Establishment 72 Hours Mortality Risk Model For Critical Pediatric Of Inter-hospital Transport

Objective The critical ill children scoring systems are widly used to assess the severity and mortality risk of critical children.However,there is no scoring system used in referral children.This study intends to establish a 72 hours mortality risk model for critical pediatrics of inter-hospital transport by analyzing the high risk factors of mortality in Hunan Children's hospital.Methods Retrospectively recruited all critical patients who were transported to the PICU of Hunan Children's Hospital from Dec 1st,2014 to Dec 31 st,2015.Randomly divided all the participants into model group and validation group.Multivariate logistic regression analysis was used to develop a preliminary model.A simple integer-based point score for each predictor variable was established by dividing beta coefcients by the absolute value of the smallest coefcient in the model and round up to the nearest integer.Discrimination of the mortality risk score between death and survival was assessed by calculating the area under the receiver-operating characteristic curve(AUC)in the derivation and validation group.Calibration of the mortality risk score was evaluatedusing the Hosmer – Lemeshow goodness-of-fit2 test in the derivation and validation group.Results(1)Basic characteristic: A total of 802 patients were included in the study,with 535 patients placed in model group and 267 patients in validation group;613(76.43%)were boys,189(23.57%)were girls,23(2.87%)died within 72 hours after referral to PICU;There were no difference between the model group and the validation group in terms of age,gender,the level of hospital,pre-transport department,mechanical ventilation,pediatric critical care transport teams,referral reason,referral distance etal.(2)Univariate analysis:Factors related to 72-month referral death in critically ill children included objective reason referral(OR=8.61,95%CI:2.44~30.37)?non-special children referral team(OR=3.91,95%CI:1.11~13.75)?adverse events during referral(OR=34.53,95%CI:11.49~103.74)?elevated TPEWS score(OR=6.54,95% CI:2.45~17.49)?the use of drugs in referral process(OR=4.70,95% CI:1.34~16.56)?referral distance(100-200 Km)(OR=2.87,95% CI:0.62~8.91)and referral distance(> 200 Km)(OR=3.27,95% CI:1.12~9.56);(3)Multivariate Logistic regression analysis: The independent risk factors associated with 72-hour death included: objective reason referral(OR=5.30,95% CI:1.33~21.12),non-special children referral team(OR=4.29,95% CI:1.05~17.44),adverse events during referral(OR=14.31,95% CI:3.73~54.98)TPEWS score increase(OR=5.06,95%CI:1.60~15.98);(4)Multivariate Logistic regression analysis showed that: 72 hours mortality risk model for critical pediatric of inter-hospital transport consisted of objective reasons of referral,non-professional children referral team,adverse events during the referral process,and elevated TPEWS scores.The ? value of the objective reasons referral was 1.74,with a score of 2 as the reference.The mortality risk score comprising non-special children referral team?adverse events during referral and elevated TPEWS score,non-professional children referral team?adverse events during referral and elevated TPEWS score was given 1?3and2points value.The range of score is from 0 to 8.(5)The AUCs(95% CI)was 0.871(95% CI:0.778 ~ 0.964)in the derivation group.Using the “optimal” cutoff point of 3,the sensitivity,specificity,positive predictive value,negative predictive value and Youden's index were 88.24%?72.39%?9.49%?99.47%?0.61 in the derivation group,respectively.The AUCs(95% CI)was 0.847(95%CI:0.00~ 1.00)in the validation group,respectively.The AUC for the mortality risk score showed no significant difference between derivation and validation groups tested by the C-statistic(Z= 0.33,P > 0.05).(6)Hosmer-Lemeshow test:In all the group,we found that there were no significant difference between the model expected and observedvalue by Hosmer-Lemeshow goodness-of-fit(model group:2=2.03,P=0.95;validation group:2=1.33,P=0.98).(7)Risk classification by the “optimal” cutoff point: In the model group478 cases(89.34%)were low-risk children,40 cases(7.48%)were high-risk children,and the proportion of high-risk and low-risk children was 10(25.0%)/7(1.46%)(2=50.73,P<0.001);In the validation group250(93.63%)were low-risk children,17(6.37%)were high-risk children,and the proportion of deaths of high and low-risk children was 4(23.53%)/2(0.80 %)(2=37.44,P<0.001);the total score>3 point is classified as high risk,and the higher the score,the higher the risk of death,score?3 point is considered as low risk.Conclusion Objective referral reason ? Non-professional children transport teams?Adverse events occurred during referral?Increased TPEWS score were included in he 72 hours mortality risk model for critical pediatric of inter-hospital transport.The mortality risk model demonstrated good discrimination and calibration in our derivation and validation groups.The mortality risk model can be used to distinguish referral patients with a high risk of death.