Data Analysis: Tomorrow’s Clinical Tool — Connecting Silos Into Searchable Databases
By David Yeager
Vol. 14 No. 10 P. 6
Digital workflow allows more information to be shared among clinicians more efficiently than ever before. Informatics advances have been a boon to radiology, but with growing data repositories spread across the health care enterprise, and by extension the entire health care system, radiologists may soon be dealing with too much of a good thing. Managing the increasing mass of data will require a new approach.
The use of informatics tools has increased productivity significantly in the past decade, and new analytics and data mining applications offer the promise of better quality care but, as data and image sharing become more tightly woven into the fabric of health care, radiologists will need to develop a more global mindset.
Gary J. Wendt, MD, MBA, a professor of radiology and the vice chair of informatics at the University of Wisconsin-Madison (UW), says tools that assist with peer, resident, technologist, and protocol reviews as well as scanner image quality offer good value to radiology departments. But he sees a need for a more cohesive approach to informatics. Rather than simply considering how a tool may help radiology, he says developers should be thinking about how it can be used by other departments as well. In other words, common functionality is the new black.
All Together Now
Wendt cites a peer review tool developed by UW’s radiology department and integrated into its PACS. The tool soon drew attention from other departments that were interested in putting it into the EHR for other clinicians to use. Wendt says these types of projects increasingly are becoming necessary.
With numerous departments relying on medical images and information from EHRs and various other departmental repositories, accessibility is no longer a perk; it’s a requirement. Data analytics and data mining tools can process more data than ever, but presenting that data in a form that’s easy to use is a work in progress. The ultimate goal is to have a single interface that can be used throughout the enterprise.
“I think, in the past, informatics has been good at developing all these tools that munch on images and munch on data, but I think the big challenge now is how do you tie all these together?” Wendt says. “I think it’s going to take a reinvented PACS. You’re going to have to have some of these mining services, or a next-generation PACS, so to speak, that actually will aggregate all of this data into one view of the world that’s accessible to not only radiologists but clinicians in general.”
Radiology is well positioned to participate in the process. Wendt points out that radiology has been fairly successful at aggregating data, moving from separate repositories for CT, MR, and CR images to today’s PACS. The expertise that’s been developed for moving images should translate well to other informatics endeavors.
But it’s not going to be easy. Departments as disparate as pathology, dermatology, cardiology, and orthopedics have specific needs that will have to be addressed. Radiology likely will be in the middle of the effort, but it will have to think about informatics with a less radiology-centric perspective. Sharing images is one thing, but scanning and viewing entire pathology slides or presenting gene sequencing in a visually useful format are problems that people are only beginning to mull.
Far Beyond PACS
“It’s not going to be as simple as the PACS world was, where it was centered on the radiologists and everyone else hung off their shirt tails,” Wendt says. “I think we’re going to have to worry about pleasing all of the clinicians without burying everybody under the radiologist’s set of tools. There may be specialty tools that different subspecialties need, like making sure that color calibration is accurate for pathology and dermatology. We need to make sure that they have their unique sets of tools, and that the tool set you give them for viewing images and data is good at viewing all types of data, not just images.”
There still are many challenges that need to be addressed before the numerous large silos of medical data coalesce into a connected enterprise tool. Paradoxically, although there’s plenty of data, there’s a need for more searchable, mineable data. Wendt says systems that automate and standardize data storage will be crucial to these efforts. Currently, analytics and data mining are targeted to specific clinical and administrative questions because there’s not enough standardized data for general inquiries.
Because of the lack of standardized data, analytics and data mining largely have been used for research, but Wendt thinks applying the tools to the clinical point of care will speed up their development. He sees the true value of this technology as a real-time adjunct to clinical care. For example, a tool that does some preprocessing of a patient’s data potentially could provide useful clinical information without being queried. Such a system may provide alerts about items in a patient’s medical history that are relevant to the symptoms but may not have been communicated by the primary care physician. Or it may provide some decision support for certain image sequences.
“It’s sort of combining CAD [computer-aided detection] and data mining and analytics into routine care so that it’s present when you open it and, likewise, making those tools available for the clinician,” Wendt says. “It’s not just meant for radiologists. When clinicians open exams and look at data or go to order an exam, make sure you give them a complete picture. If somebody has a new onset of headache, and the ER physician is going to order a routine MR scan and there was actually one done a week ago, here are the results and here’s your likelihood that the scan will add anything new. I think the big thing is that, No. 1, we have to take the analytics and mining out of being a primary research tool and make it a clinical tool. And, No. 2, we have to make sure that the tools are applicable to all clinicians across the board.”
— David Yeager is a freelance writer and editor based in Royersford, Pennsylvania. He writes primarily about imaging informatics topics for Radiology Today.