A new quantitative assessment method for predicting pneumonia caused by chest wall injury

Abstract

BACKGROUND:The severity of rib fractures has been previously evaluated by combining categorical data, but these methods have only low predictive capability for respiratory complications and mortality. This study aimed to establish a more accurate method for predicting the development of pneumonia, a frequent complication in chest injuries, using anatomical relationships. METHODS:We analyzed three-dimensional reconstructed images of 644 consecutive trauma patients who underwent whole-body computed tomography (CT) in our institution within a 36-month study period from April 2017. The anatomical relationship between the right and left thoracic volumes of non-rib fracture patients was used to estimate thoracic volume changes on the injured side in unilateral rib fracture patients. The predictive capability of changes in thoracic volume for the development of pneumonia was evaluated according to the area under the receiver operating characteristic curve and compared with that of previous chest wall severity evaluation methods. RESULTS:Of the 644 patients, 133 and 478 patients had unilateral rib fractures and non-rib fractures, respectively. The amount of change in thoracic volume due to unilateral rib fractures was significantly greater in pneumonia patients (400 mL vs. 160 mL, p < 0.01). The area under the receiver operating characteristic curve for the development of pneumonia was 0.83, which tended to be higher than that of the previous severity scoring methods. CONCLUSION:The amount of change in chest volume, which can be estimated using CT images, has better predictive capability for pneumonia than previous severity assessment methods based on categorical data. The amount of change in chest volume measured using whole-body CT can be used to rapidly determine the optimal treatment for severe chest wall injuries.

A Random Forest Model Using Flow Cytometry Data Identifies Pulmonary Infection after Thoracic Injury

“…In addition, because the logistic regression performed well, it supports that the combination of the selected features informs development of pulmonary infection after thoracic injury. Previously developed tools and models for predicting PNA use clinical factors or radiographic findings, which may provide insufficient information and can be time-consuming 29–32 . Our model can be accessed in R by running new patient data through the random forest model using the ranger package.…”

“…Previously developed tools and models for predicting PNA use clinical factors or radiographic findings, which may provide insufficient information and can be time-consuming. [29][30][31][32] Our model can be accessed in R by running new patient data through the random forest model using the ranger package. Once refined and externally validated, the model could exist on the back end of an application in which a clinician would enter the three variables into a user interface within 24 hours of patient admission and be presented with the likelihood of that patient developing PNA.…”

“…The most common thoracic injury pattern was rib fractures (52 of 70 [75.3%]), followed by pneumothorax (40 of 70 ). Median Injury Severity Score for the cohort was 27 (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34) with no significant difference between subjects with and without PNA. Median Sequential (Sepsis-related) Organ Failure Assessment and APACHE scores were significantly higher in PNA patients (7 vs. 4 and 20.5 vs. 9.5, respectively; p < 0.01).…”