Skip to main content
. 2021 Jan 30;397(10272):387–397. doi: 10.1016/S0140-6736(21)00001-5

Figure 5.

Figure 5

Capacity to rescue from major complication

(A) Multilevel logistic regression model for predictors of death after major complication in colorectal and gastric cancer. Box size proportional to group size (n). (B) Three-way decomposition mediation model of the proportion of the effect of country income group on 30-day mortality mediated by postoperative care infrastructure (the consistent presence of a designated postoperative recovery area, the availability of critical care facilities, and the existence of a working CT scanner). (C) Proportion of 30-day mortality variation explained at the level of patient or disease, hospital, country, and country income group, in patients with colorectal or gastric cancer who died after major complication. The variance explained at each of the four levels of the model (marginal pseudo R2) is expressed as a proportion of the total variance explained (conditional pseudo R2). (D) Absolute risk difference for 30-day mortality after major complication in the presence of consistently available postoperative care infrastructure. Estimates for age 55 years, ECOG performance status 1, ASA grade 2, cancer stage II, and elective surgery. WB=World Bank. OR=odds ratio. ECOG=Eastern Cooperative Oncology Group. ASA=American Society of Anesthesiologists.