“Less is more in the microbiology laboratory?”

I am by nature quite a lazy person. Don’t get me wrong, I am not afraid of working hard at times, but I am always on the lookout for ways in which I can optimise the productivity and the quality of the laboratory, whilst creating time and resource for other opportunities.

Time and effort are terrible performance metrics…

Aside from efficiencies, can doing less work in the microbiology laboratory actually lead to better patient outcomes? We know that our ultimate aim is to improve patient management. Are there circumstances where in our enthusiasm to optimise patient care, we might actually do the opposite?

Here are some examples where doing less work in the microbiology lab might actually be beneficial to patient care:

Minimising work up on probable contaminants – If coagulase negative staphylococci isolated from blood cultures are routinely reported with susceptibility profiles without any supporting clinical information that they might actually be significant, this will lead to unnecessary antibiotic use with the potential for adverse effects, along with the potential for delayed patient discharge.

Avoiding tests with low clinical utilitySputum cultures in the community setting are rarely useful, and the results may lead to undertreatment, overtreatment or simply the wrong treatment.

Reducing unnecessary microbiology tests– Rejecting urine cultures from patients where there is no evidence of UTI symptoms on the request form prevents unnecessary treatment of these patients with antibiotics.

Not processing duplicate specimens – Rejecting repeat samples (e.g. urine, sputum, stool) submitted on the same day from the one patient means that conflicting results are avoided.

Avoiding overuse of broad-range multiplex PCR panels – Running a full respiratory viral panel for a simple upper respiratory tract infection may end up delaying patient discharge from hospital. More targeted testing is often better.

Following proper sample collection and rejection criteria – Rejecting poorly collected specimens (e.g., saliva instead of sputum for pneumonia testing) avoids misleading results and unnecessary treatments.

Optimised result reporting – For example, reporting Group C/G beta-haemolytic streptococci from throat swabs in patients with acute pharyngitis may lead to unnecessary antibiotic prescribing. Along the same lines, testing and reporting unnecessarily broad antibiotics when performing susceptibility testing can lead to unnecessarily broad antibiotic coverage with concomitant side-effects on the patient and selection of antibiotic resistant bacteria.

As demonstrated above, there are lots of ways in which doing less work in the microbiology lab is not only cost-efficient, but it can also improve the overall management of the patient.

As the range of different assays we are able to offer in the microbiology lab continues to increase, we need to constantly review our current test repertoire and whether it is providing significant value to the clinicians, and ultimately the patient.

Less is often more when it comes to the microbiology laboratory.

Michael

“Bias in the diagnostic microbiology laboratory”

We are all fundamentally flawed as humans. We just have to do the best we can given our limitations.

I attended an interesting clinical Grand Round at my local hospital last week. Whilst the case presented was intriguing, it was the presenters’ focus on the different types of cognitive bias which are seen in clinical medicine which really caught my attention. 

It got me thinking… Do the same types of bias apply when reading and interpreting microbial cultures on agar plates?

The answer is of course yes.

Below are examples of the main types of cognitive bias one might be subject to when reading agar plates:

  • Confirmation Bias – Scientists may interpret bacterial growth in a way that confirms their expectations or prior hypotheses. For example, if they expect a certain antibiotic to inhibit growth, they might unconsciously downplay colonies that appear resistant.

  • Anchoring Bias – The first observation or previous experience can heavily influence interpretation. If a scientist has seen a particular growth pattern before, such as satellitism of Haemophilus influenzae around Staphylococcus aureus they might assume it’s the same species or reaction without fully considering other possibilities. 

  • Availability Heuristic – The tendency to rely on readily available examples in memory. If a scientist recently encountered an unusual bacterial isolate, they might overestimate its likelihood when analysing new plates, leading to misidentification.

  • Observer-Expectancy Effect – The scientist’s expectations may subtly influence how they interpret ambiguous results. For instance, if they believe a sample contains Streptococcus pneumoniae, they might unconsciously interpret the MALDI-TOF result as such, and ignore the possibility of Streptococcus mitis.

