Why the Data Steward’s Role Is Critical to Sustained Outcomes Improvement in Healthcare

In the field of healthcare analytics, data architects, managers, and analysts are all immediate family members. We know them by name, we know their habits, and we know their strengths and weaknesses. By comparison, the data steward is a distant, perhaps underappreciated, relative who generously supports the rest of the family. In reality, this position is as vital as a multilingual translator in the business of international diplomacy. Without her, a communications chasm exists between clinical caregivers and information technicians on the path to sustained outcomes improvement. Let’s examine the position of the data steward on the family tree, the role she plays in data governance for a healthcare system, and what personality traits (i.e., clinical and technical experience) help her to become an integrated member of the clan.

What Makes a Good Data Steward?

Stewards are not direct owners, but appointed guardians or caretakers. They are responsible for the data within their domain. An ideal data steward really understands the data. The data architect handles the ETL and pulls the data into the Enterprise Data Warehouse (EDW), but may not fully understand what he is pulling in or managing. The ideal data steward comes from the business side and can talk the business language (clinical, finance, operational, etc.). She can help with data validation, check data quality, and make sure it’s complete, accurate, and consistent. Also, because of her knowledge, she’s able to jump in and own user training and bring in the necessary SMEs to help with that knowledge dissemination. This takes a lot off the plate of the analytics vendor (i.e., saves money).

What it boils down to is that someone knows a lot about a particular system. Hypothetically, she is accustomed to answering a lot of questions because everyone goes to her for answers. If /when she doesn’t have answers, she knows who to consult to find those answers. She fills this role almost by default and she enjoys it because she is knowledgeable about it.

At Health Catalyst, when one of our partner health systems starts working on analytics applications, it’s crucial for us to have an expert on the business side. She might be on the EDW team, but she is more clinical than technical. It’s a very valuable role because she can talk clinical jargon, help find data, and verify data accuracy.

In our experience, the data steward function has been somewhat fuzzy because it’s not fully understood. Also, data-related initiatives can be very IT focused, with projects started and managed purely by the IT team. An emphasis on the technical, rather than the clinical, aspect of these projects, is also reflected in the experience of the staff working on these projects. This usually results in a struggle for clinical end-user involvement, so the project eventually fails because nobody is using it.

Data management may not be a data steward’s primary role. She will retain her primary job and her responsibilities around that job. Stewardship may be only five percent of what she does around the hospital. This number can change depending on the level of maturity around data use around the hospital. One hospital we work with has five data stewards right now, but as they mature, they might have 20 stewards a couple of years from now. Hospitals will need some consistency in standardizing this role, but they are moving in the right direction.

An Inside Look at One Data Steward Role

Danna Caillouet fills the role of data steward at Thibodaux Regional Medical Center, a patient-experience-award-winning hospital that’s about 60 miles southwest of New Orleans. She’s been there since 1989, and worked for nine years as the critical care director, managing four nursing units with 50 direct reports. She is a Six Sigma Green Belt, a Lean facilitator, and she ran the Evidence-Based Practice Council for Nursing at Thibodaux. But one day she was asked to take on the care transformation work at the hospital, her title changed to Physician Analyst, and now she fills the role of data steward.

She first started working with Health Catalyst on two care transformation projects, blood culture and wound culture contamination rates. She explains the response of her constituents as they began to work in this new realm of data stewardship:

“At that point everybody had heard of data warehousing, but nobody really understood the ramifications of the data we were going to have at our fingertips. Health Catalyst put together some things for us so we could monitor our blood and urine cultures by nurse, by unit, by date, and by contamination. Then we sat down with our “customers,” handed them this new app, and all the information that used to take hours and hours to organize was suddenly available. To see the looks on their faces was priceless. I used to be a nurse manager and I know what it was like to have to merge 20 spreadsheets. It’s very rewarding to be able to give the directors in the hospital the tools they need to take care of their patients.”

For Danna, if care transformation isn’t successful or if it doesn’t implement on time, then administration knocks on her door, so her role carries a lot of responsibility and accountability.

