Establishment And Validation Of A Clinical Prediction Model For Severe HFMD Based On Meta-analysis

ObjectiveTo establish a clinical prediction model on the basis of meta-analysis of the risk factors for severe hand,foot and mouth disease(HFMD)to provide a basis for early identification and clinical intervention in patients with severe HFMD.Methods1.Meta-analysis: According to search strategies and inclusion or exclusion criteria,we conducted a search for literature on risk factors for severe HFMD in Chinese and English databases including CBMdisc,CNKI,Weipu Chinese Science and Technology Journals Database,Wanfang Database,Pub Med,Embase,Web of Science,Cochrane Library and Scopus,and performed a meta-analysis to synthesize the identified risk factors for severe HFMD and then pooled the corresponding odds ratios(ORs).2.Establishment and validation of simple risk scoring table: Simple risk scoring table for severe HFMD was established and validated through a meta-analysis of relevant risk factors.Regression coefficients β was calculated based on the ORs of each risk factor,and scores were assigned accordingly.To validate the scoring table,clinical data from 563 patients in Guangzhou Women and Children’s Medical Center were retrospectively collected and screened,and the receiver operating characteristic curve(ROC)was used to evaluate the predictive ability of the model.The area under the curve(AUC)was calculated,and the optimal cut-off point was determined based on sensitivity and specificity.Stratification of the risk for severe HFMD in patients was performed based on the cut-point values.3.Establishment of the nomogram model: Based on meta-analysis,predictive factors for severe HFMD were selected.Univariate logistic regression and LASSO regression models were applied to the clinical data of the collected patients to screen for variables to be included in the establishment of a multivariate logistic regression model.Finally,a nomogram model of the risk of severe HFMD was constructed using multivariate logistic regression.4.Validation of nomogram model: ROC curve,C-index,Calibration curve and clinical decision curve were used to evaluate the discrimination degree,calibration degree and clinical application value of the model.Results1.The results of meta-analysis: There were 14 articles included in the meta-analysis and 14 risk factors with statistically significant differences in the pooled results of meta-analysis,namely rural living,low birth weight,body peak temperature ≥39 °C,startle,elevated white blood cell(WBC),fever duration ≥ 3 days,vomiting,enterovirus 71(EVA71)positive,elevated blood glucose,dyspnea,elevated neutrophil cell(NEUT),lethargy,convulsion and limb trembling.The pooled ORs corresponding to each risk factor were 1.55,2.05,3.36,3.43、4.55、4.92、7.55、8.03、8.46、11.74、12.11、13.86、26.38、32.84,respectively.2.Establishment and validation of the simple scoring table: According to the pooled ORs and their 95%CI of the risk factors,the corresponding regression coefficientβ is calculated and assigned a scoring system.The risk factors and their scoring criteria included in the scoring table are as follows: limb trembling,convulsion and lethargy are assigned 2 points,while elevated blood glucose,positive EVA71,vomiting,fever duration ≥ 3 days and elevated WBC are assigned 1 point.The highest possible total score is 11.The predictive ability of the scoring table was compared at different cut-off values,and a cut-off value of 3 was ultimately selected.At this point,the AUC of the scoring table and its 95%CI were 0.889(0.860,0.913),and the risk stratification of the validation population showed good discrimination.3.Establishment of nomogram model: In the basic of meta-analysis results,univariate logistic regression and LASSO regression were carried out to simplify and select variables,then the selected variables were used in multivariate logistic regression and establishment of nomogram model.Eventually,the variables included were fever duration ≥ 3 days,vomiting,dyspnea,lethargy,startles,limb trembling and EVA71 positive.4.Validation of nomogram model: The C-index of the nomogram model and its95%CI were 0.927(0.903-0.952),which displayed a good degree of discrimination.The calibration curve and the decision curve analysis showed that the model has satisfied calibration and high clinical practicability.ConclusionBased on meta-analysis,a simple risk scoring table model and a nomogram model of severe HFMD were established,which were applicable to children under 6 years old and validated to have good predictive efficacy and clinical utility.The combination of the scoring table model and a nomogram model was found to be practically feasible in clinical settings,facilitating the early identification of severe HFMD patients and providing evidence support for clinical intervention.