Time Stamps

  • 04:10 Review of Choosing Wisely Guidelines
  • 05:55 What are estimated incidences of positive blood cultures?
  • 06:17 Which clinical signs or patient history is more likely to have positive blood cultures?
  • 09:48 What is the estimated incidence of false positive blood cultures?
  • 11:01 What infections more likely to have bacteremia?
  • How do positive blood cultures affect our management?

Show Notes

  • Why do we get blood cultures? Remember, it is to specifically look for bacteremia, which has a higher mortality rate than many other infections.
  • One study found that only 7% of all cultured patients were found to have bacteremia. 
    • When these patients were stratified into high and low risk groups:
      • The “high risk” group had about 15%  positive blood cultures. 
        • High risk group included:
          • Fever over 38.3 degrees celsius or 101F 
          • Major comorbidities (bone marrow transplant, severe pancreatitis, ARDS or acute on chronic liver failure, coma, brain death, cardiac arrests, bowel perforations, multiple traumas or burns)
          • IV drug users
          • Patients with rigors
      • Those considered low risk had only 2% positive blood cultures.
  • The JAMA Rational Clinical Exam Series found
    • Fever > 100.9 F was associated with a 2.0 likelihood ratio of having bacteremia.
    • 2 or more SIRS criteria met was associated with a 1.8  likelihood ratio. 
      • Having only 1 SIRS criteria, the likelihood ratio for bacteria was 0.09.
  • It is really important to consider the pretest probability before obtaining cultures, since low risk patients can have about a 7% false positive rate.
    • Estimated pretest probabilities of bacteremia in cellulitis, ambulatory patients and community-acquired pneumonia have been identified to be 0.02, 0.02 and 0.07, respectively.
      • *It’s important to note that one of the studies estimating pretest probability patients with a fever requiring hospitalization enrolled only after blood cultures had been drawn, meaning they were not consecutive inpatients with fever. Therefore, 0.13 is likely an overestimate the true probability of bacteremia in febrile patients that get hospitalized.
    • Culturing more patients led to increased length of stay and costs.
  • For GU infections, we may not need to obtain blood cultures because we can likely identify the organism with the urine culture alone 
  • Patients who have suspected endocarditis, bacterial meningitis, or septic shock, should ALL get blood cultures to help determine appropriate antibiotic usage and coverage.
    • Estimated pretest probabilities of bacteremia in severe sepsis, bacterial meningitis and septic shock have been identified to be 0.38, 0.53 and 0.69, respectively.  
  • We have yet to define an algorithm that determines risk severity and categorizes who should be cultured.
  • Think carefully through your patient’s clinical scenario, even if you go by “the book” and culture everyone. Be thoughtful about who you’re culturing and what you expect the cultures to show and do for you.


Culture Club.

Laments of a Clinical Clerk

Of all my consultants, most easy to please

Is the fellow who comes from infectious disease.

His wants are so simple! His needs so are so few!

Just gather some sputum, blood cultures times two

X-ray the patient from guggle to zatch

Examine the urine, both cath and clean catch;

It takes but a moment to do an L.P.,

Swab wound, throat and cervix, yank out the I.V.

When all of the data at least are collected,

The last culture plated, the last slide inspected,

The attending arrives to review and recap

(While intern and student enjoy a brief nap);

He broods with the air of a scribe with papyrus

And gives his opinion: “Most likely a Virus,

Don’t bother to fix it; Can’t Treat it, Cant Cure it,

Though superinfections may later obscure it.

Should there be recurrence of fever or pain

Go back to square one and start over again!”

Frank JB. Laments of a clinical clerk. N Engl J Med 1978

J Hey Steve

S What’s that Janine

J So I saw a patient in clinic the other day that was feeling kinda lousy and reporting that they had a  fever and maybe some rigors? He said “I’ve got chills…”

S Were they multiplying?

J (laugh)  Steve. Well I was thinking the funny thing is, if that patient were admitted to the hospital I’d probably do a full workup, right?

S I guess so?

J Play along. What would you do?

S Sure, I got this. Intern bread and butter… I’d get labs, an X-ray, maybe blood cultures.

J *Maybe* blood cultures?

S I guess if your fever is high enough?

J Does fever height correlate to blood culture positivity?

S I dunno, maybe cause they were rigoring? I don’t know what’s going on but I don’t like this role reversal, Janine. The Socratic method is what I used to do to you when you were my resident…

J Haha, now you know how it feels!

S Well, all for a good cause. Lemme guess, you’re trying to get us to talk about blood cultures, right?

J How’d you know, Steve? That IS our topic for today 

S What a surprise! *said with sarcasm*

J Today we’re going to go over:

S 1) We’ll start with a basic review of the guidelines for blood cultures.

