The anticoagulant warfarin has been identified as the second most frequent drug responsible for serious, disabling, and fatal adverse drug events in the United States, and its effect on blood coagulation is monitored by the laboratory test called international normalized ratio (INR).
To determine the presence of INR policies and procedures, INR practices, and completeness and timeliness of reporting critical INR results in participants' clinical laboratories.
Participants reviewed their INR policies and procedure requirements, identified their practices by using a questionnaire, and studied completeness of documentation and timeliness of reporting critical value INR results for outpatients and emergency department patients.
In 98 participating institutions, the 5 required policies and procedures were in place in 93% to 99% of clinical laboratories. Fifteen options for the allowable variations among duplicate results from different analyzers, 12 different timeliness goals for reporting critical values, and 18 unique critical value limits were used by participants. All required documentation elements were present in 94.8% of 192 reviewed INR validation reports. Critical value INR results were reported within the time frame established by the laboratory for 93.4% of 2604 results, but 1.0% of results were not reported. Although the median laboratories successfully communicated all critical results within their established time frames and had all the required validation elements based in their 2 most recent INR calculations, those participants at the lowest 10th percentile were successful in 80.0% and 85.7% of these requirements, respectively.
Significant opportunities exist for adherence to INR procedural requirements and for practice patterns and timeliness goals for INR critical results' reporting.
In 2011, two million patients were injured seriously in the United States by prescription drug therapy, ultimately resulting in 128 000 patient deaths.1 A 2013 report summarizing patient drug safety issues found that the anticoagulation drugs dabigatran and warfarin were the 2 drugs most frequently reported to the US Food and Drug Administration (FDA) as causing serious, disabling, or fatal adverse events.2 Warfarin (Coumadin), first introduced in 1954, has been the mainstay of anticoagulation treatment for the past 60 years and requires laboratory testing to guide dosing for subtherapeutic, therapeutic, and toxic effects. In contrast, dabigatran, the leading cause of patient injury reports to the FDA, was approved for use in 2010 and its anticoagulation effect is not monitored routinely by clinical laboratory testing.
The laboratory test used to guide warfarin therapy was the prothrombin time (PT). Blood was collected in a tube containing liquid sodium citrate, which acts as an anticoagulant by binding calcium in the specimen. After centrifugation, plasma was analyzed by adding excess calcium, tissue factor, and phospholipids, followed by measurement of the clotting time. The prothrombin time will vary with the different lots of tissue factor and phospholipids used in the prothrombin reagent. More recently, the reactivity of these different lots of tissue factor reagents was calibrated against the international reference tissue factor and assigned an international sensitivity index (ISI). When the PT is controlled for the effects of various ISIs, the measurement is called the international normalized ratio (INR) and is calculated from the quotient of the patient's PT divided by the geometric mean of the control population raised to the power of the ISI. The INR reference range for healthy patients not using warfarin is 0.8 to 1.2, and for a variety of diseases, 2 general ranges are widely accepted. An INR range of 2.0 to 3.0 is required for such diseases as pulmonary embolism or atrial fibrillation, whereas a range of 2.5 to 3.5 is used for patients with mechanical aortic or mitral valves or children on ventricular assist devices whose condition is stable.3 In addition, there are a few circumstances when other INRs are desired.3
Warfarin has a narrow therapeutic window with multiple drug and food interactions and genetic metabolic variabilities, making management difficult, so that bleeding from overcoagulation, or thrombus growth from undercoagulation, are major risks. In addition, the clinical laboratory requirements for the accurate measurement and reporting of results are complex and require meticulous attention to detail. Results also are influenced by the quality and the type of the specimen required for testing. To aid in these efforts, the College of American Pathologists (CAP) Laboratory Accreditation Program (LAP) has developed several specific requirements that must be fulfilled when collecting specimens, testing, and reporting INR results.4 Adherence to these strict CAP requirements is essential for accurate INR results. This study was developed to assess the quality of clinical laboratory practices associated with the structures (policies and procedures), processes (practices), and outcomes (data of INR validation documentation as well as completeness and timeliness of result reporting) for the INR in clinical laboratories.
