10
CASE STUDY: THE SANFORD HEALTH SYSTEM AND SANFORD DATA COLLABORATIVE

Interviewees: Benson Hsu, MD, MBA, VP, Enterprise Data and Analytics (former); Emily Griese, PhD, Vice President of Population Health and Clinical Operations; and Arielle Selya, PhD, Director of Data Exchange Core

ABSTRACT

The Sanford Data Collaborative (SDC) is an enterprise-wide initiative of Sanford Health System, created in 2015, and designed to unify the data sources and data analytics capabilities of the organization in support of improving population health outcomes. The SDC responded to unmet needs of the health system and now provides an important resource for the research community at the regional and national levels. SDC developers recognized that each data element in its electronic health record (EHR) was important, singularly and in combination, for helping Sanford Health System provide higher quality health care. Building this complex resource from the ground up entailed garnering collaboration at the executive level, engaging compliance and legal and personnel at the outset, and aligning it with health system priorities—including the movement to value-based care. The architects of the SDC recognized that providing access to both internal and external researchers would support their ability to maximize the utility of the data for the good of patients and would also meet demand in the research community for high-quality health data (Griese et al., 2017).

BACKGROUND

The SDC is a dynamic data resource for researchers at the Sanford Health System as well as the academic research community outside of Sanford. The SDC integrates data from the EHR with other health system administrative data to create a secure enterprise-level data warehouse, consistent data dictionary, and data governance processes. It was established as Sanford Health consolidated its central functions after a series of mergers over multiple years, resulting in numerous data sources and silos. Creating a unified and coherent data infrastructure, including a consistent data dic-

tionary and data governance processes yielded a “single source of truth,” especially important for a health system that is geographically spread throughout a large rural region of the U.S. Consolidating the data also enabled the system to bolster and create coherence in its analytic capabilities, evolving in sophistication from descriptive analytics to predictive analytics (Hsu et al., 2017). Alongside the data resource itself, the team built the necessary processes and resources that would support its functions, including a privacy board, outreach function to promote use, and on-demand analytics tools (Griese et al., 2017).

The on-demand tools were designed to support both internal and external research. The interviewees observed that, prior to the creation of the SDC, data requests and queries submitted by a research analytic team were often lower priority, with operational analytic requests taking precedence. These query requests also could be incorrectly translated by the business analyst running a research query, or incompletely formed based on the data available in the EHR. To this end, by leveraging and augmenting the built-in analytic capabilities of the EHR, anyone in Sanford Health can now run queries on the aggregated data. External researchers submit queries to a project manager. For approved requests, data are shared via a secure file transfer protocol. Such queries can help determine the feasibility of a potential research study and answer straightforward questions about such things as prescribing trends or disease incidence, for example.

The first generation of the SDC was developed for use with aggregated data. Using it for studies that need person-level data is comparatively more complex, and the SDC team has turned its attention to developing appropriate processes and audit trails, and working through attendant ethical, regulatory, and privacy-related aspects. To support this work and sustain leadership buy-in, the group seeks to address research questions that are also of interest to the health system. This alignment with the needs of the organization is integral.

DESCRIPTION

Early work with the organization’s legal department smoothed the way for the eventual success of the SDC. With the legal team, SDC leaders were able to work through challenging data use agreement (DUA) questions and develop policies about aggregation and de-identification. An insight gleaned from devising legal and governance processes that can be useful to others is the consideration of who owns the knowledge resulting from a given project or analysis? This is often framed through the lens of data ownership, but knowledge ownership is equally important to consider for any health system aiming to create a similar data utility. The SDC director reviews data requests for alignment and appropriateness now that an established DUA template is in place.

A critical early challenge was to garner support at the highest level of health system leadership. This was achieved by helping leaders see the possibilities in sharing the data and urging that it be leveraged to its fullest potential. The axiom in informatics that “data gets better with use” was on the minds of the creators (who were research and data analytics leaders in the organization). One of the interviewees for this case study observed, “every single touchpoint provides information about our patients and populations. If we just let it sit there, we are not fulfilling part of our responsibility as a health system.” The team actively cultivated engagement from across the enterprise, ensuring that the SDC could be analyzed by business personnel, as well as research and operations teams. The SDC has been used by engineers, public health researchers, and pharmacy personnel, each of which brings a different lens and questions of interest.

An important lever used to gain organizational buy-in was related to priming the organization for future health care transformation efforts, particularly value-based care. Preparing the data infrastructure to support system-wide priorities and bolstering research capabilities were two of many benefits, and deeper understanding of the data signaled to their community that they are careful stewards of important health information. A corollary benefit has manifested from this concerted work, in that researchers

regard this resource as a signal of Sanford Health’s commitment to the community.

Significant attention was devoted to engaging and educating researchers and leaders of academic institutions in the region. Helping researchers understand the complexity of a seemingly straightforward query has helped garner bidirectional trust and has helped the SDC leaders forge meaningful connections between area researchers and service line leaders in the health system. This cycle helps improve the quality and relevance of the research and amplify the visibility of the SDC as a data resource. An important facilitator, then, is the role of a “boundary spanner,” that is, someone who understands both research and care delivery, and can serve as a translator as well as an ambassador in the organization and externally.

As described above, the SDC offers unique value to Sanford Health System, its internal researchers, the broader research community, and the patients receiving care there. Each query offers the implicit opportunity to improve on the data resource and assure that processes are meeting the needs of all stakeholders—from the researchers to the leaders to the institutional review board and compliance personnel. A consolidated and curated data warehouse can accelerate research along the continuum from idea to intervention to eventual implementation of insights.

FUTURE DIRECTIONS

The SDC team advised that others initiating a comprehensive data resource include their legal, compliance, and quality improvement leadership at the outset, along with research and clinical leaders. Despite early skepticism, the team urged a “just do it” mindset, addressing barriers as they arose, as the end result is worth the effort. One case study interviewee rebutted the notion that sharing or releasing data could diminish a system’s competitive advantage and averred that it actually enhances it.

A second dimension of the advice for others pertained to technical facets of building the data collaborative. Interviewees observed that the inconsistency in health care data and analogous inconsistency in outcome measures creates complexity and frustration.

A unifying definition across payers and providers would facilitate measurement, whereas the current idiosyncrasies from one health system to the other creates friction in measurement, data sharing, and population-level care improvement. The advice is to go into this endeavor clear-eyed about these inconsistencies and from a data governance standpoint, aiming for steps that enable data integration, whether a common data model, use of standards, or a data governance apparatus to reconcile differences across data sources.

In short, the experience of the Sanford Health System and SDC illuminate a path so that other health systems undertaking similar endeavors do not have to “go it alone.”