Data quality issues are nothing new. In fact, ever since information started flowing into patient records, healthcare professionals have grappled with poorly entered data, duplicate data, and other challenges.
The difference now is that automation has begun to expose, and in some cases, amplify the problem, according to Dick Taylor, MD, Chief Clinical Informatics Officer at BJC Healthcare. “We’re exposed to so many sources of error. We have patient-reported outcomes. We have clinician entered data. And as we look at this, we start to understand how bad our data is,” he said during a recent discussion. “If you thought you would always get accurate information from any source, you’re going to be disappointed.”
As healthcare organizations advance in their digital journeys, that disappointment is “more visible and more evident” than ever before, noted Taylor who spoke on a panel along with Rob Bart, MD (CMIO, UPMC), David Kaelber, MD (CMIO/VP of Health Informatics and Patient Engagement Technologies, The MetroHealth System) and Dale Sanders (Chief Strategy Officer, Intelligent Medical Objects) about how they’re working to reverse the trend.
Two of the biggest sources of errors, noted the speakers, are patient-entered data and health information exchanges (HIEs), both of which come with their own set of unique challenges. The former category takes on multiple forms, as patients can enter data (either discrete or indiscrete) manually or automatically through tools like remote monitoring, which opens the door to errors that can be introduced by both patients and devices.
The latter, HIE, is also fraught with opportunities for mistakes, according to Kaelber, who referenced the difficulties his team faced with external Covid immunization records. “We were getting vaccination data from other EHRs or though the HIE that hadn’t been mapped (properly), and as a result, we couldn’t process it.”
Go to the source
It’s a problem that’s getting increasingly complex. And although there’s no easy fix, there are ways to improve the quality of data, and it starts by getting to the root of the issue.
“You need to go to the source,” said Kaelber. “Sometimes it takes more effort up front, but in the long term, it’s worth it.”
For organizations like UPMC, which have used the same system for two decades, it’s going to require a lot of untangling, largely because of what he termed an “amoebic growth” that occurs with EHRs. “There’s truth to the idea that you only add things to electronic health records; you almost never remove anything,” he said. Whereas organizations tend to be “very thoughtful about the curation and implementation of the EHR, there’s a loss of discipline that occurs” following go-live, which can negatively impact the quality of the data.
“We’ve lost that focus on necessity,” Bart said, citing as an example UPMC’s nursing admission database, which had expanded to include more than 700 data elements. “We need to think about curating and getting into data of use, which is what we’re trying to achieve much more within the EHR.” His team took a key step by parsing it down to around 200 data elements, but there’s still more work to do.
The data “arms race”
Another major hurdle, according to Sanders, is the current structure for quality measures. “It’s really not doing anything, in my opinion, to dramatically improve cost or care quality, or ease the burden on clinicians,” he said, citing the inefficiency of single purpose data collection. “You have to click a box to show smoking cessation. You have to click a box to indicate there’s violence in the home. These are clinicians who have maybe 15 minutes to see a patient, and we’re asking them to do all of that.”
Sanders, who spent nearly a decade as a CIO, placed some of the blame on EHR design and user interfaces, which have failed to follow the lead of smartphones that utilize master reference data to check for errors and send alerts to correct them. “We forgot to do that with HITECH.” He envisions a master reference data library that would be accessible to everyone using an EHR through an API, which would remove the burden from CMIOs and informatics teams.
Taylor said his concern with this concept is that EHRs are often working to placate multiple masters, including billing and regulations. “It’s turned into something of an arms race,” he added. But in reality, “the data being entered is only as good as providers want it to be. If they don’t see the need to enter the data accurately, they’ll do the minimum necessary to check the box.”
If clinicians need to click boxes and note that all fields have been reviewed, just so they can satisfy quality measures, some might question whether he or she is taking the best care possible of the patient. Leaders, Taylor noted, need to focus on aligning the two. “If you take good care of the patient, that becomes self-documenting. Why do you have to click a box?” he said. “These things are simple, but we’re still at a primitive design stage where we’re trying to get overt input from the provider.”
“It’s not the technology”
Of course, technology isn’t always the culprit — far from it, Taylor emphasized. “There are a lot of things that the EHR is trying to do. And there are lots of different people who want it to work in lots of different ways that sometimes are in competition with each other.” Technology is only about 10-20 percent of the problem, with the rest coming down to people, processes, change management, and other key areas.
Kaelber concurred, especially when it comes to data exchange. “Everyone says they want it to happen, but the devil is in the details — and it’s not the technical details,” he said. Many leaders are still hesitant to share data because of fundamental business reasons. “Until those drivers change, people aren’t going to want to exchange data to the degree that we want it to happen.”
Similarly, some organizations struggle to obtain resources for improving data quality. “The normal care and feeding of data quality is one of my biggest challenges,” Kaelber noted. “I can’t make it an institutional priority because it’s very hard to explain the value proposition around doing that.”
This, noted Taylor, is where leaders can make an impact by “not chasing the bright, shiny objects and helping us understand whether we’re delivering the care we want to deliver.” If that proves to the be case, pursuing new uses for AI and machine learning makes sense, but making that distinction has become critical.
On the other hand, if data quality is a problem, leaders need to act decisively, he noted. “Have a strategy. Understand the problems, choose targets that matter clinically, financially, and operationally and focus on those. Don’t get lost in the sea of data problems.”
To view the archive of this webinar — Strategies for Improving Data Quality in Your Health System (Sponsored by Intelligent Medical Objects) — please click here.