MATERIALS AND METHODS
Participants enrolled in the CAP Q-Probes program collected data during July and August of 2010 as previously described.5 The study included 10 multiple choice questions on patient safety practices for monitoring the INR (Input Form 1) and reviewed the 2 most recent INR calculation verifications, by having participants fill in specific information about those procedures on a preprinted form (Input Form 2). Participants used preprinted Input Form 3 to collect results prospectively from 30 consecutive INR critical results. At the conclusion of the data collection, participants sent the required data to the CAP for data analysis. If they were unable to collect results from 30 consecutive results during the study dates, participants were to return the data that were collected to the CAP for data analysis. Characteristics about the test systems used in the central laboratory and throughout the organization were collected by using 28 multiple choice questions contained in Input Form 4.
The required validation components listed in the CAP LAP Hematology Checklists for clinical laboratories included the following4:
1. HEM 22748: All coagulation specimens should be collected into 3.2% buffered sodium citrate (phase II).
2. HEM 23360: The appropriate geometric mean of the PT reference interval is used in the INR (phase II).
3. HEM 23290: The calculation of the INR is appropriately adjusted for every new lot of PT reagents, changes in types of reagent, or change in instrumentation (phase II).
4. HEM 23220: For PT there is documentation that the ISI is appropriate to the particular PT reagent and instrumentation used (phase II).
5. HEM 23430: There are checks of patient reports for correct INR calculations, patient values, and reference ranges under the following circumstances: (1) change in lot or type of PT reagent, (2) change in instrument, (3) establishment of new PT reference ranges, and (4) change in INR calculation (phase II).
The percentage of checklist-required components was calculated as the number that was completed/verified divided by the total number of validation components. The successfully communicated INR critical results to the responsible caregivers were calculated as the number successfully communicated within the clinical laboratory's established time frame over the total number identified. If participants neglected to answer a question, they were excluded from the database for only the question they did not answer. A P ≤ .05 was considered significant.
RESULTS
Ninety-one participants (93%) were from the United States, their hospital type was voluntary, nonprofit in 51 (60%), and nongovernmental in 66 (77.6%) cases; 44 (53.7%) had an occupied bed size of up to 150 beds, and 53 (62.3%) were located in a city or suburb.
Table 1 describes the policies and procedures used in the 98 participating clinical laboratories for the collection of specimens for INR testing, the parameters of INR testing, and when the INR was evaluated for correctness in patient reports. In 90 participants' laboratories (92%), policies and procedures existed for all 5 CAP-required INR parameters. Some laboratories had more detailed procedures than what was required by the CAP LAP and included recalculation of the INR after instrument repair, upgrades to the INR instrument, equation used to calculate the INR, and site of electronic software where the INR was recalculated.
Table 2 describes practices in 98 clinical laboratories for measurement of the INR. On patient reports, participants in 94 clinical laboratories included an INR reference range (98.9%), in 82 clinical laboratories they incorporated an INR target range or other interpretive information (87.0%), and in 94 (98.9%) clinical laboratories they reviewed the PT reference range and geometric mean with each change in PT reagent lots. In 75 of 96 clinical laboratories (78.1%), 2 or more instruments were used in the main laboratory to perform INR measurements and in these circumstances almost all participants (98.7%) used the same lot of reagents for their instruments. Fifteen different cutoffs were used to describe the magnitude of the allowable result difference between analyzers performing the same INR measurement. Patient specimens were used for comparisons among multiple instruments in 69 of 73 clinical laboratories (94.5%) and these comparisons were conducted at least semiannually in 68 of 70 laboratories (97%). The ISIs varied widely between 0.91 and 1.97, with 6 of the participants (6%) using ISIs that exceeded 1.7 (1.73, 1.78, 1.81, 1.95, 1.95, and 1.97). In 92 of 97 clinical laboratories (94.8%) patient results were not run in duplicate, and in 43 of 96 laboratories (44.4%) delta checks for the INR were not run. In all, 86 of 97 laboratories ran 2 levels of controls every 8 hours (86%), and 11 remaining clinical laboratories (11.3%) ran 3 or more levels of controls every 8 hours.
In Table 3 are the practices in 33 institutions where INR determinations are performed at more than 1 location. The number of additional INR sites varied from 1 to 4, with 15 institutions having 1 additional site and 6 institutions having 4 or more additional testing sites. The additional sites included anticoagulation clinics, satellite laboratories, physician office laboratories, and home health agencies.
