Figure 2

Shrunken residuals (and standard errors) by PSI. y axis: random effect (standard error). x axis: hospitals sorted by random effect a. Mortality in Low-Mortality DRGs. Note: Random effect (and standard error) after modelling the cluster effect. No patient variables were adjusted as Mortality in Low-Mortality DRGs is considered a sentinel-like event. b. Decubitus ulcer. Note: Random effect (and standard error) after modelling the cluster effect. Patient variables adjusted in the model were: age, sex, paralysis, other neurological disorders, diabetes with chronic complications, weight loss and fluid and electrolytic disorders. c. Catheter-related infections. Note: Random effect (and standard error) after modelling the cluster effect. Patient variables adjusted in the model were: age, sex, peripheral vascular disease, paralysis, weight loss, fluid and electrolytic disorders. d. Postoperative PE or DVT. Note: Random effect (and standard error) after modelling the cluster effect. Patient variables adjusted in the model were: age, sex, pulmonary circulation disease, paralysis, lymphoma, metastatic cancer, solid tumor w/o metastasis, coagulopathy and weight loss. e. Postoperative sepsis. Note: Random effect (and standard error) after modelling the cluster effect. Patient variables adjusted in the model were: age, sex, congestive heart failure, paralysis, and weight loss