重症手足口病患儿死亡风险评估模型的建立

目的 探讨重症手足口病患儿实验室指标与预后的相关性，建立判断发生死亡的风险评分模型。方法 回顾性收集2012 年 1 月至 2014 年 6月湖南省儿童医院PICU收治的重症手足口病病例，其中2012年1月至2013年12月的病例数据用于建立死亡风险模型，2014年1~6月的病例数据用于模型的验证。以入住PICU首次血清N-末端脑钠素原（NT-proBNP）、乳酸、WBC计数、血糖、心肌酶谱（CK、CK-Mb、LDH、 Mb）、PCT、CRP 检查结果作为分析指标，通过受试者特征工作（ROC）曲线，筛选出曲线下面积（AUC）>0.7的实验室指标，确定各项指标的最佳界值。以 Logistic 回归分析各项实验室指标对预后的影响，以进入Logistic回归模型各项实验室指标的β值作为原始赋分，同时对β值取整后作为简化赋分，建立风险模型并验证，考察不同总评分的敏感度和特异度。结果 362例数据纳入死亡风险模型的建立，男230例（63.5%），女132例，年龄(24.0±14.8)个月，死亡25例（6.9%）；验证数据共171例，男111例（64.9%），女60例，年龄 （24.3±13.5）个月，死亡11例（6.4%）。①血 NT-proBNP、乳酸、WBC、血糖、心肌酶谱和PCT的 ROC AUC均＞0.70。②Logistic 回归分析显示血清 NT-proBNP、乳酸、WBC和血糖是死亡的危险因素，OR（95%CI）分别为30.67 (3.64~258.68)、5.22 (1.22~22.33)、10.04 (2.12~ 47.53)和10.56 (1.88~59.27)。③WBC＞16.5×109·L-1、血糖＞7.8 mmol·L-1、NT-proBNP＞10 ng·mL-1、乳酸＞3.2 mmol·L-1、NT-proBNP 1.3~10 ng·mL-1分别赋2.31、1.65、3.42、2.36和1.98分，简化赋分分别为2、2、3、2和2分。④使用原始赋分和简化赋分，以验证数据预测死亡的ROC 曲线 AUC均为 0.99；简化赋分以4分为界值时其预测死亡的敏感度和特异度分别为 100%和 95.6%。结论 4项实验室检查指标的简化赋分对预测重症手足口病患儿的预后有较好价值,总评分≥4分的患儿应引起高度重视。

Establishment of mortality risk assessment model for children with severe hand,foot, and mouth disease

Objective To explore the correlation between various laboratory parameters and prognosis, and to establish the risk scoring model of predicting adverse outcomes in children with severe hand, foot and mouth disease (HFMD). Methods All patients with HFMD consecutively admitted to the PICU in Hunan Children′s Hospital from 1st Jan 2012 to 30th June 2014 were included in the study. The data of patients between 1st January 2012 and 30th Dec 2013 were used to establish the mortality risk scoring model. The data of patients from 1st January 2014 to 30th June 2014 were used to verify the model. Receiver operating characteristic curve (ROC) was used to evaluate the cutoff value of laboratory parameters, such as N-terminal pro-brain natriuretic peptide (NT-proBNP), lactate (LAC), white blood cell (WBC), blood glucose (GLU), myocardial enzymes, procalcitonin (PCT) and C-reactive protein (CRP) from January 2012 to June 2014, and analyze the correlation between various laboratory parameters and prognosis, such as serum NT-proBNP, LAC, WBC, GLU, myocardialenzymes（CK, CK-Mb, LDH, Mb）, PCT, CRP, to screen the laboratory parameters which could predict the prognosis by ROC curve, and analyze its sensitivity and specificity and to classify each index with the critical value, analyze the influence of each index in prognosis with Logistic regression analysis, and give the grades by the power to establish and validate the risk scoring model. Results A total of 362 patients with the mean age of (24±14.8) months including 230 boys were included in the risk scoring model. 337 patients survived at the end of the PICU stay, 25 patients (6.9%) died. 171 patients with the mean age of (24.3±13.5) months including 111 boys were included as the data model of validation. 337 patients survived at the end of the PICU stay, 11 patients (6.4%) died. The area under the receiver-operating characteristics (ROC) curve was >0.7 for NT-proBNP, LAC, WBC, GLU, myocardial enzymes and PCT. Multivariate adjusted odds ratios and 95% confidence intervals were calculated using unconditional logistic regression. NT-proBNP (OR=30.67, 95%CI: 3.64-258.68), LAC (OR=5.22, 95%CI: 1.22-22.33), WBC (OR=10.04, 95%CI: 2.12-47.53), and GLU (OR=10.56, 95%CI: 1.88-59.27) were associated with increased mortality risk for HFMD. The prognostic risk model based on the results of the preliminary study showed NT-proBNP＞10 ng·mL-1, NT-proBNP 1.3-10 ng·mL-1, LAC＞3.2 mmol·L-1, white blood cell＞16.5×109·L-1, blood glucose＞7.8 mmol·L-1. According to the raw data given grades were 3.42, 1.98, 2.36, 2.31, 1.65 respectively,and they were simplified as 3, 2, 2, 2, 2 respectively in the established model. Both of the raw data risk model and simplified risk model achieved the high area under ROC of 0.99, simplified risk model showed the sensitivity and specificity were 100% and 95.6% respectivelty.Conclusion The simplified risk model consisting of four laboratory index was valuable for predicting the prognosis of severe HFMD, the patients with rating scores≥4 should be paid much more attention to.