Table 4 provides information on the documentation quality of the 2 most recent INR calculation verifications in each laboratory. The 3 most common reasons for calculation verifications were a change in reagent lot or type of reagent, required annual check, and regularly scheduled nonannual checks in 99, 37, and 31 verifications (51.6%, 19.3%, and 16.1%), respectively. In 110 of 195 verifications (56.4%), a change in the ISI occurred. There was documentation that the ISI was verified as appropriate for the instrument/reagent combination in 187 of 192 verifications (97.4%), that the ISI used was the proper ISI for the INR in 193 of 195 verifications (99.0%), and that patient reports were reviewed for correct INR calculations in 185 of 195 verifications (94.9%). Overall there were 17 documentation verification errors. Ten errors consisted of patient reports that were not reviewed for the correct INR calculation, followed by 5 errors in not documenting that the ISI was appropriate for the instrument reagent combination. In 140 of 191 verifications (73.3%), the INR verification was a manual, hand calculation, although in 27 laboratories (14.1%) the instrument performing the INR measurement, or laboratory information system in 8 laboratories (4.2%), was used for calculation verifications.
Shown in Table 5 are surveys of the policies and procedures used by participants to report INR critical values. There were 18 high critical value limits ranging from 2.6 to 10.0, with 23 (24%) choosing 5.0; 22 (23%) choosing 4.0; and 10 (10%) choosing either 4.5 or 6.0. Eighty of 96 participants (83%) remeasured specimens with critical values before reporting them. The time required to report results was variable, with 34 (37%), 26 (28%), and 10 (14%) participants allowing 60, 30, and 15 minutes, respectively. Physicians and nurses were the personnel categories that were authorized most frequently to receive results, and a minority of participants participated in a Performance Improvement Program in the previous 2 years for INR reporting.
Table 6 shows reporting data from 2604 critical INR patient specimen results from 97 institutions. Of the reported results studied, 2431 (93.4%) were received within the time frame specified by the institution, with 12 different timeliness goals ranging from 0 to 120 minutes. Of the 173 attempts (6.7%) that failed, 148 (5.7%) were not reported within the time frame and 25 (1.0%) of all critical value results were not reported. In these 25 instances where the INR critical result was not reported, the reasons for failure included the fact that the laboratory staff did not follow the critical value procedure (12 times, 52.2%) or did not recognize the result as a critical value (2 times, 8.7%). Complete reports containing all documentation elements as required in policies or procedures of the participating laboratories were present in 2386 reporting events (94.8%). In the 132 reports, 175 defective elements were found; the most common missing elements were read back, which did not occur in 49 reports (37% of the time). The other frequent missing piece of documentation was the name of the person receiving 38 reports (28.8%), followed closely by 37 failures (21.2%) to document the receiving person's title such as RN or MD.
Percentile distributions seen in Table 7 for 2 quality indicators were based upon requirements established by the CAP's LAP Hematology Checklist. Participants from 98 institutions submitted 2604 INR critical results with 93% of results reported within the time frame established by the laboratory. The rate of required INR validation elements ranged from 42.9% to 100.0% with a median rate of 100.0%. Table 8 indicates the statistical relationship between successful communication of an INR critical result and the absence of a pathology residents or fellows training program (P = .03).
COMMENT
A landmark article by Olson and colleagues6 described progress and problems associated with laboratory reporting of the INR and also provided recommendations for changes in INR laboratory practices to improve patient safety. To help improve patient safety and the precision and accuracy of the INR, the CAP LAP Checklist encompasses many of the suggestions made by Olson and colleagues6 on how INR measurements must occur. These requirements include the tube used for specimen collection, the use of geometric mean for INR calculations, the readjustment of the INR with every new lot of PT reagents, documentation that the ISI is appropriate for the PT reagent in use, and reviews of the patient reports for correct INR calculations when there is a change in the lot of PT reagents, instruments, a new reference range, or a change in the INR reagents.4 When our participants reviewed their INR policies or procedures for these 5 specifics, they found them present in 90 to 97 of the 98 clinical laboratories (93% to 97%) sampled for each of these characteristics. Many institutions had even more detailed procedures such as the specific equation used to calculate the INR and a required reverification of the INR after an instrument repair or upgrades to the instrument.
