| 摘要: | 前言：聚焦於麻醉品質之管理議題，運用麻醉手術後症狀檢視麻醉品質，其中手術後噁心嘔吐（Postoperative nausea and vomiting，簡稱PONV）是接受麻醉手術後常見的症狀，其發生率約為20-30%，高危險族群發生率高達70-80%。因此，本研究目的是找出術後噁心嘔吐的相關性危險因子，並建立高危險族群的預測模型，找出較適合的預測模型輔助臨床人員鑑別PONV高危險族群。
方法：針對新北市某準醫學中心醫院之574筆個案資料進行分析與預測，先利用SPSS 18.0之卡方檢定與羅吉斯迴歸找出術後噁心嘔吐的相關因子與影響因子，再採用R Studio1.1.456版系統介面執行R語言3.5.3版建立預測模型，以五折疊交叉驗證進行平均，計算出準確率、特異度、敏感度、陽性預測值、陰性預測值、F1值，作為演算法預測能力評估。
結果：經本研究結果發現，性別、吸菸狀況、飲酒狀況、術後暈眩或喉嚨痛、手術類型、術中麻醉藥物均與術後噁心嘔吐有顯著相關。在影響程度方面，女性有PONV的風險是男性的2.978倍。術後有喉嚨痛者有PONV的風險是無喉嚨痛者的2.305倍。術後有暈眩者有PONV的風險是術後無暈眩者的2.943倍。手術類型之減肥手術有PONV的風險是骨折復位手術的4.528倍。術中有使用類鴉片藥物有PONV的風險是無使用的1.518倍。在預測模型的準確率上，羅吉斯迴歸為64.52%、決策樹為71.83%、支持向量器為75.65%、類神經網路為75.30%。支持向量器與類神經網路較決策樹與羅吉斯迴歸有較高的預測準確率。
結論：依本研究結果顯示，性別、術後症狀、手術類型與術中麻醉藥物對於術後噁心嘔吐的影響程度也是不可輕忽的，這些都是未來我們需要注意有術後噁心嘔吐的高危險族群。資料探勘模型之決策樹、支持向量器與類神經網路的預測效果均達70%以上，具有可接受的預測力效果，而四種模型比較起來，支持向量器與類神經網路具有較高的預測能力。
Introduction: Focus on the management of anesthesia quality, use the symptoms of anesthesia to check the quality of anesthesia. Postoperative nausea and vomiting (PONV) is a common symptom after anesthesia. The incidence rate is about 20-30. %, the incidence of high-risk groups is as high as 70-80%. Therefore, the purpose of this study was to identify the relevant risk factors for postoperative nausea and vomiting, and to establish a predictive model of high-risk groups to find a more suitable predictive model to assist the trampoline personnel in identifying PONV high-risk groups.
Method: According to the analysis and prediction of 574 cases of a quasi-medical center hospital in New Taipei City, the correlation factors and influencing factors of postoperative nausea and vomiting were found by using the SPSS 18.0 chi-square test and Logistic regression, and then using R Studio1.1.456 and R language version 3.5.3 to establish a predictive model, and averages with five-fold cross-validation to calculate accuracy, specificity, sensitivity, Positive Predictive Value, Negative Predictive Value, and F1 value. .
Result: According to the results of this study, gender, smoking, drinking, postoperative dizziness , postoperative sore throat, type of surgery, and intraoperative anesthesia were significantly associated with postoperative nausea and vomiting. In terms of the degree of influence, the risk of women having PONV is 2.978 times that of men. The risk of PONV in patients with sore throat after surgery was 2.305 times that of those without sore throat. The risk of PONV in patients with dizziness after surgery was 2.943 times that of postoperative dizziness. Bariatric surgery has a risk of PONV of 4.528 times that of fracture reduction surgery. Intraoperative use of opioids has a risk of PONV of 1.518 times that of no use. In the accuracy of the prediction model, the logistic regression is 64.52%, the decision tree is 71.83%, the support vector is 75.65%, and the neural network is 75.30%. Support vector machines and neural networks have higher prediction accuracy than decision tree and Logistic regression.
Conclusions: According to the results of this study, gender, postoperative symptoms, type of surgery, and the extent of intraoperative anesthetics for postoperative nausea and vomiting are not negligible. These are high-risk groups in the future that we need to pay attention to PONV. The prediction model of the data mining model, the support vector machine and the neural network prediction effect are all above 70%, with acceptable predictive power effects. Compared with the four models, the support vector and the neural network are higher predictive ability. |
