A Least Absolute Shrinkage and Selection Operator-Derived Predictive Model for Postoperative Respiratory Failure in a Heterogeneous Adult Elective Surgery Patient Population

Figure 1 –. Conceptual framework of predictive model for postoperative respiratory failure. ASA = American Society of Anesthesiologists; PRF = postoperative respiratory failure. *Comorbid conditions included in the Elixhauser: congestive heart failure, cardiac arrythmias, valvular disease, pulmonary circulatory disorders, peripheral vascular disorders, hypertension (uncomplicated), hypertension (complicated), paralysis, other neurological disorders, chronic pulmonary disease, diabetes (uncomplicated), diabetes (complicated), hypothyroidism, renal failure, liver disease, peptic ulcer disease excluding bleeding, AIDS/HIV, lymphoma, metastatic cancer, solid tumor without metastasis, rheumatoid arthritis/collagen vascular diseases coagulopathy, obesity, weight loss , fluid and electrolyte disorders, blood loss anemia, deficiency anemia, alcohol abuse, drug abuse, psychoses, depression.25 Elixhauser comorbidity score is calculated by assigning weights to each comorbidity based on van Walraven et al.26 NOTE: Elixhauser was only used in e-Appendix 3 sensitivity analysis.

Figure 2 –. Diagram showing steps in the model derivation and validation process. M = mean; LASSO = least absolute shrinkage and selection operator.

Figure 3 –. ROC curve for fitted least absolute shrinkage and selection operator logistic regression predicting postoperative respiratory failure. This model achieved an observed area under the ROC curve (AUC) of 0.851 (95% CI, 0.824–0.878) and an optimism-corrected AUC of 0.835 (95% CI, 0.808–0.862). AUC = area under the operating curve; ROC = receiver operating characteristic.

Figure 4 –. Precision-recall curve for fitted least absolute shrinkage and selection operator logistic regression predicting postoperative respiratory failure. This model achieved an observed area under the precision-recall curve of 0.174 (95% CI, 0.123–0.221) with an optimism-corrected value of 0.156 (95% CI, 0.105–0.203).

Figure 5 –. Calibration plot for the least absolute shrinkage and selection operator logistic regression model predicting postoperative respiratory failure. To create this plot, predicted probabilities were binned into 10 equally sized groups. The mean predicted probability and 95% CI were calculated for each bin and were plotted against the observed proportion of events in each bin. Because of the very low prevalence of events, the mean predicted probability remains small (approximately 5%), even for the bin containing the largest predicted probabilities. The mean predicted probabilities are close to the 45° line, reflecting good agreement between predicted probabilities and observed probabilities, and hence good calibration.