Biomedical data science data modeling is relevant to a plethora of informatics research activities, such as natural language processing, machine learning, artificial intelligence, and predictive analytics. As Electronic Health Record systems become more advanced and more mature, with the potential to incorporate a wide and diverse array of data from genomics to mobile health (mHealth) applications, the scope and nature of the biomedical data science questions researchers ask become broader. Concomitantly, the answers to their questions have the potential to impact the care of millions of patients?getting the answers right, proactively, is high stakes. However, in data modeling currently, there is no bioethics framework to guide the process of mapping key decision points and recording the rationale for choices made. Making data modeling decision points, as well as the reasoning behind them, explicit would have a twofold impact on improving biomedical data science by: 1. Enhancing transparency and reproducibility and maximizing the value of data science research and 2. Supporting the ability to assess decision points and rationales in terms of their most crucial ethical ramifications. Research in this area is particularly timely amid the interest in, and enthusiasm for, leveraging Big Data sources in the service of improving patient population health and the health of the general public. The National Institutes of Health (NIH) recently released a strategic plan for data science; there is no better time than now to create an initial bioethical framework to inform common data modeling decision points. The improvements in data quality that will derive from decision point mapping and bioethical review will enhance efforts to apply data models across a range of high-impact areas, from predictive analytics to support clinical decision-making to robust trending models in population health to better inform local, regional, and national health policies and resource allocation. To develop this initial bioethics framework, we will use well- established qualitative research methods (interviews, focus groups, and in-person deliberation) to map the decision points in biomedical data modeling research and document the rationales invoked to support those decisions (Aim 1 key informant interviews); assess those data science decision points and decision-making rationales for their bioethical ramifications (Aim 2 focus groups); and create an initial bioethics data modeling framework (Aim 3 deliberative meeting). This study would be the first to provide a bioethics framework to meet a critical gap in biomedical data modeling activities, where the downstream consequences of developing data models without careful and comprehensive review of ethical issues can be severe. This approach directly supports core scientific values of inclusivity, transparency, accountability, and reproducibility that, in turn, foster trust in biomedical data modeling output and potential applications, whether local, national, or global.
This study would be the first to develop an initial bioethics framework to meet a critical gap in biomedical data modeling activities, where the downstream consequences of developing data models without careful and comprehensive review of ethical issues can be severe?not least because poorly developed data models have the potential to impact adversely the health of individuals, groups, and communities. Currently, there is limited conversation around potential bioethics issues in data modeling, and as yet no implementable guidance on how biomedical data science modeling research activities should occur. The initial ethics framework developed by this study would provide a roadmap to ensure that data modeling decision points are documented and their ethical ramifications considered at the outset of model creation, thus supporting core scientific values of inclusivity, accountability, reproducibility, and transparency that, in turn, foster trust in biomedical data modeling output and potential applications, whether local, national, or global.
