Reintubation After Lung Cancer Resection: Development and External Validation of a Predictive Score

Abstract

Background: Reintubation after lung cancer resection is an important quality metric because of increased disability, mortality and cost. However, no validated predictive instrument is in use to reduce reintubation after lung resection. This study aimed to create and validate the PRediction Of REintubation After Lung cancer resection (PROREAL) score.

Methods: The study analyzed lung resection cases from 2 university hospitals. The primary end point was reintubation within 7 days after surgery. Predictors were selected through backward stepwise logistic regression and bootstrap resampling. The investigators used reclassification and receiver-operating characteristic (ROC) curve analyses to assess score performance and compare it with an established score for all surgical patients (Score for Prediction of Postoperative Respiratory Complications [SPORC]).

Results: The study included 2672 patients who underwent resection for lung cancer (1754, development cohort; 918, validation cohort) between 2008 and 2020, of whom 71 (2.7%) were reintubated within 7 days after surgery. Identified score variables were surgical extent and approach, American Society of Anesthesiologists physical status, heart failure, renal disease, and diffusing capacity of the lung for carbon monoxide. The score achieved excellent discrimination in the development cohort (ROC AUC, 0.90; 95% CI, 0.87-0.94) and good discrimination in the validation cohort (ROC AUC, 0.74, 95% CI; 0.66-0.82), thus outperforming the SPORC in both cohorts (P < .001 and P = .018, respectively; validation cohort net reclassification improvement, 0.39; 95% CI, 0.18-0.60; P = .001). The score cutoff of ≥5 yielded a sensitivity of 88% (95% CI, 72-95) and a specificity of 81% (95% CI,79-83) in the development cohort.

Conclusions: A simple score (PROREAL) specific to lung cancer predicts postoperative reintubation more accurately than the nonspecific SPORC score. Operative candidates at risk may be identified for preventive intervention or alternative oncologic therapy.

FIGURE 1

Study flow diagram. Detailed patient inclusion and exclusion. The asterisk indicates that multiple criteria may apply. (ASA, American Society of Anesthesiologists; BMI, body mass index; DLCO, diffusion capacity of the lung for carbon monoxide; FEV₁, forced expiratory volume in 1 second.)

FIGURE 2

Score components. (ASA; American Society of Anesthesiologists; DLCO, diffusion capacity of the lung for carbon monoxide; PROREAL, PRediction Of REintubation After Lung cancer resection.)

FIGURE 3

Receiver-operating characteristic curves. This graph shows the score performances assessed with a receiver-operating characteristic curve for predicting reintubation by the PRediction Of REintubation After Lung cancer resection ( PROREAL) score in (A) the development cohort and (B) the validation cohort.

FIGURE 4

Calibration plots. The graphs show the calibration plots of the PRediction Of REintubation After Lung cancer resection (PROREAL) score for the prediction of reintubation in (A) the development cohort and (B) the validation cohort. Calibration plots indicate the observed over expected fraction of reintubation.