How to predict ESBL bloodstream infection?

Each day we prescribe antibiotics without knowing the specific cause of infection, yet. Some patients will have an infection caused by an ESBL-producing bug, and they would benefit from immediate treatment with a carbapenem or addition of an aminoglycoside. At the same time we don’t want to misuse carbapenems or hurt kidneys. Wouldn’t it be great if we could accurately predict who would need a carbapenem? Now you can.


The proportion of E. coli and K.pneumoniae bloodstream infections (BSI) caused by variants resistant to 3rd-generation cephalosporins (proxy for ESBL-production) in Europe varies; for instance, 5.7% and 8.6% in the Netherlands and 38% and 75% in Bulgaria (see). Interesting for many, but not very useful for physicians. They need to decide on antibiotic treatment at the time diagnostic samples are obtained. Yet, among 9,422 infection episodes in which blood cultures were obtained and that started with antibiotics covering Gram-negative bacteria, the likelihood of growing an ESBL-producing species was 0.7% (in the Netherlands, no other data known). So, how to predict these patients?

The traditional risk factors for ESBL-infections are previous culture results with these bacteria and prior antibiotic use (cephalosporins and fluoroquinolones). Applying these criteria yielded a positive predictive value of 1.8% (which means that 98.2% of those with risk factors would receive unnecessary broad-spectrum antibiotics) and a sensitivity of 56% (which means that 44% with ESBL BSI would be missed). Guideline adherence would dramatically increase unnecessary carbapenem use.

That is bad!

Can we do better? Wouter Rottier derived and validated 2 prediction rules, for community-acquired and hospital-acquired episodes, that actually did better (see here). The benefits would be that (with guideline adherence) carbapenem use could be reduced by 40% (while still treating the same proportion of patients with ESBL BSI appropriate). These are the rules:

Risk factor Community-acquired infection Hospital-acquired infection
Prior identification of 3rd Generation Cephalosporin Resistant Enterobacteriaceae (prior year)


If Yes: BSI?









Immunocompromised Yes/No
Suspected source of infection Urinary tract

Respiratory tract


Respiratory tract


Any prior use of antibiotics in the last two months? Yes/No
What medical specialty was the patient admitted to? Multiple choices Multiple choices
Cephalosporin use in the last two months? Yes/No
Renal disease Yes/No
Solid malignancy (any stage) Yes/No
Central vascular catheter present Yes/No
Signs of hypoperfusion (formerly severe sepsis) Yes/No
Surgical procedure in 30 days prior Yes/No


Last question: does it work in your place? Just try.

We created  an eCRF for the scores at the day empiric treatment starts (5 minutes work) and to enter culture results 5 days later (another 2 minutes). We will need 20,000 episodes for validation, and it would be great if we can determine accuracy at country level, resistance level, continent level. Colleagues in Germany, Belgium, Switzerland, France, already started (current count 325 episodes).

For more details (e.g., coauthor rules) contact Tim Deelen:

telephone +31 88 7569616

5 thoughts on “How to predict ESBL bloodstream infection?

  1. A very interesting artilcle
    Predicting is fine but assessing the presence of ESBL-producers in blood is better
    One of the most interestiing test for that purpose is the Rapid ESBL NP test (Nordmann et al. J Clin Microbiol, 2012)
    Results are obtained in 1 h with excellent sensitivity and specifity
    Try it as it is now in many university hospitals.


    • I know the test, don’t worry. But the prediction is needed at the time that antibiotics are prescriped and when blood cultures are obtained. The prediction rules to be evaluated are for that time point. The Rapid ESBL NP test can only be used when something has grown in the blood culture, which is at least 24 hours later.


  2. I just want to leave a general comment thanking you for taking the time to share your reflections on all these different topics. I find it both useful and entertaining. A definite highlight, I must say! Thanks again, and keep your intelligent words coming 🙂


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s