Construction And Verification A Multi-factor Predicting Score System For Moderate Severe Acute Pancreatitis Based On Decision Tree Model

Background Acute pancreatitis(AP)is a common acute digestive tract disease in clinical.It changes complex and the prognosis is quite different.In 1992,the old Atlanta AP classification criteria classified patients into mild acute pancreatitis(MAP)and severe acute pancreatitis(SAP)according to whether they have local or systemic complications or organ failure.A number of studies have shown that the old Atlanta AP classification criteria is not clear,and the incidence of SAP is overestimated,which brings certain difficulties to clinical standardized treatment and care.Therefore,in 2012,the new Atlanta AP classification standard further standardized the AP-related definition and reclassified the disease severity.This standard defines patients with no local or systemic complications or organ failure as MAP,transient organ failure(organ failure time<48h)or local or systemic complications without persistent organ failure were defined as moderately severe acute pancreatitis(MSAP),patients with persistent organ failure(organ failure time > 48 h)were defined as SAP.The condition of MSAP patients is variability in clinical.If these patients are monitored closely and intervened actively,the prognosis is better.If the treatment is not timely or inappropriate,the patient can develop into SAP,which will increase the consumption of medical resources and patients' mortality.Therefore,early detection of MSAP patients and give appropriate treatment can improve patients' prognosis and reduce the incidence of SAP.However,the current common scoring system or indicators for predicting the severity of AP patients,such as Acute Physiology and Chronic Health Evaluation(APACHE II),Ranson and C-reactive protein(CRP)can not achieve the purpose of predict MSAP.Building a new scoring system for predicting MSAP is the direction of further research.Therefore,the aim of this study was to construct a scoring system for predict MSAP based on simple and readily available clinical data within 24 hours of patient admission and to perform prospective validation.In order to provide a reference for early identify MSAP patients in clinical practice.Objectives1.According to the 2012 new atlanta AP grading standard,construct a decision tree model for early predict MSAP and form a corresponding decision tree scoring system.2.Verify the predictive value of the MSAP decision tree scoring system prospectively.Methods1.Decision tree scoring system construction phase: collect AP patients who were admitted to a tertiary hospital in Nanjing from October 2017 to November 2018,prospectively.The included subjects were randomly divided into a training set and a test set by 2:1 ratio using computer.At the same time,the general clinical data of these patients,the results of biochemical indicators in 24 hours on admission,the results of routine blood tests within 24 hours,the results of blood coagulation tests within 24 hours of admission,the blood amylase within 24 hours of admission,and blood lipase were collected.The comparability of the training sample set and the validated sample set data is compared by t-test or non-parametric test(Mann-Whitney U).Then,using single factor analysis to compare the clinical data of the MAP and MSAP patients in the training sample set,the single factor statistically significant variables were imported into SPSS Modeler18.1 software,the decision tree model for predicting MSAP was generated,and the test data set was used to verify the predict model.Finally,form a scoring system for predicting MSAP.2.Decision tree scoring system verification phase: collect general clinical data of AP patients who were admitted to a tertiary hospital in Nanjing from December 2018 to February 2019,decision tree scores within 24 h,Chinese version of GCSI-R scale score within 24 h,PASS score within 24 h,prospectively.After univariate analysis of the scores of the two groups(MAP and MSAP),the data were imported into Medcalc software,and the decision tree score,the Chinese version of the GCSI-R score and PASS score were used to draw the receiver operating characteristic curve(ROC)of three scoring systems,and comparing the predict value of these score systems.Results1.Decision tree scoring system construction results: a total of 315 patients with AP were included.4 cases of pregnancy-associated AP,1 case of pancreatic cancer and 1traumatic AP were excluded,8 cases of incomplete data and 4 cases of SAP were excluded.Finally,297 subjects were included.According to the random assignment of computer(ratio 2:1),there have 198 subjects in the training set,99 subjects in the test set.Among training set,there have 168 patients with MAP and 31 patients with MSAP.And in test set,there have 81 patients with MAP,18 patients with MSAP.The constructed decision tree scoring system includes four variables: albumin(ALB),lactate dehydrogenase(LDH),urea and calcium.The total score is 0-8,and along with the score rising,the more serious the condition.2.Decision tree scoring system verification results: The clinical data of 126 AP patients were collected,and 2 cases of pregnancy-related AP,1 case of traumatic AP,1 case of pancreatic cancer,5 cases of incomplete data and 5 cases of SAP patients were excluded.Finally,included 112 AP subjects.Among them,97 patients with MAP and 15 patients with MSAP.The results of ROC curve analysis show that the area under the ROC curve(AUC),sensitivity,specificity,yoden index,and cutoff value of the decision tree scoring system for predict MSAP are 0.868,80.00%,86.60%,0.666 and 3,respectively;the AUC,sensitivity,specificity,yoden index,and cutoff value of the Chinese version of the GCSI-R score for predict MSAP are 0.709,66.70%,76.30%,0.430 and 7,respectively;the AUC,sensitivity,specificity,yoden index,and cutoff value of the PASS for predict MSAP are 0.717,40.00%,92.80%,0.328 and 160.Conclusions1.The decision tree scoring system based on the indicators of routine test of AP patients within 24 hours of admission can predict MSAP.2.Compared with the 24 h PASS score and the 24 h Chinese GCSI-R score,the 24 h decision tree score give the greatest value for predict MSAP.