Prediction model for postoperative severe acute lung injury in patients undergoing acute type A aortic dissection surgery

This study aimed to establish a risk assessment model to predict postoperative severe acute lung injury (ALI) risk in patients with acute type A aortic dissection (ATAAD).Consecutive patients with ATAAD admitted to our hospital were included in this retrospective assessment and placed in the postoperative severe ALI and nonsevere ALI groups based on the presence or absence of ALI within 72 h postoperatively (oxygen index [OI] ≤ 100 mmHg). Patients were then randomly divided into training and validation groups in a ratio of 8:2. Univariate and multivariate stepwise forward logistic regression analyses were used to statistically assess data and establish the prediction model. The prediction model's effectiveness was evaluated via 10-fold cross-validation of the validation group to facilitate the construction of a nomogram.After the screening, 479 patients were included in the study: 132 (27.6%) in the postoperative severe ALI group and 347 (72.4%) in the postoperative nonsevere ALI group. Based on multivariate logistics regression analyses, the following variables were included in the model: coronary heart disease, cardiopulmonary bypass (CPB) ≥ 257.5 min, left atrium diameter ≥ 35.5 mm, hemoglobin ≤ 139.5 g/L, preCPB OI ≤ 100 mmHg, intensive care unit OI ≤ 100 mmHg, left ventricular posterior wall thickness ≥ 10.5 mm, and neutrophilic granulocyte percentage ≥ 0.824. The area under the receiver operating characteristic (ROC) curve of the modeling group was 0.805 and differences between observed and predicted values were not deemed statistically significant via the Hosmer-Lemeshow test (χ2 = 6.037, df = 8, p = .643). For the validation group, the area under the ROC curve was 0.778, and observed and predicted value differences were insignificant when assessed using the Hosmer-Lemeshow test (χ2 = 3.3782, df = 7; p = .848). The average 10-fold cross-validation score was 0.756.This study established a prediction model and developed a nomogram to determine the risk of postoperative severe ALI after ATAAD. Variables used in the model were easy to obtain clinically and the effectiveness of the model was good.