The Agency for Healthcare Research and Quality has granted a five-year, $2.5 million grant to the Regenstrief Institute’s Center for Biomedical Informatics to support the development and real-world testing of automated patient identification solutions to match sufferers to their electronic health information.
Precisely matching electronic health records to the correct sufferer is important to medical care and safety by decreasing the preventable errors. Although, as many as half of EHRs are mismatched when information is transferred between healthcare systems contends Shaun Grannis, MD, director of the Center for Biomedical Informatics and principal investigator for the new AHRQ grant.
“We are increasingly reliant on the disparate systems to bring data together, and so patient identification is becoming an increasingly significant topic,” claims Grannis. “We as a nation have not established any clear, comprehensive and ubiquitous guidelines for how we identify patients.”
Under the grant, Regenstrief’s CBMI will formally verify and test evidence-based best practice suggestions from the Office of the National Coordinator for Health Information Technology on how to solve the patient identification issue.
“We are going to look at thousands of datasets and analyze each of the specific suggestions from ONC,” analyzes Grannis. “We need to answer the query: Which of these suggestions actually shows good evidence for usefulness in improving accuracy and efficiency?”
Additionally, the center is going to look at betterments to some usually used patient matching algorithms to see if they can result in better accuracy. Regenstrief investigators will evaluate the newly established algorithms by leveraging the Indiana Network for Patient Care (INPC), the country’s largest inter-organizational clinical data repository, which will serve as a testing ground for their work.
“We are using our health information exchange in the state of Indiana,” adds Grannis. “We have almost 30 million unique patient registrations in the system and it goes back over 40 years, so we’ve a lot of data to study this problem with.”
Established by CBMI and now operated by the Indiana Heath Information Exchange, INPC provides an ideal atmosphere for developing and testing enhancements to broadly used patient matching algorithms, Grannis contends, adding that the software that’s developed will be open source and publicly available.
He points out that current patient matching algorithms are capable of a little more than 90% accuracy on average.
“Fundamentally, this area is going to go in one of two directions,” Grannis summarizes. “Either we’ll ultimately end up with a national unique patient identifier—such as many countries do—or we will find that algorithm and data science are sufficient to solve this issue. It’s still an open question and we need to help better understand and answer that query.”