Either a patient needs antibiotics or they don’t. If a patient needs antibiotics, a delayed prescription is harmful. If a patient doesn’t need antibiotics, a delayed prescription is harmful. Thus, delayed antibiotics are dumb.
That paragraph could have stood alone as my shortest (and perhaps best) blog post ever. However, the practice of providing patients with a prescription for delayed antibiotics has become so popular, I should probably expand on my thinking a little more. Like anything in medicine, there is likely some grey area to explore.
Proponents of delayed antibiotic prescriptions will point to a lot of evidence that they are associated with a decrease in antibiotic use. (Spurling 2017) Although they are technically correct, the decrease in antibiotics is compared to control groups that are practicing bad medicine. If doctors are giving 100% of patients with viral URTIs antibiotics, and the delayed prescription gets that rate down to 80%, that still means that 80% of patients are being harmed by bad care. I don’t think we should count that as a win, but that is exactly what is happening in this research. Depending on the study, as many as half of patients still end up filling their prescription. Of course, seeing as antibiotics don’t help viral infections, the clinical outcomes are exactly the same as giving no antibiotics at all. (Spurling 2017)
This practice is directly harmful, in that it unnecessarily exposes patients to the many harms of antibiotics. It also has probably results in many indirect harms. It is incredibly confusing to patients. We are simultaneously telling them that antibiotics are unnecessary for their condition and yet giving them a prescription for an antibiotic. How are those conflicting actions supposed to be interpreted?
The whole point of delayed antibiotic prescriptions is to decrease antibiotic use, but compared to simply not prescribing antibiotics, delayed prescriptions result in a dramatic increase in antibiotic use. (Spurling 2017) In other words, this strategy completely fails at the only objective it was designed to achieve. If you want to decrease antibiotic use, stop prescribing antibiotics when they are not necessary. Its that simple.
I think there might be a few reasonable exceptions to the black and white picture I have painted so far. However, I don’t think any of them are represented in the standard conversations about delayed antibiotics. Viral URTIs don’t need antibiotics. Otitis media doesn’t need antibiotics. Whether you give them now or later, you are probably causing your patient harm.
However, there are a number of conditions that do require antibiotics, but that aren’t always clinically obvious. Sometimes a patient comes in with hot, red, swollen skin that you are pretty sure is just a local reaction to a bug bite, but it could be cellulitis. The ideal answer is probably treating the bug bite and having the patient reassessed, but realistically, even in a country where there isn’t a direct cost to see a doctor, follow up appointments are inconvenient, and there are many indirect costs that arise from that repeat visit. Therefore, I will occasionally give a prescription for cephalexin to be started in 2-3 days if the lesion doesn’t improve with time, ice, and antihistamines. However, this scenario is drastically different from the scenarios in which delayed antibiotics are typically employed, because I am actually treating a bacterial illness for which antibiotics are the appropriate therapy. Antibiotics are not appropriate for a viral pharyngitis, no matter how long you wait.
It is true that some patients want antibiotics and are more satisfied with their visit when they receive their prescription. However, when a patient has misunderstandings about medicine, the appropriate approach is to educate and guide them, not try to trick them with a delayed prescription.
Realistically, the delayed antibiotic concept was probably more about trying to change doctors than patients. Antibiotics over-prescription is a doctor problem, not a patient problem. We are the ones writing the prescriptions. We are the ones failing to teach our patients.
Behavioral scientists may disagree with my rant. They may point out that doctors are stubborn and stuck in their ways; that doctors are unlikely to stop writing unnecessary prescriptions, but we might be able to convince them to write a different date. If you want to admit that you are this stubborn, fine, but it doesn’t change the fact that delayed antibiotics don’t make any sense.
I will close with a quote from the excellent editorial that prompted this rant: “Ultimately, delayed antibiotic prescriptions expose patients to unnecessary and potentially harmful medications and do not improve clinical outcomes… Health care professionals have a responsibility to do what is in the best interest of patients even if there are concerns that doing so will make some patients unhappy or unsatisfied. For patients with viral conditions, that responsibility means not prescribing immediate or delayed antibiotics.” (Rowe 2020)
References
Rowe TA, Linder JA. Delayed Antibiotic Prescriptions in Ambulatory Care: Reconsidering a Problematic Practice. JAMA. 2020;10.1001/jama.2020.2325. doi:10.1001/jama.2020.2325 PMID: 32297898
Spurling GK, Del Mar CB, Dooley L, Foxlee R, Farley R. Delayed antibiotic prescriptions for respiratory infections. Cochrane Database Syst Rev. 2017;9(9):CD004417. Published 2017 Sep 7. doi:10.1002/14651858.CD004417.pub5 PMID: 28881007
Find more rants about medical topics here.
