中国临床药学杂志

目的 建立基于正交偏最小二乘判别分析（OPLS-DA）算法的Ⅱ类切口围术期抗菌药物规范化评价的多变量分析模型，探讨医院围术期抗菌药物管理的新模式。方法 提取2019-2021年12个季度的全部Ⅱ类切口围术期抗菌药物病例，进行合理用药评价。以选药合理率、时机合理率和疗程合理率3个评价维度为示范，15个外科科室为代表，构建全院各外科科室用药合理率数据矩阵。建立3年的整体OPLS-DA模型，对3个年度的得分图及载荷图进行整体评估；继而分别建立2019-2020、2020-2021年的OPLSDA模型，通过构建S-plot图，对比研究影响逐年Ⅱ类切口围手术期抗菌药物使用合理率的显著变量；最后对模型进行评价。结果 在整体合理率方面，2019-2021年呈现逐年连续变化的趋势。进一步构建的影响2019-2020年、2020-2021年显著性因素的OPLSDA筛选模型结果表明，2019-2021年妇科和口腔科的抗菌药物时机合理率、骨伤一科的抗菌药物选择合理率及泌尿外科和口腔科的抗菌药物疗程合理率，可能是降低2020年整体合理率的关键因素；2020-2021年泌尿外科的抗菌药物时机合理率、抗菌药物疗程合理率，心胸外科的抗菌药物疗程合理率，骨伤二科的抗菌药物选择合理率可能是降低2021年度整体合理率的关键，需重点进行干预以提高年度总体合理率。在模型的验证的置换检验结果中，Q2在Y轴上的截距均<0，结果提示该方法具有较好的区分度及预测能力。结论 基于OPLS-DA算法的多变量监测方法在围术期抗菌药物管理中具有有效性及科学性，可为医院围术期抗菌药物管理提供新模式及新方法。

AIM To establish a multivariable analysis model for the standardized evaluation of perioperative antibacterial drugs for class Ⅱ incision based on orthogonal partial least squares discrimination analysis(OPLS-DA)algorithm, and to explore a new mode of perioperative antibacterial drugs management in hospitals. METHODS All perioperative cases of class Ⅱ incision antibacterial drugs application were extracted from 2019 to 2021, and rational drug use was evaluated, including rational rate of drug selection, timing and course as an example. Taking 15 surgical departments as representatives, the data matrix of rational drug use rate in all surgical departments of our hospital was constructed. The OPLS-DA model was established ranging 3 years, and the scores charts and loading charts of 3 years were evaluated. Then, OPLS-DA models for 2019 and 2020, 2020 and 2021 were established respectively, and S-plot plots were constructed to study the significant variables influencing the rational use rate of perioperative antibacterial drugs in class Ⅱ incisions year by year. Finally, the model was evaluated. RESULTS In terms of overall rational rate,there was a trend of continuous change year by year from 2019 to 2021. The results of the further constructed OPLSDA screening mode that affected the significance factors in 2019-2020 and 2020-2021 showed that, from 2019 to 2021, timing rational rate of antibacterial drugs from departments of gynaecology and dental, rational selection rate of antimicrobial drugs from department of 1st bone traumatology, treatment course rational rate of antimicrobial drugs from departments of urology and dental, might be the key factors to reduce overall rational rate in 2020. From 2020 to 2021, the rational rate of timing and course of antibacterial drugs in urology department, the rational rate of course of antibacterial drugs in cardiothoracic surgery department, and the rational rate of antibacterial drugs selection in the 2nd bone traumatology department might be the key factors to reduce the overall rational rate in 2021. Interventions were needed to improve the overall annual rational rate. Finally, the model was verified. The permutations analysis results showed that the intercept of Q2 on the Y-axis was less than 0 in the two models, indicating that the method had good degree of discrimination and ability of prediction. CONCLUSION This study proves the effectiveness and scientific nature of multivariable monitoring method based on OPLS-DA algorithm in perioperative antibacterial drugs management, which can provide a new paradigm and method for perioperative antibacterial drugs management in hospitals.