This is the first post in an ongoing discussion of harmful, wasteful, and useless practices.
My friend Avi Cooper recently posted a list of some medications which can cause false positive testing on urine drug screening (UDS) assays. The source for that list is here.
Just seeing this list brought to mind all the times I have seen drug screens used to support judgmental and severely cognitively biased decision making. And it made me angry to think about the implications of this testing.
For further illustration, consider this vivid example from Stellpflug, Cole, and Greller:
“The test does not discriminate; if there is a physical interaction, it turns positive, regardless of whether the interaction is truly with the drug of concern. For example, the amphetamine portion of a UDS (typically designed to actually detect amphetamine) can result positive in the presence of pseudoephedrine or phenylephrine. The PCP (targeting PCP) screen can result positive by reacting with dextromethorphan. The tetrahydrocannabinol (THC) screen can be made positive by some nonsteroidal anti-inflammatory drugs (in laboratories without up-to-date capabilities), as the TCA screen commonly turns positive in the presence of diphenhydramine and other similarly structured drugs, and the cocaine screen can turn positive with the ingestion of coca tea. These examples culminate in an odd scenario in which someone aggressively treating a cold could feasibly have a UDS “positive” for amphetamines, cocaine, opioids, PCP, THC, and TCAs.”
The authors go on in the paper to state “False positives, false negatives, broad time frame of detection, and delayed confirmation all make the UDS inapplicable to real-time clinical decision making.”
The calculus is not improved in the ICU. The UDS almost never provides actionable information and serves to bias our opinion of the patient and their clinical syndrome.
A few examples of the ways this testing can lead us astray:
- “The UDS is positive for [X], so we can stop looking for causes of the encephalopathy”
- “They’re probably just delirious because they haven’t had [X] in a while, it was on their admission UDS”
- “UDS was negative, I doubt this was a toxidrome at all”
- “If we start them on this medication, will they even follow-up? Their UDS was positive for THC”
Bad data is worse than no data. I’m struggling to remember a time when a UDS had a net positive impact in the care of a patient.
If ordering a UDS is a common practice for you, I ask you to reflect on the risk:benefit ratio and whether it makes sense at all.
- Further reading: a fun note “breaking up with the UDS” by Greller at Tox and the Hound (hosted on Emcrit).
- Header image by Steve Buissinne from Pixabay
I am a huge fan of the UDS, and I order them frequently, almost always with positive effect. Working for 7+ years in a community ICU where drug OD or drug related issues were the primary cause of admission, they were indispensible. Frequently the tyros who came to work with us would sign out bizarre cases after going on a big Zebra Safari and I’d say “Meth. It’s Meth. Have you gotten a drug screen?” Very often that was the neglected piece of the puzzle. Recently a person with near-fatal asthma and recurrent admissions was referred for biological therapy, to which they never showed. Denied drug use. I told team to order UDS. “It’s going to be Meth.” I was wrong. It was cocaine, and it has been positive on every subsequent admission. Not knowing that would have led to harm and waste with expensive biological therapy and would have prevented identifying the true culprit and counselling in that regard.
The list of rarefied substances that cause false positives is not meaningful without a statement about specificity. No test is 100% specific, and a Bayesian approach can guide the use of a test with inperfect sensitivity. Missing from the discussion is the woeful sensitivity of patient history for use of substances. So we actually want a very sensitive test, and we can sort out the false positives with highly specific serum assays when there is unceratinty. Also, there are instances where the use of substances that can lead to false positives are far less likely than the actual substances we’re looking for.
Finally I resist the notion that I must remain ignorant of substance use because that is the only way patients may be protected from my ineluctable biases resulting from their unsavory habits. I prefer to work with full information, so long as it may be useful, and given the very high prevalence of alcohol, drug use in ICU populations, and their contribution to mortality and morbidity, I think I would be negligent to NOT get UDS whenever the suspicion arose – if only for the counselling opportunity.
Scott, thanks as always for a thoughtful take on things.
One issue I see, which is not unique to the UDS: the discernment of the clinician ordering the test plays a huge role. There are plenty of diagnostics and tests which I think are useful in the right hands, which are useless or harmful in the hands of many. The pulmonary artery catheter, or even right atrial pressure measurements come to mind. Sadly, the same may happen someday with echocardiography for the same reasons.
The Bayesian approach for all tests is essential, but oft not employed. The issue is the lack of discerning thought that goes into this very broadly ordered test.
Your counter argument is useful, but I still think it would support much more narrow ordering of the UDS than is currently employed.