When data are not missing at random: implications for measuring health conditions in the Behavioral Risk Factor Surveillance System.
When data are not missing at random: implications for measuring health conditions in the Behavioral Risk Factor Surveillance System.
复制标题
DOI：
10.1136/bmjopen-2011-000696
复制
发表时间：
2012
期刊：
影响因子：
2.9
通讯作者：
Strine T
中科院分区：
文献类型：
--
作者：
Frankel MR;Battaglia MP;Balluz L;Strine T
为了检查从大规模调查中产生的健康状况的估计水平的影响，无论是列表式的受访者删除或标准的人口统计项目级插补。评估在项目插补过程中纳入相关辅助变量导致偏倚进一步降低的程度。大型横断面（美国一级）家庭调查。2006年行为风险因素监测系统调查中的218 726名美国成年人（18岁及以上）。 这项调查是美国疾病控制和预防中心进行的最大规模的电话调查。美国成年人严重抑郁症的估计比率。使用列表式应答者删除和/或人口统计学插补导致美国成年人严重抑郁症的低估。列表式删除产生低估9%（8.7% vs 9.5%）。人口统计学插补低估了7%（8.9% vs 9.5%）。这两个差异在0.05水平上都是显著的。在估计某些健康状况的国家水平时，使用列表删除和/或仅按人口统计数据进行估算可能会产生严重的失真。本文讨论了与以下事实相关的问题：当使用横断面调查来估计公共卫生状况和行为时，一些受访者没有回答所有问题。这被称为项目无响应。虽然“加权”是用来解决总体（单位）不答复的问题，但为回答每个问题的答复者子集确定权重是不切实际的。根据回答问题的人将（与问题有关的）具体估计数制成表格可能会导致调查偏差。已经开发了许多插补技术，以解决与仅将表格限制为问题应答者相关的偏倚。将调查估计仅限于总体调查应答者（排除特定问题的无应答者）可能会产生有偏差的调查估计。特定问题插补的标准方法可能会消除或减少一些这种偏倚。强烈建议在所有变量中进行系统搜索，以确定与插补目标变量之间的密切关系。涉及基本人口统计学的项目插补的标准方法可能无法最大限度地减少偏倚。如果存在应答者报告的其他（非人口统计学）相关性，则可用于改进无应答插补模型。本文的重点是自我报告的焦虑和抑郁水平的行为危险因素监测系统，一个随机数字拨号电话调查。非随机拨号调查中的焦虑和抑郁以外的其他情况报告可能不适合这种无应答模型进行插补。
To examine the effect on estimated levels of health conditions produced from large-scale surveys, when either list-wise respondent deletion or standard demographic item-level imputation is employed. To assess the degree to which further bias reduction results from the inclusion of correlated ancillary variables in the item imputation process. Large cross-sectional (US level) household survey. 218 726 US adults (18 years and older) in the 2006 Behavioral Risk Factor Surveillance System Survey. This survey is the largest US telephone survey conducted by the Centers for Disease Control and Prevention. Estimated rates of severe depression among US adults. The use of list-wise respondent deletion and/or demographic imputation results in the underestimation of severe depression among adults in the USA. List-wise deletion produces underestimates of 9% (8.7% vs 9.5%). Demographic imputation produces underestimates of 7% (8.9% vs 9.5%). Both of these differences are significant at the 0.05 level. The use of list-wise deletion and/or demographic-only imputation may produce significant distortion in estimating national levels of certain health conditions. The article addresses issues associated with the fact that when cross-sectional surveys are used to estimate public health conditions and behaviours, some respondents do not answer all the questions. This is referred to as item non-response. While ‘weighting’ is used to address overall (unit) non-response, the development of weights for the subset of respondents answering each question is impractical. The tabulation of specific estimates (related to a question) based on persons responding to the question may result in survey bias. A number of imputation techniques have been developed that address the resulting bias associated with the restriction of tabulations to question responders only. Restricting survey estimates to overall survey responders only (eliminating question-specific non-responders) may produce biased survey estimates. Standard methods of question-specific imputation may eliminate or reduce some of this bias. A systematic search among all variables for strong relationships with the target variables for imputation is strongly recommended. Standard methods for item imputation involving basic demographics may fall short of maximum possible bias reduction. If additional (non-demographic) correlates of reporting among responders are present, these may be used to improve non-response imputation models. The article is focused on the self-reporting of anxiety and depression levels in the Behavioral Risk Factor Surveillance System, a random-digit-dialling telephone survey. Reports of conditions other than anxiety and depression and in non-random-digit-dialling surveys may not be amenable to this non-response modelling for imputation.
登录
查看更多内容
影响因子：
4
作者：
Shrive FM;Stuart H;Quan H;Ghali WA
通讯作者：
Ghali WA
DOI：
10.1080/13645570802394003
发表时间：
2009-01-01
影响因子：
3.3
作者：
Durrant, Gabriele B.
通讯作者：
Durrant, Gabriele B.
影响因子：
120.7
作者：
Spitzer, RL;Kroenke, K;Williams, JBW
通讯作者：
Williams, JBW
影响因子：
4.4
作者：
Lin, Ting Hsiang
通讯作者：
Lin, Ting Hsiang
影响因子：
2
作者：
Shlomo, Natalie
通讯作者：
Shlomo, Natalie