Can Anatomical Morphomic Variables Help Predict Abdominal Injury Rates in Frontal Vehicle Crashes?

<b>Objective:</b> Abdominal injuries resulting from vehicle crashes can be significant, in particular when undetected. In this study, abdominal injuries for occupants involved in frontal impacts were assessed using crash and medical data.<b>Methods:</b> Injury rates and patterns were first assessed with respect to thoracic injuries. A statistical analysis was then conducted to predict abdominal injury outcome using 18 covariate variables, including 4 vehicle, 4 demographic, and 10 morphomic, derived from computed tomography (CT) scans. More than 260,000 logistic regression models were fitted using all possible variable combinations. The models were ranked using the Akaike information criterion (AIC) and combined through the model-averaging approach to produce the optimal predictive model. The performance of the models was then assessed using the area under the curve (AUC).<b>Results:</b> The rate of serious thoracic injury was 2.49 times higher than the rate of abdominal injury. The associated odds ratio was 2.31 (P &lt;.01). These results suggest a strong association between serious abdominal and thoracic injuries.The optimal model AUC was 0.646 when using solely vehicle data, 0.696 when combining vehicle and demographic data, 0.866 when combining vehicle and morphomic data, and 0.879 when combining vehicle, demographic, and morphomic data. These results suggest that morphomic variables better predict abdominal injury outcomes than demographic variables. The most important morphomics variables included visceral fat area, trabecular bone density, and spine angulation.<b>Conclusion:</b> This study is the first to combine vehicle, demographic, and anatomical data to predict abdominal injury rates in frontal crashes.**研究目的：** 机动车碰撞导致的腹部损伤往往后果严重，尤其是未被及时发现的病例。本研究利用碰撞及医学数据，对正面碰撞中乘员的腹部损伤情况进行评估。 **研究方法：** 首先针对胸部损伤分析了损伤发生率与损伤模式。随后采用18项协变量开展统计分析以预测腹部损伤结局，这些协变量包括4项车辆相关变量、4项人口统计学变量以及10项源自计算机断层扫描（computed tomography, CT）的形态学变量。本研究通过所有可能的变量组合构建并拟合了超过26万个逻辑回归模型，以赤池信息准则（Akaike information criterion, AIC）对模型进行排序，并通过模型平均法组合得到最优预测模型。随后采用曲线下面积（area under the curve, AUC）评估模型性能。 **研究结果：** 严重胸部损伤的发生率是腹部损伤的2.49倍，对应的优势比为2.31（P < 0.01），提示严重腹部损伤与胸部损伤之间存在显著关联。仅使用车辆数据时，最优模型的AUC为0.646；结合车辆与人口统计学数据时，AUC为0.696；结合车辆与形态学数据时，AUC为0.866；同时结合车辆、人口统计学及形态学数据时，AUC为0.879。上述结果表明，形态学变量在预测腹部损伤结局方面优于人口统计学变量。其中最重要的形态学变量包括内脏脂肪面积、骨小梁密度及脊柱成角。 **研究结论：** 本研究首次结合车辆、人口统计学及解剖学数据，对正面碰撞中的腹部损伤发生率进行预测。