Department of Health Education and Promotion, College of Health and Human Performance, East Carolina University, Greenville, NC.
Department of Health Behavior, UNC Gillings School of Global Public Health, Chapel Hill, NC.
Nicotine Tob Res. 2018 Sep 25;20(11):1353-1358. doi: 10.1093/ntr/ntx149.
In the United States, tens of thousands of inspections of tobacco retailers are conducted each year. Various sampling choices can reduce travel costs, emphasize enforcement in areas with greater noncompliance, and allow for comparability between states and over time. We sought to develop a model sampling strategy for state tobacco retailer inspections.
Using a 2014 list of 10,161 North Carolina tobacco retailers, we compared results from simple random sampling; stratified, clustered at the ZIP code sampling; and, stratified, clustered at the census tract sampling. We conducted a simulation of repeated sampling and compared approaches for their comparative level of precision, coverage, and retailer dispersion.
While maintaining an adequate design effect and statistical precision appropriate for a public health enforcement program, both stratified, clustered ZIP- and tract-based approaches were feasible. Both ZIP and tract strategies yielded improvements over simple random sampling, with relative improvements, respectively, of average distance between retailers (reduced 5.0% and 1.9%), percent Black residents in sampled neighborhoods (increased 17.2% and 32.6%), percent Hispanic residents in sampled neighborhoods (reduced 2.2% and increased 18.3%), percentage of sampled retailers located near schools (increased 61.3% and 37.5%), and poverty rate in sampled neighborhoods (increased 14.0% and 38.2%).
States can make retailer inspections more efficient and targeted with stratified, clustered sampling. Use of statistically appropriate sampling strategies like these should be considered by states, researchers, and the Food and Drug Administration to improve program impact and allow for comparisons over time and across states.
The authors present a model tobacco retailer sampling strategy for promoting compliance and reducing costs that could be used by US states and the Food and Drug Administration (FDA). The design is feasible to implement in North Carolina. Use of the sampling design would help document the impact of FDA's compliance and enforcement program, save money, and emphasize inspections in areas where they are needed most. FDA should consider requiring probability-based sampling in their inspections contracts with states and private contractors.
在美国，每年都会对成千上万的烟草零售商进行检查。各种抽样选择可以降低旅行成本，强调在违规行为较多的地区加强执法，并允许各州之间以及随着时间的推移进行比较。我们旨在为州烟草零售商检查开发一种模型抽样策略。
使用 2014 年北卡罗来纳州 10,161 家烟草零售商的名单，我们比较了简单随机抽样、按邮政编码分层聚类抽样和按普查地段分层聚类抽样的结果。我们对重复抽样进行了模拟，并比较了各种方法的精度、覆盖率和零售商分散程度。
在保持适当的设计效果和适合公共卫生执法计划的统计精度的同时，基于邮政编码和基于地段的分层聚类方法都是可行的。邮政编码和地段策略都比简单随机抽样有所改进，分别为零售商之间的平均距离（减少 5.0%和 1.9%）、抽样社区的黑人群体比例（增加 17.2%和 32.6%）、抽样社区的西班牙裔居民比例（减少 2.2%和增加 18.3%）、抽样零售商靠近学校的比例（增加 61.3%和 37.5%）以及抽样社区的贫困率（增加 14.0%和 38.2%）。
各州可以通过分层聚类抽样使零售商检查更加高效和有针对性。各州、研究人员和食品药品监督管理局应考虑使用这些具有统计学意义的抽样策略来提高计划的影响力，并允许随着时间的推移和各州之间进行比较。
作者提出了一种促进合规和降低成本的烟草零售商抽样模型策略，美国各州和食品药品监督管理局（FDA）可以使用。该设计在北卡罗来纳州是可行的。使用抽样设计将有助于记录 FDA 合规和执法计划的影响，节省资金，并在最需要的地方加强检查。FDA 应考虑在与各州和私营承包商的检查合同中要求基于概率的抽样。
