In 2012, the American Board of Internal Medicine (ABIM) Foundation introduced the Choosing Wisely initiative, with “the goal of advancing a national dialogue on avoiding wasteful or unnecessary medical tests, treatments and procedures.”(1)   I have had the privilege of working closely in this effort with the American Society for Clinical Pathology (ASCP), the organization that represents the specialty of Pathology and Laboratory Medicine to the Choosing Wisely initiative.

Since its inception, the Choosing Wisely project has produced a large series of specific recommendations, each released as 5-item lists, from each of the sponsoring societies spanning a diverse array of medical specialties.  ASCP has released a total of 10 specific recommendations, with the third list of 5 coming soon in 2016, and other lists to follow.  In addition, many of the recommendations made by other clinical medical societies are centered on clinical laboratory ordering practices, further emphasizing the central role the clinical laboratory plays in the effective practice of medicine and the judicious use of medical resources.

The Choosing Wisely lists are generated from recommendations drawn from the individual contributing societies’ members, and are then subject to peer review of available evidence.  After a thorough vetting process they are released in the form of the above-mentioned lists.  The specific lists of recommendations—from ASCP and from many other medical organizations—are available at www. choosingwisely.org.

For the clinical laboratory, recommendations range from guidance on the prudent use of routine testing (“Avoid routine preoperative testing for low risk surgeries without a clinical indication”) to guidance on more esoteric laboratory testing (“Only order Methylated Septin 9 (WEPT9) to screen for colon cancer on patients for whom conventional diagnostics are not possible”), and recommendations to end certain practices entirely (“Don’t use bleeding time test to guide patient care.” “Don’t perform population based screening for 25-OH-Vitamin D deficiency.” “Don’t perform low-risk HPV testing.”)

Sunsetting Obsolete Tests

In my experience it can be more difficult to sunset an obsolete clinical laboratory test than it is to onboard a new test.  Many of us remember the persistent demand for “LE cell preps” for years after the discovery that the root cause of the formation of “LE cells” in vitro was the presence of anti-nuclear antibodies that could be evaluated directly with a targeted, more accurate, more precise ANA assay.(2)

Likewise, the subjective and labor-intensive leukocyte alkaline phosphatase (LAP) score for the prediction of chronic myelogenous leukemia (CML) was replaced by the highly sensitive and specific BCR-ABL fusion study;(3) but demand for the LAP score persisted, and many physicians in training were counseled on its use long after it became obsolete.   The Choosing Wisely initiative gives us a toolbox for the data-driven, peer-reviewed discontinuation of tests that may have persistent champions in medicine, but that are simply not indicated in modern practice.

Utilization Management

The increased attention in recent years to utilization management in clinical laboratory ordering practices has been a very positive development toward the “triple aim” of improving the patient experience, improving the health of populations, and reducing the per capita cost of health care.(4)  Such efforts are moving beyond the ineffective gatekeeping that traditionally took place at the level of the clinical laboratory after a given test had been ordered, and are now moving toward more effective institutional utilization management committees, laboratory formularies, and other institution- level (and in some larger health systems, enterprise-wide) control models.(5)   But effective laboratory test utilization transcends the recommendation of specific ordering practices.

In his book TheSeven Habits of Highly Effective People, author Steven Covey famously advised us to “begin with the end in mind.”(6) The same is true for the ordering of clinical laboratory tests. Ordering large panels or disparate arrays of numerous laboratory tests is very likely to yield one or two abnormal results even in healthy individuals, and those abnormal results may have very little medical value in the absence of pre-test suspicion for disease.

In an editorial in a previous Warde Report, I outlined the perils of “shotgun testing,” and emphasized that the ability of an abnormal laboratory test to predict the presence of disease goes down measurably in proportion to the decrease in the pre-test probability of disease.(7)

There does not exist a test that is exempt from the possibility of false positive results— each of us has the potential to harbor a cross-reactive antibody, a normal concentration of a given analyte that is out of sync with most of our fellow beings, or a transient but clinically insignificant abnormality that over time will normalize.

For some analytes, even the most cutting- edge methodology may lack precision and yield high coefficients of variation (CVs).  For these reasons, as the pre-test probability of disease decreases, the relative proportion of false positives as a percentage of total positives increases, reducing the predictive value of the test.  At the extreme, a positive test for a disease that no longer exists has essentially a 100% probability of being a false-positive, and a predictive value of zero.  Likewise, the predictive value of a positive test ordered for a patient who is extremely unlikely to have a given disease (say, the ordering of an ovarian cancer marker in a male), is likely to be essentially zero.

In essence, clinical laboratory results should be viewed as tools by which to provide likelihood ratios for adjusting pre-test probability.  If a patient’s likelihood of having a particular medical condition (based on history, physical examination, perhaps radiographic information, etc) is deemed to be, say, 5%, then a clinical laboratory result that increases that likelihood by 400% still only yields a 20% probability for that particular condition.

This explains the surprisingly low predictive values of some of the best-performing laboratory tests when ordered in low prevalence populations. Modern HIV screening methods, for example, have stellar performance characteristics when measured over entire populations (typically with sensitivity and specificity exceeding 99%), but may have low positive predictive values (perhaps dipping below 50%) for individual results within very low prevalence populations.(8) This in turn explains the apparent paradox of the need to perform additional confirmatory testing for certain conditions for which screening methods demonstrate outstanding performance.

We need to evolve to a culture in which tests that simply do not perform to modern standards of precision and accuracy quickly yield to more current and predictive tests.  But beyond that, every test that is ordered should be ordered with a clear view of how any particular result will affect clinical decision-making and diagnostic probabilities. Otherwise the value of laboratory data is measurably diminished. It is the solemn obligation of each member of the medical community to choose wisely.