Fig. 2

Association of septic shock case volume and the ln(log)-transformed in-hospital mortality rate. Predictions of ln(log)-transformed in-hospital mortality rate are indicated by solid lines and 95% CI by shaded areas; dashed vertical line indicates the threshold before which an increase in volume resulted in a substantial reduction in estimated mortality. Knots were placed at 5th, 35th, 50th, 65th, and 95th centiles of septic shock case volume distribution. Model A was unadjusted (R² 0.141, p for nonlinearity < 0.0001). Model B was adjusted for type of hospitals, geographic location, adherence to 3-h SSC bundle, site of infection (sites of bloodstream, urinary, skin, central nervous system, lung, bone, abdominal, gastrointestinal), microbiology specimen collection before antibiotic therapy, infection management, infection training, infection monitor, infection contingency plan, infection performance, and the proportion of APACHE II score ≥ 15 and proportion of age ≥ 60 (R² 0.17, p for nonlinearity < 0.0001)