Morgenstern, J. Delayed antibiotics are (mostly) dumb, First10EM, October 5, 2020. Available at:
https://doi.org/10.51684/FIRS.46294
Image by mohamed Hassan from Pixabay
14 thoughts on “Delayed antibiotics are (mostly) dumb”
”It takes 10 minutes to educate your patients about why they shouldn’t get antibiotics , but it takes 3 minutes to prescribe it”
My dad (private practise , Denmark)
This , I think, is the core of the problem with overtreatment
The current default is sadly to treat (more is more – mentality) , and most forces in this process are against the doctors interest (maybe unsatisfied patients , takes more time , requires a confrontation , if we are wrong there’s the “risk proximity problem” https://www.stemlynsblog.org/risky-business-risk-proximity/ etc)
How to fix that problem ?
A good start is reminding us of our duty as doctors (I.e Hoffman’s “fiduciary” argument 10-17:00 (https://m.youtube.com/watch?v=1p-UwJXNwwY)) and reminding us that extra time spend on communication/ education / compassion and validation is time won (I.e fatovich : https://research-repository.uwa.edu.au/en/publications/the-time-paradox-of-emergency-medicine-another-inverted-u-curve). That might not be your time won, with this particular patient but overall it is time won (less crowding (emcases ep 129) , less overtreatment and overdiagnosis , less re-visits etc)
However I think it has to be more system-level than saying some doctors are bad apples (certainly this is not untrue , but I’d argue it’s more because of the pressures on that doctor stated above , than ignorance ). Stuff like
– expectation management I’m society
– lessening the outright propaganda and harm that the Medicine-industrial complex (big pharma) is putting on guidelines , priorities in healthcare , politicians etc
– high quality knowledge sites (like you do or like Casey Parker has done on the sore throat aspect of this topic – https://broomedocs.com/2014/09/on-sick-kids-sore-throats-swabs-and-such/)
– probabilistic / Bayesian diagnostics
– And so on
I think you would agree on most of this if I’ve read your blogs right (please correct me if I’m wrong )
I do however disagree on one thing:
You write
“follow up appointments are inconvenient, and there are many indirect costs that arise from that repeat visit“
I work in a system (Denmark and Sweden ) where appointments are free(ish). I think one of the strongest part of our system is the bond that the family doctor has to his patients
– he / she knows them (therefore has a larger interest in educating them for future appointments )
– he / she can by saying “come back in a few days” (I.e. your post test probability is currently below a treatment threshold ), give a chance for disease to regress to the mean (Where the patient will not come back) or evolve so that a better harm:Benefit ratio now so that treatment is more indicated (kind of like “time as a test when the post test probability is low at first visit) . Few times they will have to come to me at the ED in this waiting period .
I think this is a method that decreases overdiagnosis and overtreatment
I think the notion that everything “should be fixed today “ is driving some of the overdiagnosis and overtreatment , not allowing the powerful “time as a test”, especially if people come early
I might be totally off , though , and can only speak of our system in Denmark and Sweden
Again I truly appreciate your inspirational- and thought provoking posts and podcasts
All the best
Peter , EM resident , Denmark / Sweden
Thanks for one of the best written, most thorough comments the site has ever had.
In terms of repeat visits, a lot depends on your baseline – and that one throw away comment certainly can’t cover the full breadth of the topic. I start at a baseline of using LOTS of follow up appointments, even in the ED. Ultrasound inconclusive for appendicitis – the answer is a repeat exam in 12 hours rather than an immediate CT almost every time. CHF that is borderline for outpatient management: give them a day to try at home and recheck. Corneal abrasions don’t need opthalmologists, they just need another emerg doc to recheck them if there are any ongoing symptoms.
All hospital and doctor visits are free in Canada. However, when making patients come back to see us, we tend to focus only on the medical stuff and forget about the other expenses of a doctor’s visits – parking, days off work, childcare, etc. The comment was mostly aimed at making people consider all factors in these decisions. I have a great relationship with my family doctor, and would be happy to be reassessed, but who has time for extra unnecessary appointments in their life? I would take every opportunity to avoid an extra visit, which is why the delayed antibiotics may occasionally have a role.
Thank you so much, Justin! And thank you for your thorough and quick answer!
