Data. Governance. Two words. Fourteen letters. Loaded with history and pain.
The label “data governance” carries with it a lot of baggage in healthcare organizations. Most agree that healthcare data is poorly understood and inadequately documented, and that the resulting reporting and analysis is unreliable and can’t be trusted. This causes frustration among managers and executive decision-makers. Instead of being viewed as a valued asset, data becomes just one more input into ad hoc decision-making based on gut instinct and historical precedent.
Most healthcare organizations have attempted to establish data governance programs at least once in the past and have failed to meet, much less surpass, already low expectations. The common pattern for these efforts often begins with a top-down, executive-driven edict to once-and-for-all get data right.
These programs start with great fanfare. They quickly establish a rigorous hierarchical steering committee structure with senior leaders participating. Regular meetings are held sharing current state, challenges, and recommendations on how to proceed. But pretty soon, key steering committee members start missing more and more meetings because they never truly bought into their role and felt accountable, instead focusing their energies on other critical priorities. Because no one has ultimate accountability, the data governance effort inevitably achieves little and the committees fade away after nine to 18 months or so.
So many of these efforts struggled because their advocates and champions—with the best of intentions—believed that centralized, top-down data governance was required to tell people in the organization what their data meant and how it should be used. Another contributing factor is a boil-the-ocean approach, focusing the data governance effort on a broad array of data domains, paired with an understandable but flawed belief that each data term must only have a single definition adopted across the enterprise.
These three contributing factors—top-down, one-directional data governance; a govern-everything approach; and an expectation that single data definitions can support the entirety of your business—are the most significant stumbling blocks to establishing effective data governance. Here are three steps you can take to avoid them and follow a high-value, low-risk path to success:
View data governance as a collaborative endeavor. Every one of your employees and key stakeholders creates, manipulates, and consumes data, and they have very clear insights and an understanding of its value to them in supporting their processes and decisions. They know the limitations of their data and the assumptions that underlie their analysis and have spent countless hours mitigating data quality issues. They have resolved the ambiguities that inevitably occur when data is used for different reporting and analytic purposes than was originally intended and are intimately familiar with regulatory metrics, as well as the filtering and aggregating that are required to meet reporting requirements.
These people are your data heroes. They certainly do not need to be told what their data means. Rather, the data governance program needs to engage these experts to share and document their knowledge. You need to solicit their help to capture the high priority business processes and decisions that the data supports for the benefit of the extended enterprise in a manner that raises the data literacy of the entire organization.
Embrace “just enough” data governance. This approach suggests that you govern only the data you are using, and only to the extent you are using it. Only once you have delivered the anticipated business value from the scope of that initial data governance effort will you have the buy-in, sponsorship, and support to build the momentum for your next areas of focus.
The path to this initial value is to focus on governing the data required to solve a business or clinical use-case. This use case may be an executive dashboard for a command center, a financial model for forecasting revenue, or a predictive model for risk of readmission.
Focusing on a targeted use case has several advantages. First, it allows you to concisely scope the data and terms that need to be governed to support a business or clinical use case that will deliver measurable value. Second, it provides context to the extent that the data will need to be governed. For example, if a report that you’re targeting to deliver for this initiative groups data by month and year, we don’t need to govern the data to minutes and hours. (And trust me, there will be impassioned arguments from people who want to govern the data to minutes and hours—and milliseconds!—based upon notional use cases that don’t exist, or only exist as a rare exception.) Expending effort to govern data that is not being used is a wasted effort that will derail the focus on your stated objective.
Old school data governance looked to categorize enterprise data into domains (billing data, clinical data, patient administrative data, etc.) and then set up a multi-year project plan to attack each domain in sequence. This approach was doomed from the beginning since it focused on the data rather than how (and why) the data was being used: Few, if any, business or clinical reporting or analytics questions are solved using a single domain of data (e.g., ask yourself how often your providers look at patient administrative data WITHOUT clinical data, and vice versa?); and much of the data in any individual domain is seldom if ever used. Don’t apply enormous effort and resources to try to govern an individual domain in its entirety if it’s never going to deliver any direct, measurable value.
Accept that there are multiple definitions of a term. While this may seem counterintuitive, the goal of data governance isn’t to come up with a single agreed-upon definition for every term used in the enterprise. Rather, the goal is to understand all of the appropriate definitions of a term, based on the context of how that data is being used. You should then capture these variations and ensure there is a consistent, standardized way to discover and analyze the right version to deliver the right insights in the right context.
A favorite example here is length-of-stay, where there are legitimately a dozen or more appropriate and useful definitions depending on the use of the term—regulatory reporting, quality of care, specific reimbursement rules, etc. What is important is to accurately describe each use, document the definition of the term, and then assign the term a unique name. In this manner, it is clear and unambiguous what the data means when it is used, thereby avoiding the problem of the same term (and label) being used for multiple definitions.
A modern data governance approach avoids the mistakes of the past and sets the enterprise on a fast path to value. Although you’ve heard it before, the concept of think big but start small is an important mantra to embrace since enterprise data governance is a critical program enabling digital transformation of the enterprise. Starting small and focusing on delivering value will concentrate resources and attention on delivering real, short-term business value, thereby building necessary momentum.
But thinking big is absolutely required. Enterprise data governance is critical to delivering the reliable and trustworthy data essential to the desperately needed digital transformation of healthcare. But like all big, hairy, audacious goals, they need to be achieved by building on success. This success is achieved by thinking small and solving meaningful business and clinical use cases that delivery measurable value.
To learn how one healthcare organization built a sustainable data governance program, watch this on-demand webinar, “Navigating L.A. Care Health Plan’s Data Governance Odyssey“