Objective: To explore the sampling method in China National Human Biomonitoring Program (HBP) and the related errors, so as to calculate and evaluate the study design in sampling. Methods: The sampling method of HBP is of multistage nature. Taking the results of sampling method from Guizhou province as an example, results related to sampling error and variation coefficient were calculated, using the multistage unequal probability sampling error method. Results: The HBP covered 152 monitoring sites in 31 provinces (autonomous regions and municipalities) and with 21 888 residents selected. The replacement rates at various stages were 5.26%, 6.35% and 40.6% respectively. The sampling error in Guizhou province was 3 207 594, and the coefficient of variation was 0.097. Conclusions: According to the multi-stage unequal probability sampling method, the sampling coefficient variability appeared small with high precision, in Guizhou province. However, this method did not consider the weight adjustment of non-sampling errors such as population missing rate and response rate. Methods related to the calculation on multi-stage sampling error among large-scale public health monitoring projects need to be further studied.
目的： 探讨国家人体生物监测项目抽样方法，并通过计算抽样误差来评估抽样设计。 方法： 该监测采用多阶段复杂抽样方法，并以贵州省抽样结果为例，应用多阶段不等概率抽样误差计算方法计算贵州省抽样误差及变异系数，探讨多阶段抽样设计下误差估计方法。 结果： 该监测覆盖全国31个省（自治区、直辖市）的152个监测点，抽取21 888名3~79岁城乡居民，三个阶段置换率分别为5.26%、6.35%及40.6%。经计算贵州省抽样误差为3 207 594人，变异系数为0.097。 结论： 根据多阶段不等概率抽样方法计算，贵州省抽样变异系数较小，精度较高，但该计算方法未考虑人群缺失率、应答率等非抽样误差的权重调整，大型公共卫生监测项目多阶段抽样误差计算方法有待继续研究。.
Keywords: Human biomonitoring; Multistage sampling; Sampling error.