文献速递烧伤

儿童意外烧伤相关因素:逻辑回归与数据挖掘算法的比较

Factors Related to Pediatric Unintentional Burns: The Comparison of Logistic Regression and Data Mining Algorithms.
PubMed 文献 · J Burn Care Res · 2019-01-01

该医院病例对照研究在伊朗克尔曼沙赫省纳入15岁以下烧伤患儿与年龄性别频数匹配的对照,比较支持向量机、人工神经网络、随机森林与逻辑回归识别烧伤相关因素。ANN算法表现最优,BMI、社会经济状况、无人看护时长、母亲年龄与教育、家庭规模、父亲职业与年龄、多名看护者及石油储存为最重要因素,多数变量与社会福利和环境相关。

要点

  • ANN算法在识别儿童烧伤相关因素上优于其他算法。
  • 多数烧伤相关变量归因于社会福利状况与环境因素。
  • 减轻这些因素影响或可降低儿童烧伤发生率。

核心结论

  • ANN算法表现优于其他算法,多数烧伤相关变量与社会福利状况和环境有关,减轻这些因素影响可降低儿童烧伤发生率。

研究设计与样本

  • 伊朗克尔曼沙赫省医院病例对照研究,数据收集15个月,病例为 Imam Khomeini 医院烧伤病房15岁以下患儿,对照为 Dr. Mohammad Kermanshahi 医院同龄同性别频数匹配儿童。

主要发现

  • 烧伤患儿平均年龄4.29 ± 3.51岁,58%为男孩。
  • ANN算法表现优于支持向量机、随机森林和逻辑回归。
  • BMI、社会经济状况、无人看护时长、母亲年龄、母亲教育、家庭规模、父亲职业、父亲年龄、多名看护者及石油储存为最重要相关因素。

临床怎么用

  • 临床与公共卫生可关注社会经济状况、看护安排及石油储存等环境因素以预防儿童烧伤。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

Burn injuries are one of the traumas seen in all parts of the world and children are usually one of the vulnerable groups. The aim of this study was to determine the factors related to unintentional burns in children, using data mining algorithms. In this hospital-based case-control study conducted in Kermanshah province, Iran, data were collected over a period of 15 months. Children under the age of 15 years old who were referred to the burn ward of Imam Khomeini Hospital, the only burn referral in Kermanshah province, were included as cases. For the control group, children who were admitted to Dr. Mohammad Kermanshahi Hospital, the only specialist and subspecialist pediatric center in this province, were included. Frequency matching was performed for age and sex. Support vector machine, artificial neural network (ANN), random forest, and logistic regression were employed to determine the factors related to burns in children. The mean age of children with burn injuries was 4.29 ± 3.51 years and 58% of them were boys. The ANN algorithm had better performance than other algorithms. Body mass index (BMI), socioeconomic status, hours without a watchful, mother's age, mother's education, household size, father's job, father's age, having more than one watchful, and petroleum storage were the most important factors related to pediatric burns. The majority of the burn-related variables were related to individuals' social welfare status and their environments. Lessening the effects of these factors could reduce the incidence of pediatric burns.

来源:J Burn Care Res,查看原文。版权归原作者所有。

引用

  1. Aghaei A, Soori H, Ramezankhani A, et al. Factors Related to Pediatric Unintentional Burns: The Comparison of Logistic Regression and Data Mining Algorithms. J Burn Care Res. 2019. PMID: 31116850. doi:10.1093/jbcr/irz066

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。