There is a wealth of data within the healthcare industry that can be used to drive innovation, direct care, change the way systems function, and create solutions to improve patient outcomes. But with all this information coming in from multiple unique sources that all have their own ways of doing things, ensuring data quality is more important than ever.
The COVID-19 pandemic highlighted breakthroughs in data sharing and interoperability advances in the past few years. However, that does not mean that there aren’t challenges when it comes to data quality.
“As we have seen, many organizations have created so many amazing solutions around data,” Mujeeb Basit, MD, associate chief medical informatics officer and associate director, Clinical Informatics Center, University of Texas Southwestern Medical Center said. “COVID really highlighted the innovations and what you can do with sophisticated data architectures and how that flow of data really helps us understand what's happening in our communities. Data has become even more important.”
Dr. Basit shared some of his organization’s experiences in creating strategies to improve data quality while making the process as seamless as possible for all stakeholders.
The medical center had four groups working together on solution co-development, including quality, clinical operations, information resources and analytics.
“It is the synergy of working together and aligning our goals that really helps us develop singular data pipelines as well as workflows and outcomes that we're all vested in,” Dr. Basit said.
One of the problems the organization previously faced was that errors would slowly accumulate in their systems because of complicated processes or frequent updates. When an error was found, Dr. Basit noted it was usually fixed as a single entity, and sometimes a backlog is fixed.
“But what happens is, over time, this error rate redevelops. How do we take this knowledge gained in this reported error event and then make that a sustainable solution long term? And this becomes exceedingly hard because that relationship may be across multiple systems,” Dr. Basit said.
He shared an example of how this had happened while adding procedures into their system that become charges, which then get translated into claim files.
“But if that charge isn't appropriately flagged, we actually don't get that,” Dr. Basit said. “This is missing a rate and missing a charge, and therefore, we will not get revenue associated with it. So, we need to make sure that this flag is appropriately set and this code is appropriately captured.”
His team created a workaround for this data quality issue where they will use a user story in their development environment and fix the error, but this is just a band-aid solution to the problem.
“As additional analysts are hired, they may not know this requirement, and errors can reoccur. So how do you solve this globally and sustain that solution over time? And for us, the outcome is significantly lost work, lost reimbursement, as well as denials, and this is just unnecessary work that is creating a downstream problem for us,” Dr. Basit said.
Their solution? Apply analysis at regular intervals to keep error rates low.
“This is not sustainable by applying people to it, but it is by applying technology to it. We approach it as an early detection problem. No repeat failures, automate it so we don't have to apply additional resources for it, and therefore, it scales very, very well, as well as reduced time to resolution, and it is a trackable solution for us,” Dr. Basit said.
To accomplish this, they utilized a framework for integrated tests (FIT) and built a SQL server solution that intermittently runs to look for new errors. When one is found, a message is sent to an analyst to determine a solution.
“We have two types of automated testing. You have reactive where someone identifies the problem and puts in the error for a solution, and we have preventative,” Dr. Basit said.
The outcome of this solution means they are saving time and money—something the leadership within the University of Texas Southwestern Medical Center has taken notice of. They are now requesting FIT tests to ensure errors do not reoccur.
“This has now become a part of their vocabulary as we have a culture of data-driven approaches and quality,” Dr. Basit said.
Another challenge they faced was streamlining different types of information coming in through places like the patient portal and EHR while maintaining data quality.
“You can't guarantee 100% consistency in a real-time capture system. They would require a lot of guardrails in order to do that, and the clinicians will probably get enormously frustrated,” Dr. Basit said. “So we go for reasonable accuracy of the data. And then we leverage our existing technologies to drive this.”
He used an example from his organization about a rheumatology assessment to determine the day-to-day life of someone with the condition. They use a patient questionnaire to create a system scoring system, and providers also conduct an assessment.
“Those two data elements get linked together during the visit so that we can then get greater insight on it. From that, we're able to use alerting mechanisms to drive greater responsiveness to the patient,” Dr. Basit said.
Conducting this data quality technology at scale was a challenge, but Dr. Basit and his colleagues utilized the Agile methodology to help.
“We didn't have sufficient staff to complete our backlog. What would happen is somebody would propose a problem, and by the time we finally got to solve it, they'd not be interested anymore, or that faculty member has left, or that problem is no longer an issue, and we have failed our population,” Dr. Basit said. “So for us, success is really how quickly can we get that solution implemented, and how many people will actually use it, and how many patients will it actually benefit. And this is a pretty large goal.”
The Agile methodology focused on:
They began backlog sprint planning, doing two-week sprints at a time.
“We want to be able to demonstrate that we're able to drive value and correct those problems that we talked about earlier in a very rapid framework. The key to that is really this user story, the lightweight requirement gathering to improve our workflow,” Dr. Basit said. “So you really want to focus as a somebody, and put yourself in the role of the user who's having this problem.”
An example of this would be a rheumatologist wanting to know if their patient is not on a disease-modifying anti-rheumatic drug (DMARD) so that their patient can receive optimal therapy for their rheumatoid arthritis.
“This is really great for us, and what we do is we take this user story and we digest it. And especially the key part here is everything that comes out for the ‘so that,’ and that really tells us what our success measures are for this project. This should only take an hour or two, but it tells so much information about what we want to do,” Dr. Basit said.
Acceptance criteria they look for include:
“And we try to really stick to this, and that has driven us to success in terms of leveraging our data quality and improving our overall workflow as much as possible,” Dr. Basit said.
With the rheumatology project, they were able to reveal that increased compliance to DMARD showed an increase in low acuity disease and a decrease in high acuity.
“That's what we really want to go for. These are small changes but could be quite significant to those people's lives who it impacted,” Dr. Basit said.
In the end, the systems he and his team have created high-value solutions that clinicians and executives at their medical center use often.
“And over time we have built a culture where data comes first. People always ask, ‘What does the data say?’ Instead of sitting and wasting time on speculating on that solution,” Dr. Basit said.
The views and opinions expressed in this content or by commenters are those of the author and do not necessarily reflect the official policy or position of HIMSS or its affiliates.
December 14–15, 2021 | Boston & Digital
Machine learning and AI are full of possibilities to address some of healthcare’s biggest challenges. Learn how leading healthcare organizations have leveraged the power of machine learning and AI to improve patient care and where they see real ROI—better care, cost containment, and operational improvements and efficiencies.