J 2) We’ll look at what studies say about which patients are more likely to have positive blood cultures

S 3) Even if you can guess who has a positive blood culture when is it a good idea to get it?

J And this point really has three smaller considerations. Baby points, lets call them 3A through 3c.

S But all super important. So let’s start with 3A) What are general incidences of positive blood cultures. And the much hated false positives… 

J 3B) are some infections more likely to have positive blood cultures, also known as bacteremia

S Leading us to 3C) even if the culture is positive, are all positive results created equal? How do positive blood cultures affect our management?

J Such good baby points!

S Sure, lastly to point 4) this is going to be a lot of information, so we’ll try to give some advice on who we should culture and who we should leave alone.

J We’re going to challenge the medical culture 

S Pun definitely intended

J of culturing every inpatient with a fever and see where the data leads us 

S And hopefully taking a deeper dive into this topic will help all of us become more informed practitioners

J And reassure all those poor night float interns that are out there culturing patients at 2am…

S Well I’m not sure we’ll do that

J Just trying to give them some hope, Steve!


J Hi I’m Janine Knudsen

S And I’m Steve Liu

J Welcome to mind the gap

S A Core IM podcast

J We’d like to thank Dr. Aditya Shah, Chief infectious disease fellow at the Mayo Clinic, for peer reviewing this episode 

S Subscribe for our show notes at CoreIMPodcast.com

J And follow us on insta and twitter

S So let’s start off with what the guidelines say about getting blood cultures.

J For this we’ll be looking to the Choosing Wisely campaign – my favorite!

S From the Choosing Wisely campaign: “Don’t perform urinalysis, urine culture, blood cultures, or c dif testing unless patients have signs or symptoms of infections. Tests can be falsely positive leading to over diagnosis and overtreatment.”

Well that was pretty simple.

J That’s what the Choosing Wisely Campaign is all about! For those of you who don’t know, it’s an initiative of the ABIM

S For our listeners who aren’t familiar with them, that’s the The American Board of Internal Medicine 

J They aim to steer providers away from unnecessary or harmful testing, towards more high value care. 

S Check them out!

J Ok, but for the purposes of this episode, we’re going to get just a little more nuanced

S Just a little

J So to start, let’s focus on who is the most likely to have a positive blood culture?

S Like does a high fever, or fever at all, correlate with a blood stream infection? AKA bacteremia? Because that’s what blood cultures are looking for, right?

J Yes, let’s be clear about that – the reason for getting blood cultures specifically is to look for bacteremia, which is associated with a higher risk of mortality than other infections.

S So some folk, actually the same ones cited in that choosing wisely recommendation, decided to look into who was more likely to have bacteremia in 1989. Their paper was called: “Predicting Bacteremia in Hospitalized Patients: a Prospectively Validated Model

J That title pretty much sums up what they tried to do. They compared patients that had bacteremia to those who didn’t to try to figure out if there were particular risk factors associated with bacteria in the blood.

S Ah, a case control study.

J Well sort of. They then used this analysis to create a predictive model which they applied to a second cohort of patients in order to validate it.  

S So what did they find? Which patients were most likely to have positive blood cultures?

J The authors found that only 7% of all cultured patients ended up having bacteremia, but the rates were higher in the high risk group (up to 15%) and low – only 2%! – in the low risk group

S Really, even in this high risk group only 15% of cultures were positive?

J Yep!

S Wait, who was more likely to be high vs. low risk? 

J That’s what they used multivariate analysis for. They identified that the following characteristics were associated with a higher risk of bacteremia: 

S IV drug use, high mortality rates within a month, fevers over 38.3 degrees, rapidly fatal disease (defined as death within a month), an acute abdomen, having a major comorbidity, and rigors. 

J I feel like this still brings up more questions than it answers! What is a major comorbidity? Diabetes? And what is 38.3 degrees? An even less popular boy band from the late 90s?

S Man don’t come after my boy Nick like that, they were legit R&B and signed with Motown records.

J Ah your true passion – overanalyzing 90s boy bands. But to clarify, comorbidities were not the typical list in a mundane HPI. Instead they referred to more serious stuff, like coma or brain death, bowel perforations, multiple traumas or burns, cardiac arrest in the last 24 hours, recent cardiac or bone marrow transplant, severe pancreatitis, ARDS, or acute or chronic liver failure.

S And as for why they chose 38.3 degrees? Otherwise known as 101 degrees fahrenheit?

J Well, The authors contended that higher fevers correlated to a higher risk of positive blood cultures, and in their case this correlated specifically to a cutoff of 38.3 degrees.

S: What else did you come across in your reading about blood cultures?

