Laboratories are one of the few areas in medicine that can claim to approach six sigma standards regarding error rates. Regulatory standards from CMS, the Joint Commission, CLIA, CAP, and various state health departments set the bar high and most labs have very robust quality assurance, proficiency testing, and performance improvement programs in place. However, most of those standards and statistics apply to the analytical phase of laboratory work. The pre-analytical and post-analytical phases are the areas today where the majority of lab errors occur (Hammerling 2012).
Much of the best formal literature on laboratory errors comes from Croatia and Italy. Two of the most prolific writers on errors pertaining to laboratory medicine, Mario Plebani and Guiseppe Lippi discuss the “brain to brain loop” in lab testing (Plebani 2011). This emphasizes the need to consider the entire loop from the time a test is ordered (eg. is it the correct test?) to the time the results are interpreted and used to apply to a plan of care for a patient. They note that currently that “loop” is never closed. That is, we don’t have metrics that measure how the lab is part of patient outcomes. In fact, patient care involving non-laboratory personnel accounts for 95% of the extra-analytical mistakes involved in “lab” errors (Lippi 2010). To use a sports analogy, statisticians in baseball assign an error to an individual even though some errors are truly “team” errors. Plus many of the errors that we will uncover are really symptomatic of flawed systems rather than individual errors. The National Quality Forum (NQF 2009) also recognizes the importance of the pre-analytical and post-analytical phases in its 6 preferred practices for measuring and reporting patient safety, including items like accuracy of ordering for lab tests and communicating the results.
But, lacking good “loop” metrics, how do we get our arms around this problem? One potential method would be to use the tracer methodology, akin to the way Joint Commission uses tracers during their surveys. You can “trace” a test through its full loop in hospitalized inpatients. If you employ physicians on the outpatient side you can also trace a test through the full loop for outpatients or emergency department patients.
Below are the steps we’d recommend in doing a tracer on laboratory testing:
Where would you start? Which tests should you look at? One option would be to take a look at your highest volume tests, since statistically most errors in the loop would occur for these tests. However, you might also consider looking at tests you already know may be “abused” or of controversial value. Or you might look at tests for which errors would be likely to have the most serious patient consequences.
After you choose a test on which to run a tracer, let’s start at the beginning: the ordering of the test by a clinician. Is the rationale for ordering the test clear from the medical record? Is it for diagnosis related to current patient symptoms? Is it for screening or risk factor management? Is it a necessary follow up to a prior abnormal test result? Is it for monitoring treatment (eg. serum drug levels) or assessing for treatment efficacy or side effects?
But there are other questions you should ask. Was there a prior result of that test that might have sufficed? Was that result known? Could it have been known? Was that result available on the hospital IT system or the regional RHIO? Did the provider attempt to see if a prior result was available?
If the rationale for the test is not obvious, also look to see if there were circumstances that “nudged” the provider to order the test. Was the test part of a “panel” or was ordering the test influenced by its appearance on a standardized order set or clinical protocol or the way the lab requisition was formatted (some commercial labs use the requisition form in a manner that tends to “market” certain tests).
Was the timing of the ordered test appropriate? For example, if the test was for a serum anticonvulsant level was the test likely ordered before a steady state level would have been achieved? Or if it is an HbA1C level has enough time elapsed since the change in management that the HbA1C level would reflect the overall glycemic status resulting from that change?
The patient interaction must be considered as well. Was the reason for the test discussed with the patient? Was special preparation for the test (eg. fasting) discussed with the patient? Most importantly, did the provider discuss with the patient how long it would be before the test results come back and how the result will be communicated with him/her (more on that on the post-analytic phase)?
Lastly, an most importantly, before ordering a test the clinician should ask him/herself, and discuss with the patient (1) what will we do if the test result is normal? (2) what will we do if it is abnormal as we expect it to be? (3) what will we do if it shows us something unexpected? For example, do you really need to order that C-reactive protein (CRP) in your patient who has multiple CAD risk factors and a high LDL who you are going to treat with statins regardless of the CRP result?