Stewards Help Users Understand Information in a Clinical Context

When a non-clinician is looking at the data, he will pull in the data steward and ask her to walk through when the data is being entered, where it’s being entered, who is entering it, and why the data looks like this versus why it looks like something else when its being entered somewhere else. The steward is usually the one to answer all these questions. In our experience, every organization has these domain-rich, business-experienced individuals, even if, at first blush, they can’t identify them. The person in this role at one of our health systems calls herself the data translator. She attends the meetings with us and then she helps the clinical people understand why certain things looks the way they do in the application. And then she helps us to understand the clinical workflow so we can develop properly. She works back and forth between the clinical and analytics teams and she works with the data architects to make sure data is flowing properly into the EDW.

Finding the Data Steward Diamond in Your Rough

Rarely does a healthcare system already have this specific data steward role. Mostly, the function is identified, but it comes via a data architect, system analyst, or a system source manager. Some organizations will specifically hire a data steward, but others will adapt another role from the internal ranks.

One of the things we’ve noticed with our health systems is how the data steward can hide in plain sight. This is not a new person brought in from the outside. She is already embedded in the system, but she has another role. She is often a business user who has expert knowledge of a certain business process and she knows how data flows through a specific system to impact care. She is recognized by her colleagues as the go-to expert, so it just becomes a matter of deputizing her or officially calling her out as functioning as the data steward.

Data sources and their potential data stewards.

Successes in the Data Steward Role

When Thibodaux Regional Medical Center did a sepsis improvement initiative, it tapped Danna’s experience. She was already familiar with the lab work, which is a key component for identifying septic patients. She was not a net new person and already had this background. She was familiar with the workflow, with the doctors who touch this process, and she had the respect of this group. She was already the go-to person. Like Danna, this person could be a nurse on the floor or an impassioned doctor who wants to really focus that passion around whatever domain is in question.

Another example is a health system in the eastern United States with 17 hospitals and over 3,000 beds that converted a clinician, a registered nurse, who wanted to move beyond clinical practice. She moved over to the EDW team and, although she doesn’t see herself as a data steward at this point, she’s filling the role nicely. She has filled the gaps for Health Catalyst in terms of working with the clinical teams on some advanced patient safety apps, as well as helping with translation and validation. She’s stepped into this role by default and helped fill that data steward need. All we have to do is help her balance her project manager focus with a data steward focus.

In yet another example, a few years ago, we worked with a west coast health system to setup a community coordinated care project. Part of the work was to build a cross-functional team to take on medically complex patients and manage their care through this team. Being able to understand the data, knowing where it came from, and knowing how the data would be used was very important, so having a data steward there was critical. Not only did the team have physician leads, registered nurses, and data architects, but the knowledge provided by the data steward was essential to the success of the project.

Challenges with Assigning the Data Steward

The most pressing challenge for the data steward position is that because everybody wants a piece of her, she is spread really thin. She is accessed by everybody at the hospital so it’s hard to get her time. This is symptomatic across all of our health systems. These individuals are very busy because they are one of the few people who have vast knowledge of the domain landscape.

When assigning roles, the healthcare improvement team will go through each role and ask, who is the physician lead? Okay, it’s Dr. Smith. Who is the data manager? Well, it’s this nurse who has some analytics background. But when they get to role of data steward, that’s when they pause and ask what this person does. Then they say, for this project, we’ll bring in this person on an as-needed basis because she is so busy.

