Shebl Fatma, Poppell Carolyn E, Zhan Min, Dwyer Diane M, Hopkins Annette B, Groves Carmela, Reed Faye, Devadason C, Steinberger Eileen K
Department of Epidemiology and Preventive Medicine, University of Maryland, Baltimore, School of Medicine, Baltimore, MD 21201, USA.
Public Health Rep. 2009 Jul-Aug;124(4):495-502. doi: 10.1177/003335490912400406.
Population-based landline telephone surveys are potentially biased due to inclusion of only people with landline telephones. This article examined the degree of telephone coverage bias in a low-income population.
The Charles County Cancer Survey (CCCS) was conducted to evaluate cancer screening practices and risk behaviors among low-income, rural residents of Charles County, Maryland. We conducted face-to-face interviews with 502 residents aged 18 years and older. We compared the prevalence of health behaviors and cancer screening tests for those with and without landline telephones. We calculated the difference between whole sample estimates and estimates for only those respondents with landline telephones to quantify the magnitude of telephone coverage bias.
Of 499 respondents who gave information on telephone use, 80 (16%) did not have landline telephones. We found differences between those with and without landline telephones for race/ethnicity, health-care access, insurance coverage, and several types of cancer screening. The absolute coverage bias ranged up to 6.5 percentage points. Simulation scenarios showed the magnitude of telephone coverage bias decreases as the percent of the population with landline telephone coverage increases, and as landline telephone coverage increases, the estimates from a landline telephone survey would approximate the estimates from a face-to-face survey.
Our findings highlighted the need for targeted face-to-face surveys to supplement telephone surveys to more fully characterize hard-to-reach subpopulations. Our findings also indicated that landline telephone-based surveys continue to offer a cost-effective method for conducting large-scale population studies in support of policy and public health decision-making.
基于人群的固定电话调查可能存在偏差，因为只纳入了有固定电话的人群。本文研究了低收入人群中电话覆盖偏差的程度。
开展查尔斯县癌症调查（CCCS），以评估马里兰州查尔斯县低收入农村居民的癌症筛查行为和风险行为。我们对502名18岁及以上的居民进行了面对面访谈。我们比较了有固定电话和无固定电话居民的健康行为和癌症筛查测试的患病率。我们计算了全样本估计值与仅针对有固定电话的受访者的估计值之间的差异，以量化电话覆盖偏差的程度。
在499名提供了电话使用信息的受访者中，80人（16%）没有固定电话。我们发现，有固定电话和无固定电话的人群在种族/民族、医疗保健可及性、保险覆盖范围以及几种癌症筛查类型方面存在差异。绝对覆盖偏差高达6.5个百分点。模拟情景显示，随着有固定电话覆盖的人口百分比增加，电话覆盖偏差的程度会降低，并且随着固定电话覆盖范围的扩大，固定电话调查的估计值将接近面对面调查的估计值。
我们的研究结果强调了需要进行有针对性的面对面调查来补充电话调查，以便更全面地描述难以接触到的亚人群。我们的研究结果还表明，基于固定电话的调查继续为开展大规模人群研究以支持政策和公共卫生决策提供了一种具有成本效益的方法。
