV-2 antibody level. Antibody levels remained inversely correlated with risk of death after adjusting for all other variables, most evidently in patients not receiving mechanical ventilation.