Johnson & Johnson recently sponsored a colloquium at The New York Academy of Sciences presented in collaboration with the NYU School of Medicine. The event discussed the opportunities and challenges of health care in the era of big data.
To further discuss the issue, Outsourcing-Pharma (OSP) caught up with Dr. Patrick Ryan (PR), head, epidemiology analytics, Janssen R&D, office of the chief medical officer, Johnson & Johnson.
OSP: What are the key ethical concerns of integrating big data into the health care landscape?
PR: Generating reliable evidence to inform future clinical practice based on the past experience of patients is essential to advancing better health outcomes. However, we need to do this in a manner that protects patient privacy and respects the interests of those who contribute to the data collection or could be impacted by the data use.
In the context of observational data, such as administrative claims and electronic health records, a strategy that is used by groups such as OHDSI (Observational Health Data Sciences and Informatics) and FDA-Sentinel is to establish data networks that conduct analysis using a distributed model, whereby patient-level data are maintained securely behind the firewalls of the data-holding organization and only aggregate summary statistics are shared between participating sites.
A counter-balancing concern is ensuring the reliability of the evidence generated from observational data. Consensus on proper standards for evidence quality has not been established, which raises questions of how evidence generated from observational data should be integrated with evidence from other sources, including weighing against evidence from randomized clinical trials.
To address this, some have expressed a desire to ensure data provenance and auditing through source record adjudication, which is not always possible. Further, if centralization of data were required, it could potentially compromise patient privacy.
OSP: How are these being addressed? And what are the main challenges?
PR: A key challenge is achieving consensus across scientific disciplines (epidemiology, statistics, medical informatics, clinical sciences) and stakeholders (industry, regulatory agencies, health systems, academia) for what constitutes ‘reliable evidence.’
That consensus is a necessary pre-requisite before determining appropriate governance for how and when these observational data can be used by which analytical methods to meet the scientific needs of the community.
To address this challenge, OHDSI has developed an open-science community that allows academia, industry, and government to collaborate in the design and conduct of observational analyses across its international network of databases.
It is also conducting methodological research that aims to empirically evaluate analytical methods and data standards to quantify the reliability of evidence generated using alternative approaches.
Janssen contributes to OHDSI’s research agenda, because it recognizes the need to perform this foundational ‘meta-science’ that can be used to establish scientific best practices that can be consistently implemented for specific clinical questions of interest.
OSP: Specifically, what considerations must be taken into account to navigate recruitment and consent?
PR: Retrospective data analysis of de-identified patient-level data requires different governance structures from prospective data collection in the context of clinical research – which requires informed consent.
Observational data can be used to support clinical research in the following ways: First, they can support clinical trial design to ensure that studies are feasible and sufficiently generalizable to be informative.
Second, they can be used for site identification – to find the providers who may be interested in engaging in clinical research because they treat patients who would qualify for a given study – and patient identification – allowing for authorized clinicians to raise awareness of patients of their opportunities to participate in clinical research for which they may qualify if they are interested.
Note that the ability to identify sites and patients is limited to the providers of the data, e.g., a health insurance company, and is not available to end users who pay a fee to access these datasets.
OSP: How has the entrance of new players (from tech, etc.) created a period of disruption?
PR: New players have introduced exciting innovations in data collection platforms, data aggregation systems, and analytic ecosystems.
Technological advances, coupled with scientific progress across multiple disciplines, have enabled analyses to be designed and executed by a broader community, with new technical infrastructures making it feasible to conduct large-scale analyses at a more reasonable cost point.
These advances are also introducing opportunities to extend the breadth and depth of analyses performed beyond what was imaginable less than a decade ago.
OSP: How is the pharma industry working with these companies?
PR: Pharma is partnering with big tech to learn about innovations in technical infrastructure, analytical solutions, and novel scientific approaches, and actively applying these new technologies to attempt to advance their data science capabilities and expand the role of data to inform their decision-making.
Another innovation is to make data collection more patient-centric. In this model, patients are asked to consent to providing access to all of their medical records and claims, and may also get asked to complete questionnaires, e.g., on symptoms they may be experiencing (which are not generally available in any of the otherwise available data sources). These systems also provide patients with access to their own data in a useable form.
OSP: Looking ahead, five, ten years, how do you expect the industry to look? What may be some of the biggest changes?
PR: Over the next 10 years, I expect we will finally realize the untapped potential of a ‘learning health care system,’ whereby data captured during the course of routine clinical practice can be continuously used to generate evidence that can directly inform future clinical practice.
The data captured will be systemically used to characterize disease and treatment utilization patterns, to study the safety and comparative effectiveness of alternative medical interventions, and to learn patient-level prediction models that allow for personalized medicine and disease interception.
I hope we will see greater international collaboration across industry, academia, and government, so that learnings from one organization or health system can be proactively used to inform policy and care decisions around the world.
Technology will enable not only ‘data liquidity’ but also ‘evidence liquidity’ that will more seamlessly fill the current gaps in our understanding of the effects of medical interventions and encourage greater dissemination of real-world evidence so that clinicians and patients can be more informed and empowered in their shared decision-making.