  • Hindsight Bias – After identifying bacterial species via additional testing, e.g. MALDI-TOF, scientists might believe the identification was more obvious than it actually was when first observing the plate, leading to overconfidence in future interpretations.

…and then of course there is the classical clinical judgement bias, that of “Premature Closure” where an “easy” or quick diagnosis is made, and further investigations into a more challenging/important secondary diagnosis are withheld because of this. It happens not infrequently in clinical medicine. An example of this in the microbiology laboratory might be a patient with crystals in a synovial fluid sample leading to a diagnosis of gout, where sufficient duration of culture was not performed to pick up the secondary diagnosis of septic arthritis!

Now that we can simply google the key biochemical reactions of E. coli, maybe cognitive bias is the sort of thing we should be teaching microbiology students, so that they are aware of the sub-conscious ways we can slip up when reading agar plates, no matter how good our intentions… A nice exam question perhaps!

Michael

 

 

“Kiestra TLA and the impending Artificial Intelligence revolution”

We are now into our 10th year of having Kiestra TLA at the laboratory where I work in New Zealand. I think it is fair to say that once you have worked in a laboratory with bacterial culture automation (i.e. Kiestra TLA, WASPLab) in place, you would never go back! We certainly don’t intend to.

I am a firm believer in optimising the quality of results generated by the microbiology lab. From a quality perspective, the advantages of automated bacterial culture systems over traditional manual-based methodologies are very impressive.

Here are ten important benefits in terms of quality that result from having a Kiestra TLA in place:

  • Improved Standardization – Automates streaking, incubation, and imaging, reducing variability between technicians and ensuring consistent results.
  • Enhanced Sample Traceability – Uses barcoding and digital tracking to prevent sample mix-ups and ensure a complete audit trail.
  • Optimized Culture Conditions – Automated incubation ensures optimal temperature and humidity, leading to better microbial growth and more reliable colony morphology.
  • Higher Reproducibility – Robotics ensure that plating and streaking techniques are performed identically every time, minimizing human error.
  • Faster Turnaround Times – Automation accelerates the workflow by processing and incubating samples continuously, leading to earlier pathogen detection and reporting.
  • Advanced Digital Imaging – High-resolution imaging captures colony growth at multiple time points, allowing for early detection and remote review without disturbing culture plates.
  • Reduced Contamination Risk – Minimizes human handling of samples, lowering the risk of cross-contamination and false-positive results.
  • Integration with LIS (Laboratory Information System) – Enables seamless data transfer, reducing transcription errors and improving result accuracy.
  • Enhanced Quality Control – Automated processes ensure that each step is performed according to predefined parameters, improving compliance with laboratory standards (e.g., ISO, CLSI).
  • Improved Staff Efficiency and Safety – Reduces manual labor, decreases repetitive strain injuries, and allows microbiologists to focus on complex tasks like interpretation and antimicrobial susceptibility testing.

It is important to note that the list above is Artificial Intelligence (AI) generated. It would take me much, much longer to generate such a list myself! I have however reviewed it and agree with all the points mentioned.

And it is due to the impending AI revolution, that systems such as Kiestra TLA are really going to come into their own over the next 10 years.

The Kiestra TLA system generates thousands of images of cultured agar plates each day, which are ripe for machine learning approaches. AI assisted applications, such as for MRSA identification and identification of urine pathogens are already available on the BD Kiestra platform.

I have no idea what the researchers at BD Kiestra are currently up to (!), but one could envisage that there is a lot of development work going on to further extend these AI-assisted apps into pathogen identification for general wound swabs, sputum samples, etc.

I observe with interest what the Kiestra TLA will be capable of by 2035. One would think that a lot of the routine microbiology culture results will be generated with very little human intervention, leaving the laboratory scientists to focus on the more complex (and interesting) samples.

Undoubtedly, by 2035, we will have new Kiestra TLA hardware in place in our laboratory, but it is in the AI-assisted software where the real revolution is coming…

Michael