Ask AI
Back Back
BSI mRDTs
Using Local Data to Optimize Molecular Rapid Diagnostic Tests for Bloodstream Infections

Released: July 10, 2025

Expiration: July 09, 2026

Emily Heil
Emily Heil, PharmD, MS, BCIDP, AAHIVP
Activity Information

Activity

Progress
1
Course Completed
Key Takeaways
  • The absence of resistance genes can potentially allow for therapy de-escalation, but decisions should be made based on an institution’s antibiogram and treatment algorithms.
  • Local epidemiology and institutional data are needed to develop treatment algorithms that reflect the area’s antimicrobial resistance patterns.

When an antimicrobial resistance marker is detected in a blood culture molecular rapid diagnostic test (mRDT), what to do is relatively straightforward. In the setting of a resistant organism, we often need to escalate therapy to target that marker of resistance (eg, if Escherichia coli and Klebsiella pneumoniae carbapenemase are detected, a patient’s antibiotics would need to be escalated to something like meropenem-vaborbactam or ceftazidime-avibactam).

When a blood culture mRDT detects a gram-negative organism without any resistance markers, it is not always as clear what to do. Can this information potentially be used to de-escalate a patient's therapy? For example, if a blood culture mRDT detects E coli with no resistance mutations on the panel identified, can you de-escalate therapy before receiving full antimicrobial susceptibility testing results?

The Potential for Therapy De-escalation
E coli and K pneumoniae are 2 of the most commonly detected gram-negative bacteria in bloodstream infections, and CTX-M is the most common extended-spectrum β-lactamase in the United States. If an mRDT for bloodstream infections detects CTX-M, you can confidently escalate the patient’s therapy to a carbapenem, but if it is not detected, can you confidently say that the patient’s broad-spectrum gram-negative coverage can be de-escalated to ceftriaxone?

I would say that it depends! Although far less common, there are other extended-spectrum β-lactamases that may be present, so there is still potential for ceftriaxone resistance. Your local data can help answer the question of whether you can safely de-escalate in this scenario. You can evaluate data from your microbiology laboratory and look over the course of a year or 2 to see what percentage of CTX-M–negative E coli was resistant to ceftriaxone. If it is a very low percentage, you could develop an algorithm recommending de-escalation to ceftriaxone earlier based on this result. You could also review the data for ceftriaxone-resistant non–CTX-M producing E coli to see if any risk factors exist that might be incorporated into your algorithm to guide situations where you would not de-escalate.

Using Local Antibiogram Data
You need local antibiogram data to craft treatment algorithms. What the algorithms at 1 institution suggest might not be applicable to institutions in other geographic areas where antimicrobial resistance patterns may differ. It is also helpful to consider patient-specific risk factors when using these treatment algorithms. A patient with a history of infection with resistant organisms or a patient with significant immunosuppression might be situations where you would await full antimicrobial susceptibility testing results before adjusting the patients’ therapy.

What I recommend for smaller hospitals, especially those working within a healthcare system, is to share data within their system and to pool data to put together treatment algorithms. At my institution, we share our data across our hospital system. However, there are some differences between the algorithms at my institution and those of smaller community hospitals in our network that tend to see less antimicrobial resistance. They have slightly more aggressive de-escalation recommendations than we do, but they used our algorithms as a starting point while collecting their own internal data. It is also important to regularly re-evaluate your algorithms to ensure that your recommendations align with your most recent epidemiologic data.

If you are newly implementing a novel mRDT for bloodstream infections, you might not have local data on the incidence of specific resistance markers. In this case, you might not use your mRDT platform for de-escalations until you have used the platform for awhile to collect data and tailor recommendations to your institution.

The algorithms that my institution currently use are not the same as those we started with, and we are continually going through a quality improvement process. We update the algorithm based on our internal data from the microbiology laboratory, examining the frequency of resistance in bloodstream infections. We map that back to our treatment algorithms every year or 2 to ensure that everything we recommend is still up to par.

Your Thoughts
What challenges are your institutions facing in developing treatment algorithms for blood culture mRDTs? Leave a comment to join the discussion!