Analysis Of Risk Factors For The Sepsis-associated Encephalopathy And Its 30-day Mortality:A Single Center Retrospective Cohort Study

【Objectives】1.To evaluate the impact of sepsis-associated encephalopathy（SAE）on the short-term prognosis of patients with sepsis,and to determine whether SAE is an independent risk factor for the short-term mortality of sepsis;2.To investigate the potentially remediable risk factors for the development of SAE in patients with sepsis;3.To construct a prediction model for the 30-day mortality of patients with SAE;4.To investigate the potentially modifiable risk factors for the 30-day mortality of patients with SAE in patients with sepsis.【Methods】1.To evaluate the impact of sepsis-associated encephalopathy（SAE）on the short-term prognosis of patients with sepsis,and to determine whether SAE is an independent risk factor for the short-term mortality of sepsis:1)The data from Metavision electronic medical record（EMR）system in mimic III clinical database was used for data analysis and the data from Carevue EMR system was used for sensitivity analysis;2)Sepsis was diagnosed according to the Sepsis-3 criteria.After excluding patients with diseases that affect or damage brain function,the data in the final cohort were extracted,they are demographic data,comorbidity,mean value of vital signs on the first day of ICU admission,results of the first laboratory test in ICU,infection-related data,medical interventions within 24 hours before and after ICU admission（including mechanical ventilation,vasoactive drugs,antibiotics and analgesics）,length of ICU stay and hospital stay,and 30 and 90-day outcome;3)The patients with GCS score<15 or delirium in the final study cohort were diagnosed as SAE,and the other patients were defined as sepsis without SAE（Non-SAE）;4)Kaplan-Meier（K-M）curve and log rank test were used to analyze the 30-and90-day survival of SAE and Non-SAE patients.Cox proportional hazards regression was used to identify the independent risk factors related to the 30-and 90-day mortaility of patients with sepsis so as to determine whether SAE is one of the independent risk factors.2.To investigate the potentially remediable risk factors for the development of SAE in patients with sepsis:1)The data included were the same as"1-2)"except for the prognosis-related data;2)Univariate Logistic regression combined with multivariate stepwise regression was used to screen independent risk factors associated with SAE and then verified by a sensitivity analysis;3)Stacked bar graph exhibiting the correlation between the risk factors and incidence of SAE was used to recognize the potentially remediable risk factors.3.To construct a prediction model for the 30-day mortality of patients with SAE:1)After excluding patients with metastatic cancer,solid tumor or lymphoma,SAE patients in"1-3)"were included.The data included were the same as"1-2)"except for the data related to infection and clinical intervention;2)Patients in Metavision EMR were randomly divided into a training set and an validation set with a ratio of 7:3.The prediction model was constructed in the training set and then conducted internal and external validation in the validation set and Carevue EMR,respectively;3)Logistic regression,Gradient boosting（GBoost）and random forest were used to build three prediction models,respectively.Then,AUROC was used to evaluate their discrimination to select the best prediction model;4)The prediction performance of the best prediction model was evaluated in the training,validation and external validation sets,respectively from three aspects:discrimination（assessed by AUROC and IDI）,calibration（assessed by calibration curve and Brier score）and clinical usefulness（assessed by DCA curve）.4.To investigate the potentially modifiable risk factors for the 30-day mortality of patients with SAE in patients with sepsis:1)The data included were the same as"3-1)".Besides,the data related to infection and clinical intervention were also included;2)Univariate Logistic regression combined with multivariate stepwise regression was used to screen independent risk factors associated with 30-day mortality of SAE and then verified by a sensitivity analysis;3)Stacked bar graph exhibiting the correlation between the risk factors and the 30-day mortality of SAE was used to recognize the potentially remediable risk factors;4)Replacement test combined with boosting method was used to screen continuous variables related to the red blood cell distribution width（RDW）,one of the independent risk factors picked out above,and then verified by the sensitivity analysis.Relative weight was calculated and visualized to show the importance of continuous variables related to RDW;5)Mann Whitney U test was used to screen the category variables related to RDW,and then to recognize the potentially modifiable factors.