Using Matched Cohort Methods to Investigate the Risk of Respiratory Diseases in Survivors of Adult Cancer

越来越多的研究正在使用患者与任何医疗保健服务（例如全科医生或医院就诊）互动时产生的健康数据进行。这些数据可用于开展比较流行病学研究，以调查患病个体（暴露组）与无病史个体（未暴露组）之间不良健康结果的风险。在这种情况下，匹配队列方法通常用于选择未暴露组。在这种方法中，暴露个体与一个或多个未暴露个体在一个或多个患者特征（例如年龄和性别）上相匹配。从理论上讲，这将允许在暴露和未暴露的队列之间进行较少偏见的比较，因为匹配将使组之间更加相似。然而，这一假设尚未得到验证；对于何时应该使用匹配队列设计，如何进行匹配以及应该使用哪些特征来匹配个体，人们知之甚少。该项目的主要目的是调查一个现实世界的流行病学研究问题。特别是，我们的目标是量化有癌症病史的个体患呼吸系统疾病的风险，与那些没有癌症病史的个体相比，这是利用大数据确定癌症幸存者疾病预防领域的更大雄心的一部分。第二个目标是评估进行和分析匹配队列研究的最佳方法，用于长期结果的临床研究。进行描述性系统文献综述，系统地回顾癌症与随后的呼吸系统疾病之间关联的证据。这一步将用于确定感兴趣的具体主要研究呼吸结果。步骤2：进行叙述性回顾，了解匹配队列方法在流行病学研究中的使用情况，以及选择该研究设计的当前原因，并确定当前研究的空白。步骤3：进行模拟研究。我们将生成一个实验模拟数据集，它将作为一个“受控环境”，在这个环境中，我们将能够进行实验，以测试支撑匹配队列设计的假设。将解决的具体研究问题将从叙述回顾（步骤2）中确定。本阶段的目的是为匹配队列研究创建最优匹配策略。这一阶段的结果将用于开发开放获取的编码宏，其他研究人员可以使用这些宏进行匹配的队列研究。这一步将通过开发可应用于许多不同研究领域的数据分析技术来满足跨学科技能。此外，这一步骤将需要发展高级信息学技能，因为需要先进的计算知识来生成模拟数据。步骤4：将步骤2和3的结果应用于现实世界的流行病学问题。特别是，我们想在一个真实世界的健康数据集中调查癌症及其治疗对不断增长的癌症幸存者中发生后续不良呼吸结局风险的影响，该数据集收集了来自数百万人的数据。我们将使用匹配队列方法来解决研究问题，使用步骤3中构建的编码宏。这一步涉及开发各种优先技能，如数据分析、大数据可视化和高性能计算。最后，这一阶段的一个关键目标是让患者群体参与到与他们自身健康有关的研究中来。为此，我们将与癌症幸存者一起设计一系列研讨会，重点是为患者和研究人员获得互惠互利。

An increasing number of studies are being conducted using health data generated when a patient interacts with any healthcare service (e.g a GP or hospital visit). This data can be used to conduct comparative epidemiological studies to investigate the risk of adverse health outcomes in diseased individuals (exposed group) compared those without a history of disease (unexposed group). Matched cohort methods are commonly used in this setting to select the unexposed group. In this method an exposed individual is matched with one or more unexposed individuals on one or more patient characteristics (e.g age and sex). In theory, this will allow a less biased comparison between the exposed and unexposed cohorts as matching will make the groups more similar to each other. However, this assumption has not been tested; little is known about when one should use a matched cohort design, how the matching should be conducted and what characteristics should be used to match individuals.The main aim of this project is to investigate a real-world epidemiological research question. Particularly, we aim to quantify the risk of respiratory diseases in individuals who have a history of cancer, compared to those with no history of cancer as part of a greater ambition of using big data to identify areas of disease prevention in cancer survivorsThe secondary aim is to evaluate the best methods to conduct and analyse matched cohort studies for clinical research of long-term outcomesThis project will be conducted in four steps:Step 1: Conduct a descriptive systematic literature review to systematically review the evidence on the association between cancer and subsequent respiratory disease. This step will be used to define the specific primary research respiratory outcomes of interest.Step 2: Conduct a narrative review to understand how matched cohort methods are used in epidemiological research and the current reasoning behind selecting this study design as well as identify the current research gaps.Step 3: Conduct a simulation study. We will generate an experimental simulated dataset which will act as a "controlled environment" in which we will be able to conduct experiments to test the assumptions that underpin the matched cohort design. The specific research questions that will be addressed will be identified from the narrative review (step 2). The aim of this stage is to create the most optimal matching strategies for matched cohort studies. The results of this stage will be used to develop open-access coding macros that can be used by other researchers to conduct matched cohort studies. This step will meet interdisciplinarity skills by developing data analytics technology that can be applied to many different research areas. Furthermore, this step will necessitate the development of high-level informatics skills by requiring advanced computational knowledge in order to generate simulation data.Step 4: Apply the findings of step 2 & 3 to a real-world epidemiological question. Particularly, we would like to investigate the impact of cancer and its treatment on the risk of developing subsequent adverse respiratory outcomes in the growing population of cancer survivors in a real-world health dataset that captures data from millions of individuals. We will use matched cohort methods to address the research question, employing the coding macros built in step 3. This step involves the development of a variety of priority skills such as data analytics, big data visualisation and high-performance computing. Finally, a key aim of this stage will be the involvement of patient groups in research that concerns their own health. To this end, we will design a series of workshops with cancer survivors focussed specifically on obtaining a mutual benefit for both patients and researchers.