Patient Matching: Why it is Critical in Radiology

Three years ago — in a worst-case scenario for any physician — a surgeon at St. Vincent’s Hospital in Worcester, MA, removed a healthy kidney from a patient after mistakenly reading the CT scan of another patient. Both patients had the same name and received the same CT scan on the same day. The surgeon logged into the electronic records system at UMass Memorial Medical Center, Worcester, MA, (where he also practiced), to access the patient’s scan using the first and last name.

Unfortunately, the doctor did not use a second identifier, pulled up the wrong scan, and incorrectly diagnosed the patient with a kidney tumor. More than a week later, the surgeon removed a healthy kidney from his patient based on his analysis of the wrong CT scan, according to reports from the state Department of Public Health and Centers for Medicare and Medicaid Services. This type of patient misidentification doesn’t have to happen. What if the encounter between the physician and the patient began with a biometric scan — for example, a scan of a patient’s face or retina? Would that have changed the outcome? According to Eliot Siegel, MD, professor of diagnostic radiology and nuclear medicine and vice chair of information systems at the University of Maryland School of Medicine, such a scan, combined with a patient matching technology such as referential matching, could vastly improve matching accuracy. Nevertheless, the technology has yet to be adopted by a wide segment of health care.

According to a 2018 report by Pew Charitable Trusts, matching accuracy within health care facilities is as low as 80%, while accuracy between organizations can be as low as 50%, even when those organizations share the same electronic health records (EHR) vendor. For its report, Pew commissioned studies of key topics, interviewed hospital and clinician executives, held focus groups with patients, consulted experts and examined the existing literature.

“Patient matching has been an ongoing challenge in health care,” said Ben Moscovitch, Pew’s health information technology project director. “And that’s for many reasons.”

Mistakes can happen, for example, when patients change their name, address or phone number, among other demographic information. And a simple typo in entering that information can cause havoc when trying to match patients with their records. “In radiology that challenge is being exacerbated by the fact that we are getting imaging studies from outside facilities that we need to put into our own image management systems,” Dr. Siegel said. “If you read the Pew report, you see how incredibly important it is to match and link records correctly, and how abysmal the track record is for doing the matching from one institution to another. The state of the art is pretty bad.”

Data Standardization Necessary

However, the Pew report offers several promising strategies for overcoming this challenge. According to Pew, the most promising near-term approach involves refining demographic data standards.

“Standardization of data has been recommended by our organization for several years,” Moscovitch said. “Through data standardization, organizations would format information in the same way.”

Pew joined with the Regenstrief Institute at Indiana University, Bloomington, to collaborate on a project testing the standardization of data using real-world records and found that standardization of addresses and last names significantly improved patient matching. For example, standardizing addresses according to the format used by the U.S. Postal Service can improve match rates by approximately 3%, and standardizing addresses in conjunction with last nameswould yield even greater improvements, Moscovitch said.

“Better demographic data standardization could leverage data already being used and could be implemented in the near term,” Moscovitch said.

In the long term, approaches such as biometrics could result in significant improvements in patient matching accuracy, he said. Referential matching, however, can be implemented in the short term and is already being implemented across the country. Another benefit of a biometric approach is that patients like it, Moscovitch said.

“In focus groups we conducted, patients overwhelmingly wanted to use biometrics to match their records,” he said. “The patients we talked to said they use biometrics to unlock phones or go on airplanes and asked why such an approach couldn’t be used in health care to link their records.”

Referential Patient Matching Effective

Referential matching represents another promising approach. With referential matching, instead of directly comparing demographic data from two different patient records to see if they match, referential patient matching technologies will match the demographic data from a patient record to a third-party reference data base.

According to Verato, a Virginia-based technology vendor in this space, reference matching can yield patient matching accuracy rates as high as 98.5%. While promising, there are currently limitations to each of these technologies.

For example, said Moscovitch, there are privacy and technical challenges to the use of biometrics, while referential matching would be problematic in the case of the pediatric population whose demographic information would not be included in these third-party databases.

“No single solution is likely to solve the patient matching problem,” Moscovitch said. Simultaneously, he noted that approaches like data standardization, biometrics and referential matching aren’t mutually exclusive.

Combining Technologies

Dr. Siegel believes that using a combination of biometrics and referential matching technologies offers the best potential solution for solving the seemingly intractable problem of patient matching.

“Patient matching is a major problem, and it is something that is extraordinarily near and dear to radiologists’ hearts,” he said. “The potential for disaster if a patient were to receive an incorrect scan leading to a potentially unnecessary interventional procedure or surgery is an issue that commands attention from radiology administrators, radiologists and hospital administrators.” “It looks like referential matching is becoming more common,” Dr. Siegel said. “For me, the ideal would be a combination of referential and biometric matching – fingerprint, retinal scan, palm scan or a facial scan. The combination of the two would, in my opinion, provide something close to 100 percent accuracy.”