5
Public-Private Partnerships
Atul Butte, University of California, San Francisco (UCSF), discussed his experience within the University of California Health System (UC Health), a health enterprise that aims to become the single accountable care organization for the entire University of California system. UC Health combines 20 health professional schools, including six medical schools, 12 hospitals, and 1,000 care delivery sites.
UC Health has a centralized electronic health record (EHR) database that uses the Observational Medical Outcomes Partnership Common Data Model with data elements that are continuously being harmonized. The database includes EHRs from the six academic health centers: UCSF, UCLA, UC Irvine, UC Davis, UC San Diego, and UC Riverside. The database also includes additional information, such as California regulatory data, pathology and radiology text elements, and death index data.
Butte noted that initially the UC Health data are identifiable, and include geographic location, such as home address, as this information is needed for ensuring and improving the quality of medical care delivered. The address information can be linked to indices such as the Area Deprivation Index, Social Vulnerability Index, and the California Healthy Places Index, and contributes to an emerging understanding of social determinants of health. Butte also pointed out that all University of California academic medical centers make health data accessible to patients following federal standards such as Fast Healthcare Interoperability Resources.
Butte said that the database “pays for itself” because it benefits the health system’s operations by improving the quality of care, decreasing specific unnecessary inpatient drug use, helping with managing costs in the
self-funded health plans, and assisting with centralized population health management. The same database can also be used for research, once the data are de-identified. For example, UC Health recently conducted several COVID-19 related projects in collaboration with the Food and Drug Administration.1
The challenge highlighted by Butte to undertaking collaborations and data sharing is the competitive health care environment. While the benefits of data sharing are evident, health systems are cautious because the data can also potentially be used against them, leading to revenue loss. As an example, he mentioned Stanford University Medical Center health care workers’ use of data on infection rates as a contract negotiation tactic. Butte argued that despite these concerns, leveraging scale for value (as in the case of UC Health) has clear advantages for health systems. Once a database is developed for operational purposes, the additional cost of making it available for research purposes can be relatively small.
Vincent Mor, Brown University, discussed his experiences with building a public-private data sharing cooperative focused on long-term care. The project was part of the National Institute on Aging (NIA) Imbedded Pragmatic Alzheimer’s Disease and AD-Related Dementias Clinical Trials Collaboratory (IMPACT Collaboratory), which builds capacity to conduct pragmatic clinical trials of interventions embedded within health care systems for people living with dementia and their care partners.
Mor noted that COVID-19 has disproportionately affected long-term care residents, and the need for rich real-time data on nursing homes quickly became evident at the beginning of the pandemic. To respond to this need, Mor and his colleagues partnered with Genesis Health Care to obtain real-time data, and later expanded collaborations with the American Healthcare Association, Acumen LLC, and Exponent to build a data repository involving a broader range of nursing homes that have electronic medical records. Mor said that the database can become useful in monitoring future pandemics and policy changes. The database can also make it possible to selectively recruit facilities to participate in embedded randomized controlled trials of interventions.
Mor said that the initial Brown-Genesis COVID-19 partnership was possible because of prior collaborations that laid the foundation for this work. He echoed Butte’s comments that trust between the partners is essential, particularly when there is a sharing of large volumes of real-time data. In the case of this project, there was also a need to access identifiable data
___________________
1 R. Vashisht, A. Patel, B.O. Crews, O.B. Garner, L. Dahm, C. Wilson, and A.J. Butte. (2021). Age- and sex-associated variations in the sensitivity of serological tests among individuals infected with SARS-CoV-2, JAMA Network Open 4(2), e210337. doi:10.1001/jamanetworkopen.2021.0337.
to enable linkages to information that was not available from the electronic medical records. The researchers were sensitive to the risks for Genesis and have agreed to provide input focused on solving operational challenges as a priority over publications. However, the project also produced important research that contributed to knowledge on the COVID-19 outbreak.
Ultimately, the success of the initial collaboration led to the expansion of the project to include additional nursing home companies and partners. The resulting nursing home data-sharing collaborative now serves as the basis for several new studies, including randomized controlled trials, an analysis of vaccine effectiveness, and tracking breakthrough infections. Mor noted that the American Health Care Association, an industry association, played an important role in recruiting providers and negotiating agreements.
