Scientific Institute of Public Health, Brussels, Belgium.
Hasselt University, KU Leuven Leuven, Hasselt, Leuven, Belgium.
Int J Public Health. 2017 Nov;62(8):949-957. doi: 10.1007/s00038-017-0976-3. Epub 2017 May 3.
Substitution of non-participating households is used in the Belgian Health Interview Survey (BHIS) as a method to obtain the predefined net sample size. Yet, possible effects of applying substitution on response rates and health estimates remain uncertain. In this article, the process of substitution with its impact on response rates and health estimates is assessed.
The response rates (RR)-both at household and individual level-according to the sampling criteria were calculated for each stage of the substitution process, together with the individual accrual rate (AR). Unweighted and weighted health estimates were calculated before and after applying substitution.
Of the 10,468 members of 4878 initial households, 5904 members (RRind: 56.4%) of 2707 households (RRhh: 55.5%) participated. For the three successive (matched) substitutes, the RR dropped to 45%. The composition of the net sample resembles the one of the initial samples. Applying substitution did not produce any important distorting effects on the estimates.
Applying substitution leads to an increase in non-participation, but does not impact the estimations.
在比利时健康访谈调查（BHIS）中，使用替代不参与家庭的方法来获得预设的净样本量。然而，应用替代对应答率和健康估计的可能影响仍不确定。本文评估了替代的过程及其对应答率和健康估计的影响。
根据抽样标准，计算了替代过程每个阶段的应答率（RR）——家庭和个人层面的应答率——以及个体累积率（AR）。在应用替代之前和之后，计算了未加权和加权的健康估计值。
在 4878 个初始家庭的 10468 名成员中，有 5904 名成员（RRind：56.4%）的 2707 个家庭（RRhh：55.5%）参与了调查。对于连续三个（匹配的）替代者，RR 下降到 45%。净样本的组成与初始样本相似。应用替代对估计值没有产生任何重要的扭曲影响。
应用替代会导致不参与率增加，但不会影响估计值。
