Oak Ridge Probed Health IT Infrastructure to Solve VA Problem

June 13, 2019 - Oak Ridge National Laboratory took a look at the underlying health IT infrastructure when the Department of Veterans Affairs (VA) was having trouble with thousands of automatic cancellations of diagnostic test orders at VA hospitals.

As reported by USA Today, patients were showing up for diagnostic tests and finding out that the test orders had been cancelled, not by their doctor.

Working with the VA, Oak Ridge National Laboratory conducted an investigation into why these cancellations were happening.

The researchers discovered that the cancellations were the result of machine errors, explained Olufemi Omitaomu, Ph.D., a senior researcher at the Computational Sciences and Engineering Division at Oak Ridge National Laboratory,

“We found out from talking to people that machine-induced errors are a more challenging problem than human-induced errors because you don’t really know where to look,” Omitaomu told HITInfrastructure.com.

READ MORE: GAO: IT System Issues Persist with VA’s Family Caregiver Program

According to the research paper summarizing the undertaking, the team developed an end-to-end framework for improving the reliability and performance of the health IT infrastructure.

“For us to have a framework, we first of all developed a data correction and processing step. The next step was what we called the hazard monitoring and detection phase. For that phase we had two levels: a specific hazard detection approach and a system-level approach,” Omitaomu said.

The research team defined a hazard as any condition, event, or circumstance that could cause patient harm.

“The specific hazard detection approach looks for things out of the ordinary. For example, you can collect history data that gives you a baseline so you can see an abnormal number of cancellation, and you can quickly flag that,” he explained.

“The other threat is the system level. In this case you don't have a particular hazard you are going after but you are monitoring the entire system. So if you see an emerging trend that is different from the normal, you can investigate what is causing it. You detect it before it becomes a problem," he observed. 

READ MORE: Byrne Says Aging Health IT Infrastructure Complicates VA EHR Rollout

“There is also the reporting phase. This is where we have a surveillance system that takes all the data from the other phases. Once events are detected they are reported to safety managers and analysts. Then, the analysts have a feedback component where they can investigate and report whether that detection is a false positive or a false negative," Omitaomu said. 

Feedback from the VA's review informs improvements to the surveillance tool. The next phase for the framework will involve machine learning techniques for smarter, faster error detection, he noted.

Omitaomu's team encountered challenges in data processing for the model. “It has to do with the way the data is collected, the way the system is developed or designed to collect the data, so that was one of the biggest challenge,” he said.

“In terms of methodology, there are machine learning approaches that are applicable to this application, but getting the data in a form in which you can easily develop an algorithm is a challenge,” he added.

Oak Ridge Team Takes $16B VA EHR Modernization Into Account

The VA is moving head with plans to replace its legacy VistA electronic health record (EHR) with a $16 billion Cerner EHR system, which is expected to take 10 years to implement.

READ MORE: VA’s IT Management Lapses Could Jeopardize EHR Modernization

Omitaomu said the Oak Ridge team took that into consideration in its development of the framework. He said that the team talked to the VA several times about the new EHR system. As a result, it will be able to extend and modify the framework as the new system comes online.  

“We don’t know how it’s going to change the configuration and how it’s going to change some of the current issues … but we are going to make adjustments as needed,” he said.

Oak Ridge wants the framework to be applicable beyond the VA -- to the Department of Defense and even the private sector. He expects the surveillance component in particular to have wide health IT application.

“These health IT systems need a surveillance tool that can survey what is going on and alert somebody that there is something out of the ordinary. From that perspective, we think it can be extended to the private sector as well,” he said.

“This area is new in terms of the problem of looking at machine error, so we believe this can help the healthcare industry, especially as Internet of Things devices become more integrated and people use wearable devices more,” Omitaomu concluded.