Approaches and Strategy for Cancer Research and Surveillance Data: Integration, Information Pipeline, Data Models, and Informatics Opportunities

Thesis (Ph.D.)--University of Washington, 2016-04The advancement of cancer research, patient care and public health currently rely on acquisition of data from a variety of sources, information-processing activities, and timely access to data that is of acceptable quality for investigators, clinicians and health officials. With cancer patients living longer and undergoing multiple rounds of treatment, as well as the rise of molecular data that characterize individual patient tumors, there are challenges across all aspect of cancer data collection, integration and delivery. Although there have been advances in deployment of electronic medical records (EMRs) and use of data from EMRs and related systems to support cancer research and patient care, most data needs are still met through costly project specific manual abstraction and project specific databases. This dissertation builds on my previous work on the Caisis cancer research database at Memorial Sloan-Kettering Cancer Center, and my assessment of trends in information technology (IT) and informatics through site visits and interviews at 60 cancer centers. My hypothesis for this dissertation was that new tools and methods from biomedical informatics could improve the availability of data for cancer research if they were applied thoughtfully and strategically. Within the context of experimenting with the application of selected informatics tools and methods in a cancer center, my overarching research question was: how can we improve access to clinical and related data about cancer patients for research? The first aim of this dissertation was to develop and assess a modern integrated data platform to support a wide variety of cancer research, which explored the following questions: What are the challenges and opportunities for informatics at Fred Hutch? Do these challenges and opportunities align with my previous work and lessons learned? Are there tools and methods that others or I have used before that could be successfully applied at Fred Hutch? How can we enable new types of research, faster results, and better quality of research data at Fred Hutch? How can we improve access to data with tools and methods that are scalable, extensible to new projects, sustainable and portable to other centers? What is the impact of the data platform developed at Fred Hutch? The second aim was to develop and characterize a clinical data processing pipeline that is scalable to the cancer center enterprise level, is well-supported and sustainable, and can complement or streamline existing manual data abstraction and information-processing activities. It explored the following questions: How can we improve the quality, speed, and economics of the acquisition, processing, and delivery of clinical data to support cancer researchers? How can we make clinical data processing a core competency of Fred Hutch at an enterprise scale? The fourth aim was to develop, model, and assess database frameworks for cancer, which explored the following question: How can we best characterize and compare database models and big-data technologies to inform a sensible strategy for cancer research database design and technology? The fifth aim was to characterize the big-data needs and informatics opportunities for cancer surveillance at the national level, which explored the following questions: What informatics tools and methods have already been applied successfully in the cancer surveillance domain? What are the current and coming opportunities for applying or developing new tools and methods for advancing biomedical informatics in this domain? The research includes four related papers. The first paper describes the design, development and results of the Hutch Integrated Data Repository and Archive (HIDRA), a modern data platform that provides data feeds, a high security operational and hosting environment, and a self-service data access tool for exploring clinical data as well as associated biospecimen, study and molecular or other assay data. An additional chapter following the first paper describes the impact of HIDRA. The second paper describes the development, implementation and usage of an enterprise pipeline to facilitate the transition from manual data collection and information processing to broad use of clinical data processing and machine- learning methods. The fourth paper characterizes cancer database approaches and big-data technologies to support current and next-generation cancer research. The third paper reviews and summarizes the need for informatics, clinical data processing and machine learning to advance cancer registries at the hospital, state and national level, and reviews current informatics research related to cancer surveillance. The results and contribution of this dissertation are new examples of approaches to data platforms, data models and a pipeline for clinical data processing for cancer research and cancer surveillance, as well as explanations of underlying motivations, concepts, and tradeoffs for these informatics tools and methods. The contribution of the fourth paper is a potential roadmap for application of informatics, clinical data processing and machine- learning tools and methods to cancer registries and national cancer surveillance. The generalizable contributions of Chapter 2 are a working a comprehensive database model and associated web-based tool for data abstraction that is temporally organized and has the ability to stack into an analytic structure for predictive modeling. This system is freely distributed under an open-source license, meets common requirements for IT security, extensibility and supportability, and it has already been adopted and extended by numerous other cancer centers in the United States and internationally. The generalizable contributions of Chapter 3 are the following. First, the volume and variety of data elements that can practically be collected through clinical templates is limited. Second, given the importance of research and collaboration networks, cancer centers should adopt or at least be interoperable with common platforms like REDCap, i2b2, OpenSpecimen and OnCore so that we can wrestle with common issues as a community. Third, due to limited and variable funding for research, solutions need to scale down to affordable levels for individual researchers and labs. Fourth, site visits and active cross-pollination of tools and methods across center must extend deeper into all levels of IT and informatics staff rather than just connecting senior IT leaders and informatics researchers. Finally, centers should spend time and effort resolving social and organizational barriers to progress in informatics and IT. The generalizable contributions of Chapters 4–5 are the following: First, the legal, IRB, and security framework for HIDRA is relevant to other centers and has been applied to at least two similar efforts. Second, HIDRA provides an example of leveraging a clinical data repository at a broader academic medical center to support a cancer-specific data repository. Third, HIDRA provides an example of adopting and extending the IT and informatics work of other groups to solve local issues economically. Fourth, HIDRA provides an example of an overall strategy for clinical data acquisition, processing, storage and self service data access. Fifth, HIDRA identified the need for a realistic and de-identified testing dataset to facilitate software development and system implementation. Sixth, the HIDRA work found that lack of federated security for a consortium or matrix cancer center is a critical barrier to progress on an integrated data repository. Finally, the HIDRA project found that the Agile approach to software engineering and system implementation was critical for project momentum and success. The generalizable contribution of Chapter 6 is a tool for shifting the work of manual data abstraction so that it generates training and validation data suitable for development of automated clinical data processing algorithms (e.g., statistical algorithms, NLP, machine learning). This tool was built with a technology partner, LabKey Software, so that it can be portable to other clients, scalable, and extensible to different clinical data sources and databases. Other groups are already adopting this tool. The generalizable contributions of Chapter 8 are the following. First, it provides a description of cancer registries and cancer surveillance from an informatics perspective, including the case for automation. Second, it contributes a review of informatics tools and methods applied to cancer registries that indicates potential for automation of clinical data processing. Third, it identifies cancer registries and cancer surveillance as an area for funding and advancing biomedical informatics research