Because PT INR is a test that has profound influence on patient care, 75 of 96 clinical laboratories (78%) had more than 1 instrument to perform these tests, assuring that the laboratories would make important INR measurements for patient care without delay. To make it easier to maintain their instruments, almost all participants used the same lot of reagents on their primary and secondary instruments. Having additional instruments complicates testing, as instrument results must be compared at regular time intervals to assure that they are equivalent. Although there are no CAP LAP requirements for the frequency of INR comparisons,4 routine laboratory practices for other laboratory tests require that comparisons of results among instruments occur every 6 months. Almost all laboratories compared results semiannually or more frequently. When results were compared among instruments in the same laboratory, there was little intralaboratory agreement of what precision specification was acceptable. In a minority of clinical laboratories the absolute difference in INR values was used, with most of these choosing up to 0.4 absolute units as acceptable. In contrast, in most clinical laboratories a percentage difference was used with the largest group choosing values within 10%.
The influence of the ISI on participant results was great, as there was more than a 2-fold difference in the activity of this material among some of our participants. In 1998, Fairweather and colleagues7 recommended that ISIs be between 0.9 and 1.7, and between 1997 and 2003, the percentage of users who had ISIs greater than 1.7 dropped from more than 60% to 22%.6 Our data indicating that 6% of users have ISIs greater than 1.7 demonstrate that this percentage has continued to decrease and manufacturers are close to achieving the 1998 recommendation.7 In addition, all ISIs in our participants' INR test systems outside of the clinical laboratory were below 1.7.
In almost one-third of laboratories, INR testing was performed in close proximity to patients in locations such as anticoagulation clinics, satellite laboratories, physician office laboratories, and home health agencies; it is common practice to use point-of-care testing with instruments that require whole blood specimens rather than plasma specimens for these types of testing locations. Karon and colleagues,8 as well as others,9,10 have demonstrated that marked differences in INR results occur between central laboratory and point-of-care instruments, and these biases among locations change over time, complicating physician interpretation of results.8 Hence, it is likely that the INR for most testing performed outside the clinical laboratory was different from that of the clinical laboratory, requiring additional education of physicians using this type of testing in addition to the INR testing performed in the central laboratory. A CAP LAP procedure for point-of-care devices requires that when 2 or more devices are used for the same test, they must be checked against each other at least twice a year for correlation of results.11
Guyatt and colleagues3 have reviewed the INR therapeutic range and found that the same range may not be optimal for all indications. They also point out that an equally important goal is that the INR targeted range be achieved promptly and maintained. Key to maintaining accurate INR results over time is that appropriate practices be applied when introducing new lots of reagents when current lots are replaced. Review of the practices indicated that the change in reagent lots or types was the overwhelming reason why the INR calculation verifications were again performed, with the routine check at a specific time interval the second most common reason. Almost all calculation verifications (97%–99%) included an appropriateness check for the ISI, a verification that the INR calculation was the ISI assigned, and a review of patient reports for correct INR calculations. Also, in more than half of the verifications, a change was required in the INR. The percentage of INRs that required a change in ISI appears low; however, approximately 45% of the INR calculations were performed because of routine procedures, such as annual or scheduled checks, software upgrades, or proficiency testing survey challenges, and these most likely did not involve changes to the reagents or the instrument performing the INR measurement. We suspect that many laboratorians remembered a well-publicized and unfortunate situation in a clinical laboratory in Philadelphia, Pennsylvania, whereby 2146 patients' INR results were released when an incorrect ISI was used during a 7-week period for the INR reagents, causing the death of patients.12,13 Other reports of patient injuries or deaths have resulted in recalls of INR reagents because of errors in calculations of the ISI, the INR, or manufacturer errors.14–17 Such reports support the meticulous performance of the INR and the need to follow CAP LAP guidelines.
