Development and validation of a pulmonary complications prediction model based on the Yang's index.

BACKGROUND

Blunt chest trauma patients with pulmonary contusion are susceptible to pulmonary complications, and severe cases may develop respiratory failure. Some studies have suggested the extent of pulmonary contusion to be the main predictor of pulmonary complications. However, no simple and effective method to assess the severity of pulmonary contusion has been available yet. A reliable prognostic prediction model would facilitate the identification of high-risk patients, so that early intervention can be given to reduce pulmonary complications; however, no suitable model based on such an assumption has been available yet.

METHODS

In this study, a new method for assessing lung contusion by the product of the three dimensions of the lung window on the computed tomography (CT) image was proposed. We conducted a retrospective study on patients with both thoracic trauma and pulmonary contusion admitted to 8 trauma centers in China from January 2014 to June 2020. Using patients from 2 centers with a large number of patients as the training set and patients from the other 6 centers as the validation set, a prediction model for pulmonary complications was established with Yang's index and rib fractures, etc., being the predictors. The pulmonary complications included pulmonary infection and respiratory failure.

RESULTS

This study included 515 patients, among whom 188 developed pulmonary complications, including 92 with respiratory failure. Risk factors contributing to pulmonary complications were identified, and a scoring system and prediction model were constructed. Using the training set, models for adverse outcomes and severe adverse outcomes were developed, and area under the curve (AUC) of 0.852 and 0.788 were achieved in the validation set. In the model performance for predicting pulmonary complications, the positive predictive value of the model is 0.938, the sensitivity of the model is 0.563 and the specificity of the model is 0.958.

CONCLUSIONS

The generated indicator, called Yang's index, was proven to be an easy-to-use method for the evaluation of pulmonary contusion severity. The prediction model based on Yang's index could facilitate early identification of patients at risk of pulmonary complications, yet the effectiveness of the model remains to be validated and its performance remains to be improved in further studies with larger sample sizes.