我国免费艾滋病抗病毒药物治疗一线人数ARIMA模型预测

Application of the ARIMA Model in predicting the number of ART-naive patients on first-line treatments

目的探讨ARIMA模型在我国成人艾滋病一线治疗人数预测中的应用,阐述建模过程,建立预测模型并预测2011年至2012年治疗人数,分析预测影响因素,为今明两年艾滋病治疗资源配置提供参考。方法数据来自国家免费艾滋病抗病毒药物治疗信息管理系统2003年至2009年成人(>15岁)一线新治疗者按感染途径、基线CD4水平分组的季报人数,通过模型识别、参数估计、检验诊断、模型评价,建立艾滋病一线治疗人数ARIMA模型,预测2010年、2011年治疗人数。结果我国免费艾滋病一线成人治疗累计人数与在治人数预测模型均为ARIMA(1,1,0)(0,1,0)4,预测2010年与2011年底累计人数分别为9.9万、12.1万,在治人数分别为7.3万、8.9万,模型预测误差率分别为0.26%、0.67%。我国免费艾滋病一线成人新治人数预测模型为(1,0,0)(0,1,0)4,预测2010年与2011年一线新治人数分别为2.1万、2.4万,每季度平均新治人数为5.2万,新治人数模型预测误差率为2.39%。我国艾滋病一线新治人数表现逐年上升趋势和季节变化特点,而且受到感染途径、入选治疗标准等因素显著影响。结论 ARIMA模型适用于艾滋病治疗人数早期预测。

Objective To explore the application of ARIMA model in predicting the number of Chinese patients who will start first line antiretroviral therapy during 2010-2011,through describing the modeling process,building the prediction model and verification of the model＇s applicability,with the aim to facilitate the management and planning of Chinese national free ART program.Methods Data were extracted from China＇s National Free ART program datasbase.The quarterly reported numbers of new patients enrolled for ART from 2003 to 2009 were stratified by routes of transmission and CD4 counts.The ARIMA models were built by means of model identification,parameter estimation,inspection and diagnosis and model appraisal.The data of 2003 to 2008 were used to build the time sequence model,with those from 2009 used for model verification.Results With ARIMA（1,1,0）（0,1,0）4,the cumulative numbers of Chinese patients given free first-line ART will be 99 000 and 121 000 by the end of 2010 and 2011 respectively,while active patients on ART will be 73 000 and 89 000 by the end of 2010 and 2011 respectively.Error rates for the prediction were 0.26% and 0.67%.With ARIMA（1,0,0）（0,1,0）4 model,the number of new patients enrolled were predicted to be 20 000 in 2010 and 22 000 in 2011.The error rate of this prediction was 2.39%.The number of patients on first-line ART in China have been increasing with clear seasonal variation,and were impacted by routes of HIV transmission routes and change of treatment criteria.Conclusions This study shows that the ARIMA model can effectively predict the number of patients starting free ART in China.