CHICAGO -- Clinical data analytics holds a lot of promise for health care organizations: It can help a hospital bill and code services more effectively and accurately. It can yield insights into other operational processes, such as staff utilization and skills evaluation. Just as important, the access it gives many practitioners to deep patient-data search tools across many departments can help uncover problems, lead to new treatments and streamline workflows.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Clinical data analytics systems, however, typically are limited to large academic facilities. Most hook into data warehouses -- systems that just aren't affordable for many providers.
Facilities without data warehouses still can tackle analytics, but the process poses challenges to the in-house informaticists charged with stitching together databases from disparate applications, ones that often reside in different departments.
All this is about to change, however. In a presentation about radiology-specific search tools here at the Radiological Society of North America (RSNA) 96th Scientific Assembly and Annual Meeting, Dr. Woojin Kim and Dr. Keith Dreyer predicted that federally mandated meaningful use rules will make sophisticated decision support tools feasible for many more organizations.
Search is the key to clinical data analytics functioning
All analytics projects start with searching through organizational data. Two meaningful use requirements in particular will drive clinical data analytics: the adoption of electronic health record (EHR) systems themselves, and the standardized terminology contained in the Systematized Nomenclature of Medicine (SNOMED) and the International Statistical Classification of Diseases and Related Health Problems, 10th Edition (ICD-10).
EHR systems will access many data sources -- labs, radiology and pathology, to name just three -- and aggregate them into one spot. This aggregation will mean conducting just one search instead of many that involve separate hospital departments running their own proprietary applications.
Right now, searches across multiple departments require the extra step of comparing and aggregating results. That can take weeks. Once EHR systems offer an on-ramp into all of a hospital's data systems, that step will be automated too.
While hospitals already use ICD-9, the ICD-10 code set features tens of thousands more descriptions for patient conditions. They not only will yield more detailed data to analyze, but also will make databases more searchable. When everyone uses the same expressions and terminology, search results will be more comprehensive, accurate and useful for quality-improvement initiatives.
"[Search] will get smarter and smarter; the engines will get smarter; and you'll be able to develop even more useful tools when there's standardization," Kim, chief of radiography modality at the University of Pennsylvania in Philadelphia, told SearchHealthIT.com after the presentation.
I don't know anyone who doesn't use data mining in our department.
Dr. Keith Dreyer, vice chair of radiology informatics, Massachusetts General Hospital
Once states build their health information exchanges and hook up to each other in a national network, clinical data analytics really will take off. "The only way to really see epidemiological data is through data sharing," Kim said. "Right now, hospitals are so tight with security. But eventually there will be a need to share data, and this will allow it to happen."
While it's not related to meaningful use, natural language processing technology will further enable clinical data analytics, said Ricky Taira, radiology professor at the University of California at Los Angeles and Dreyer's and Kim's co-presenter. He demonstrated a project in which software compared free-text notes in radiology studies. As natural language processing evolves, it will match search queries with appropriate results better, by recognizing structure where there is none.
How clinical data analytics improves care, research
Dreyer, vice chair of radiology informatics for Massachusetts General Hospital (MGH) in Boston, demonstrated his department's in-house clinical data search tools. The challenges have been many -- everything from complying with the Health Insurance Portability and Accountability Act (HIPAA) to working with different database schemas across applications, to generating complete results ("We learned we have about 45 ways to spell tuberculosis," he said, laughing). The system inspires medical residents to come up with new and innovative research ideas that drive new and better patient care, however.
They've done research examining data by patient population; by disease; by physician; and by method (for example, computed tomography, or CT, scan; positron emission tomography, or PET, scan; or magnetic resonance imaging, or MRI). Moreover, especially enterprising residents create their own software tools to work with MGH's data search, Dreyer added.
"I don't know anyone who doesn't use data mining in our department, of one sort [or another]," Dreyer said.
The search tool Kim demonstrated combined data from the pathology and radiology departments and de-identified it for the sake of HIPAA compliance. The tool lets medical residents search data for potential clinical research studies without having to get their ideas vetted by the institutional research board (IRB). The Food and Drug Administration would require IRB vetting if there were HIPAA-protected patient data included in the results. Using the tool saves everyone time -- the school, the residents, and the IRB -- because in the past, residents would have to get IRB permission just to pore through patient records to see if their study ideas were even feasible, he said.
Opening up the search to everyone, including administrators, has reaped dividends for the teaching hospital, Kim said. The more people can use search, the more creative ways they have discovered to take advantage of clinical data analytics, he said.
"You give that ability to pretty much everyone in your practice, [and] the kinds of things they will come up with are absolutely fantastic," Kim said. "You will get tremendous ideas, tremendous advancements in research and operational processes."
Clinical data analytics as BI tool on the operations side
Probably more useful to non-research institutions were Kim's and Dreyer's examples of how clinical data analytics also can be used as a health care business intelligence tool for the following tasks:
• Sharpening billing and coding processes.
• Checking radiology against pathology to make sure that codes are accurate according to the diagnosis; that everything was coded; and, as a final check, that nothing was coded that wasn't billable (therefore, potentially fraudulent).
• Building profiles of the kinds of patients, diagnoses and scanning methods individual radiologists were involved in, as a resource utilization project. (In an earlier RSNA session on meaningful use, Dreyer suggested that such an analysis could qualify more radiologists in a given facility for federal incentives by ensuring they were seeing Medicare patients).
• Checking individual radiologists' accuracy against each other, as well as against pathology results.
• Streamlining operations by examining how necessary tests were. In one example, MGH compared the symptoms of patients who received CT scans of the head against the resulting diagnoses, to determine whether or not doctors were over-ordering the scans.
Analytics can also be wired into decision support processes, such as those for recommending tests or medications, when based on the results of a patient's radiology study, the presenters said.
Meaningful use rules mandate that health care providers incorporate decision support into their workflows. Clinical data analytics, then, not only could benefit from meaningful use rules -- it also could help satisfy compliance with them.
Let us know what you think about the story; email Don Fluckinger, Features Writer.