Study On Risk Factors And Predictive Models For Prolonged ICU Stay In Obese Sepsis Patients

Object:Obese sepsis patients are a special group in the Intensive Care Unit(ICU).The aim of this study was to investigate the risk factors for prolonged stay in theICU for obese sepsis patients and to develop a simple and efficient risk prediction model and validate it.Method:Clinical data from 23,136 patients admitted with a diagnosis of sepsis between 2014 and2015 were retrospectively analysed in the large real-world database eICU Collaborative Research Database(EICU).Excluded cases: 75 cases of pregnancy sepsis,3699 cases with missing Acute Physiology and Chronic Health Evaluation IV(APACHE IV),561 cases with missing height or weight,3528 cases with ICU length of stay less than or equal to 24 hours3528 cases,789 cases with an admission SOFA score of less than 2,and 1 case with missing gender.Of the 14,483 sepsis cases included,5,206 were obese patients.Some of the variables had missing values,but all were less than 20%,so multiple interpolation was used to fill in the missing values.Sepsis patients in the obese group were divided into two groups in a 7:3 ratio using the random number method,with the data set containing 3606 cases as the model training set and the remaining data set containing 1600 cases as the internal model validation set.The threshold for prolonged ICU stay was defined based on the third quartile of ICU length of stay for all sepsis patients in the model training set,and the model training set was subsequently divided into a non-prolonged group and a prolonged group based on whether the patients’ ICU stay was prolonged or not.Comorbidities,physiological indicators and laboratory findings measured on the day before and after ICU admission,clinical interventions,length of ICU stay,length of hospital stay,ICU mortality,hospital mortality and other clinical variables were analysed to further screen for independent risk factors affecting prolonged ICU stay in obese sepsis patients using logistic regression models and to develop corresponding risk prediction models.Subsequently,cases meeting the corresponding criteria and variables in the risk prediction model were extracted from the Medical Information Mart for Intensive Care IV database(MIMIC IV),another large realworld database,as an external model validation set,using the subject working The prediction model was evaluated using the Receiver Operating Characteristic(ROC),Calibration correction curve,Decision Curve Analysis(DCA),and Clinical Impact Curve(CIC).model’s predictive performance and net clinical benefit for both the internal and external model validation sets.Result:A total of 14,483 sepsis ICU patients were included in this study and divided into four groups according to the World Health Organization’s classification criteria for BMI.Analysis of the general information of the four sepsis groups showed that the difference in ICU length of stay between the obese group and the remaining three non-obese groups was statistically significant and that patients in the obese group had longer ICU stays than those in the nonobese group.Subsequently,extended length of stay ICU stay was defined as ICU stay greater than 5 days or more based on the third quartile of overall ICU stay for 5206 sepsis patients in the obese group,and logistic regression results in the original cohort and the post Propensity Score Matching(PSM)cohort indicated that obese sepsis patients’ prolonged ICU length of stay was significantly associated with increased ICU mortality.In the model training set,a multifactorial logistic regression analysis was used to identify 13 independent factors associated with prolonged ICU stay in obese sepsis patients based on the clinical variables extracted previously,namely: age,chronic kidney disease,hypertension,chronic obstructive pulmonary disease,mean blood pressure max,white blood cell count max,white blood cell count min,use of ventilation,use of Glasgow Coma Scale(GCS),blood albumin minimum,serum creatinine minimum,respiratory rate maximum and red blood cell distribution width minimum.All were independent risk factors for prolonged ICU stay in obese sepsis patients,except age,GCS,white blood cell count minimum and blood albumin minimum,which were protective factors.Subsequently,based on this,a risk prediction model was developed using the remaining seven predictor variables according to Occam’s razor’s law,excluding chronic kidney disease,serum creatinine concentration minimum,mean blood pressure maximum,age,hypertension and chronic obstructive pulmonary disease,which had good discrimination in the model training set,internal and external model validation sets.The calibration curves showed that the prediction curves were close to the standard curves,and that the prediction models obtained from the surface were well calibrated.Furthermore,the DCA and ICI clinical impact curves visually demonstrate the high net clinical benefit of the prediction model.Subsequently,the prediction model was converted into a more intuitive nomogram and a web-based version of the dynamic nomogram to facilitate clinical utilisation.Conclusion:Obesity sepsis patients have longer ICU stays than non-obese patients and prolonged ICU stays were an independent risk factor for ICU mortality.Chronic kidney disease,hypertension,chronic obstructive pulmonary disease,mean blood pressure maximum,white blood cell count maximum,use of ventilation or not,blood creatinine minimum,respiratory rate maximum and red blood cell distribution width minimum were independent risk factors for prolonged ICU stay in obesity sepsis.The risk prediction model and corresponding nomogram constructed in this study have been internally and externally validated to have high predictive performance and stability.Based on predictive,preventive,personalised and participatory medicine in theICU,this modelling tool can predict the risk of prolonged ICU stay in obese sepsis patients,enabling clinicians to identify this group of patients early and achieve personalized medicine.