Construction Of A Predictive Model For Severe Acute Pancreatitis Based On Bedside Nursing Evaluation

OBJECTIVESThe prevalence of severe acute pancreatitis has increased in recent years,with rapid progress,multiple complications,and high mortality.The overall mortality rate is 5-10%.Early and accurate assessment of the severity of acute pancreatitis is of great significance for the intervention and treatment of acute pancreatitis and the reduction of mortality.At present,the scoring of the severity of acute pancreatitis needs to wait for the interpretation of the test or image results,and the evaluation cycle is long and cannot be completed immediately.Based on the characteristics of nurses’ bedside work,this study searched for quick,accurate and reliable simple indicators,constructed a nursing prediction model for severe acute pancreatitis,verified it,and developed a web calculator to explore the predictive value of the model in the tendency of severe acute pancreatitis.METHODSA total of 496 in patients diagnosed with acute pancreatitis in a third Grade hospital in Guangdong Province in recent four years were analyzed.According to the requirements of Logistic regression,the sample size was divided into internal validation and external validation,internal validation included modeling set and validation set,which are retrospective analysis;and external validation included test set,which is a prospective study.The sample size ratio of the three sets was about 2:1:1,which was 244 cases,134 cases and 118 cases,respectively.Each data set was further divided into severe acute pancreatitis group and non-severe acute pancreatitis group.Firstly,the data of the severe acute pancreatitis group and the non-severe acute pancreatitis group in the modeling set were analyzed by univariate analysis,and the indicators that could be quickly obtained at the bedside were selected to be included in the multivariate analysis,according to β coefficient value to build prediction model.The area under the ROC curve(AUC)can be used to evaluate the discrimination of early warning models,and the Hosmer-Lemeshow goodness-of-fit test was used to evaluate the model calibration.The application efficiency of the model was evaluated by sensitivity,specificity and accuracy.The model information was imported into the internal validation set to generate the AUC.The MedCalc software was used to compare the AUC with the recognized BISAP score and APACHE II score for predicting severe acute pancreatitis,and the predictive efficiency was compared in terms of sensitivity,specificity and accuracy.The AUC was generated by substituting the model information in the externally verified test set again,and the Z-test of AUC was compared with BISAP score and APACHE Ⅱ score again,and the AUC generated by internal verification and external verification was compared.Finally,we use R language to develop a web calculator of nursing prediction model for severe acute pancreatitis.RESULTSA total of 6 indicators were identified and included in the forecast model,including heart rate(OR=1.040,95%CI:1.010～1.072,P=0.010),respiratory(OR=1.269,95%CI:1.022～1.575,P=0.031),pulse oxygen saturation(OR=0.824,95%CI:0.691～0.981,P--0.030),weak bowel sounds(OR=2.383,95%CI:1.080～5.262,P=0.032)or disappearance of bowel sounds(OR=3.227,95%CI:1.169～8.907,P=0.024),pulmonary auscultation(OR=2.185,95%CI:1.072-4.452,P=0.031)and peritoneal irritation(OR=2.097,95%CI:1.048～4.195 P=0.036).Hosmer-lemeshow goodness-of-fit test of modeling set P=0.278,AUC was 0.840(95%CI:0.788～0.892,P<0.001),sensitivity was 68.30%,specificity was 85.80%,the AUC of internal validation set was 0.802(P5%CI:0.714-0.891,P<0.001).The sensitivity(70.50%vs 84.10%vs 79.50%)and accuracy(79.80%vs 85.10%vs 82.10%)were lower relative to those by BISAP and APACHE Ⅱ scoring methods,while the specificity was higher than the other two methods(90.00%vs 84.40%vs 83.30%).The external test set AUC is 0.787(95%CI:0.708～0.859,P<0.001),the sensitivity(71.00%vs 83.70%vs 80.40%)and accuracy(76.30%vs 84.70%vs 81.20%)were lower relative to those by BISAP and APACHE II scoring methods,while the specificity was higher than the other two methods(88.60%vs 84.10%vs 83.60%).The AUC of the three models was tested by using MedCalc software.The pairwise comparison of the three models in the internal validation set showed no statistical significance in the pairwise comparison between the forecast model and BISAP score,the forecast model and APACHE II score,the BISAP score and APACHE II score(P>0.05).There was no statistical significance in the pairwise Z-test comparison of the three models,indicating that the forecast model had a certain predictive efficiency and was similar to the two scoring models.CONCLUSIONSThe nursing prediction model for severe acute pancreatitis constructed in this study is presented in the form of a web page,which is easy to obtain and measure,simple and fast,and convenient for nurses to apply.It is similar to BISAP score and APACHE Ⅱ score in predicting the severity of acute pancreatitis,which highlights the advantages of rapid bedside assessment for nurses,and can be used as a supplement to the assessment tool for acute pancreatitis.