Evaluation of Developmental Metrics for Utilization in a Pediatric Advanced Automatic Crash Notification Algorithm

ABSTRACT<b>Objective</b>: Appropriate treatment at designated trauma centers (TCs) improves outcomes among injured children after motor vehicle crashes (MVCs). Advanced Automatic Crash Notification (AACN) has shown promise in improving triage to appropriate TCs. Pediatric-specific AACN algorithms have not yet been created. To create such an algorithm, it will be necessary to include some metric of development (age, height or weight) as a covariate in the injury risk algorithm. This study sought to determine which marker of development should serve as a covariate in such an algorithm and to quantify injury risk at different levels of this metric. <b>Methods</b>: Retrospective review of occupants &lt;19yrs within the MVC dataset NASS-CDS 2000–2011 was performed. R² values of logistic regression models using age, height or weight to predict 18 key injury types were compared to determine which metric should be used as a covariate in a pediatric AACN algorithm. Clinical judgment, literature review and Chi Square analysis were used to create groupings of the chosen metric that would discriminate injury patterns. Adjusted odds of particular injury types at the different levels of this metric were calculated from logistic regression while controlling for gender, vehicle velocity change (delta V), belted status (optimal, sub-optimal or unrestrained) and crash mode (rollover, rear, frontal, near-side or far-side). <b>Results</b>: NASS-CDS analysis produced 11,541 occupants &lt;19yrs with non-missing data. Age, height and weight were correlated with one another and with injury patterns. Age demonstrated the best predictive power in injury patterns and was categorized into bins of 0–4 years, 5–9 years, 10–14 years and 15–18 years. Age was a significant predictor of all 18 injury types evaluated even when controlling for all other confounders and when controlling for age and gender specific BMI classifications. Adjusted odds of key injury types with respect to these age categorizations revealed that younger children were at increased odds of sustaining AIS 2+ and 3+ head injuries and AIS 3+ spinal injuries, while older children were at increased odds of sustaining thoracic fractures, AIS 3+ abdominal injuries, and AIS 2+ upper and lower extremity injuries. <b>Conclusions</b>: The injury patterns observed across developmental metrics in this study mirror those previously described among children with blunt trauma. This study identifies age as the metric best suited for use in a pediatric AACN algorithm and utilizes 12 years of data to provide quantifiable risks of particular injuries at different levels of this metric. This risk quantification will have important predictive purposes in a pediatric-specific AACN algorithm.

摘要
<b>研究目的</b>：在指定创伤中心（Trauma Centers, TCs）接受恰当救治，可改善机动车碰撞（Motor Vehicle Crashes, MVCs）后受伤儿童的预后。高级自动碰撞通知（Advanced Automatic Crash Notification, AACN）已被证实可优化创伤中心的分诊适配性，但目前尚未开发出儿科专用的AACN算法。若要构建此类算法，需将发育相关指标（年龄、身高或体重）作为协变量纳入损伤风险预测模型。本研究旨在明确此类算法中应选用何种发育指标作为协变量，并量化该指标不同水平下的损伤风险。
<b>研究方法</b>：本研究对2000–2011年国家汽车采样系统碰撞数据档案（National Automotive Sampling System Crashworthiness Data System, NASS-CDS）数据集内年龄小于19岁的机动车碰撞乘员开展回顾性分析。通过比较以年龄、身高或体重作为协变量的逻辑回归模型的R²值，筛选出最适合作为儿科AACN算法协变量的发育指标，该模型用于预测18种关键损伤类型。结合临床判断、文献复习与卡方检验，对选定的发育指标进行分组以区分不同损伤模式。在控制性别、车辆碰撞速度变化量（delta V）、安全带使用状态（规范使用、不规范使用或未使用）及碰撞类型（翻滚、后方碰撞、正面碰撞、近侧碰撞或远侧碰撞）的前提下，通过逻辑回归计算该指标不同水平下特定损伤类型的校正比值比。
<b>研究结果</b>：通过对NASS-CDS数据集的分析，共纳入11541名年龄小于19岁且数据完整的乘员。年龄、身高与体重三者之间以及它们与损伤模式均存在相关性。年龄在预测损伤模式方面展现出最优的预测效能，因此将其划分为0–4岁、5–9岁、10–14岁及15–18岁四个年龄段。即使在控制所有其他混杂因素以及按年龄和性别分层的身体质量指数（Body Mass Index, BMI）分类后，年龄仍是本次评估的18种损伤类型的显著预测因子。针对上述年龄分组的关键损伤类型校正比值比分析显示，低龄儿童发生简明损伤定级（Abbreviated Injury Scale, AIS）2+及3+头部损伤、AIS 3+脊柱损伤的风险更高，而大龄儿童则更易发生胸部骨折、AIS 3+腹部损伤以及AIS 2+上下肢损伤。
<b>研究结论</b>：本研究中不同发育指标对应的损伤模式与既往钝性创伤儿童的报道结果一致。本研究明确了年龄是最适合纳入儿科AACN算法的发育指标，并利用12年的数据集量化了该指标不同水平下的特定损伤风险。该风险量化结果将为儿科专用AACN算法提供重要的预测支撑。