Blood cultures and GIRFT - some suggestions for comparative metrics

The Get It Right First Time programme in the UK is about to start the pathology disciplines. In essence, GIRFT looks to take clinically relevant data on performance to demonstrate unwarranted variation. Data can be used by local teams to

• show excellence
• show practice in line with peers
• show performance below that of peers

In this latter scenario, this can be used to show the need for changes in practice or resourcing, or may be used as part of an conscious decision of risk management based on local exigencies.

In order for GIRFT to be successful it is essential that clinically relevant metrics are chosen by the profession. It is not always clear what data will be available, but it is useful to start by considering what might be useful and then working from there.

Blood cultures are a mainstay of microbiology and are seen as a core part of a ''diagnostic stewardship" programme in which culture and sensitivity information is used to optimise patient infection management. Recent papers have shown the importance of paying attention to blood culture pathways (eg. Weinbren et al 2018; Lamy and Sundqvist 2018). This blog explores some of the issues raised in these papers in relation to possible metrics, and how this data can be accessed from the laboratory and hospital data of our hospital in North Devon. We are a relatively small DGH with about 250 beds and 3863 acute medical admissions in 2017 (number needs checking - seems too low). Note that similar data may already be collected as part of enhanced surveillance programmes (eg. E coli and S. aureus bacteraemia). We need to be careful to avoid duplicating effort here, and may want to either use this as is, or lobby for changes in what is collected). Note that we have a rubbish lab system and no electronic patient record. So here goes…

All this data relates only to blood cultures taken in the ED. I have done this as it is most likely to be comparable between sites, is at the front end of any sepsis pathway, and as a result is probably the data that most cuts through any site-specific noise. If this pathway is not working well, then there are likely to be other issues with the approach to blood cultures.

1. Number of episodes with blood cultures sent; and number of sets per episode

Why is this important? You can’t detect organisms if you don’t take a blood culture! This number needs standardising to eg. number of medical admissions. There is also some debate about number of sets of blood cultures that need to be taken. Surviving Sepsis says take 2. This might be expensive and may not add much value...I will do a separate blog on what our data says. But we have had a minor push on this since 2015.

2. Time to load blood cultures

Why is this important? Reducing the time to put a blood culture bottle on the machine reduces overall time to positivity (see later). It is part of the national standard.

3. Percent loaded on same day as taken

Another way to look at the data on loading could be the proportion of blood cultures that are loaded on the same calendar day that they are taken.

4. Time to negative result

Why is this important? A negative report is often used by clinicians to step down treatment, and it is a part of the NICE guidance on management of suspected neonatal sepsis, where the standard is for negative results to be available within 36 hours of taking.

5. Calendar days to negative result

Again, a perhaps more clinically relevant way to look at this is the number of calendar days taken to release a negative report.

6. Episode positivity rate

Why is this important? We need to make sure we are targeting blood cultures to patients who need them, and we need to make sure these are done properly, whether this be through multiple sets, or optimisation of blood volume, or whatever.

The ratio of positive to negative tests has been used by others as a marker of appropriateness of test usage (eg. O Sullivan et al Scientific Reports 2018).

This looks at how many episodes with at least one blood culture taken result in isolation of a probable pathogen. Probable contaminants are excluded. Note that this data is purely based on speciation, and does not take in any clinical details. So for instance, I have called all viridans streptococci contaminants - they clearly are not! My feeling is that as a marker of performance these details are not important, and we could spend too much time splitting hairs.

This graph shows the positivity rate in all ED taken blood cultures :

7. Contamination rate

Why is this important? Contaminated blood cultures significantly impact on patient management. They may lead to inappropriate treatment pending final results. They may also lead to significant results being ignored, either because a pathogen may look like a contaminant initially (eg. S aureus) or because an organism that is typically a contaminant may on occasion be significant.

This graph shows the contamination rate in all ED taken blood cultures :

8. Time to final positive result (E coli only)

Why is this important? It’s what clinicians see, and it is probably a number that most of us can collect from the LIMS. I have focussed only on E coli, as this is already a key point of focus for enhanced surveillance, there is probably little debate about clinical relevance, and we all call it the same thing.

9. Calendar days to final positive report (E coli only)

Again, another way to look at this is how many days does it take to get out a report.

10. Time to gram stain (E coli only)

Why is this important? The gram stain is probably the thing that changes management more often than any other part of the process, and in important ways. For instance, we are all well aware of the gram negative rod in the patient with “pneumonia”.

We cannot easily get this data - it needs a manual review of the LIMS record.

11. Days to gram stain (E coli only)

Again, another way to look at this is calendar days to gram stain

12. Time to preliminary sensitivities (E coli only)

Why is this important? Most laboratories have some idea on sensitivities before the final report is released. This will usually be communicated verbally to teams, especially if clinically important. (eg. allowing stepdown treatment; or escalation if resistance detected). Again, a focus on E coli seems sensible as resistance in gram negative organisms is probably what causes most problems in everyday infection management.

13. Days to preliminary sensitivities (E coli only)

Again, we can show this data by calendar day.

Finally, what is the impact of all this on patient outcome? Here are 2 suggested metrics…

14. Mortality (during in patient episode)

In hospital Mortality 2017
No growth E coli Any significant growth Contaminant All medical admissions	8.7% 5.7% 10.2% 11.8% 6.6%

15. Length of stay

Mean length of stay 2017
No growth E coli Any significant growth Contaminant All medical admissions	6.1 days 8.1 days 8.9 days 7.2 days 4.8 days

So there you go. Send me your thoughts. And I’ll write another piece now on why I’m not so sure that we need 2 sets any more, but why we definitely need more than 1 bottle!