口腔癌游離皮瓣重建患者氣管切開術后肺炎風險 預測模型的構建

[中圖分類號]R473.78[文獻標志碼]A[doi]10.7518/gjkq.2025095

Development of a risk prediction model for pneumonia after tracheotomy in patients with oral cancer undergoing free flap reconstruction

Chen Yu^I，2 ，LeiYinfu，WuHemei，DuFulan4，DangJie4，ChenYongmei，TaoMing，ShiXinglian

1.Dept.ofNusing，ZuedicalUvesityAfilatedHospital，ui6oina;eptoOraldill cialSurgerydialialsita;ti HospitalofouthwestdicalUivesityuzhou646ohina;4DeptofOradaxilofacialurgerid toNorthSichuanMedicalColege，Nanchong67oo，Chna;5.DeptofNursingAfliatedStomatologicalHospialof ZunyiMedicalUniversity，Zunyi563ooo，China Supportedby：Guizhou Provincial Scienceand Technology Plan Project （QiankeSyntheticFruitLC[2022]006）Correspondence： Tao Ming， Email： 2284142542@qq.com

[Abstract]ObjectiveThis studyaimsto explore the risk factors for postoperative pneumonia （POP） in patientswith oral cancerundergoing free flapreconstructionafter tracheostomyand developarisk prediction model forPOPin patients withoral cancer undergoing freeflap reconstructiononthebasisofalogisticregression modelandadecisiontreemodel.MethodsRetrospectivecollctionof clinicaldatawasperformedon299 patientswithoralcancerwhounderwenttracheostomywith freeflapreconstruction in threehospitals in southwestern China from January 2017 to September 2023.Logistic regression analysis was conducted toidentifytherisk factors forPOPinpatients withoralcancer freeflapreconstruction.SPSsModeler18.0software was usedto developadecision tree model forPOPin patients with oralcancerfreeflapreconstruction，andthe predictive performanceofthetwomodels was evaluated.ResultsAtotalof299 patients，including56POPpatients （POPincidence rateof 18.73% ），were included inthis study.Logisticregresson analysis revealed thatthe time of tracheal cannula retention（OR _：=1.097 95%CI [1.039，1.158]）， jawbone resection （OR=2.439， 95%CI [1.084，5.484]）， number of anastomotic veins （OR =3.345 ， 95%CI [1.449，7.719]），and postoperative antibiotic typesgt;3 （OR=14.732，95%CI [4.405，49.273]） were independent risk factors for POP after tracheostomy in patients with oral cancer free flap reconstruction （Plt;0.05） ： The decision tree model showed that postoperative hospitalization time （ gt;14 days）was the main risk factor for the occurrence ofPOP inoralcancer，followedbythenumberofanastomotic veinsand the typeof postoperative antibiotics.The receiver perating characteristiccurve revealed thatthedecision tree model hadslightly beter predictive abilitythan the logistic regresion model. ConclusionPatients with oralcancerundergoing free flapreconstruction face manyrisk factors forPOP.The decision tree model demonstrated a slightly beterpredictive performance for POPin this study.Itcanbe used to understand therelevant characteristicsof POPafter tracheostomy inpatients with freeflapreconstructionof oral cancer，and its prediction performance should be verified.

[Key words] oral cancer;flapreconstruction；tracheostomy；postoperative pneumonia；Logistic regression model; decision tree model

游離皮瓣修復口腔癌術后缺損已成為常見的治療方式，患者術后氣道塌陷，皮瓣、舌體、咽后壁腫脹，易導致氣道梗阻，預防性氣管切開常用于減少患者圍術期氣道梗阻的風險。近年來，國內外對于口腔癌大手術是否需要行預防性氣管切開存在爭議[14]，有學者[建議移植過厚過大的皮瓣時，如術中直視口咽腔狹窄或者預估術后皮瓣腫脹可能封閉口咽腔等特殊情況時，可能需要預防性氣管切開外，其他情況可不用。但臨床大部分醫生還是會選擇氣管切開。研究顯示，口腔癌皮瓣重建患者術中預防性氣管切開比例為 60.2% 。氣管切開增加了術后肺炎（post-operativepneumo-nia，POP）等并發癥的發生風險，使患者住院時間延長，費用增加，發病率和病死率增加[3-46。目前關于頭頸領域的研究多關注于術后肺部并發癥或術后感染的危險因素，鮮有研究僅關注于POP^[7-8] 。研究證實，游離皮瓣、手術時間、術后住院時間和氣管切開等因素可能與POP有關。但上述研究多采用Logistic回歸篩選危險因素，Logistic回歸在預測單個風險因素對POP影響方面有一定的預測效能，但在預測多個因素交互作用下患者的感染特征方面效果欠佳。決策樹對數據大小、類型和特征要求不高，能從海量數據中快速有效挖掘出潛在風險因素，并以樹模型清晰呈現多個自變量之間的交互作用，其學習速度快、分類耗時短、效率高等優點而廣泛用于危險因素的預測和篩選[10-1]。且有研究[1]發現，2種模型聯合使用能實現優勢互補，提高預測效能。……