The frequency and consequences of critical INR values make the critical values system a very important quality attribute for clinical laboratories, their physicians, and patients.18–20 We found clinical laboratory policies and procedures of reporting critical value results highly variable. For example, the INR critical limit used ranged from 2.6 to 10.0 with a median of 4.5, and there were 18 different limits, with the most common limits being 5.0 (24%), 4.0 (23%), 4.5 (10%), and 6.0 (10%). These results were similar to those found by Pai et al21 in a small sample of 21 specialized hematology laboratories where the median INR limit was 5.0 with a range of 2.0 to 6.0. Although the high critical value limit varied from 2.6 to 10.0 in our study, it is unlikely that the patient populations had a major influence on the choice of the critical value limit because for most patient conditions the INR target is 2.0 to 3.0. Exceptions include children with central venous access devices, with a suggested INR target of 1.5 to 1.9; patients with acute deep vein thrombosis, with a recommended target of 2.0 or higher; and patients with mechanical aortic, mitral, or both aortic and mitral values, with a recommended INR target of 2.5 to 3.5.3
The time required by participants for reporting of the INR critical values varied from 0 to 120 minutes. The employees receiving the results were almost always physicians, nurses, and mid-level providers such as nurse practitioners and physician assistants.
Despite The Joint Commission's National Patient Safety Goals,22 and the CAP's Laboratory Patient Safety Goals,23 only 93.4% of the critical values were reported within the clinical laboratories' required time, and 1.0% of the results were not delivered to the appropriate caregivers. Of the 25 results that were not delivered, most were because of errors by laboratory staff in either failure to recognize the critical result or failures to follow established polices for critical value notification. These easily can be improved, as knowledge of the clinical laboratory policies and the critical value limit can be achieved by persistent reinforcement and thorough education of the clinical laboratory staff. Fully automated systems for notification of critical results can reduce the number of times physicians are unable to receive results and are now being used in a few institutions.24,25
Documentation was defective in 6.7% of critical value calls with read back, the name of the person receiving the results, the result elements, and the title of the person receiving the results not being recorded. These errors also can be reduced with re-education of both the laboratory staff and those approved to receive results about how important it is to follow policies.
The one policy that was in need of improvement in many laboratories was the use of 2 patient identifiers, used in only 70% of laboratories. Safe laboratory practices require accurate patient identification; adverse events may occur when a patient has identifiers similar or identical to those of another patient (a “doppelgänger”), when a patient is doubly registered (a “duplicate registration”), or when registration details are derived from 2 or more separate sources (a “hybrid” registration).26 Such errors, once thought of as coincidences, are in fact not extraordinary, but ordinary; 2 identifiers are required to reduce the chance of an action occurring with an incorrect patient. We recommend that similarly to phlebotomy practices, 2 patient identifiers be used to report critical values.
It is common for physicians in general practice to admit their patients to more than 1 hospital; for example, if physicians practice in hospitals that use the critical value limit of 2.6 or 10.0, the physicians may quickly become confused by their expectation of when they will be called for a critical INR result. Efforts aimed at harmonizing results such as the critical value limits for the INR should be encouraged, as it will lead to less confusion for busy physicians treating patients at multiple institutions, all with different critical value limits. Such discussions have started worldwide27 with recommendations by Kost and Hale28 widely circulated. As a first step, we would recommend that those with critical INR limits far from the median reconsider their limits and chose a limit closer to our median or to the median described by Pai et al.21
International normalized ratio is one of the most important laboratory tests, as the reproducibility of the result is one of the most important quality attributes of a clinical laboratory. Not only the usage is complex, but also the establishment and management of the test are unusual for the clinical laboratory. The CAP requirements have been valuable in providing the best practices for the laboratories they inspect, and the activities listed in the checklists are implemented almost entirely within clinical laboratories. However, like all issues, there are some that must be improved. Greater than 50% of participants were able to communicate all their critical INR values within their timeliness goals, and more than 75% of laboratories had the required checklist elements within their most recent calculations. Also, the success in communicating critical INR results was significantly better in those institutions without a pathology training program. We suspect that this may be related to the size of the institution, as hospitals that have residency programs generally are larger than hospitals that do not train residents. However, those laboratories that were not able to communicate all their critical INR values within their own timelines, and those laboratories that did not have the required checklist elements within their most recent INR calculation verifications, although in the minority, must improve their performance.
Although errors in the structure, process, and outcomes of INR results appeared to be fewer than for most laboratory tests, because of the criticality of the INR results, goals for all aspects of this test must be at complete compliance; indeed, experience has shown that an error, such as not using the correct ISI, has resulted in the death of patients. Hence, until the INR is performed in an error-free environment, improvement must continue. To reach this goal, we suggest that the clinical laboratory improvement programs include evaluations of the INR performance so that INR results can be available error-free.