【Results】1.To evaluate the impact of sepsis-associated encephalopathy（SAE）on the short-term prognosis of patients with sepsis,and to determine whether SAE is an independent risk factor for the short-term mortality of sepsis:1)The K-M survival curve showed that the 30-and 90-day survival rate of SAE patients was significantly lower than that of Non-SAE patients（log rank test:p<0.0001in both）;2)The results of multivariate Cox regression showed that the risk factors significantly increasing the 30-day mortality of sepsis（HR>1.5）were SAE（HR:2.10（95%CI:1.69-2.59））,metastatic cancer,mechanical ventilation and morphine.The risk factors significantly raising the 90-day mortality of sepsis（HR>1.5）were SAE（HR:1.76（95%CI:1.49-2.09））,lymphoma,metastatic cancer,solid tumor,mechanical ventilation and morphine.2.To investigate the potentially remediable risk factors for the development of SAE in patients with sepsis;1)The results of multivariate Logistic regression showed that age,systolic blood pressure,heart rate,blood oxygen saturation,blood sodium,hemoglobin and neutrophils were continuous variables positively correlated with SAE（OR>1）,while diastolic blood pressure,respiratory rate,creatinine and bilirubin were continuous ones negatively correlated with SAE（OR<1）.Among the categorical variables,female and comorbidity including cardiovascular disease,diabetes,chronic liver disease and hypertension were associated with reduced risk of SAE（OR<1）,while mechanical ventilation,cephalosporins,vancomycin,pulmonary and/or urinary tract infections were related to increased risk of SAE（OR>1）;2)Based on the current data,the correlation between the above risk factors and the incidence of SAE revealed that the following factors may lead to a significantly increased incidence of SAE,they are:systolic blood pressure>140mm Hg,serum sodium≥145mmol/L,increased neutrophils percentage,cephalosporin use and urinary tract infection.3.To construct a prediction model for the 30-day mortality of patients with SAE:1)No difference existed in the AUROC of the prediction model constructed by the Logistic regression,GBoost and random forest in predicting the 30-day mortality risk of SAE.As the Logistic regression nomogram is intelligible and availiable,it was the best prediction model in this study;2)Discrimination comparison:the AUROC of the prediction model was significantly larger than SOFA score in training set,internal validation set and narrow external validation set,and the IDI index of the prediction model compared with SOFA score in three validation sets was 0.202[0.161-0.244],0.117[0.058-0.175]and 0.125[0.101-0.148]（p<0.001 in all of them）;3)Calibration comparison:the calibration curve indicated that the prediction model was not over-fitted,and its prediction on the 30-day outcome was consistent with the actual 30-day prognosis.Besides,the Brier index of the prediction model was significantly lower than that of SOFA score;4)Clinical usefulness:according to the DCA curve,medical intervention guided by the prediction model can obtain more net benefits than SOFA score within a certain threshold probability range.4.To investigate the potentially modifiable risk factors for the 30-day mortality of patients with SAE in patients with sepsis:1)The results of multivariate Logistic regression showed that age,respiratory rate,heart rate,BUN,sodium,RDW,mechanical ventilation and morphine sulfate were positively correlated with the 30-day mortality of SAE,while GCS score,temperature,oxygen saturation,PH value and PCO₂were negatively correlated with it;2)Based on the current data,the correlation between the above risk factors and the30-day mortality of SAE revealed that the following factors may lead to a significantly increased 30-day mortality,they are:（a）decreased GCS score;b.tachypnea(>20 min^-1),tachycardia(>100 min^-1)or low oxygen saturation（Sp O₂<90%）;（b）decompensated acidosis（p H<7.35）or hypothermia（<36℃）;（c）hypernatremia（>145 mmol/L）;（d）RDW≥16%or BUN>27 mg/d L;（e）PCO2<35 mm Hg;（f）adoption of mechanical ventilation or morphine sulfate on the first day of ICU admission;3)RDW has the greatest impact on the 30-day prognosis of SAE in the continuous variables mentioned in the“a-f”above.Moreover,hemoglobin is the continuous variable that has the greatest impact on RDW in patients with SAE;4)RDW in SAE patients were significantly increased by the adoption of vasoactive drugs,carbapenems,quinolones and vancomycin,G+bacteria,G-bacteria and fungal infection,as well as bacteremia and urinary or pulmonary infection.【Conclusion】1.SAE is an independent risk factor for the increased short-term mortality of sepsis under the Sepsis-3 criteria;2.Hypertension,hypernatremia,high percentage of neutrophils,cephalosporins adoption and urinary tract infection are the potential remediable factors contributing to SAE in patients with sepsis;3.Logistic regression nomogram including age,GCS score,mean values of heart rate,respiratory rate and temperature on the first day of ICU admission,and the first value of PCO₂,PH,BUN,RDW and sodium since ICU amission is the best model for the prediction of 30-day outcomes in patients with SAE based on the current data;4.Hypoperfusion,especially cerebral hypoperfusion,which was reflected in this study by tachypnea,tachycardia,hypooxia and decompensated acidosis,was an important modifiable risk factor for the 30-day prognosis of SAE patients.Besides,hypernatremia was also an modifiable risk factor;5.Infection and the subsequent inflammation was a key factor for the increase of RDW value.Therefore,early control of the infection source,retention of microbial culture specimens and standardized antibiotic treatment are of great significance to improve the prognosis of SAE patients;6.As hemoglobin was negatively related to the RDW value,it is interesting to evaluate whether seting a certain RDW value as the target for blood transfusion in SAE patients whose hemoglobin is lower than the threshold can improve the short-term prognosis of SAE.