Catching ID errors where it
  counts—in the lab

September 2005
Feature Story

Anne Ford

Ask any flight attendant: Most airline passengers pay little or no attention to the safety information presented at the start of a flight, because they know the odds of an accident are slim. For the same reason, laboratories run the risk of becoming blasé about the possibility of patient identification errors.

"Most patient errors are caught inside the laboratories, only some make it out the door, and only a fraction of those cause adverse events for patients," says Paul N. Valenstein, MD, president of Pathology and Laboratory Associates, Ann Arbor, Mich., and chair of the CAP Quality Practices Committee. "Those sorts of odds breed a culture of complacency, because most errors don’t come back to bite you. They’re like the fire alarms in my hospital. I ignore them because they’re all tests. Someday," he adds darkly, "I might burn up."

So what are laboratories doing to detect and avert patient misidentification disasters, and how well are they doing it?

A recent CAP Q-Probes study, "Identification Errors," asked just that. In the study, authored by Dr. Valenstein and Stephen S. Raab, MD, chief of pathology at the University of Pittsburgh Medical Center Shadyside and a member of the Quality Practices Committee, 120 laboratories recorded patient and specimen identification errors for five weeks. Identification errors could involve misidentified specimens or patients. To qualify, a test result either must have been released for the wrong specimen or patient, or some intervention must have taken place that prevented the wrong result from being released. Nearly all of the laboratories were in the United States. Thirty-six percent of participants were teaching hospitals, and 18 percent of participants had a pathology residency program. Before the study began, 88 percent of participants had a monitoring program in place to track identification errors.

In addition to noting identification errors, the laboratories also recorded whether the errors were spotted before or after results were released and whether they resulted in adverse patient events. Examples of adverse events—defined as identification errors that resulted in patient harm or significant changes in patient care—included, among others, patients who received incorrect medication or patients who were mistakenly admitted to the hospital or for whom hospital admission was delayed. Billing problems that stemmed from identification errors but did not affect patient care were not classified as adverse events, nor were identification errors that resulted in only a second phlebotomy and had no other patient consequences. Also, adverse events that resulted from a laboratory problem other than an identification error were not recorded.

Laboratories could, as an option, classify an error as an initial registration or order-entry error; a primary specimen label error; an error identifying an aliquot, block, or slide; a result-entry error; a different type of clerical error; or other error. Laboratories were asked to report their institutional practices for avoiding and detecting patient identification errors in five areas: clerical and accessioning, chemistry/ hematology, microbiology, transfusion medicine, and anatomic pathology. The study did not track whether the person responsible for a particular identification error was a member of the laboratory staff or worked outside the laboratory.

Drs. Valenstein and Raab found that of all the identification errors reported, about 85 percent were discovered before results were released. But that percentage varied from laboratory to laboratory. One-fourth of laboratories detected 95 percent of errors before results were verified, while another one-fourth detected less than 70 percent. The rate of identification errors that were detected post-verification was 60 per 1 million billable tests. That rate, too, varied among laboratories: One-fourth reported fewer than 20 errors per 1 million billable tests, and another quarter reported error rates of at least 130 per 1 million billable tests. Laboratories reported 345 total adverse events, most of which resulted in no permanent patient harm. "On average, adverse events resulted from one out of every 18 identification errors," the study says.

How surprising are the study’s findings? It’s hard to say, since so little previous investigation has been conducted in this area. "I think a good thing is that this establishes a baseline," Dr. Raab says. He agrees with Dr. Valenstein, however, that there’s room for improvement. "We extrapolated the error rate to the national population," Dr. Valenstein says by way of illustration. "Our best estimate is that about 160,000 adverse patient events occur yearly in the U.S. because of patient or specimen ID errors involving the laboratory.

"In this study, we didn’t characterize the full range of adverse events, but we know from isolated reports that they range from a patient receiving the wrong medicine to something more serious, such as a patient undergoing a procedure or operation the patient didn’t need. On a national scale, there are a lot of patients being hurt by identification mixups."

