Identifying Cancer Recurrence with Novel Data Linkages with a Cancer Registry

摘要 对于今天在美国估计有1700万癌症幸存者来说，对复发的恐惧是一个重要的来源 压力和一个驱使生存护理的问题。了解癌症复发的范围 幸存者可以为临床实践提供信息，改善患者的健康，并允许对治疗进行真实的评估 有效性。关于癌症复发的人群水平的数据很难获取，需要对 多个数据来源，以准确识别癌症复发。犹他州癌症登记处(UCR)，SEER 自1973年以来，这一登记制度在查明人口环境中的复发方面具有很强的地位。注册处 数据链接到犹他州人口数据库(UPDB)，其中包括电子病历(EMR)， 全州医疗机构数据(SHFD；住院、门诊手术和急诊科)，以及 索赔数据(所有付款人索赔数据库(APCD)，Medicare)。我们建议评估使用数据的效用 所有州癌症登记的共同来源，并调查新的数据联系的附加值 可在犹他州癌症登记处获得。我们还建议对最近开发的算法进行扩展和验证 确定个体水平的乳腺癌复发以确定其他癌症类型的复发以估计 人口层面的复发负担。我们的具体目标如下：1)确定可预测的 使用乳腺癌和癌症登记中心目前可用的数据来识别复发的性能 前列腺癌。这些将包括癌症复发变量委员会、电子病理学 报告和死亡证明。2)评估预测性能的改进，以确定 通过纳入乳腺癌和前列腺癌的新的管理数据关联而复发。3)评估 递归识别算法跨设置和移植的可扩展性和可移植性 用于研究的人口。我们将通过估计正数和正数来验证算法的预测性能 在西雅图不同种族和民族的癌症病例集合中的负面预测价值- Puget Sound SEER注册，包括跨种族/民族、年龄、阶段和 农村/城市地位。此外，我们还将验证最近开发的乳房复发识别算法 在西雅图登记的犹他州乳腺癌人群中。目前还不存在评估数据的算法 根据癌症登记和管理确定复发事件的单独和组合来源 数据。我们的结果将为常规可用数据的预测性能和 行政数据来源，这可能是不同的完整和/或昂贵的采购。我们的工作将 为癌症复发的人群水平跟踪建立一条前进的道路，并促进数据的优先顺序 可根据不同情况下可用的数据进行定制的生成工作和算法。 1

ABSTRACT For the estimated 17 million cancer survivors in the US today, fear of recurrence is a substantial source of stress and an issue that drives survivorship care. Understanding the scope of recurrence among cancer survivors can inform clinical practice, improve patient health, and allow for real-world assessment of treatment effectiveness. Population-level data on cancer recurrence are difficult to capture, and require evaluation of multiple data sources to accurately identify cancer recurrences. The Utah Cancer Registry (UCR), a SEER registry since 1973, is strongly positioned to identify recurrences in a population-based setting. The registry data are linked to the Utah Population Database (UPDB), which includes electronic medical records (EMR), statewide healthcare facility data (SHFD; inpatient, ambulatory surgery and emergency department), and claims data (All Payer Claims Database (APCD), Medicare). We propose to assess the utility of using data sources common across all state cancer registries and to investigate the added value of novel data linkages available at the Utah Cancer Registry. We also propose to extend and validate a recently-developed algorithm to identify individual level breast cancer recurrence to identify recurrence for other cancer types to estimate the population-level burden of recurrence. Our specific aims are as follows: 1) Determine the predictive performance to identify recurrence using data currently available to cancer registries for breast and prostate cancer. These would include Commission on Cancer recurrence variables, electronic pathology reports, and death certificates. 2) Estimate the improvements in predictive performance to identify recurrence by inclusion of novel administrative data linkage for breast and prostate cancer. 3) Evaluate the scalability and transportability of recurrence identifying algorithms across settings and populations for research. We will validate the algorithms’ predictive performance by estimating positive and negative predictive values among a racially and ethnically diverse collection of cancer cases from the Seattle- Puget Sound SEER registry, including comparisons of performance across race/ethnicity, age, stage, and rural/urban status. In addition, we will validate the breast recurrence identification algorithm recently developed in the Seattle registry in the Utah breast cancer population. No algorithms currently exist to evaluate the data sources individually and combined to identify recurrence events based on cancer registry and administrative data. Our results will inform the predictive performance for routinely available data and the value added of administrative data sources, which may be differentially complete and/or costly to procure. Our work will establish a path forward for population-level tracking of cancer recurrence and facilitate prioritization of data generation efforts and algorithms that can be customized based on the data available in different situations. 1