Risk Prediction Model Of Wound Complications Post Modified Radical Mastectomy For Breast Cancer

Objective:(1)To retrospectively investigate the common complications within90 days after modified radical mastectomy for breast cancer.(2)To identify the predictors of wound complications within 90 days after modified radical mastectomy for breast cancer among preoperative and intraoperative variables(3)To develop and internally validate a Logistic regression prediction model of wound complications within 90 days after modified radical mastectomy for breast cancer,and visualize the model with a nomogram to provide reference for risk assessment of wound complications and perioperative management of modified radical mastectomy for breast cancer.Method:(1)A retrospective study of the inpatient electronic medical record database was used to analyze the data of all patients who underwent modified radical mastectomy for breast cancer in our hospital from January2019 to October 2020.(2)Use the literature review and expert discussion method to determine the data to be collected.(3)Three methods of univariate-multivariate Logistic regression,LASSO-Logistic regression,and professional knowledge-based methods were used to screen variables.(4)Logistic regression was used to model different variable combinations.(5)The Bootstrap method is used for internal validation,and the discrimination and calibration of different models are compared and evaluated,and the optimal model is selected.(6)Evaluate the clinical use of the optimal model to draw the decision analysis curve.(7)A nomogram was drawn for model visualization.Results:(1)The incidence of wound complications within 90 days post modified radical mastectomy for 844 patients with breast cancer was44.9%.(2)Univariate-multivariate Logistic regression and LASSO-Logistic regression variable screening methods showed that six variables including age,BMI,liver disease,preoperative chemotherapy,number of lymph node resections,and stratified decompression suture were the independent influencing factors of wound complications post modified radical mastectomy.(3)The AUC of the 6-variable Logistic regression model constructed based on statistical methods was 0.664(95% CI: 0.628-0.700),and the Bootstrap optisim adjusted C-index was 0.672;the AUC of the 8-variable Logistic regression model based on professional knowledge was 0.665(95%CI: 0.629-0.701),and the optisim adjusted C-index was 0.675.The ROC curve of the two models by De Long test was not statistically significant(Z=-0.220,P=0.826).(4)Hosmer-Lemeshow goodness-of-fit test and calibration plot showed that the prediction consistency of the 8-variable model(P=0.92)was higher than that of the 6-variable model(P=0.528).The 8-variable model was chosen as the best model.(5)The decision curve analysis of the 8-variable model shows that the model has certain clinical utility within the threshold probability range of14%-76%.When the threshold probability is 0.40,the clinical net benefit obtained by the model is 15%.Conclusion:Risk prediction model of wound complications post modified radical mastectomy for breast cancer,with age,BMI,hypertension,diabetes,liver disease,preoperative chemotherapy,number of lymph nodes removed ≥14and layered relaxation suture as predictors,has a certain degree of discrimination,calibration and clinical utility and can provide a clinical reference for wound complication risk assessment and management in patients with modified radical mastectomy.