J Well, here’s my favorite paper of them all from the JAMA Rational Clinical Exam series. It has a subtle title: “Does This Adult Patient With Suspected Bacteremia Require Blood Cultures?”

S And well like they always do in that series, they really looked at a number of things including what we just were discussing: fever severity.

J Yep – with semi interesting results. Apparently fever >100.9 was associated with a likelihood ratio of 2.0 for having bacteremia. Not a lot, but it’s still something.

S Ok, how about rigoring? That’s what I really want to know. 

J You and the rest of the research world, Steve! So it’s been studied, and the general consensus is that rigoring does correlate more with bacteremia.

S And what about SIRS criteria? Was that a useful predictor?

J Can’t let that SIRS criteria go! But yes, this paper was written before qSOFA came out, and was very pro SIRS. They found that having 2 or more SIRS criteria increased the likelihood ratio to 1.8, whereas meeting only 1 criteria or fever decreased the likelihood ratio to 0.09

S Buuut, not surprisingly they found that SIRS criteria were sensitive without being specific.

J So that was a whole lot of information. I’m a little overwhelmed. There are so many of them risk factors. How do I remember them?

S You’re not alone Janine. Researchers have proposed some scores to simplify or risk stratify which patients with fever or other infectious symptoms may have bacteria in their blood. 

J But I’ve never seen anyone use them in clinical practice – why is that?

S Well, part of it may be that a lot of these scores have poor receiver operating curves, which means they just aren’t that accurate.

J More specifically, they don’t perform well when we consider both sensitivity and specificity together.

S The other reason might be that they’re just too cumbersome to use.

J So where does that leave us?

S I dunno? We could look to better technology?

J Well there are things like PCR for bacteremia – much faster than a culture!

S But even they have shortcomings. PCR isn’t as readily available for most people and may be too specific for widespread use.

J How about using machine learning to create a better predictive algorithm?

S Now you just sound like a bad episode of Silicon Valley.

J But those are still pretty far out from becoming used in standard clinical practice. And you may have noticed that while we’ve named a number of factors that correlate to positive blood cultures, none of these dramatically increase the likelihood of someone having bacteremia – the likelihood ratios just aren’t that good.

S Speaking of which, weren’t we supposed to answer how common bacteremia is in specific infections? Or specific situations?

J So let’s go back to the JAMA Rational Exam article. I really liked their conclusion! “Blood cultures should not be ordered for adult patients with isolated fever or leukocytosis without considering the pretest probability.” 

S Yeah, that really matters. Because if you have a low risk patient who ends up with a positive test, that doesn’t 100% mean they have bacteremia. It could be a false positive.  Due to reasons like contamination from skin flora.

J And in the JAMA article, they cite that there’s a 7% rate of false positives! Due to reasons like contamination from skin flora.

S That leads them to warn that, in really low risk patients, cultures may do more harm than good. False positives lead to further workup and possible overtreatment.

J Yep, in an almost perfect example of choosing the wrong population, in the ED an outpatient populations, bacteremia rates have been consistently shown to be around 1 and 2%. And only a fraction of that percentage had changes in clinical management because of their culture findings.

S That could because this population has higher rates of false positives.

J Yeah, one paper argues that because of all these issues, we’re probably culturing too many patients in the Emergency Department. Trends show that culture rates have gone up, but that doesn’t mean we’re actually helping more patients. 

S Ok, so moving away from pretest probability in the ED,  let’s talk about clinical management.

J We already went over what clinical features may help predict which patients are bacteremic. But what types of infections are more likely to cause bacteremia, and when does it actually change management?

S Back to our friends at the JAMA Rational Clinical Exam. They tried to identify which patients were low, medium, and high pretest probability for bacteremia by reviewing 35 studies.

J Uncomplicated community acquired pneumonia, cellulitis, and outpatients are generally low risk for bacteremia.

S Patient with pyelonephritis are considered medium risk. Compare that to folks with endocarditis, septic shock, or bacterial meningitis who are high risk.

J Besides risk, what also differs by infection type is how much that bacteremia matters. Does it change clinical management?

S Yeah, does it change what antibiotics you use or how long you use them for?

J And that’s where the bacteremia seen in urinary infections or pyelonephritis isn’t the same as those high risk infections In GU infections, it’s pretty easy to identify the right organism to treat by getting a urine sample. Different than meningitis, where you can’t always get a positive culture on an LP and are left treating empirically.

S So it goes to say that since you’ll have positive urine cultures in most urinary infections, you don’t *usually* need a blood culture to identify the causative organism

J But if patients with UTIs or pyelo are bacteremic, doesn’t that mean they’re sicker? Do you culture them just to see if they need more antibiotics

S Sure they might be sicker, but bacteremia is only one signal to tell you that. Their vital signs will be worse, or their labs, or their overall clinical presentation. It’s not just the bacteremia in and of itself.