Laboratories concerned about detecting identification errors may be puzzled about how to measure their success in this area. As Dr. Valenstein points out, "We don’t know how often identification errors really occur—we just know how often they’re detected."

So does a higher detection rate mean that a laboratory is doing a good job of ferreting out errors, or that it’s simply committing more errors in the first place? The study addresses this conundrum by recommending that laboratories focus not on their total error rates, but on the percentage of their identification errors that are detected before results are verified: "Errors detected post-verification have more potential to cause patient harm than errors that are corrected within the laboratory before a result is released.... Calculating the percentage of identification errors you detect before results are released, and comparing this percentage to the fraction found in other facilities, will help you assess whether a ’culture of safety’ has taken hold within your facility."

Of course, many factors contribute to a culture of safety, and staff vigilance is only one of them. "In addition to vigilance, we need systematic controls or practices that reduce the likelihood that identification errors will occur in the first place and increase the likelihood that they’ll be detected," Dr. Valenstein says. "For critical problems, you don’t rely on just one system to prevent errors. You want defense in depth. We wouldn’t rely only on pilot vigilance to keep planes airborne. We want multiple interlocking defenses."

For example, he says, check digits—which a computer can use to check the accuracy of identification numbers—are highly effective at catching errors made by a human operator when keying in specimen numbers, such as juxtaposition of adjacent digits. "But it’s not a perfect defense," he says. "If somebody attaches the wrong label to a tube, the tube will have a perfectly printed ID number and a perfectly calculated check digit. But it will be someone else’s ID number and check digit. You can’t rely completely on check digits. It’s a good defense against key entry errors, but not a good defense against mislabeling a tube." Other, supplementary checks are needed.

The study examined which identification error detection practices are associated with higher or lower percentages of errors detected before results are released. Chemistry or hematology sections that regularly check patient identification numbers against a database of existing patients, for example, detect a higher percentage of identification errors before results are verified. Meanwhile, transfusion medicine sections that used more than two data elements (something in addition to the patient’s name and identification number) to verify identification were associated with lower rates of errors detected post-verification. Likewise, institutions that had a monitoring program in place to track identification errors before the study began enjoyed lower rates of post-verification error detection.

Interestingly, the study found that several laboratory methods and practice variables commonly used to detect identification errors had no significant association with lower error rates. They include, among others, using the same medical record number for subsequent patient visits, and automatically rejecting a specimen or investigating the specimen’s identity when the specimen label or number doesn’t match that on the requisition. "Either these practices have no impact on error rates or the study failed to detect an impact because the sample size wasn’t sufficiently large," Dr. Valenstein says.

Interested readers are referred to several other studies, among them "Comparing Near Misses with Actual Mistransfusion Events: A More Accurate Reflection of Transfusion Errors" (Ibojie J, Urbaniak SJ. Transfusion. 2002;32:601-608) and "Identification Errors in Pathology and Laboratory Medicine" (Valenstein P, Sirota R. Clin Lab Med. 2004;24:979-996).

The Q-Probe study’s official recommendations to laboratories? Monitor identification errors continually, and make sure to properly identify patients who don’t already appear in the laboratory or hospital database. In transfusion medicine, use multiple identifiers—such as first name, last name, middle initials, date of birth, sex, medical record number, and/or phlebotomist—on both the specimen tube and the requisition. Before releasing results, match requisitions to computer orders if necessary instead of waiting for final results. And, of course, think about adopting methods other laboratories use, such as double-checking the identity of a lost specimen or of a specimen not accompanied by a requisition. After all, "This Q-Probes is a means to set the stage for how labs can learn from each other," Dr. Raab says. "Any checks laboratories put in place are beneficial."

Attention to potential identification errors, Dr. Valenstein adds, "should be an animating force in the laboratory. There is no point to getting the right answer on the wrong patient. It’s worse than no answer at all."