The image of “precision medicine” is a dream of what medical care could become— driven by data analysis and tailored to individual patients. Access to data, specifically large volumes and varieties of health data, could help health care providers intervene and begin to address some of our health care ills. However, this dream should also prompt us to question how precision medicine might not develop in the ways that we think, and how bringing the increasing power of computing to bear on more and more kinds of health data could have unintended consequences.
To better understand how bias might impact precision medicine in the sphere of biomedical research, we conducted a qualitative study to identify the tensions and frameworks of diverse medical stakeholders. Our goal was to understand how precision medicine research projects are developing, as these will shape the future directions of clinical precision medicine. Building on what we learned from the research community, we define precision medicine as: the effort to collect, integrate, and analyze multiple sources of data in order to develop individualized insights about health and disease.
Health data is subject to bias in multiple ways, and the increasing quantity and types of data that are available today can make it hard to identify where bias can emerge. In an ecosystem in which research can inform clinical guidelines and treatments, biases can have potentially life-threatening impacts. Our findings identify two main types of bias in precision medicine: 1) bias in the building and analyzing of datasets, and 2) bias as the result of precision medicine research.