While researchers at Brown University and others who are part of the IMPACT Collaboratory will be the first users of the data, the data will later become available for other NIA investigators. The data from the electronic medical records can be particularly useful for recruiting facilities for trials and clinical research, while the electronic medical records data linked to claims data can further expand the possibilities to areas such as pharmaco-epidemiological research, public health surveillance research, and studies of the impact of treatments or policies.
Marc Overhage, Anthem, shared his perspectives on what makes public-private partnerships work based on his experience in a variety of settings, including academia, an academic medical center, an electronic medical records vendor, and currently a payer (Anthem). He emphasized that it is important to think of opportunities for collaboration as more than just data sharing, and consider the sharing of analytic techniques, computational resources, knowledge, and opportunities to commercialize and create value in other ways.
Overhage said that a typical public-private partnership might involve a collaboration between a government entity and a private organization, with the objective of building infrastructure or other services. The partners share the investment, the risks, and the rewards. He underscored the importance but also the challenge associated with establishing the mutual benefit in a partnership.
Overhage echoed previous comments about the particular concerns for private entities, including the potential disclosure of private, sensitive, or proprietary information; risks to brand reputation; and financial risks. He also highlighted the opportunity cost associated with the time invested in the collaboration. The fragmented nature of regulations, including state and local regulations that might apply to data sharing, also represents a risk. There are also security risks, both real and perceived.
Ethical risks were another area of concern Overhage discussed, specifically those associated with the loss of control over how the data might be used in the future. He said it is important to consider potential public
perception, even if regulations permit data sharing. The public might have particular concerns about sharing data that involves both public and private entities, because their trust in the data being used appropriately has to extend beyond the original interaction. Overhage reiterated that the alignment of interests is especially important in these types of collaborations. He noted that while there is value in data, it is important to keep in mind the roles that domain knowledge, analytical expertise, and computational capabilities can play in partnerships.
The fact that data can be reused in ways that many other types of property cannot be raises additional ownership considerations, Overhage said. For example, analyzing the data in ways that someone else might not have analyzed them does not diminish the data’s usefulness to someone else. This is important to consider, particularly for data that resulted from an effort that was publicly funded.
In summary, Overhage listed five key enablers that he believes can make public-private partnerships work:
- Achieving clear alignment of stakeholder interests at the outset of the partnership;
- Establishing responsible data governance;
- Putting processes in place to ensure that the insights that are created are accurate, unbiased, and, where appropriate, explainable;
- Providing decision makers with the tools, processes, and support to act on the insights resulting from the work; and
- Ensuring the long-term economic sustainability of the partnership.
DISCUSSION
A central theme of the discussions that followed the presentations was the risks associated with data sharing. Speakers noted that resistance to transparency is common among all types of entities. For example, despite recent regulations focused on price transparency, it is still not possible to compare costs across health care providers.
Participants also discussed ways of overcoming resistance to data sharing. A potentially compelling argument could be that there is a wealth of information in the various health datasets, and the organizations that own them do not have the capability to take full advantage of the data. Even in organizations with thousands of analysts, the demand for using the data to answer questions far exceeds the capacity to produce answers. Collaborations with academic groups, other commercial entities, government organizations, and others are necessary to take full advantage of the data, despite short-term risks, but the conversation further highlighted the important role trust plays in establishing collaborations that involve data sharing.
Finally, the challenges associated with navigating the process of establishing agreements were discussed, along with the challenges of coordinating with Institutional Review Boards across multiple institutions. At the same time, participants underscored how collaborations, once they are established, can lower data access barriers for early career researchers.
CONCLUSION
Discussions in this session echoed some of the themes that emerged in prior sessions, including barriers to accessing data. While the benefits of data sharing are clear, the workshop also highlighted some of the reasons behind reluctance to share and underscored the risks involved for the organizations providing the data. Successful data sharing agreements can be established when these factors are taken into consideration.
CONCLUSION 5-1: Successful partnerships across health care systems require participant trust, clear evidence of mutual benefit, and the ability to control risk.
This page intentionally left blank.