I agree! 🙂
“who has time for extra unnecessary appointments in their life?” – Who can disagree with that! (I think it’s hard though – modern society and busy living often demands things to be fixed now. Even if that often is impossible, and even harmfull, and better to wait)
If it’s “time” (and low risk of dangerous progression) VS radiation (or antibiotic side-effects or any other harm), I think I’ll choose time, but maybe it should be a shared decision making thing (unless the condition is below the 2% threshold of “meh”). As long as you are informed that your NNT might be much lower if you are treated early, rather than late, but your NNH is the same (i.e lower harm . Benefit ratio – like the figure from this St Mungos blog https://stmungos-ed.com/blog/noteverythingthatcounts )
Simon Carley made this great illustration, with an inverse relationsship between “probability of error” and “disease manifestation” (from what I believe to be one of the best blogs ever from St emlyns. I Use it and your “why pretest probability is essential”, and “EBM is still the best..” and the guideline-rant blogs, frequently in my danish FOAM blogs, akutmedicineren.dk): https://www.stemlynsblog.org/making-good-decisions-in-the-ed-rcem15/ (slide 13-15).
I think it, in a picture / model, describes why we have to wait if treatment right now is not needed. The error here is, as you’ve written about several times as well, not just omission but also comission (i.e giving antibiotics).
But for the low risk stuff, as long as the antibiotic stewardship is not totally off, I think your wise suggestion of getting the “patient values” part of EBM (i.e do you want to wait or not?) into the conversation as well. That may be the one time for delayed antibiotics.
Cant predict future! Pt may need antibiotics in a few days.. Sx may become more evident then with wound infection, sinusitis etc..
How do u know “URI” is viral for sure or wont go into a bacterial atypical bronchitis that would benefit from antibiotics?..I rarely give delayed antibiotic rx but sometimes I feel its prudent esp in these corona days with poor access to pcp..
I think the reason for the post was to try to divide some of those conditions out more. If the condition is something that should be treated with an antibiotics, but there is clinical uncertainty – then a delayed prescription in place of a repeat visit may be reasonable. However, if the condition does not require antibiotics, the delayed prescription is just harmful. I would argue that most of the delayed antibiotics prescribing, including two of the examples you give, fall into the later category. Sinusitis and bronchitis essentially never required antibiotics and are exactly the kind of conditions this post was aimed at.
I guess it’s all about what Harms-Benefit ratio you, and your patient is comfortable with, given the current situation. In my oppinion these decisions are all about numbers – how many of this particular patient case, would progress to somewthing worse, considering my assesment of this patient (and if it progresses, would it really have been wiser with “earlier is better”? – I think we have a lot of pressures on us, to say yes to that question. But I don’t think the evidence backs it up (please correct me if I’m wrong)), with Vs without antibiotics? What is an “acceptable missrate” to unneccessecary antibiotic use? 1:10? 1:100? 1:1000?
Gilbert Welsh talked about cancer treatment and overdiagnosis in terms of three animals, and our treatment as a “fence” to try to catch them:
– Turtles:diseases that don’t need fences / low risk of development
– Rabbits: the ones we want to catch (reversible diseases if caught early) with our fence
– Birds: no matter what you do, they are gone. Fences don’t catch birds
I think much of our angst / declining inability to tolerate uncertainty as doctors and as a system (https://areasoci.sirm.org/uploads/Documenti/SDS/828b6893d23b878ccb2be267ac322d6a1c91085c.pdf ) is because we hear statistics on birds, and apply it to turtles (indication creep). Most turtles are young, most birds are old and frail, and rabbits are everyone inbetween
Even Mervyn Singer (the sepsis 3.0 guy), wrote about what he called “sepsis hysteria”, and how it’s harmfull (https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(19)32483-3/fulltext?utm_campaign=lancet&utm_source=twitter&utm_medium=social )
I share your sense of uncertainty with these patients. But I think we need to be better as a system, as a culture and as individuals at handling this uncertainty. For me the best way is through shared decision making and informing the patient (hard though, if the reason why you are prescribing antibiotics in the first place is because you don’t have 10 min to discuss)
You mentioned otitis media does not require antibiotics. This is the only situation where I will sometimes utilize a delayed rx approach. I have done this based on the 2016 CPS guideline. Am I out of date? If I do it, it’s in the well appearing child who has already had a few days of fever and instruct to give based on objective measures (ie persistent fever at the 48h mark). I instruct parents it’s possibly viral, possibly bacterial but either way likely to resolve on its own. Only select cases that will go on to require antibiotics. Happy to hear your approach to moderately ill confirmed AOM kids over 6 months old.
Yeah – sounds like that need a review in itself, but I am pretty convinced by the studies that there is no benefit.
There have been a few “positive” studies, but with a few red flags. They essentially all come from the same lead author, and the outcome that is positive in each trial changes, but is essentially never something I care about. For example, there might be a decrease in asymptomatic ear effusion 3 month after the AOM if you use Clavulin. Does doesn’t sound like an important outcome, and almost certainly doesn’t justify the side effect of the antibiotics.
