Demystifying Healthcare Data Governance (Executive Report)
As the age of analytics emerges in healthcare, health system executives find themselves increasingly challenged to define a data governance strategy that maximizes healthcare data’s value to the mission of their organizations. “Data is the new oil,” Andreas Weigend, former chief scientist at Amazon.com recently said. That statement might be a bit dramatic—data can’t heat homes or power cars— but there is no denying the dramatically growing importance of data to business success. This is especially true for those businesses based in knowledge management delivery, which is certainly the case with healthcare.
The rising hype of the data warehouse and analytics market has resulted in significant noise from vendors and consultants who promise to help health systems develop their data governance strategy. The purpose of this paper is to share my observations and lessons learned in data governance, as a CIO, analytics specialist, and vendor; and in so doing, remove the mystery and confusion about how and where to start the data governance journey in healthcare and provide a roadmap to evolve that journey incrementally over the lifespan of the organization.
An Elusive Target: The Perfect Data Governance Environment
It is difficult to find a perfect data governance environment. Over the course of my career, starting in the early 1980s when I was a young Air Force information systems officer, I’ve seen attempts at data governance swing between too much and too little, rarely finding equilibrium. Until the early 1990s, the term data governance wasn’t commonly used, and the body of knowledge around data governance was almost non-existent.
The chart depicts a small but significant sampling of projects I was involved in and the initial state of the data governance for each. While each project has a unique story of data governance, there are common themes and patterns that emerge as indicators of success and failure. The point of this paper is to expose those common patterns.
The Maintenance Management Information and Control System (MMICS, 1987) is an Air Force-based system that can be thought of as an EMR for aircraft. Continuing that analogy, Air Force crew chiefs function as the primary care physicians for their assigned aircraft. These crew chiefs are responsible for maintaining and optimizing the health of their aircraft while at the same time, consuming as few supplies and resources as possible. The development of MMICS was prompted by a 1983 GAO report criticizing the Air Force for its poor ability to accurately track and manage the health and maintenance costs for individual aircraft, as well as entire fleets of aircraft. It is easy to see the parallels between patient care and population health management.
In my career, MMICS still stands out as the only project that achieved a perfect balance of user interface, data collection, and analytic governance, from its origins. MMICS was designed from the back, forward. That is, the Air Force first decided what type of data was needed to be collected to effectively manage the health of its vast variety of aircraft and other assets. Then they designed the MMICS user interface to support the collection of that data, as well as the workflow efficiency of the crew chiefs using the system. The efficiency of user interface and data collection was a very important attribute because those crew chiefs had to operate in less-than-ideal conditions on active flight lines in scorching heat and subzero cold. The overriding design motives were analytic output and workflow efficiency.
It is not hard to imagine the benefits of designing an EMR with those same two design motives: analytics and clinician workflow efficiency.
The Sanders Philosophy of Data Governance
My philosophy of data governance is to be as lean as possible—govern to the least extent necessary in order to achieve the greatest common good. Healthcare organizations tend to govern too much too soon, which results in unnecessary constraints on data and wasted labor. These health systems govern data that, in practice, doesn’t need any governance yet. Pairing the data governance function with overseeing the development and evolution of an enterprise data warehouse (EDW) gives the data governance committee something tangible to govern. A Late-Binding™ data engineering architecture and an EDW, combined with lean data governance, work very well together. Bind no data before its time, and govern no data before its time.
Data Governance Cultures
Data governance tends to mirror societal governance in the three categories shown in the diagram: authoritarian, democratic, and tribal.
Authoritarian data governance cultures are typically associated with a centralized and closely managed EDW. These data warehouses are usually designed around a monolithic, early-binding data model, and the data governance culture tends to believe that most, if not all, analytic use cases are well known and persistently agreed upon. Access to the EDW is tightly controlled through a top-down bureaucratic approval process, and there is very little—if any—tolerance for unconstrained and unsupervised data exploration within the EDW. In healthcare, the dominant users of the data warehouse in an authoritarian governance culture are typically from finance. Occasionally, users are data aggregators in the organization responsible for submitting reports to external agencies, but they are not data analysts who play an active role in process improvement or cost reduction.