Demystifying Healthcare Data Governance: An Executive Report
The tribal form of data governance is characterized by the lack of a centralized data governance function as well as the lack of an EDW. Analytics and reporting functions are resourced from decentralized business and clinical units, resulting in significant resource inefficiencies and redundant, confusing reports. Without a centralized EDW, there is no ability to fuse data together in such a way that supports a better understanding of longitudinal care and population health. The value of data to the organization is much lower than it could be, if that data were used more synergistically.
The balanced, democratic governance model is characterized by a centralized data warehouse, and these data warehouses typically follow a Late-Binding™ data engineering architecture. This type of governance culture offers the ability to produce meaningful, actionable analytical insights across the continuum of care. A centralized data governance body exists and is staffed by executive leadership from across the organization (such as the COO, CFO, CIO, CMIO, CMO, and CNO). This group establishes overall principles of data management and analytics, then entrusts adherence to those principles to the data analysts that access the EDW. There is a cultural commitment to participate in data governance according to shared values, policies, and procedures. Significant emphasis is placed upon the protection of sensitive patient information, but with that exception, there is a high degree of tolerance on the part of the data governance committee for unconstrained data access and exploration. Access to the EDW is encouraged, as is cultural data literacy and data-driven decision making.
Too Little Data Governance: What Does It Look Like?
Too little data governance is marked by analytic inefficiency and uncoordinated data analysis resources. Analytic redundancy and expenses are high, in both human labor and technology. Multiple, redundant resources result in inconsistent analytic results when attempting to answer the same question. Poor data quality runs rampant. When there is a data quality problem, there is no formal body for appeal, and no process exists for fixing data quality problems. Thus, data quality problems tend to languish. Finally, there is an inability to respond to analytic use cases and requirements. Without a data governance function in place, when new analytic requirements emerge, there is no appropriate method for discussing and supporting the new requirements. The net business impact of these characteristics is lengthy and inaccurate decision cycles.
Too Much Data Governance: What Does It Look Like?
As expected in any culture characterized by highly centralized and authoritarian decision-making, too much data governance is also very inefficient. One of the vital sign indicators of a burdensome and overbearing data governance culture is an unhappy collection of data analysts and their internal customers. Approving and loading new data content into the EDW takes too long. Making changes to data models to support analytic use cases takes too long. Getting access to data takes too long. Resolving data quality problems takes too long. Developing new reports takes too long. In sum, everything takes too long. Interestingly, the consequences of too much data governance are the same as too little data governance—decision-making cycles are longer than they should be, the decisions are less accurate, and the mean time to improvement for the organization is longer than it should be.
The Triple Aim of Data Governance
Borrowing from the familiar IHI Triple Aim initiative for healthcare improvement, the Triple Aim of Data Governance is: (1) Ensuring data quality; (2) Building data literacy; and (3) Maximizing data exploitation for the organization’s benefit.
Ensuring data quality is the first step in a data governance mission. In this context, data quality is defined by the completeness of the data times its validity (Data Quality = Completeness x Validity). That is, collecting all the needed data for a particular analytic use case and ensuring that the data is valid. The data governance committee and function must have strategies to support and improve data quality—ensuring completeness and validity of the data to support analytics.
The second aim is building data literacy throughout the organization, and the data governance committee should champion this initiative. It makes no sense to build a library in an illiterate community. Similarly, it makes no sense to invest in the technology and data content of an EDW in an organization that suffers from a lack of data literacy. The data governance committee must sponsor training, education, and hiring practices that build the data literacy of the organization.
Finally, the third aim is data exploitation—maximizing the value of data to the organization, creating a data-driven culture that lowers costs, improves quality, and reduces risk. It’s not enough to support data quality and data literacy. Those attributes alone will not serve the betterment of the organization. That data and those skills must be put to good use by creating a culture that constantly seeks self-improvement through the spotlight of data.
Mindset, Skillset, and Toolset
Another useful three-part paradigm to guide the data governance committee is: mindset, skillset, and toolset, in that order of importance. The data governance committee must play an active executive leadership role in the development of a data-driven mindset throughout the organization. This is an important first initiative for the data governance committee—simply communicating from the executive level that the organization is, from this point forward, becoming a data-driven culture, constantly searching for ways to reduce their mean time to improvement. The next step is the development of the skillset among the employees to support this data-driven mindset. Finally, the data governance committee is the most logical choice for executive sponsorship of the toolset, such as an EDW, necessary to support the analytics journey.
The Data Governance Layers
As seen in the diagram, there are multiple layers in the data governance process, flowing down from the executive and board leadership, then the data governance committee, data stewards, data architects and programmers, the database and systems administrators, and finally the technical data access control system that surrounds the data warehouse and the analytics platform.
Combined effectively, these layers of data governance result in happy data analysts. At first glance, these layers might appear bureaucratic, but if implemented properly, the layers efficiently complement one another.
From the high-level guidance and aspirations of the executive and board layer down to the low-level, embedded technology of the information system supporting the EDW, every layer plays an important role in data governance. Below is a description of the typical functions and behaviors that occur in each layer.
Data Governance Role: Executive and Board Leadership
The executive and board leadership sets strategic goals for analytics, which in turn initiates an implementation strategy by the data governance committee. For example, leadership of a large ACO may decide that their healthcare organization needs a longitudinal analytic view across the ACO of patient treatments and costs, as well as all similar patients in the population served.
Data Governance Role: Data Governance Committee
The data governance committee’s job is to take that aspirational goal and translate it into analytic skillsets and toolsets. The committee evaluates options to achieve that goal. In this example, for the continuum-of-care-analysis and population health, a data governance committee would likely conclude the need for an EDW. They would also recognize that, in this data warehouse, the organization will need to consolidate all clinical and financial data associated with the beneficiaries in the ACO, as well as implement a master patient identifier and common coded terminologies to tie it all together.