Keep in mind that some lab results may be “abnormal” by chance. If you have a 5% chance that a test result will fall outside the “normal range” statistically and you order a comprehensive metabolic profile of 18-20 tests you are very likely to have one test result that is “abnormal”. Interestingly, when we talk to lab directors we often get responses like “it is less expensive and more efficient for me to run the panel than the individual test”. That, however, fails to take into account the expense and inconveniences that will be generated following up on such an “abnormal” result.
Was it written out on a prescription form? Was a lab requisition used? Was CPOE used? Regardless of the method used, was the intent of the order clear? Were there handwriting issues? Inappropriate abbreviations used? Was it clear who was ordering the test? (Ever get a test report for a patient who was not your patient because the lab could not read the name of the actual ordering physician?)
Where, when and how was the specimen obtained? Were the appropriate patient identification procedures used prior to obtaining the specimen? Were the correct tubes or other containers used for collecting the specimen? Were they correctly labeled? Are all specimens labeled immediately and individually? How did they get to the lab (collected at the lab, sent by courier to the lab, transported from a hospital unit to the lab, etc.)? Do you have a system that actually tracks the specimen on its way to the lab? How do you know if a specimen never reached the lab? If the specimen and test were time-sensitive, did the specimen get to the lab within the appropriate time frame?
See our Patient Safety Tips of the Week for October 9, 2007 ““ and November 16, 2010 “ ” for discussions on specimen identification, labeling, etc.
We recommend particular attention be paid to sites doing point of care (POC) testing, whether in the office or at the bedside in the hospital. Our experience is that procedures for identification and labeling of specimens in those settings are more prone to “workarounds” and thus more errors.
Actually, we don’t know how many steps are in this phase and it is not the focus of today’s column. But see our Patient Safety Tips of the Week for October 9, 2007 “LEAN in the Lab” for details on this phase.“, November 16, 2010 “ ” and October 11, 2011 “
Was the report accurate? Was the report done promptly? Was the report made available promptly? Did you send the report back to the ordering provider? Should it have been sent to anyone else (eg. was it requested that a copy go to the patient’s PCP?), keeping in mind that studies show sending a result to 2 clinicians actually doubles the likelihood that no one will follow up on the result? If the result was a “critical value” (or a finding requiring urgent response), how and to whom and in what timeframe was that result communicated?
When did the physician (or other ordering provider) receive the report? Did he/she acknowledge reading the report? If the ordering physician is no longer “in the loop” (eg. the ED physician who only worked one shift) did the report get to another physician who acknowledged the report and followed up appropriately? Do the lab, the ordering physician, and the patient each have systems in place to ensure that any test ordered actually completes the entire loop (i.e. that no test “falls through the cracks”)?
Perhaps the most important part of the “loop” is the responsible physician acknowledging the test result and using that result in clinical management of the patient. A couple striking statistics (Plebani 2007) are that clinicians ignore or overlook 25-60% of abnormal laboratory results and 45% of results to urgent laboratory tests requested by the emergency department were never accessed!
We’ve talked on numerous occasions about “pending” tests falling through the cracks. The result that comes back after an inpatient is discharged is one example and we’ve suggested that you always include in discharge summaries a “tests pending” section so the next provider for that patient is at least alerted to a result he/she needs to check. We’ve also suggested that you use your EHR’s to send a “results returned after discharge” to the alert list for your inpatient providers.
But the Emergency Department is a particularly vulnerable area. That’s for two reasons: (1) the patients stay for short periods and (2) the providers work in shifts and may not return to that hospital for long periods (or not at all). So you need to have in place some system in which test results returning after the patient has left are reviewed by a clinician and important results get conveyed to the patient and the provider who will next care for that patient.
The February 2012 issue of AHRQ Web M&M has a case (Mohta 2012) that illustrates an error in a situation that likely occurs frequently and may be underappreciated: the amended lab report. In that case a young pregnant woman was admitted to a hospital with hypertension and protein on urinary dipstick testing. A 24-hour urine for protein was initially (erroneously) reported as negative for significant proteinuria. A resident was preparing the patient for discharge when the attending physician fortunately double-checked the result. In fact, the initial erroneous report had been amended and did, in fact, show significant proteinuria. That confirmed a diagnosis of pre-eclampsia and significantly altered the clinical management of the patient.
