March 2011
Feature Story

Karen Lusky

A recent Q-Probes study, “Patient Safety Practices for Monitoring PT/INR,” provides a warning about the potential for dangerous lab errors and presents a case for standardizing coagulation laboratory practices.

Remember St. Agnes Hospital in Philadelphia where patients died in 2001 because of a laboratory error in PT/INR testing and reporting? Some of the laboratories in this study were vulnerable to having that same problem, because they did not verify that the International Sensitivity Index of a reagent in use was the ISI employed in the corresponding INR calculation, warns Q-Probes co-author Frederick Meier, MD, director of regional pathology services, Henry Ford Health System, Detroit.

In the Philadelphia case, says Q-Probes co-author Teresa Darcy, MD, the ISI used in the INR calculation was not the ISI assigned to the reagent actually used in testing. The problem “came to light when the coroner found patient deaths due to unexpected hemorrhagic complications. Doctors who received the erroneous INR results acted on the results and increased patients’ warfarin doses,” says Dr. Darcy, medical director of clinical laboratories, University of Wisconsin Hospital.

To prevent that type of mistake, Dr. Meier says, the CAP Laboratory Accreditation Program checklist requires laboratories to validate, and document the validation, that they used the correct ISI in the INR calculation and that the calculation is performed correctly. Laboratories are required to perform INR calculation “verification events” after a change in the reagent lot or instrument, he adds.

In the Q-Probes study, 98 mostly hospital-based labs compared their practices to the CAP checklist requirements, which are designed to ensure accurate PT/INR testing. The data analysis notes that “none of the participants reported finding discrepancies in either the ISI used or the INR calculation in the calculation verification events reported.”

So far so good. Yet when they probed further, laboratories uncovered 17 instances of documentation shortfalls involving their two most recent INR calculation verification events.

The study included a total of 195 verification events. In two of those events, says the Q-Probes analysis, “laboratories failed to verify the ISI in use was the ISI used in the INR calculation,” and this was the critical check that St. Agnes Hospital did not carry out. “In five events, there was no docu­mentation that the ISI in use was appropriate for the reagent/ instrument combination,” the analysis says, “and in 10 events, there was no documentation that patient reports had been reviewed,” that is, they did not check to confirm that the correct result had made it to the patient record.

The maxim, of course, says Q-Probes co-author Peter Howanitz, MD, is, “If it isn’t documented, it ­wasn’t done.”

Five percent of the laboratories reported that they didn’t have a procedure requiring staff to review patient lab result reports to make sure the INRs appearing in the record that clinicians regularly consult were correct. And as Dr. Darcy points out: “Your risk of an error is higher if you don’t even have a requirement in the lab’s procedures.”

Only 39 percent of labs had a procedure for rechecking the ISI after instrument repair; 62.2 percent required doing so after software and instrument upgrades.

The CAP checklist doesn’t require labs to perform either of those tasks, but the CAP Quality Practices Committee, which is the group that develops the Q-Probes and of which the co-authors are members, decided to include that question, says Dr. Darcy, because one of its members knew of a lab that made an ISI error because of instrument repair. “The ISI loaded in the calculation was correct, but after the [instrument vendor] technical specialist came, it was reset to something different. The lab noticed the INR calculation was off and tracked it back to an instrument error.”

In Dr. Howanitz’s view, one of the most serious defects the Q-Probes study discovered was the failure of laboratories to document that they used the correct ISI in calculating the INR. The other one was failure to report critical value INR levels, says Dr. Howanitz, director of clinical laboratories at State University Hospital and vice chairman and professor of pathology, State University of New York Health Science Center, Brooklyn.

Laboratories indicated that they did not report one percent, or 25, of a total of 2,604 critical values for emergency department patients and outpatients. “In 14 cases, laboratory staff did not follow their procedures or did not recognize the result as a critical value,” the study’s authors wrote. “In only three cases was the caregiver not reachable.”

Dr. Howanitz suspects that the Quality Practices Committee may probe again sometime soon the performance of laboratories in reporting critical values. “Studies we have done in the past indicated the major reason for not reporting critical values is that the lab sometimes finds it nearly impossible to contact the appropriate clinician within the allowed time frame. And given the relatively small size of the Q-Probes study sample, it might be that the stars and sun were just aligned here. We need to see if this is a real finding or a statistical issue.”

Although the most common critical value thresholds selected were INRs of four and five, labs in the study showed a wide range of critical value thresholds—INRs between 2.6 and 10—above which the lab indicated that it felt a call to the clinician was warranted. One lab used a threshold critical value of 10, which Dr. Darcy describes as “really high.”

In addition, for point-of-care testing, 23 percent of laboratory participants set an INR of 6.1 or higher for checking whether INR values produced by a POC device should be confirmed by a value determined by the central lab testing method. This finding concerned Dr. Darcy, who notes “the differences between a POC INR result and one produced by a main lab method are pretty striking the higher the INR result gets.”

Dr. Meier says the study also found “a surprising lack of consensus” about what constitutes an acceptable level of variation in test results produced by different instruments reporting PT/INR results from the same lab. Four laboratories accepted a 20 percent difference between instruments, and one lab, a 25 percent difference. “Eight labs had a five percent difference, 22 were at 10 percent, and 10 at 15 percent.”

There was even variation, Dr. Meier says, on how such differences were to be expressed, with some using percentages and others using tenths of an INR.