My approach is pretty simple. Otitis media is viral and doesn’t get antibiotics. I may break this rule for high risk children, such as the immunosuppressed of the unvaccinated, but I don’t know that I am actually helping there.
Interesting. That may be your interpretation of the literature, but definitely deviates from any major guidelines on AOM Management. Do you think the guidelines will catch up to your understanding with future revisions or ever become the standard of care? Or if this is the approach taken, you have to be comfortable practising different than the majority of EM providers and therefore open yourself up to some liability?
Guidelines seem rather resistant to evidence. Some are better than others, but based on the topics where I have reviewed the evidence myself, one cannot expect the guidelines to adequately reflect the evidence.
Luckily, guidelines do not set the standard of care. If you practice based on the best available science for the good of your patients, liability shouldn’t be a concern. The best defense against being sued is involving your patients in decision making. The only way to really do that is to understand the evidence. For otitis media, I can tell them that antibiotics don’t decrease pain or fever (the only things they really care about). There may be a small amount of decrease in the amount of fluid behind the ear drum if someone looks in 2-3 months, but that doesn’t seem to affect children at all. On the other hand, antibiotics are associated with a relatively high rate of side effects like allergies and diarrhea. Overall, the evidence suggests your child is much more likely to get diarrhea from an antibiotic prescription than be helped by it.
When presented that way, I have never had a parent request antibiotics (so much so that I have mostly stopped having the conversation). But if you are concerned, have a quick conversation, and you can document that it was the patient’s (parent’s) decision, not yours independently.
I’d say it takes an impressive level of confidence to be able to definitively say that certain inflammatory things aren’t infections. Questionable cellulitis, strep? (yes, that one is a long discussion), bronchitis v pneumonia, early ear infection (v resolving spontaneously). Are you a proponent of testing for such things regularly and accepting the limitations that the tests hold as well as the expense. I don’t think your rant on this one is realistic or necessarily represents better medicine.
I think the point is not to distinguish between inflammatory and infection. Delayed antibiotics are almost always used in scenarios where there is an infection, but in which antibiotics don’t work. Your example of ear infection is perfect – even if the diagnosis is definitively otitis media, the treatment is time and analgesia. Antibiotics have never been shown to have an important benefit, but do have lots of side effects. So the delayed antibiotic doesn’t make sense. Cellulitis on the other hand needs antibiotics. Uncertainty between inflammation and infection in that context is a great reason to use delayed antibiotics, and is in fact the example I use in the post.
Hi.
If you go deep on this it concerns accountability. Physicians should not be free to prescribe wherever they want, they should be limited to scientific guidelines. In these modern times, with electronic prescribing, we know what each doctor prescribes, and if that’s wrong he should be held accountable. This should be even more stringent in public ERs/Hospitals, why should you have 100 options when 2 are enough, why should you prescribe this when nothing is needed?
I always scream inside when another doctor tells me “Curtail my freedom, NEVER” YES, CURTAIL YOUR FREEDOM WHEN ITS BASED ON IGNORANCE OR ADVERTISING
ER physician from Portugal
Where are these “scientific guidelines” you speak of 😛 ….
Guidelines usually are biased by conflicts of interest ; are often overstating their recommendations (i.e no wiggle room); and often don’t show NNT / NNH, or describe what was the motivation for their recommendation (so it’s hard to utilize it, if your patient is atypical) (Jeanne Lenzer and Jerome Hoffman have written about this for a long time, Justin too) – https://static1.squarespace.com/static/58d417c8c534a598c8c22474/t/59df8d8cc534a50f7d6e2106/1507822989928/BMJ+Why+we+can%27t+trust+clinical+guidelines.pdf + https://www.youtube.com/watch?v=5DoORqZYYkY
Gary Klein (decision making psychologist) would say that guidelines and algorthims only work in “simple” envoirenments, and we usually work in complex ones. So for guidelines to work, the physician would need to be able to reduce the patients problem (through communication and clinical assesment) – I think this sometimes can be done for infectious diseases, but often we need to keep it in the “complex” realm (where guidelines often overstate their visit).(https://www.researchgate.net/publication/300343833_Can_We_Trust_Best_Practices_Six_Cognitive_Challenges_of_Evidence-Based_Approaches )
I agree that optimal guidelines (if they will ever excist), where there’s a list of NNT / NNH for different conditions, and their the choice is left up to the doctor (choosing mavbe from a top 2 or 3 list) would be preferable for antibiotics. For antibiotics I think we are pretty close to IRL optimal guidelines in scandinavia, and people are using very few antibiotics https://pro.medicin.dk/Artikler/Artikel/6#a000
Ususally it’s set up like this, with three choices:
– Primary treaetment option
– Secondary treatment option (if failure)
– Allergic to primary and / or secondary option
All the best
Peter