Kadija Ferryman: “Fairness in Precision Medicine”

January 3, 2018
Kadija Ferryman: “Fairness in Precision Medicine”

Kadija Ferryman’s talk on November 30, 2017 for the Precision Medicine: Ethics, Politics, and Culture CSSD working group drew from her post-doctoral project, “Fairness in Precision Medicine,” a study on which she is co-PI with danah boyd at the Data and Society Institute.

The question that motivates Ferryman’s work is: How do ethical and moral frames change the way we understand health data and outcome? Using content analyses of policy documents, observations of conferences, a mapping of major precision medicine projects, and interviews with 21 experts, Ferryman honed in on two sets of biases that various stakeholders recognized: embedded biases and biases in outcome. Regarding embedded biases, experts were concerned about biases in sampling of research data such as electronic health records. For biases in outcome, the stakeholders interviewed were worried about how precision medicine can exacerbate already existing inequalities.

Crucially, Ferryman emphasized that these biases should be thought about in relation to genomic data, but also the various data types that precision medicine relies on, such as electronic medical records, the “Internet of Medical Things,” and mobile and digital technologies. As such, Ferryman argued that those concerned about precision medicine should pay attention to discussions in “big data” and “algorithmic bias,” and that bioethics and “data ethics” could learn from each other.

In the meeting of the Precision Medicine working group the next day, several themes emerged from our discussion:

Correcting for Bias
A question raised during the meeting touched on how experts who recognize that bias exists can come up with strategies to correct these biases. For example, policy makers and researchers worried about diversity in precision medicine have made the recruitment of minority subjects a centerpiece of All of Us. This is also an instance of agreement on the existence of bias between different experts in precision medicine. Thus, finding more areas of agreement between different stakeholders is crucial in building the alliance of political capital, policy know-how, and technical expertise necessary to correct for biases that may arise with the introduction of precision medicine.

Different Data Types
From the standpoint of social scientists and humanists, the inclusion of different types of data in precision medicine efforts is definitely welcomed, as decades of public health research has recognized the importance of environmental and social factors in shaping health outcomes. Nonetheless, important questions here remain regarding the ability of precision medicine to reconcile the characteristics of different data types. For instance: How do biomedical researchers view these types of more qualitative data versus more quantifiable and “scientific” data types? How are different types of evidence evaluated by scientists? Relatedly, the work of linking disparate data types and recognizing patterns between them requires complex technical expertise. As such, more work should be devoted to thinking through how to integrate these various types of data to create a precise, but complete picture of an individual’s health.

Ethics in the Health Industry: “Precisely” Where Are We Headed?
Health and the data it generates are increasingly commodified. From private tech companies to healthcare providers, precision medicine ushers in greater opportunities to wield personalized health data for commercial use. This raises parallel concerns regarding the ethical use and handling of our personal information. From targeted Facebook shopping ads to Netflix recommendations, we trade our information and data privacy for access to services and convenience. Mass personalization at its current stage generally produces innocuous, if eerie, results. We retain a sense of autonomy and choice to partake in these services and disengage if we choose. As health and genomic personalization approaches arrive in the healthcare space, however, the ability to opt-out becomes much more constrained. Health is foundational in enabling meaningful engagement and participation in society. Greater integration of individual data into the healthcare system provides an opportunity for better care, but brings into question the genuine ability to opt-out of such a system in the future.

With the rise in personal health data spurred by the “Internet of Medical Things” (IoMT) and devices, we are afforded insight into not only genetic profiles, but behavioral, lifestyle, and environmental dimensions of individuals. Their implications extend beyond clinical contexts. Employers, not unreasonably, seek employee health data in pursuit of optimizing efficiency and a more productive workforce. More sinisterly, employment discrimination based on health is the next addition to contemporary concerns that include disability, race, gender, and sexual orientation.

Other ethical concerns flow more directly from technology and automated algorithms we increasingly use to analyze data. Our artificial intelligence and neural networks pick up the deeply ingrained racial and gender prejudices concealed within patterns of language, imagery, and social cues in our datasets. If we are not vigilant about policing these embedded beliefs, algorithmic bias may result in and reinforce discriminatory and exclusionary practices.

Involving the Community and Public Voice
Part of guarding against bias and discrimination involves engaging the communities directly impacted by this research. This may come in the form of Institutional Review Board (IRB) assessments or consulting local Community Board representatives drawn from the affected population. Even the selection of chosen representatives to give voice to a community, however, can be fraught with complications. How are such representatives selected — by appointment or election, and by whom? Are those who end up on the Community Board truly representative of the community’s views? What are the power dynamics and hierarchies within that community influencing who is selected? In any structure, the intricacies of human relational and power dynamics play a tangible and meaningful presence, impacting the strength of community voice in discussion and decision-making. We need to be cognizant of such complexities when implementing structures and ensure they embody the representative democratic principles we value.

While the day-to-day responsibilities of IRB members largely involve checking off applications, on the macroscale, the arc and pattern of their decisions set precedents. As Ferryman poignantly questioned in discussing her role on the IRB board, “Are we the ethical conscience of a project?” A concern present in these circles is that passing IRB review or consulting Community Board representatives may become an ethics “check-off,” rather than a genuine partnership in understanding and appreciating the potential impact of their research on populations. We want and encourage research investigators, however, to consult ethics reviews and boards, recognizing they may not have the expertise to deal with these issues. “Seeking ethical assistance” is instinctive behavior we want to standardize in future precision medicine research.

As AI and health technology increasingly infiltrate daily life outside clinical contexts and the definition of health data is expanding, the modern role of bioethics may also need to evolve and cross traditional disciplines. Precision medicine is a collaborative effort that requires multiple perspectives. If this discussion imparted one actionable recommendation, it is that the scientific fields must call upon their ethical counterparts. Ethics is not an ancillary component of precision medicine, but a fundamental one in actualizing our communal vision for precision medicine.

Building Public Trust and Responsibility
The success of the All of Us study and other human genomic research requires the generous contribution of personal health and genomic data from individuals. This partnership between the public and science is needed to realize the network effects of a robust genetic database, and usher in a new model of precision healthcare that generations will benefit from. Building public trust is critical to these efforts, and without it, achieving a precision medicine approach will be a long and arduous process. While the U.S. culture naturally lends itself towards great suspicion of state power in these contexts, government imposes desirable safety regulations and constraints on profit-maximizing corporations. Designing ethical guidelines and a comprehensive regulatory landscape is important to enable proper oversight.

Conclusion: Ethics as a Partnership
Our unfolding discussion on the array of challenges that precision medicine poses increasingly points towards a more active and potent role of modern ethics in both industry and academic research. Precision medicine and our advancing abilities to arrange massive amounts of data herald great promises for our capacity to improve human health, behavior, and lifestyles. We must ensure ethical and regulatory safeguards keep pace with these abilities and align them with our core values on equity, fairness, privacy, autonomy, etc. Protecting these rights and evolving policy to reflect these ethical principles is key to ensuring our society does not stray onto a dystopic path.

Contributed by Larry Au and Jade H. Tan