Data Stewardship as Part of Data Governance

In the overall data governance model, senior leadership is there to empower, the data governance committee is there to coordinate, and the data steward is there to inform and optimize. Additionally, she:

  1. Reinforces the global data governance principles. She would be familiar with these, espousing and evangelizing them.
  2. Helps develop and refine details of local data governance practices. Whatever is established at a system level, she makes sure it’s done at a local level.
  3. Is the eyes and ears of the organization with respect to data governance and the governance committee.
  4. Provides direction to peers, architects, the data governance committee, and data analysts regarding appropriate data definitions, usage, and access.
  5. Anticipates local consequences of global changes proposed by the data governance committee (e.g., if the committee decides to change the definition of LOS, at a central level what impact does that have at local grassroots level?).
  6. Addresses local data issues and escalates them to the data governance committee when necessary. For example, with sepsis care, there may be multiple ways to define LOS. The data steward could argue that this needs to be standardized across the system or that it needs to be unique for this particular domain and argue for an exception.
  7. Identifies data literacy gaps and is able to escalate those when appropriate. If someone doesn’t understand guiding principles or data use policy or certain definitions, she would be there to look at those gaps and help fill them.
  8. Analyzes and resolves data accuracy and integrity issues.

The data steward is there to establish trust with the end users of the data. If the trust is absent, the data remains dormant. The steward’s role is crucial to identifying dirty data, fixing it with the cross functional team, and then reestablishing the trust and continuing use of the data. In essence, the data steward’s role is critical to becoming a data-driven organization.

A steward should never reside exclusively in IT while playing the roles around data quality, access, and utilization. This is a common mistake that we see, which creates a problem where the business is isolated from the data. We’ve seen the most success when the steward is embedded in the business, represents the business viewpoint, and can help with education around access and use of data.

Tools of the Trade

A great data steward has the ability to manage and maintain the visualizations of the data under her management. If she is the sepsis data steward, like with our example of Danna at Thibodaux, then she has ownership of that app, she is granting access, and she is helping to train. Perhaps unique to the industry, Health Catalyst has a clinical hierarchy that they rolled out to their health systems to help scale this and other roles. For the example of sepsis, the data steward might take on other care processes around infectious disease, which has proven successful time and again.

The data steward needs access and ownership of Atlas, the freestanding, web-based metadata tool developed by Health Catalyst that connects end users with data. It’s like an index card for a library to see where everything is located within the data warehouse. So the data steward who is helping to answer questions on CLABSI or sepsis, is the same person giving access to Atlas to define LOS for the population of CLABSI.

In addition to the clinical skills and background mentioned before, the data steward needs to be knowledgeable with her data and how to navigate with it on the backend, so the technical expertise needs to be there, as well.

Prepping Your Organization for Data Stewardship

When Health Catalyst does an assessment, our health systems will grade themselves on their maturity by answering many different questions and, without fail, there’s a gap between the ideal state and where they currently are. There may be a little bit of recognition and understanding that they aren’t as advanced as they think they are. That’s just human nature.

Health Catalyst conducts assessments for both existing and prospective health systems around three systems: Analytics (where we get a feel for data stewardship), Best Practices, and Adoption. We do comprehensive assessments where we interview a lot of folks and ask a lot of questions to assess the organization’s overall maturity level (not just with data stewardship). Ultimately, what we’re trying to answer is whether or not the organization is able to transform care. If not, we offer solutions for improvement in each of the three systems.


If you were to ask your EDW team, they might have a totally different idea of what they want for their data stewardship environment. Some organizations are so IT focused in building the EDW, creating a self-service model, and generating reports, that they aren’t quite at the outcomes improvement level where they should be. Their perception of a data steward in the IT department is someone who owns the actual data movement instead of someone who is the knowledge expert and overall manager of a particular care domain.

Data stewardship is really very important, even though it’s very hit-and-miss right now. When this cog is missing in the very complex machinery that defines outcomes improvement projects—data quality, governance, access, navigation, visualizations, to name a few processes—the output required for improvement fails to materialize. Data stewardship is not standardized among our health systems. We have yet to see the principles executed perfectly, but any company hoping to evangelize data stewardship should recognize that this is a journey. It’s not just about pulling people in and moving things around in an inefficient way. It’s about recognizing where you are along the journey and understanding the vital and numerous responsibilities of data stewards.

Powerpoint Slides

Would you like to use or share these concepts? Download this presentation highlighting the key main points.

Click Here to Download Slides

Loading next article...