ISSN 2097-6046(网络)
ISSN 2096-7446(印刷)
CN 10-1655/R
主管:中国科学技术协会
主办:中华护理学会

中华急危重症护理杂志 ›› 2024, Vol. 5 ›› Issue (10): 882-888.doi: 10.3761/j.issn.2096-7446.2024.10.003

• 论著 • 上一篇    下一篇

天疱疮患者皮肤多重耐药菌感染两种预测模型的构建及验证

张肖莹, 李柏樟, 方卉, 杨婷婷, 冯佳星, 王璐   

  1. 710032 西安市 空军军医大学第一附属医院皮肤科
  • 收稿日期:2023-10-20 发布日期:2024-10-21
  • 通讯作者: 王璐,E-mail:wanglulu84@163.com
  • 作者简介:张肖莹:女,本科,护师,E-mail:start1yingying@126.com
  • 基金资助:
    国家自然科学基金(82173410)

The development and validation of two prediction models of skin infection affected with multidrug-resistant organism in patients with pemphigus

ZHANG Xiaoying, LI Baizhang, FANG Hui, YANG Tingting, FENG Jiaxing, WANG Lu   

  1. Department of Dermatology,The First Affiliated Hospital of Air Force Medical University,Xi‘an, 710032,China
  • Received:2023-10-20 Published:2024-10-21

摘要: 目的 构建并比较天疱疮患者皮肤多重耐药菌感染Logistic回归模型和分类决策树模型。方法 纳入2015年10月—2023年8月西安市某三级甲等医院收住院的165例天疱疮患者,根据皮肤分泌物培养和药敏结果,将患者分为多重耐药菌组和对照组,比较两组的各项指标并构建Logistic回归预测模型及分类决策树模型,通过区分度(曲线下面积)及准确度(灵敏度、特异度)指标对两种模型进行评价及比较。选择2023年9月—2024年3月住院的42例天疱疮患者对模型进行验证。结果 天疱疮皮肤多重耐药菌感染率为38.8%。Logistic回归模型纳入中性粒细胞绝对值、低蛋白血症及住院季节3个影响因素,模型公式为P=1/[1+exp(-4.168+0.207 × 中性粒细胞绝对值+2.913 × 春季住院+1.256 × 夏季住院+1.508 × 秋季住院+1.340 × 低蛋白血症)];决策树模型纳入中性粒细胞绝对值、系统激素治疗、住院季节3个因素,其中中性粒细胞绝对值是最重要的预测因子。两种模型比较,曲线下面积(0.820和0.795)差异无统计学意义(P=0.438);Logistic回归模型灵敏度为0.609,特异度为0.871,决策树模型灵敏度为0.422,特异度为0.970。模型验证结果:Logistic回归模型曲线下面积为0.784,灵敏度为0.625,特异度为0.769;决策树模型曲线下面积为0.804,灵敏度为0.500,特异度为0.885。结论 基于Logistic回归和决策树构建的模型均有较好预测效能且二者差异无统计学意义,护理工作中可联合使用两种模型以提高预测准确度。

关键词: 天疱疮, 多重耐药菌, Logistic回归, 决策树, 影响因素分析, 护理

Abstract: Objective To construct and compare Logistic regression model and classification decision tree model of skin infection affected with multidrug-resistant organism(MDRO) in patients with pemphigus. Methods A total of 165 inpatients with pemphigus admitted to a tertiary class A hospital in Xi‘an from October 2015 to August 2023 were included and divided into the MDRO group and the control group according to the results of secretion culture and drug sensitive test from skin lesions. The factors were compared between the two groups and were used to established Logistic regression model and decision tree model. The discrimination(area under the curve) and accuracy(sensitivity and specificity) were committed to evaluate and compare between the two models. The two prediction models were validated by a verification group of 42 patients with pemphigus from September 2023 to March 2024. Results The skin infection rate of MDRO in pemphigus was 38.8%. Absolute neutrophil count (ANC),hypoproteinemia and season of hospitalization were included in Logistic regression model. The model formula was P=1/[1+exp(-4.168+0.207 × ANC+2.913 × hospitalization in spring+1.256 × hospitalization in summer+1.508 × hospitalization in autumn+1.340 × hypoproteinemia)]. ANC,systemic glucocorticoid treatment and season of hospitalization were included in decision tree model,among which ANC was the most important predictor. There was no significant difference in area under the curve(AUC)(0.820 vs. 0.795) between the two models(P=0.438). The results of accuracy showed that the sensitivity and the specificity of Logistic regression model was 0.609 and 0.871,and that of decision tree model was 0.422 and 0.970 respectively. The model validation results showed that AUC of Logistic regression model was 0.784,with a sensitivity of 0.625,and a specificity of 0.769 and that AUC of decision tree model was 0.804,with a sensitivity of 0.500,and a specificity of 0.885. Conclusion Both Logistic regression model and decision tree model have good predictive value and there is no significant difference between them. The two models should be combined in nursing application to improve the accuracy of prediction.

Key words: Pemphigus, Multidrug-Resistant Organism, Logistic Regression, Decision Tree, Root Cause Analysis, Nursing Care