Establishment And Application Of Predictive Model For Purulent Meningitis In Infants

PART ONE CLINICAL ANALYSIS OF INFANTILE PURULENT MENINGITISBackground: Purulent meningitis(PM)usually caused by a variety of pyogenic infection,is a kind of central nervous system(CNS)infectious disease most common in children.It is easily misdiagnosed and its symptoms are varied.Excessive application of broad-spectrum antibiotics results in increased sickness and death of infants and young children.In this study,clinical data of 430 cases of PM in infants were analyzed to summarize the clinical experiences so as to achieve early diagnosis and early treatment of PM.Objective: To investigate the clinical characteristics of purulent meningitis in infants.Methods: The clinical data of 430 infants with purulent meningitis admitted to the Children's Hospital of Medical University of Chongqing from January 2004 to 2013 were collected and analyzed retrospectively.Results:(1)Male-to-female ratio was 1.61:1.The May was the modal month of onset(11.7%).(2)Main clinical manifestations were fever(89.3%),vomiting(67.2%),mental fatigue(62.1%),anterior fontanelle full/ bulging/ high tension(54.2%),seizure(52.6%),and meningeal irritation sign(24.7%).(3)Cerebrospinal fluid(CSF)bacterial culture was done for 420 cases,of which 1.2% cases were positive.Blood culture was done for 146 cases of which 15.1% were positive.(4)175(40.7%)cases had complications,among which 133 cases(76.0%)were subdural effusion,21 cases(12.0%)were epilepsy.Conclusion:(1)Infantile PM is common in Spring and Winter.(2)The CSF/blood pathogen detection rate is very low and it is difficult to find evidence of cause.(3)Fever,vomiting,mental fatigue and anterior fontanelle full/ bulging/high tension,seizure are the main clinical manifestations on which diagnosis depends.(4)For those children diagnosed as PM and still having recurrent fever and prominent anterior fontanelle after treatment,clinicians should consider the probability of subdural effusion and treat them with brain imaging test on time.PART TWO ESTABLISHMENT OF PREDICTIVE MODEL FOR PURULENT MENINGITIS IN INFANTSBackground: PM in infants needs emergency diagnosis for early beginning of treatment.The fact that it should be distinguished from viral meningitis(VM)make the diagnostic challenge more complicated since these two kinds of diseases can have similar characteristics.It is usually difficult to diagnose whether the patients suffer from PM if their cerebrospinal fluid of Gram staining and culture do not show positive.Modeling with the results of positive or negative results and clinical manifestations,and using the model to accurately predict other possible causes of PM(especially VM)have become a research hotspot.There are no reports on the prediction models for 29 days ~3 age group.And markers related to inflammatory,such as serum C-reactive protein(CRP)and procalcitonin(PCT)are not included at the same time in the calculations of existing scoring models.Objective: To establish a predictive model for PM in infants.Methods:The clinical data(including whether having fever,vomiting,seizure,mental fatigue,disturbance of consciousness,meningeal irritation,and initial serum CRP,serum PCT,protein in CSF,glucose in CSF,CSF WBC counts,and blood WBC counts are taking as variables in the prediction model of PM in infants)of 403 infantile PM and VM cases admitted to the Children's Hospital of Medical University of Chongqing from January 2004 to December 2013 were collected and analyzed statistically.Results:(1)The transformed regression equation is: y=-2×Days of fever+23×Mental fatigue +19.5×Meningeal irritation sign+20×Seizure+17.5×Disorder of consciousness +0.03×WBC counts in CSF-13.5×Glucose in CSF +41.5×Protein in CSF +53×Serum CRP score +58×Serum PCT score+3.5×Blood WBC counts-100.(2)When the score is greater than 100,the patient is identified as suffering from PM with the higher the score,the greater the probability of PM diagnosis.When the score is less than 100,it is identified as VM.(3)The regression equation was used to test the model data.The prediction accuracy rate of VM was 100%,the accuracy rate of PM was 94.6%,and the overall accuracy rate was 97.3%.Conclusion:(1)In the present study,using the clinical features,including fever,vomiting,seizure,mental fatigue,disturbance of consciousness,meningeal irritation,and initial serum CRP,serum PCT,protein in CSF,glucose in CSF,CSF WBC counts,and blood WBC count as statistic variables in the regression equation of prediction model is feasible and convenient for clinical operation.(2)VM and PM can be distinguished effectively by a retrospective analysis using the regression equation with the data of cases.PART THREE APPLICATIONS OF PREDICTIVE MODEL FOR PURULENT MENINGITIS IN INFANTSBackground: Some scoring and predictive models containing CSF or serum and /or cell chemical parameters have been reported for the PM diagnosis or distinguishing the VM and PM.The ability to distinguishing them correctly of these models has been verified by prospective or retrospective studies.However,the widespread use of predictive models is not always appropriate,and these models are also limited by the age of patients.This dilemma is a greater challenge for clinicians.Based on the regression equation of the prediction model for PM in infants established in the second part,in this part,prospective applications in clinical cases for predictive diagnosis are done for the purpose of helping early clinical diagnosis and standard treatment.Objective: To further study and correct the prediction model established in this research,and to point out the direction of further clinical research.Methods: Taking the 32 suspected infantile PM cases admitted to the Children's Hospital of Medical University of Chongqing from January 2016 to December 2016 as research object,their prospective scores according to the prediction model established in the second part were given.And the final diagnosis was followed up and analyzed.Results:(1)The results on initial CSF and PCR of serum samples for 3 cases in hospital: the serum PCR are all negative;cerebrospinal fluid PCR of 1 case are meningitis diplococcus and Klebsiella pneumoniae;cerebrospinal fluid PCR of 1 case is meningitis diplococcus.(2)Predictive scores and final diagnosis results for 32 suspected 32 PM cases are: 27 cases with scores more than 100 have been predictively diagnosed as PM,24 cases are finally diagnosed as PM,2 cases finally diagnosed as tuberculous meningitis(TM),1 case finally diagnosed as VM,the prediction accuracy rate is 88.89%;5 cases with scores less than 100 have not been predictively diagnosed as PM,all cases are finally diagnosed as VM,the prediction accuracy rate is 100% and the overall accuracy rate is 94.45%.Conclusion:(1)This prediction model is mainly suitable for the identification of PM and VM,but the applicability to TM is poor.(2)Further clinical studies are needed to further expand the sample size and establish a reliable model to identify PM and TM in infants,and other causes of infection.