Hybrid program successfully de-identifies patient info in EHRs

A hybrid computer program developed to de-identify patient information will allow faster, more collaborative clinical research than current out-of-the-box de-identification programs, according to new study published this week in the Journal of the Medical Informatics Association.

The researchers, from the University of Utah and elsewhere, noted that the Health Insurance Portability and Accountability Act requires patient data to be de-identified before it can be used for research, but called de-identifying data manually is "tedious."

To make the job easier, the researchers developed a hybrid program, which they call a "best-of-breed" or "BoB" system, using automated natural language processing focused both on improved sensitivity and precision to de-identify clinical information for the Veterans' Health Administration. They tested the program to flag and de-identify patient names in 275 clinical documents, comparing its record to five existing "out-of-the-box" de-identification programs.

The researchers found that their hybrid de-identification program was more successful than the other programs and had fewer false positives. They concluded that their hybrid design "demonstrates robustness and efficiency, prioritizing patient confidentiality while leaving most clinical information intact."

Other studies have recognized that EHRs can be an invaluable tool to faster and more comprehensive research.  At least one recent study also has noted the success of custom designed software programs can make better use of data stored in EHRs for research than existing systems, such as extracting free text from EHRs, enabling researchers to access much more data than before.

To learn more:
- here's the study's abstract

Suggested Articles

Roche, which already owned a 12.6% stake in Flatiron Health, has agreed to buy the health IT company for $1.9 billion.

Allscripts managed to acquire two EHR platforms for just $50 million by selling off a portion of McKesson's portfolio for as much as $235 million.

Artificial intelligence could help physicians predict a patient's risk of developing a deadly infection.