Refining Postoperative Intensive Care Triage After Anatomical Lung Resection: A Retrospective Cohort Study of Perioperative Reassessment

Figures

• 

  Figure 1. Patient selection and postoperative triage flow. Institutional lung cancer patient records were retrospectively reviewed, and 1060 patients who underwent elective anatomical lung resection for primary non-small cell lung cancer between 1 January 2019, and 1 December 2025, were identified as the screened population for the present analysis. Among these, 159 patients had a documented preoperative recommendation for postoperative ICU monitoring recorded in the routine preoperative anaesthesia assessment and constituted the analytical cohort. All 159 patients fulfilled the prespecified eligibility criteria, none met the predefined exclusion criteria, and no additional patient-level exclusions were applied after cohort identification. The remaining 901 screened patients did not have a documented preoperative ICU monitoring recommendation and were therefore not included in the analytical cohort. Of the 159 included patients, 28 were admitted directly to the ICU from the operating room, whereas 131 were initially monitored in the PACU. Among patients initially monitored in the PACU, 17 were subsequently transferred to the ICU and 114 were transferred to the surgical ward. Overall,…

• 

  Figure 2. Receiver operating characteristic curves for the three ICU triage approaches. AUC, area under the receiver operating characteristic curve; ICU, intensive care unit.

• 

  Figure 3. Calibration plot for the primary perioperative multivariable model. Each point represents a decile of predicted probability plotted against the observed proportion of actual postoperative ICU admission. The diagonal line indicates perfect calibration. Because each decile included a limited number of patients, the plot should be interpreted as a preliminary assessment of calibration. ICU, intensive care unit.

• 

  Figure 4. Decision curve analysis for the primary perioperative multivariable model, the exploratory perioperative ICU triage score, and the preoperative-only model. The analysis compares the net benefit of each approach against treat-all and treat-none reference strategies across threshold probabilities. Decision curve findings were derived from the same cohort used for model development and should therefore be interpreted as exploratory. ICU, intensive care unit.

• 

  Figure 5. Actual postoperative ICU admission rates by exploratory perioperative ICU triage score category. Numbers above the bars indicate the proportion (%) and count (n/total) of patients with actual postoperative ICU admission within each score category. The score 4 category included only three patients; therefore, this estimate should be interpreted descriptively. ICU, intensive care unit.