Construction And Verification Of A Prediction Model For Chronic Critical Illness After Elderly Sepsis Patients Based On MIMIC-Ⅳ Database

ObjectivesBased on MIMIC-Ⅳdatabase,explore the relevant risk factors of chronic critical illness（CCI）after elderly sepsis patients,and screen the risk factor indicators of CCI in elderly patients with sepsis.On this basis,the prediction model of CCI was established to provide basis for medical staff to evaluate high-risk patients with CCI.MethodsIn this retrospective study,the elderly patients with sepsis in MIMIC-Ⅳdatabase were selected as the screening population.Structured query language（SQL）was used to select the subjects according to the inclusion and exclusion criteria.Finally,5908patients were enrolled.The general information,vital signs,complications and laboratory indicators of patients admitted to ICU within 24 hours were combined,sorted out and analyzed by Stata.1.R software was used to randomly divide all the included patients into the training group and the validation group in a ratio of 3:1.The feature importance of all variables in the training group data was sorted based on the Random forest（RF）algorithm,and the best variables were selected.2.The independent variables obtained from the above screening were analyzed by univariate and multivariate logistic regression to explore the independent risk factors of CCI after elderly patients with sepsis.3.The risk prediction model of CCI after elderly patients with sepsis was constructed using independent risk factors,and the results were presented in the form of a nomogram.At the same time,internal validation was conducted in the validation group,and the ROC,DCA and Calibration Curve were used to evaluate the discrimination,clinical usefulness and calibration of the models,which were presented as nomograms.Results1.A total of 5908 elderly patients with sepsis were included in this study,including1496 patients with secondary CCI,accounting for 25.32%.The average age of patients was 74.93±9.22 years old,and the male proportion was 55.75%.In the validation group（n=4431）,the average age was 74.95±9.21 years old,and male accounted for 55.22%.In the validation group（n=1477）,the average age was 74.84±9.26 years old,and the proportion of males was 57.35%.All variables between the two groups have no statistical difference（P>0.05）,which conforms to the complete random distribution,indicating that the training group can represent the overall situation.2.The RF algorithm was used to rank the importance of all the included variables in the training group.There are 22 variables including age,HR,RR,MAP,TEMP,SPO₂,WBC,PLT,HB,Scr,BUN,Lac,GLU,AG,HCO₃^-,Cl^-,Na⁺,Ca²⁺,PT,RDW,HCT,INR.The above variables were included in the logistic analysis.The results showed that age,HR,RR,MAP,TEMP,SPO₂,WBC,PLT,HB,Scr,BUN,Lac,GLU,AG,HCO₃^-,Cl^-,Ca²⁺,PT,RDW,HCT and INR were statistically significant（P<0.05）.3.Multivariate logistic regression showed that nine variables including HR（OR=1.215,P=0.001）,TEMP（OR=2.595,P<0.001）,SPO₂（OR=0.582,P<0.001）,BUN（OR=1.407,P<0.001）,Lac（OR=1.330,P=0.002）,AG（OR=1.270,P<0.001）,Ca²⁺（OR=0.893,P=0.034）,RDW（OR=1.240,P<0.001）and HCT（OR=1.830,P<0.001）were independent risk factors associated with CCI after sepsis.4.The above independent risk factors are used to build a prediction model.The AUC of the model in the training group is 0.792,the AUC of the model in the validation group is 0.746,which were higher than the SOFA score,indicates that the model has a high discrimination.The DCA curve indicates that the prediction model has good clinical usefulness,and the calibration curve also proves that the model is well calibrated.Conclusion1.High HR,high TEMP,low SPO₂,high BUN,high Lac,high AG,low Ca²⁺,high RDW,and high HCT indicate that elderly patients with sepsis have a higher risk of CCI.2.The prediction model was constructed with 9 independent risk factors,including HR,TEMP,SPO₂,BUN,Lac,AG,Ca²⁺,RDW and HCT,with good differentiation and clinical practicability.3.Internal validation shows that the model has good stability.It can be used as a tool to assess the risk of CCI after elderly sepsis patients and provide reference for clinical medical staff to identify high-risk groups.