Epidemiological Study Of Antimicrobials Use In Critically Ill Patients And Early Predicting Multivariate Models For Sepsis(EPMMS)

Part 1 Epidemiological study of antimicrobials use in critically ill patientsObjective:Global public health is severely threatened by multidrug resistance,which is due to the abuse of broad-spectrum antimicrobials.Overuse of antimicrobials results in increasing resistance and its infection.The high intensity antimicrobial use and its consequent high prevalence of multidrug-resistance have become an important cause of increased mortality and burden for critically ill patients in intensive care unit(ICU).Although "the management of clinical use of antimicrobials" in China was formally implemented in 2012,high resistance was frequently reported recent years,it is urgent to understand the actual antimicrobial use in clinical practice.The objective of this part was to get the data of antimicrobials use and resistance pattern from critically ill patients in ICU,which would be the base of further study on optimization of antimicrobial therapy.Methods:According to the inclusion and exclusion criteria,we investigated patients admitted to 8 ICUs from March 1,2014 to July 30,2014.The main targets were antimicrobials used and antimicrobial resistance in critically ill patients.The difference between sepsis and non-sepsis cases was compared by Chi-square test or Mann-Whitney test.Then,using multivariate logistic regression,we analyzed the risks factors for in-hospital death of critically ill patients.Results:A total of 644 critically ill patients were included in this study,230(35.7%)patients with sepsis and 414(64.3%)without sepsis.The rate of antimicrobial use in critically ill patients was 89.6%(577),and about 60.1%(347)critically ill patients used antimicrobials were non-sepsis.The β-lactam/β-lactamases inhibitor compound preparations(BLICs)were the most frequently used(47.0%)and most consumption antimicrobials(2540.49 DDD,43.8%).The rates of BLICs(43.5%vs.52.2%,P=0.051),≥3rd generation cephalosporins(13.5%vs.8.3%,P=0.061)and fluoroquinolones(4.0%vs.8.3%,P=0.129)used as initial empirical antimicrobial therapy in non-sepsis patients had no significant difference from infected patients;the average consumptions of carbapenems {4.0[interquartile range(IQR),2.0-10.0]vs.4.5[IQR,2.0-10.0],P=0.936},≥3rd generation cephalosporins {3.8[IQR,2.4-7.0]vs.3.0[IQR,1.5-4.0],P=0.057},≤2nd generation cephalosporins {2.0[IQR,1.5-3.4]vs.1.5[IQR,1.0-9.5],P=0.839},anti-MRSA {4.0[IQR,2.5-9.5]vs.4.5[IQR,2.6-9.8],P=0.642} and antifungal agents {8.0[IQR,3.8-41.0]vs.12.0[IQR,4.0-18.0],P=0.904} in non-sepsis patients had no significant difference from infected patients.Three hundred and sixty-four pathogens were isolated from patients,of which 190(52.2%)were multi-drug resistant organisms(MDROs),and 103(28.3%)were carbapenems/vancomycin resistant MDROs.Total MDROs(70/85 vs.95/115,P=0.962)and carbapenems/vancomycin resistant MDROs(36/85 vs.57/115,P=0.833)infected/colonized in non-sepsis patients had no significant difference from sepsis patients.Multivariate analysis indicated sepsis at admission(OR=1.855,95%CI 1.161-2.963,P=3.010)was an independent risk of in-hospital death,and effective initial empirical antimicrobial therapy(OR=0.171,95%CI 0.107-0.272,P<0.001)was a protect factor.Conclusion:Broad-spectrum antimicrobials overused in critically ill patients without sepsis resulted in the high rate of antimicrobial use in critically ill patients,it might be associated with the high rates of multidrug resistance organism(s)infection/colonization in critically ill patients.Sepsis at admission was an independent risk of in-hospital death,while effective initial empirical antimicrobial therapy was a protect factor.Therefore,it is critical to use antimicrobials rationally to improve of the prognosis in critically ill patients with sepsis.Part 2 Construction of early predicting multivariate model for sepsis(EPMMS)incritically ill patientsObjective:Sepsis is common in critically ill patients,with high mortality and expediture.The appropriate use of antimicrobial is critial to improve the prognosis of critically ill patients with sepsis,however,overuse of antimicrobial induces resistance.Our previous study ’Epidemiological study of antimicrobials use in critically ill patients’ showed effective initiation of antimicrobial therapy was a protect factor for in-hospital death.So,it is critical to discriminate sepsis from non-sepsis for appropriate antimicrobial therapy.The day-by-day used culture was a time-consuming and low positive method to diagnosis,and the biomarkers like procalcitonin to diagnose sepsis was also in debate.Therefore,it is urgent to explore an early predicting model of early and accurate diagnosis for sepsis.The objective of this part was to construct an early predicting model for sepsis(EPMMS)that combined multiply routine clinical variates in critically ill patients.Methods:We conducted a case-control study at 9 university hospital intensive care units(ICUs).Patients’ baseline data were collected at admission.Group 1 included 573 patients from eight ICUs and group 2 included 200 patients from nineth ICU.Multifactor predicting models were developed using logistic regression analysis in group 1.Receiver operation characteristic(ROC)curves and Hosmer-Lemeshow test were used to assess the discrimination and calibration of the