On the upside, he says, labs enrolled in the study complied 92 percent of the time with the CAP checklist requirements put in place to ensure accurate testing and timely critical result reporting. So you can look at the findings as a “glass nine-tenths full or one-tenth empty,” as he puts it.

But from a Lean point of view, Dr. Meier adds, “the idea is to be consistent by having standardization and stipulating what degree of variation you are willing to accept.” (The study did not ask if the labs had implemented Lean or Six Sigma.)

Dr. Darcy, in fact, sees a number of issues examined by the Q-Probes study for which the CAP Coagulation Resource Committee and the College could develop standard recommendations for laboratories that perform PT/INR testing. One is where to set the critical value. Another is specifying the events that should trigger verifying the POC method with the method used in the main laboratory. A third might be to identify what labs can view as a “reasonable difference” in test results between two instruments in use in the same laboratory at the same time.

Charles Eby, MD, chair of the Coagulation Resource Committee, says committee members are concerned not only about the lack of uniformity of INR critical values among laboratories in general but also the maximum values reported.

He points to the CAP 2010 CGL-B Survey in which 4,837 participants were asked how they reported PT and INR results. “The critical value cut-offs for 77 percent of respondents were between four and six inclusive, with five to 5.49 the most common,” he says. However, the remaining 23 percent were widely distributed above and below an INR range of four to six, including 54 laboratories whose critical INR value was 10 or higher.

In addition, he says, 4,516 participants provided the highest INR reported from their laboratories, “and again we observed marked variation”: 41 percent reported a maximum INR between five and 10.49, 44 percent between 10.5 and 19.99, and 13 percent reported a maximum INR of 20 or higher.

The variation isn’t news to John Olson, MD, PhD, vice chair and professor of pathology and director of the clinical laboratory at the University of Texas Health Science Center at San Antonio. Dr. Olson is first author of a study reported in a 2007 issue of the Archives of Pathology & Laboratory Medicine which he notes also showed “considerable variation in the critical values and analytical measurement range used by laboratories” (Olson JD, et al. Arch Pathol Lab Med. 2007;131:1641–1647).

So where should the critical value ideally be set? Dr. Eby notes that clinical guidelines for managing excessively anticoagulated patients taking warfarin are based on INRs ranging from five to nine, which “would be reasonable for critical and maximum INR reporting, respectively.” He and Dr. Olson point to difficulties in ensuring the accuracy of higher PT/INR values.

Dorothy (Adcock) Funk, MD, a member of the Coagulation Resource Committee and medical/laboratory director at Esoterix Coagulation in Englewood, Colo., advises repeating critical values for PT/INR before reporting them—a step that 16.7 percent of labs in the Q-Probes said they did not do. When repeating the test, Dr. Funk adds, the lab should make sure the sample doesn’t have a clot in the tube, or a “short sample,” and that the sample is in a sodium citrate (light blue) tube.

Should labs identify a critical value on the lower-end of the spectrum? “There is significant risk at both ends of the spectrum,” Dr. Olson says. Thus, in his view, low INRs may merit a critical value in an anticoagulation clinic setting where clinicians order the testing solely to monitor patients on oral anticoagulant therapy.

Most labs, however, including his, do not have a “lower limit” critical value for INR, Dr. Olson says, adding it’s probably because they don’t have ready access to the indication for the test. “Thus, in most laboratories, an INR of one would be normal for many patients but a problem for someone on oral anticoagulation.” Yet, in all instances, a high INR is a critical value, he points out.

While Dr. Olson’s institution uses a critical value of five to reflect clinical guidelines, he doesn’t think there’s any single correct answer to the question of where to set the critical value. Instead, clinicians and laboratories should work together to make that decision, he says. “If labs set the critical value too low, below three or 3.5, they are going to be calling clinicians on values that are actually within the therapeutic intervals.”

Dr. Olson also notes there are no clear guidelines that address what level of agreement labs should accept when comparing patient test results obtained from different instruments. If the laboratory is “harmonizing among instruments,” essentially using the same methods, it might apply a different set of criteria for values within the reference interval than for an INR of five and above, he suggests. In either case, “a variability of 15 percent in either interval seems more generous than most laboratories would be satisfied with while a value of five percent may be difficult to achieve.”

Esoterix Coagulation uses a 10 percent cutoff for imprecision between instruments doing INR testing using the same reagent lot, says Michael Taylor, MS, laboratory manager at Esoterix. “It’s more standardized to use a percent to express the difference,” he says. But the question of what to use “really comes down to understanding the test system and the data it’s providing. We use a percent difference, but other labs may base this decision on the INR itself rather than a percent difference,” Taylor says.

If the lab does use the INR, Dr. Funk says she would not want to see more than a 0.5 difference between machines, noting that the therapeutic range for warfarin is generally an INR of 2.0 to 3.0.

Dr. Olson believes it’s important to set a threshold at which the POC INR needs to be confirmed with a laboratory INR. However, the answer is not the same with all POC and laboratory systems. “Thus, an external agency should not provide a guideline that states the level,” he says. He thinks guidelines could provide the laboratory with a method to determine where to set the level rather than dictate what the value should be. For example, the University Health System clinical laboratory recently changed its point-of-care INR method, and, he says, “when compared to the lab method, it is actually reliable to a slightly higher INR than the previous method.”