Classification Tree Analysis to Enhance Targeting for Cancer Screening Programs

描述（由申请人提供）：利用癌症筛查测试的差异的证据在国家癌症预防和控制的总体努力中起着核心作用。行为风险因素监测系统（BRFSS）和国家健康访谈调查（NHIS）是两个关键资源。确定人群中筛查使用率较低的人群，通常首先对一组潜在协变量与筛查试验使用率之间的关联进行单变量分析，然后进行多变量分析，最常使用逻辑回归，以确定最终达到统计显著性的一组变量。然后将这些具有统计意义的变量转化为需要有针对性的干预措施以减少观察到的差异的人群。这些干预措施的目标是提供必要的资源，以改善筛查试验的获得和利用。这项拟议的研究旨在检验分类分析的潜力，以便为筛查项目和服务确定高危人群的过程提供信息。分类分析使用数据集的渐进分割（即递归分割）来识别在因变量（例如，癌症筛查）上具有较高和较低比率的组。这些群体中的每一个都是由特定的特征组合定义的。这个为期两年的二次分析R21提案的具体目标是：目标1。将分类方法应用于2006年和2008年的BRFSS，以及2005年和2008年的NHIS，以确定基于多变量分类的低筛查利用率风险人群。筛查领域包括乳腺癌、宫颈癌、结直肠癌和前列腺癌。目标2。)检查不同分类方法获得的结果，以便更好地了解如何最好地利用使用分类方法的潜力来确定低癌症筛查风险人群中的群体。目标3。)利用分类方法研究BRFSS状态可选模块的内容是否可以补充核心BRFSS的信息，从而识别出可能具有低利用率和复杂服务需求的群体。根据CTA方法对风险群体进行多变量分类，可以加强针对重要的干预前任务的受众，包括需求评估、障碍/促进者识别和整理所需资源，以改善这些群体的筛查状况。

DESCRIPTION (provided by applicant): Evidence of disparities in the utilization of cancer screening tests plays a central role in the nation's total set of efforts for cancer prevention and control. The Behavioral Risk Factor Surveillance System (BRFSS) and the National Health Interview Survey (NHIS) are two key resources. Identifying groups in the population who are at risk of lower screening utilization typically starts with univariate analyses of associations between a set of potential covariates and utilization rates of a screening test, followed by multivariable analysis, most often using logistic regression, to determine the final set of variables that achieve statistical significance. These statistically significant variables are then translated into population groups who need targeted interventions to reduce the observed disparities. Those interventions have the objective of providing resources necessary for improving access to, and utilization of, screening tests. This proposed research is intended to examine the potential for Classification Analysis to inform the process of identifying at-risk groups for screening programs and services. Classification Analysis uses progressive segmentation of a dataset (i.e., recursive partitioning), to identify groups with higher versus lower rates on a dependent variable (e.g., cancer screening). Each of these groups is defined by a specific combination of characteristics. The Specific Aims of this two-year, secondary analysis R21 proposal are: Aim 1.) To apply classification methods to the 2006 and 2008 BRFSS, and to the 2005 and 2008 NHIS, in order to identify groups in the population at risk of lower screening utilization based on multivariable classification. Screening domains will be for breast, cervical, colorectal, and prostate cancer. Aim 2.) To examine the results obtained across classification methods, in order to better understand how best to capitalize on the potential of using classification methods to identify groups in the population at-risk of low cancer screening. Aim 3.) To use classification methods to investigate whether content from BRFSS state-optional modules can supplement information from the core BRFSS to identify groups likely to have low utilization and complex service needs. Multivariable classification of at-risk groups, informed by CTA methods, could enhance the targeting of audiences for the important pre-intervention tasks of needs assessment, barrier/facilitator identification, and marshalling the resources needed to improve the screening status of those groups. PUBLIC HEALTH RELEVANCE: National-level surveys are used to identify groups in the population who need extra resources and services in order to improve their utilization of cancer screening tests. It is important to continue to develop ways of identifying these groups, so that those at risk of low use can be targeted for screening programs and services. This research will examine how "classification analysis" methods can improve the identification of groups in the population who have low utilization of cancer screening tests (breast, colorectal, cervical, prostate).