Establishment Of A Predictive Model Of Systemic Inflammatory Response Syndrome After Percutaneous Nephrolithotomy Based On Machine Learning

Objective Based on machine learning algorithm,the prediction model of systemic inflammatory response syndrome(SIRS)after percutaneous nephrolithotomy(PCNL)was constructed,and the diagnostic predictive ability and clinical value of the model for SIRS were analyzed,so as to provide a basis for clinicians to make disease diagnosis decisions.Methods A total of 444 patients with upper urinary tract calculi who underwent PCNL were enrolled in this study.According to the occurrence of postoperative SIRS,the patients were divided into SIRIS group and non-SIRS group.The clinical data of the two groups were compared to analyze the risk factors,and binary Logistic regression was used to establish a predictive model.Then the best machine learning algorithm is selected to establish the SIRS machine learning prediction model after PCNL.The performance of the prediction model was evaluated by drawing the subject working characteristic curve(ROC curve)and calculating the area under the curve(AUC),and the clinical efficiency of the final machine learning prediction model was evaluated.Results Among the 444 patients in the study group,131 patients developed SIRS after PCNL.A total of 68 clinical factors were included in this study,and the results of univariate analysis of 24 factors were statistically significant.Eight independent risk factors were identified by multivariate analysis:history of hypertension(OR=1.93,P=0.021),preoperative nitrite(OR=7.453,P < 0.001),preoperative bladder urine culture(OR=2.341,P=0.002),postoperative urinary leukocyte count(OR=1.001,0.034),postoperative nitrite(OR=6.775,P < 0.001)and postoperative interleukin-6(OR=1.001).P = 0.028),postoperative hypersensitive C-reactive protein(OR=1.014,P = 0.027),postoperative nephrostomy(OR=3.004,P < 0.001).Combined with 8 independent risk factors,a binary Logistic regression model with an AUC of 0.827 was established.Machine learning XGBoost is used to establish a prediction model with an AUC of0.941.The decision analysis curve verifies that the machine learning model has better clinical benefits.Conclusion The machine learning prediction model of SIRS after PCNL is successfully established,and the machine learning prediction model is higher than the traditional prediction model AUC.Machine learning has a strong ability to predict SIRS after PCNL,and the model has good credibility and good clinical benefits.To a certain extent,it can provide auxiliary decision-making basis for doctors to predict the occurrence of SIRS early and accurately.