作者：张雪琼，刘斯巧，白晓静，邓英

单位：

626000四川省康定市，甘孜藏族自治州人民医院儿科

Units：

Department of Pediatrics, People's Hospital of Ganzi Tibetan Autonomous Prefecture, Kangding 626000, China

关键词：

呼吸道感染；反复呼吸道感染；儿童；高原地区；列线图；模型；预测

Keywords：

Respiratory tract infections; Recurrent respiratory tract infections; Child; Plateau area; Nomograms;Model; Forecasting

CLC：

DOI：

10.12114/j.issn.1008-5971.2022.00.217

Funds：

四川省卫生厅科研课题（130528）

摘要：

目的　建立高原地区儿童反复呼吸道感染（RRTIs）发生风险预测列线图模型，并验证其临床适用性。方法　选取2019年1月至2020年10月甘孜藏族自治州人民医院收治的因初次呼吸道感染就诊的高原地区（海拔3 000 m以上）患儿221例为研究对象。采用留出法将患儿按照7∶3的比例分为训练集（n=154）和测试集（n=67）。收集训练集患儿一般资料，检测训练集患儿营养相关指标｛锌、铁、钙、镁、铁蛋白、25羟维生素D〔25（OH）D〕、维生素A、维生素E、血红蛋白｝；对所有患儿进行随访，记录RRTIs发生情况，并根据RRTIs发生情况将训练集患儿分为RRTIs组和普通感染组。采用多因素Logistic回归分析探讨训练集患儿发生RRTIs的影响因素；基于多因素Logistic回归分析结果，建立高原地区儿童RRTIs发生风险预测列线图模型；采用Bootstrap抽样法重复抽样1 000次，计算一致性指数（CI ），同时绘制校准曲线以评估该列线图模型预测训练集患儿发生RRTIs的效能；采用ROC曲线分析该列线图模型对测试集患儿发生RRTIs的预测价值。结果　截至2021-11-01，共5例患儿失访，最终本研究共纳入216例患儿，其中训练集153例〔发生RRTIs 51例（33.3%）〕、测试集63例〔发生RRTIs 20例（31.7%）〕。训练集RRTIs组患儿出生体质量低于普通感染组，母乳喂养时间短于普通感染组，被动吸烟率高于普通感染组（P ＜0.05）。训练集RRTIs组患儿维生素A、维生素E低于普通感染组（P ＜0.05）。多因素Logistic回归分析结果显示，出生体质量增加、母乳喂养时间延长及维生素A、维生素E升高是训练集患儿发生RRTIs的保护因素，被动吸烟是训练集患儿发生RRTIs的危险因素（P ＜0.05）。基于多因素Logistic回归分析结果，构建高原地区儿童RRTIs发生风险预测列线图模型。该列线图模型预测训练集患儿发生RRTIs的CI 为0.889〔95%CI （0.837，0.942）〕；校准曲线分析结果显示，该列线图模型预测训练集患儿RRTIs发生率与训练集患儿RRTIs实际发生率基本吻合。ROC曲线分析结果显示，该列线图模型预测测试集患儿发生RRTIs的AUC为0.866〔95%CI （0.757，0.939）〕，最佳截断值为0.490，灵敏度为80.0%，特异度为79.1%。结论　出生体质量增加、母乳喂养时间延长及维生素A、维生素E升高是高原地区儿童发生RRTIs的保护因素，被动吸烟是高原地区儿童发生RRTIs的危险因素，基于上述因素构建的高原地区儿童RRTIs发生风险预测列线图模型对高原地区儿童RRTIs发生风险具有一定预测价值。

Abstract：

Objective　To establish a nomogram model for predicting the occurrence of recurrent respiratory tractinfections (RRTIs) in children in plateau areas, and to verify its clinical applicability. Methods　A total of 221 children inplateau areas (over 3 000 meters above sea level) who were admitted to the People's Hospital of Ganzi Tibetan AutonomousPrefecture from January 2019 to October 2020 for initial respiratory infection were selected as the research objects. The childrenwere divided into training set (n=154) and test set (n=67) according to the ratio of 7∶3 using the set-out method. General dataof the children in the training set were collected, and the nutrition related indicators {zinc, iron, calcium, magnesium, ferritin,25-hydroxyvitamin D [25 (OH) D] , vitamin A, vitamin E, hemoglobin} of children in the training set were detected. All thechildren were followed up, the occurrence of RRTIs was recorded, and the children in the training set were divided into the RRTIsgroup and the common infection group according to the occurrence of RRTIs. Multivariate Logistic regression analysis was usedto explore the influencing factors of RRTIs in children in the training set; based on the results of multivariate Logistic regressionanalysis, a nomogram model for predicting the risk of RRTIs in children in plateau areas was established. The Bootstrap samplingmethod was used to repeat sampling 1 000 times, and the consistency index (CI ) was calculated. At the same time, a calibrationcurve was drawn to evaluate the effectiveness of the nomogram model in predicting the occurrence of RRTIs in children in thetraining set. The ROC curve was used to analyze the predictive value of the nomogram model for the occurrence of RRTIs inchildren in the test set. Results　As of 2021-11-01, a total of 5 children were lost to follow-up. In the end, a total of 216 childrenwere included in this study, including 153 cases in the training set [51 cases (33.3%) with RRTIs] and 63 cases in the test set [20cases (31.7%) with RRTIs] . In the training set, RRTIs group had lower birth weight, shorter breastfeeding time, and higher passivesmoking rate than the common infection group (P < 0.05) . In the training set, the levels of vitamin A and vitamin E in the RRTIsgroup were lower than those in the common infection group (P < 0.05) . The results of multivariate Logistic regression analysis showedthat increased birth weight, prolonged breastfeeding time, and elevated vitamin A and vitamin E were protective factors for RRTIs inchildren in the training set, and passive smoking was a risk factor for RRTIs in children in the training set (P < 0.05) . Based on theresults of multivariate Logistic regression analysis, a nomogram model for predicting the risk of RRTIs in children in plateau areaswas constructed. The CI of the nomogram model for predicting the occurrence of RRTIs in the training set was 0.889 [95%CI (0.837,0.942) ] . The results of calibration curve analysis showed that the predicted incidence of RRTIs in the training set by the nomogrammodel was basically consistent with the actual incidence of RRTIs in the training set. The results of ROC curve analysis showed thatthe AUC of the nomogram model for predicting the occurrence of RRTIs in children in the test set was 0.866 [95%CI (0.757, 0.939) ] ,the best cutoff value was 0.490, the sensitivity was 80.0%, and the specificity was 79.1%. Conclusion　Increased birth weight,prolonged breastfeeding, and elevated vitamin A and vitamin E are protective factors for RRTIs in children in plateau areas, andpassive smoking is a risk factor for RRTIs in children in plateau areas. Based on the above factors, the nomogram model for predictingthe risk of RRTIs in children in plateau areas has a high predictive value for the risk of RRTIs in children in plateau areas.

ReferenceList：