Chinese Journal of Child Health Care ›› 2021, Vol. 29 ›› Issue (4): 425-429.DOI: 10.11852/zgetbjzz2020-0103

• Clinical Research • Previous Articles     Next Articles

Application of auto regressive integrated moving average model in predicting the incidence of congenital hypothyroidism in Guangzhou

JIANG Xiang, XU Ai-jing, TANG Fang, JIA Xue-fang, TANG Cheng-fang, CHEN Qian-yu, ZHENG Rui-dan, LIU Ji-lian, HUANG Yong-lan   

  1. Neonatal Screening Center,Guangzhou Women and Children Medical Center,Guangzhou,Guangdong 510180,China
  • Received:2020-01-19 Revised:2020-02-03 Online:2021-04-10 Published:2021-04-27
  • Contact: HUANG Yong-lan,E-mail:xxhuang321@163.com

自回归求和移动平均模型在广州市新生儿先天性 甲状腺功能减低症发病率趋势预测的应用

蒋翔, 徐爱晶, 唐芳, 贾雪芳, 唐诚芳, 陈倩瑜, 郑锐丹, 刘记莲, 黄永兰   

  1. 广州市妇女儿童医疗中心,广州市新生儿筛查中心,广东 广州 510180
  • 通讯作者: 黄永兰,E-mail:xxhuang321@163.com
  • 作者简介:蒋翔(1979-),男,湖南人,副主任技师,硕士学位,主要从事新生儿疾病筛查工作。
  • 基金资助:
    广东省医学科研基金项目(C2018035);广州市卫生计生科技一般引导项目(20171A011259);广州市妇女儿童医疗中心儿科研究所项目(IP-2018-026)

Abstract: Objective To analyze the feasibility of applying auto regressive integrated moving average model (ARIMA) to predict the incidence of congenital hypothyroidism (CH) in Guangzhou,so as to provide references for the prevention and control of congenital hypothyroidism. Methods The incidence rate of CH in Guangzhou from 1991 to 2018 was collected,and the ARIMA model was established according to the CH incidence rate from 1991 to 2016.The ARIMA model was tested by comparing the predicted incidence and actual incidence in 2017-2018,and the trend of CH incidence in the next three years was predicted. Results The overall incidence of CH showed an upward trend from 1991 to 2018.ARIMA (0,1,0) was the optimal model based on 1991-2016 CH incidence rate,which was better fit the time trend of CH incidence.The maximum relative error between the predicted and actual incidence of CH in 2017-2018 was 7.9% in the model fitting,providing a relatively accurate prediction.The predicted CH incidence in 2019-2021 was 86.1 per 100 000 births,90.7 per 100 000 births and 94.3 per 100 000 births,respectively. Conclusion ARIMA (0,1,0) model has a good fitting effect on the trend of CH incidence and can be used for the prediction and dynamic analysis of CH incidence in Guangzhou.

Key words: congenital hypothyroidism, autoregressive integrated moving average model, prediction, neonates

摘要: 目的 分析构建并应用自回归移动平均模型(ARIMA)对广州地区先天性甲状腺功能减低症(CH)发病率进行预测的可行性,为合理制定 CH 的防治、保健的策略及措施提供科学依据。方法 汇总广州市1991-2018新生儿CH筛查发病率数据,基于1991-2016年CH发病率数据建立最优ARIMA模型,通过比较2017-2018年预测发病率和实际发病率对模型预测性能进行检验,并预测未来3年CH发病率趋势。结果 1991-2018年CH发病率总体呈现上升趋势。基于1991-2016年CH发病率数据进行模型拟合,ARIMA(0,1,0)为最优模型,可以较好的拟合CH 发病率的时间变化趋势。模型拟合的2017-2018年CH预测发病率和实际发病率相对误差最大值为7.9%,提供较准确的预测。预测的2019-2021 CH发病率分别为86.1/105、90.7/105、94.3/105结论 ARIMA(0,1,0)模型对新生儿CH发病率的拟合效果较好,可用于广州地区CH发病率的短期预测及动态分析。

关键词: 先天性甲状腺功能减低症, 自回归移动平均模型, 预测, 新生儿

CLC Number: