Crowd Sourcing for Healthcare Innovation
By Benson Hsu, VP-Data and Analytics, Sanford Health
Healthcare delivery as a whole, with all its technological advances in patient care, has seemingly been inattentive in applying advanced analytics onto the endless amounts of data being collected.
Historically, one could argue that healthcare data was difficult to access. It resided within paper charts with unintelligible provider handwriting. For instance, laboratory results were often printed, faxed, hand copied, and stapled into these same charts. During my medical training, charts at times even had scrapes of paper with reminders and notes tucked in for quick reference during office visits. Culling any useful population analysis from this haphazard process proved to be nearly impossible.
But it wasn’t without trying. Academic centers and countless researchers have devoted thousands upon thousands of hours abstracting information from charts. The focus of these research projects often conformed to a data definitions standard that was researcher or project specific. Thus, it rarely had an impact at the population level.
“Data and analytics, through a collaborative approach with our academic partners, has given us another vehicle to better care for our communities’’
At Sanford Health, a $4+ billion integrated healthcare delivery system in the upper Midwest, our goal in analytics is not only to improve our internal operational and clinical care processes but to, more importantly, improve the health of the population and communities we serve.
Please don’t get me wrong, Sanford Health has moved to leverage data and analytics in the traditional way. This foundation is the basis for our analytics innovation.
Our analytics team, named Enterprise Data and Analytics, has brought together functional and operational expertise that were traditionally siloed across our six-state geographic footprint. In doing so, we have centralized and standardized the approach to data, from the request process to the data and analytic output. This was a vital first step given our large rural geography.
Additionally, we have established a resilient data warehouse plan. Contrary to our history of having multiple data silos, whether clinical, operational, or financial, we have intentionally brought data together through virtualization. This foundation lends itself to not just accurate representation of data but outputs that are consistent and reproducible, regardless of the end user or the analyst. This platform also allows us to anticipate future data and analytics needs, crucial in the era of fast moving data from personal health devices and social media feeds.
Lastly, we have created a robust data governance structure. In doing so, we now have a common language across our four major regional centers and 40+ critical access hospitals. Strategic decisions that cross multiple function and geographic regions are being made using data aligned and understood across our whole organization. In the same vein, we’ve established benchmarking standards so that comparisons are also consistent throughout our organization, a contrast to past days when everyone brought their own comparative measures.
Despite these rapid advances in our data and analytics capabilities, we realized we were always thirsty for more analytic talent. One approach is to continuously hire data scientists into our team. With the clear limitation in the supply of data scientists coupled with a natural organizational limitation on continuous expansion, Sanford Health solved this problem by looking external towards the vibrant academic communities in the region.
Luckily, Sanford Health is situated in a region with multiple data science programs, some within the Department of Mathematics (South Dakota State University) some through Schools of Business (Beacom School of Business, University of South Dakota), some through the School of Medicine (University of North Dakota), and some through informatics and computer science (Dakota State University.)
Taking a page from other organization’s approaches in crowd sourcing including Nextflix’s “Netflix Prize” for the best predictive algorithm, we established the Sanford Data Collaborative. The Sanford Data Collaborative, housed within our research center, looked to pave the way in data sharing by offering regional researchers access to timely healthcare data with the ultimate goal of improving healthcare delivery.
To do so, the Collaborative functions through a four-step process. Step one requires our organization to identify high priority issues that require advanced analytics for solutions. For instance, can an innovative algorithm-based solution help identify patients at risk for lack of engagement utilizing existing data, such as clinical history? Step two is to partner with local institutions with data expertise through a Request for Proposal (RFP) process. Step three is to leverage real life data through these collaborations to develop innovative solutions to our healthcare priorities. Step four is the implementation and validation of the analytic solution through an A/B test environment to improve algorithm development.
Although housed within research, Collaborative is far from being an academic exercise. At its core, the Collaborative strives to inform innovation in the delivery of care. Sanford Health understands that our responsibility is to continuously improve the way we care for our communities, within and outside of our four walls.
We are in the first year of this collaborative and already, we have seen the fruits of this labor. Proposals have taken revolutionary approaches to analytics not before seen in healthcare analytics, swiped from the ivory towers of academia and experiences from other data heavy industries such as finance.
Sanford Health is dedicated to improving the human condition. Data and analytics, through a collaborative approach with our academic partners, has given us another vehicle to better care for our communities.