The Prediction Model Of Community Acquired Pneumonia Complicated With Adult Acute Respiratory Distress Syndrome According To Artificial Neural Network

Objective:To investigate the independent risk factors of community-acquired pneumonia(CAP)complicated with acute respiratory distress syndrome(ARDS),and the accuracy and prevention value of ARDS prediction based on artificial neural network model in CAP patients.Methods:A case-control study was conducted.Clinical data of 989 patients with CAP who met the criteria from the comprehensive intensive care unit and respiratory department of Changzhou Second People’s Hospital Affiliated to Nanjing Medical University,Wuxi Fifth People’s Hospital,Jiangsu Provincial People’s Hospital,Nanjing Military Region General Hospital from February 2020 to February 2021 were analyzed.They were divided into two groups according to whether they had complicated with ARDS.The clinical data of the two groups were collected within 24 hours after admission,the influencing factors of ARDS were screened out by univariate analysis and multiple factors,the artificial neural network model and the logistic regression model was constructed.Through the artificial neural network model,the importance of input layer independent variables on the output layer dependent variables was drawn.The artificial neural network modeling data were divided into a training group(n=701 patients)and a testing group(n=288 patients)in a ratio of 7:3,and the sensitivity,specificity and overall prediction accuracy of the two groups of models were calculated respectively.At the same time,the receiver operator characteristic curve(ROC)was drawn,and the area under the ROC curve(AUC)was calculated.Compare this model with the validation queue logistic regression model.Results:A total of 989 patients were included,including 323 patients with ARDS and 666 patients without ARDS.Univariate analysis showed that gender,age,maximum heart rate(MHR),maximum systolic blood pressure(MSBP),maximum respiratory rate(MRR),source of admission(emergency,outpatient),hypertension,,procalcitonin(PCT),neutrophil count(NEUT),bilirubin(TBil),eosinophil count(EOS),fibrinogen equivalent unit(FEU),c-reactive protein(CRP),serum creatinine(SCr),activated partial thromboplastin time(APTT),lactate dehydrogenase(LDH),erythrocyte sedimentation rate(ESR),hemoglobin(HBG),albumin(ALB),blood glucose(GLU),total serum kalium(K~+)level and total serum natrium(Na~+)level,both ARDS and non-ARDS groups were significant(P<0.05).22 risk factors are analyzed by artificial neural network,and the topological hierarchical structure is drawn.and these risk factors are included in the multifactor analysis,and the logistic regression model is established.Among the independent variables,the top nine indicators with the largest influence weight on the neural network model were LDH(100%),APTT(84.6%),PCT(83.8%),age(77.9%),MRR(76.0%),NEUT(75.9%),source of admission(68.9%),K~+(61.3%),TBIL(50.4%),all of which are more than 50%important,indicating that that these nine indicators had a greater impact on the occurrence of ARDS in patients with CAP.The sensitivity of the artificial neural network model training group was 88.9%,the specificity was 90.1%,and the overall prediction accuracy was 89.7%,and that of the testing group,its sensitivity was 85.0%,specificity was 87.3%,and overall prediction accuracy was 86.5%,When using the artificial neural network model to predict ARDS,The AUC predicted by the aforementioned artificial neural network model for ARDS in CAP patients was 0.943(95%confidence interval was 0.918-0.968).In this study,there is no significant difference between artificial neural network model and logistic regression model in predicting and evaluating CAP concurrent ARDS.Conclusion:The prediction model of ARDS in CAP patients based on artificial neural network model has good prediction ability.There was no significant difference in predicting the incidence of ARDS patients compared with logistic regression model.