Mining Electronic Health Data Algorythmically To Make It More Productive

The biggest problem with big data is not acquiring it - most organizations are drowning in information - but accessing it from the disparate silos in which it resides and is jealously guarded by those who claim to 'own' it.

Once it is stored in ways that permit critical analysis, more effective, efficient and productive solutions can be derived and applied to optimal outcomes. JL

Sara Castellanos reports in the Wall Street Journal:

75 to 80% of the information in electronic health records is unstructured and can’t be stored in a conventional database. That limits the ways the data can be searched, analyzed and put to use. (But) algorithms can “read” the information and load it into a database (to) be mined for insights. It is especially important now in the era of “value-based care,” where providers are paid higher reimbursements for delivering high-quality, efficient care instead of being paid on volume of patients.
Health-care companies are collecting more data on patients yet, they are struggling to realize its full value because much of it can’t be analyzed in a traditional database.
McKesson Corp., the drug distributor and health-care technology firm, says it is developing a new analytics tool that could help solve that problem for oncologists by using advanced algorithms and natural-language processing technologies to read and reformat such “unstructured” information.
In recent years, health-care organizations have been prioritizing the adoption of new health information technology, such as analytics. The shift gained momentum in 2009 under former President Barack Obama and amid widespread pressure to reduce health-care costs, said Jiban Khuntia, an assistant professor of health administration and information systems at the University of Colorado Denver.
Health-care organizations, trying to gain an edge, will turn to technology for knowledge that can be acted upon, said Brian Hopkins, a principal analyst at Forrester Research Inc., who specializes in big data and data management.
”Think about figuring out when and if a patient is in danger…and sending specific information about the danger to the floor nurse so she can act,” he said in an email. “It’s the digital insight implemented in the software. That’s what’s important.”
McKesson executives say that 75 to 80% of the information in electronic health records is unstructured and can’t be stored in a conventional database. That limits the ways in which the data, including doctor’s notes and pathology reports, can be searched, analyzed and put to use.
McKesson’s new tool will be made available to a select group of oncologist practices in a pilot program beginning this spring. Later this year, it is expected to be integrated into McKesson’s “Practice Insights” analytics platform for oncologist practices, which launched in 2015.
A team of about 10 in-house data scientists, project managers and cancer experts at McKesson has been working since last July to develop the technology, said Dan Lodder, vice president and general manager of technology solutions for McKesson Specialty Health, an outpatient health care division of the company that is spearheading the development of the analytics tool.
Right now, information is in disparate digital sources within oncologist practices. Matching cancer patients with the right clinical trial, for example, can take days, weeks or more following a doctor’s visit, because the data isn’t easily searchable, Mr. Lodder said.
“All this information from electronic health records is an untapped treasure trove of information,” said Kathy McElligott, McKesson’s chief information officer and chief technology officer.
Structuring the data requires advanced algorithms and natural language processing technologies that can “read” the information in the documents, identify and understand specific terms, pull that information out and load it into database tables so it can be mined for insights, Mr. Lodder said.
That is why the McKesson team is working with an undisclosed third-party startup that has invented an algorithm to process health-care-specific data in natural language, and understand its context. McKesson will process raw unstructured information from oncology practices using the startup’s natural language processing technology, converting it into structured data that’s stored in McKesson’s analytics database in its private cloud.
Ms. McElligott said there’s a continuing emphasis at the company to develop new technology, particularly by making better use of data and analytics that can help improve the outcome for patients.
It is especially important now, Ms. McElligott said, in the era of “value-based care,” where health-care providers are paid higher reimbursements for delivering high-quality, efficient and affordable patient care instead of being paid based on the volume of patients they treat.
Shares of McKesson were down $11.96 to $139.14 in after-hours trading Wednesday after the company reported that revenue for its third quarter, ending Dec. 31, 2016, was $50.1 billion, up about 5% from the same period in the prior year. Net income was $633 million, down slightly from $634 million a year ago. Earnings per diluted share from continuing operations were $2.86, up from $2.71.
McKesson announced Wednesday it would acquire CoverMyMeds for $1.1 billion, a company that says it automates prior authorizations online through electronic health records systems.