Assessing the Impacts of Epidemiologic Biases in WTC Health Studies

描述(由申请人提供)：虽然几乎所有的流行病学研究都涉及一定程度的偏差，但WTC响应者队列(纽约/新泽西州WTC联合会)的独特特征引起了人们对其用于流行病学研究的偏见的潜在影响的关注。该队列是在9/11事件(2002年4月)后不久建立的，目的是为WTC反应人员提供健康筛查，并建立一个快速医学监测系统。当时，对于如何快速实施医疗筛查系统，以一种可以稍后用于研究的方式收集数据的方式，没有明确的程序性指导方针。从那时起，当研究人员试图利用这些队列数据进行流行病学研究时，出现了许多问题。这些问题包括：在后续健康监测访问期间对问卷进行了多次修订，数据可靠性较差(后续访问期间数据缺失和不匹配)，后续访问的保留率较低(丢失跟踪)，以及五个临床中心通过健康监测计划转诊到治疗率的显著差异， 还有其他的。前面的例子可能是偏见的根本来源。WTC健康研究经常讨论一些潜在的偏差来解释他们的发现，如健康工作者效应、自我选择偏差和回忆偏差。然而，这些研究都没有提供偏见存在的证据，也没有量化偏见对因果推理的影响。这项拟议研究的主要目标是通过确定偏见的存在，然后通过量化和调整偏见的影响来评估流行病学偏见在WTC健康研究中的影响。为了实现这一目标，我们将评估偏见对四种不同潜伏期和严重程度的健康结局的影响：鼻窦炎、哮喘、结节病和创伤后应激障碍。这些健康结果中的每一项都显示WTC应答者的发病率升高，尽管与WTC暴露的相关性相对较弱。我们假设，通过识别和调整偏差，WTC健康研究的准确性可以得到改善，潜在地加强了与健康结果的关联，就像我们建议研究的四个结果。在本研究中，将使用两种方法来评估偏差效应：直接量化偏差效应的内部验证数据和利用已发表的方法或在没有验证数据时量化偏差效应。虽然不能确定偏差的“真实”大小，但可以确定合理的范围，并使用它来生成经偏差调整的影响估计。这些经偏差调整的效果估计更有可能反映真实的健康影响。成功完成拟议的偏差分析将 帮助其他研究人员通过提供建议来识别偏见的存在并对其进行调整，从而得出关于WTC健康影响的可信推论。此外，这项研究将为未来的灾害研究推荐指导方针，通过数据解释将研究设计的各种偏差来源降至最低。

DESCRIPTION (provided by applicant): While nearly all epidemiologic studies involve a certain level of bias, the unique characteristics of the WTC responder cohort (NY/NJ WTC Consortium) have raised concerns regarding the potential impacts of bias in its use for epidemiologic research. The cohort was established shortly after 9/11 (April 2002) to provide health screenings for WTC responders and to create a rapid medical surveillance system. At that time, there were no clear procedural guidelines for how to quickly implement a medical screening system where data are collected in a manner that can later be used for research. Since then, numerous problems have arisen as researchers attempt to utilize this cohort data for epidemiologic research. These problems include: multiple revisions of questionnaires during follow-up health monitoring visits, poor data reliability (missing and unmatched data across follow-ups), low retention rate for follow-up visits (loss to follow-up), and significantly differet rates of referral to treatment from the health monitoring program across the five clinical centers, among others. The previous examples may be fundamental sources of bias. Often WTC health studies discuss some potential biases to explain their findings, such as the healthy worker effect, self-selection bias, and recall bias. However, none of these studies have provided evidence of the presence of bias, and none have quantified the effect of bias on causal inference. The main goal of the proposed study is to assess the impacts of epidemiologic biases in WTC health studies by identifying the presence of bias and then by quantifying and adjusting for the bias effects. To achieve this goal, we will evaluate the impacts of biases on four differen health outcomes with different latencies and severities: sinusitis, asthma, sarcoidosis, and post-traumatic stress disorder. Each of these health outcomes have shown elevated rates among WTC responders, though the associations with WTC exposure were relatively weak. We hypothesize that by identifying and adjusting for bias, the accuracy of WTC health studies can be improved, potentially strengthening associations with health outcomes like the four we propose to study. In this study, bias effects will be evaluated using two approaches: internal-validation data for direct quantification of bias effects and the utilization of published methods or quantifying bias effects when no validation data are available. While the "true" magnitude of a bias cannot be established with certainty, plausible ranges can be determined and used to generate bias-adjusted estimates of the effects. These bias-adjusted effect estimates are more likely to reflect true health effects. The successful completion of the proposed bias analysis will assist other researchers in drawing plausible inferences on WTC health effects by providing recommendations to identify the presence of, and adjust for bias. In addition, the study will recommend guidelines for future disaster studies which may minimize various sources of bias from the study design through data interpretation.