We’ve discussed on multiple occasions the issue of test results reaching the chart after a patient is discharged. However, this one is even more concerning because the initial report was negative and one would not have left himself reminders to follow up on “pending” tests (because this test was no longer considered “pending”).
Proteinuria may not be on the “critical results” list in all hospital labs. Many labs we have seen do include it as a “critical result” for pregnant patients. However, most IT systems lack foolproof ways of identifying pregnant women. So they tend to use as a workaround the service of the attending physician. That works fine if the patient is on the obstetrics service and has an attending clearly identifiable as an obstetrician. But what if the patient is instead on the service of a family medicine physician (who may have obstetric privileges but is listed on the computer system under family medicine rather than obstetrics)? Or on another service all together? Or in the emergency room?
The case, though, also has important implications from the perspective of our prior discussions on diagnostic error.
Did the physician appropriately interpret the result of the test?
Did the physician document the result in the patient record? Did the physician indicate his/her interpretation of the test results as it pertains to the patient? Did the physician develop an action plan to deal with the test result? Note that an action plan may be required both on tests with abnormal results and tests with normal results.
Was the patient notified of the result? Was the patient made aware of the implications of the test result and the next steps? Was notification timely? Who did the notification and how was the notification done and was that appropriate? Could that notification have taken place via other means that might have still been appropriate but more timely or more convenient?
Did the test address the reason for which it was originally ordered? Did the test have to be repeated? Did the test lead to further tests? Would those tests have been done anyway if the abnormal test had not been done in the first place? Did the test result (normal or abnormal) have meaningful impact on the patient’s care?
We don’t have time in today’s column to address all the cognitive factors to be considered in assessing the impact of lab tests on diagnosis. In our prior discussions of diagnostic error we have talked about many of the processes that can promote diagnostic error (anchoring, premature closure, confirmation bias, etc.). But looking at how a test result influenced (or did not influence) the diagnostic reasoning may be illuminating for both the physician and others.
If an error did occur, was the patient be informed of the error? What if no patient harm occurred? Who was involved in disclosure and apology to the patient? We’ve done numerous columns on disclosure and apology. A good discussion on communicating pathology and laboratory errors from the perspective of the pathologist and laboratory was published recently (Dintzis 2011).
You can see that using the tracer methodology can reveal a lot about your vulnerabilities in the “lab error” loop. It is, however, a very labor-intensive and time-consuming methodology. Ultimately we would hope to be able to computerize or otherwise automate parts of this process so that it could be utilized more efficiently.
The vast majority of occurrences called “lab errors” are obviously errors in our overall systems of care. Many of the same considerations apply equally well to radiology and diagnostic imaging. We need to do a better job of assessing where those potential vulnerabilities in our systems are and take steps to close those gaps.
Some of our other columns on errors related to laboratory studies:
See also our other columns on communicating significant results:
Hammerling JA. A Review of Medical Errors in Laboratory Diagnostics and Where We Are Today. Lab Medicine 2012; 43: 41-44
Plebani M, Lippi G. Closing the brain-to-brain loop in laboratory
testing. Opinion Paper. Clinical Chemistry & Laboratory Medicine 2011;
49(7): 1131-1133, July 2011.
Lippi G, Simundic A-M, Mattiuzzi C. Overview on patient safety in healthcare and laboratory diagnostics. Biochemia Medica 2010; 20(2): 131-143
National Quality Forum. Preferred Practices for Measuring and Reporting Patient Safety and Communication in Laboratory Medicine: A Consensus Report. April 2009 last updated May 6, 2011
Plebani M. Laboratory errors: How to improve pre- and post-analytical phases? (Editorial). Biochemia Medica 2007; 17(1): 5-9.
Mohta VJ. Amended Lab Results: Communication Slip. AHRQ Web M&M February 2012
Dintzis SM, Stetsenko GY, Sitlani CM, et al. Communicating Pathology and Laboratory Errors. AJCP 2011 135:760-765