Fig. 2
Plot of percent overall agreement and free-marginal kappa (κfree) for diagnostic method comparisons stratified by site of infection. For each infection site, the following comparisons were performed: (B) initial assessment by attending physician vs. consensus discharge assessment by site investigators; (C) initial assessment by attending physician vs. external RPD; (D) initial assessment by site investigator vs. consensus discharge assessment; and (E) initial assessment by site investigator vs. external RPD. As a control, the following comparisons were performed for respiratory infection samples including pneumonia (N = 49): (F) consensus discharge assessment vs. external RPD; (G) RPD panelists 1 vs. 2; (H) RPD panelists 1 vs. 3; (I) RPD panelists 2 vs. 3. Note: the number of subjects in the various categories add up to 250 (not 249) because the infection site for one sepsis case was diagnosed as both abdominal and pneumonia. “SIRS” indicates that no site of infection was identified. Horizontal blue bars indicate average values for the free-marginal kappa statistic, over the indicated comparisons