models,rspectively.Group 2 was used to validate the models.Results:In group 1,with good calibration(Hosmer-Lemeshow test:χ2=5.886,P=0.660)and discrimination[areas under the ROC curve(AUC)0.801,95%confidence interval(CI)0.764-0.837,P<0.005],EPMMS model adjusted for covariates showed that the neutrophil-to-lymphocyte ratio(NLR)[odds ratio(OR)=1.023,95%CI 1.009-1.039,P=0.002],C-reactive protein(CRP)level(OR ＝1.013,95%CI 1.009-1.017,P<0.001),and Acute Physiology and Chronic Health Evaluation(APACHE)II score(OR=1.339,95%CI 1.120-1.601,P=0.001)were the independent risk factors for sepsis-1.The validation of these variates in group2 showed excellent discrimination(Sepsis-1:AUC 0.801,95%CI 0.764-0.900,P<0.001;Sepsis-3:AUC 0.848,95%CI 0.794-0.901).The EPMMS score could be calculated by 0.292x(ln APACHE Ⅱ score)+0.023x(ln NLR)+0.013x(ln CRP,in milligrams per milliliter),and In denoted natural logarithm.Four risk levels were classified based on the EPMMS score.The occurrence rates of sepsis-1/-3(compared to the low-risk group,the medium-,high-and ultra-high-risk groups faced 2.468/3.369,3.544/5.514 and 6.991/11.515 folds the risk of getting sepsis)and mortality rose increasing risk level(compared to the low-risk group,the medium-,high-and ultra-high-risk groups faced 3.450,12.267 and 15.300 folds the risk of achieving death).Conclusion:Including NLR,CRP and APACHE Ⅱ score,EPMMS model could effectively discriminate sepsis from non-sepsis at admission in critically ill patients.The EPMMS score=0.054x(ln APACHE Ⅱ score)+0,023×(ln NLR)+0.013x(ln CRP,in milligrams per milliliter),and the stratification of EPMMS score was associated with the mortalility within 28 day.Part 3 Basic leucine zipper transcription factor ATF-like(BATF)combined multiply variates improvingEPMMS model and its application in critically ill patientsObjective:Our previous study ’Constuction of EPMMS model in critically ill patients’constructed EPMMS model could effectively discriminate sepsis from non-sepsis.The variates included in EPMMS model were based on the known mechanisms of sepsis,however,the pathophysiology of sepsis was not completely exlpored,and new biomarkers introduced in the model might improve the diagnosis for sepsis.A study using bioinformatics methods found that an 11 genes set expressed significantly between sepsis and non-sepsis.To elevate the efficiency of discrimination,we explored 6 genes whose expressions were significant higher in sepsis,and then combined with previous constructed EPMMS model to make improvement,and then,we applied the EPMMS model in critically ill patient to validate its discrimination.Methods:The study was conducted in the critical care unit(ICU)and emergency critical care unit(EICU)of the First Affiliated Hospital,School of Medicine,Zhejiang University.We used group 1 critically ill patients to test and validate the different expression level of biomarkers in white blood cells between sepsis and non-sepsis.And then,using logistic regression,we combined the biomarkers with previous constructed EPMMS model to make improvement in group 2 patients.Hosmer-Lemeshow test was used to assess the calibration which measured the ability of the model to generate predictions.AUC were used to assess the discriminatory ability of the model.In addition,improved model was applied in group 3 patient to validate its discrimination.Results:The level of Carcinoembryonic Antigen Related Cell Adhesion Molecule 1(CEACAM1)(P=0.010)and Basic leucine zipper transcription factor ATF-like(BATF)(P=0.024)mRNA were significantly different between sepsis and non-sepsis in white blood cells in group 1;however,only the level of BATF(P<0,001)mRNA was validated significantly different between sepsis and non-sepsis in group 2.In group 2,factors P<0.1 were enrolled in logistic regression model after univariate analysis,multivariate analysis identified acute physiology and chronic health evaluation(APACHE)Ⅱ score[(Odds rate,OR)=1.192,95%confidence interval(CI)0.997-1.425,P=0.054],Neutrophil-to-lymphocyte ratio(NLR)(OR=1.057,95%CI 1.009-1.106,P=0.019),level of C-reactive protein(CRP)(OR=1.011,95%CI 1.000-1.023,P=0.043)and the level of BATF(OR=5.344,95%CI 1.410-20.246,P=0.014)were the independent risks for sepsis in critically ill patients.The Hosmer-Lemeshow test showed that improved EPMMS model was well fitted(χ2= 1.077,P=0.197).After introducing BATF into EPMMS model,the ability of discriminationg had made some improvement(AUC:0.905 vs.0.856;sensitivity 84.6%vs.84.6%;specificity 82.4%vs.67.6%;Youden index 0.670 vs.0.523).The discrimination of the improved EPMMS model was validated to be good in group 3(AUC:0.813;sensitivity 70.8%;specificity 86.5%;Youden index 0.573),and the increasing risks grade were significantly and positively associated with the increasing trends rates of sepsis(compared with the low-risk group,the medium-,high-risk and ultra-high-risk groups faced 1.625,5.056 and 26.000 times the risk of sepsis diagnosis,respectively).Conclusion:Among 6 candidate genes,the level of BATF in white blood cells was significant difference between sepsis and non-sepsis.BATF was an independent risk factor for sepsis in critically ill patients.The improved EPMMS model including BATF,NLR,CRP and APACHE II score had a higher ability of diagnosis.The risk stratification of improved EPMMS model was an efficient tool to semi-quantitatively diagnose sepsis in critically ill patients.