J Also, most GU infections are caused by gram negatives. And there was a recent trial suggesting that in uncomplicated gram negative bacteremia (mostly urinary infections)  7 days of antibiotics were non-inferior to 14 days of antibiotics.

S But we’re still waiting to see how the guidelines interpret the findings of that study. Technically one of the subgroups did show inferiority and there were some short-comings to the trial – if you want to know more, read the discussion section of that paper because the authors do a pretty good job talking about the study’s limitations. 

J Ok, so we have to say that bacteremia still does impact management of GU infections… for now. How about those infections that were higher risk for bacteremia, like septic shock. 

S Don’t over think that, remember Nike… Just do it.

J It’s pretty well accepted that we should definitely obtain blood cultures in patients with those 3 conditions:  suspected endocarditis, bacterial meningitis, and like we mentioned before septic shock. 

S In these situations any amount of data that can guide you to appropriate antibiotic usage is imperative. 

J Furthermore, endocarditis in part is defined by the presence of bacteria in the blood, so if you don’t get blood cultures, you’re gonna have a hard time figuring out what bacteria to treat. 

S That was a loaded point! So unfortunately you gotta get those needles out…

J Too many puns, Steve!

S I can’t help myself – That point bears repeating! studies support the importance of obtaining blood cultures in patients with suspected endocarditis, bacterial meningitis, and septic shock. They’re more likely to have bacteremia, and it’s more likely to impact clinical management.

J On to one of our last points. How do blood cultures impact patient outcomes and the use of resources? There’s a NEJM article showing that culturing more patients led to false positives that increased use of hospital resources.

S This is from that same Harvard group that tried to create a predictive model for bacteremia. They used their same data set, but now compared the cost of the nearly 1100 negative blood cultures versus the nearly 100 false positives.

J Perhaps unsurprisingly, they found on average there was an increase in length of stay from 8 to 12.5 days, and an increase in cost from $8700 to $13000. The cost increase was driven by multiple factors including length of stay and increased lab and pharmacy charges.

S Which according to google is after inflation from 1990 to now is an increase in cost of almost $11000 dollars per hospital stay!

J A QI professional’s nightmare.

S Not to mention countless exposures to unnecessary antibiotics.

J An antibiotic stewardship’s nightmare.

S To paraphrase an old commercial

When you don’t pay attention to who you should get blood cultures on, you get them on everyone. 

When you get them on everyone, you get a lot of false positives. 

When you get a lot of false positives, people get more antibiotics. 

When people get more antibiotics, nice, happy bacteria probably die.

J That was great Steve

S So don’t kill nice, happy bacteria. Get blood cultures when it clinically makes sense.


S So here are our big takeaways from this deep dive into the wonderful world of microbes, fevers, and blood cultures.

J Stop, you sound like Ms. Frizzle.

S That is a compliment Janine! She was a beautiful human being.

J You just wish you had a magical school bus don’t you.

S Yes, yes I do. I would paint it orange. So point 1: a fever starts at 38 degrees celsius, not 38.3

J Wait wait Steve, that’s not our main takeaway.

S Just kidding!

J So let’s leave Steve to his daydreams and review what we’ve discussed today. 1) The guidelines according to the folks at Choosing Wisely are quite clear. Don’t get blood cultures, or really any infectious workup unless you suspect an infection.

S Furthermore, 2) being super sick, having high fevers, and rigors along with other factors probably increase your risk of having bacteremia. 

J And there may even be some helpful technology driven tools that help you predict this. 

S But  you can’t get away from the clinical side of things.

J So point 3, pretest probability really matters because false positives can happen. some infections and even some settings, are more likely to have bacteremia. 

S Uncomplicated CAP, cellulitis, and being an outpatient are associated with lower pretest probability of bacteremia, and higher rates of false positives. So blood culture mayyy not be useful in those cases.

J GU infections like UTI and pyelo are medium risk for bacteremia, and the jury is still out on how much that affects management. For now it does – you treat patients for longer.

S In other severe infections like endocarditis or meningitis or septic shock

J Don’t overthink it, just get the blood cultures.

S But for this final point, we want to be clear – it’s probably still a good idea to culture a patient when they are admitted with presumed infection. 

J Even if their risk of bacteremia is low, never say never.

S We just suggest that you always think carefully through your patient’s clinical scenario, even if you go by the book and culture everyone. Just remember, false positives are real and have their own complications.

J So ultimately we’re not advocating for a change in typical practice – sorry night float interns – but we want you to be thoughtful about who you’re culturing and what you expect the cultures to show and do for you.

S Told you we’d disappoint them.


Tags: , , , ,