Sepsis: Can we pick who’s sick?

I have been going on about sepsis for a while now around the hallways.  So – what is all this talk about?  I think we need to have a robust system in place that allows a small hospital to detect early (occult) sepsis, intervene early and manage it in a way that is evidence-based and inline with the ‘best practice’ out there.  

If you are like me you find new information hard to process – so I think the best analogous process we have to “map” this onto is the “Chest Pain Pathway“.  The principle is the same:

  • Acute coronary syndromes and sepsis are both severe life-threatening problems
  • They can be subtle / difficult to diagnose, especially in the early stages, or in patients with atypical presentations
  • They both benefit greatly from early aggressive intervention (thrombolysis vs.resus & IV ABs)
  • We need to cast the net wide to ensure we “catch” all the badness – this means we will find nothing in 99 out of 100 people we screen – but that is OK, if we catch the sick one (eg. we do ECG and troponin on anyone with pain between their head and bum “just in case” – it is a low-cost, low-risk test to do)
  • The protocol is “nurse-based” – it kicks off at triage, therefore gets done early and is not dependent on the whim of the doctor or the busy-ness of the department – it just happens.

So how do I think it should work?  well, similar to the Chest Pain pathway.  Here goes:

IDENTIFY Patients at risk

This is where we cast the net wide.  Essentially this is the same as using the current MET criteria or COMPASS tool that has been rolled out in the last year or so.  

I have hijacked a simple set of criteria out of the Greater NY Hospital Sepsis Collaborative – with a few modifications which I think are a reasonable screening tool for triage nurses to use in identifying patients who should be in the 100 we look at to find the 1 sick one.  Here are my proposed Sepsis Screening criteria –

STEP  1  Does the patient have any 3 of the following?

  • Suspected infection
  • Temp > 38 deg  or under  35.5 deg,  or history of rigors
  • Heart rate > 90  (or over age adjusted max.)
  • RR > 20  (or over age adjusted max.)
  • Any altered mental state
  • SpO2 < 90
  • Systolic BP < 90 mmHg (or under age adjusted minimum)
  • Prolonged > 2 sec central capillary refill
  • Any child under 3 months of age
  • Any indwelling invasive devices (PICC, IDC, Hickman etc)
  • Dialysis patient (Haemo or peritoneal)

So if your patient meets these criteria, then you go to step 2 – activate the screening tool:


If 3 criteria met => Activate the Sepsis screening panel

  1. Notify senior Doctor
  2. IV access and take blood set:
  • FBP, UECr, CRP, VBG / lactate, blood cultures, coags – URGENT to lab / ED gas analyser
  • Catch urine culture

     3.   Commence frequent obs in monitored bay.

So now – if you screen include 100 patients at Step 1, you can hopefully narrow it down to 10 using Step 2.  The bloods will either be OK, or not OK.  Your best indicator of early sepsis if you had to pick 1 is the lactate.   Lactate can be < 2: normal;  2 – 4: grey zone,  > 4:  bad – needs resus

Lactate is difficult to understand for the uninitiated – so here is a nice summary (Click) (Emcrit)


So now we have narrowed it down to a handful of patients whom might have sepsis from the general ED population.  Armed with a bit more info we can now treat them and  split them into 2 groups –

  1. Fluid responders:  those who get a bolus of fluids and they get better – BP comes up, pulse down, maybe their lactate clears to normal – you are winning.  These patients are good to go to the ward with “normal care”.  They need monitoring for subsequent deterioration on the charts but are not looking likely to “go south”.
  2. Non-responders:  these patients still ahve dodgy looking Obs and unwell after a generous fluid resuscitation over 30 – 60 minutes.  This is the 1 out of 100 whom is going to need the full treatment – and probably invasive monitoring


Of course no system is perfect so there are a few Safety net concepts I would like to introduce.  We can also catch patients at the backdoor of the ED.  Those who did not get identified at triage, got the usual workup then came to a point where they need admission or further investigation for PUO, or other infection.  So I think we should runa VBG (lactate) on any body who meets any of the following:

  • Doctor takes blood cultures for any reason – if you do a BC, then do a lactate at the same time.
  • Admission diagnosis = pyelonephritis, pneumonia, severe skin infection, PUO, diabetic foot etc… anyone with a possible severe infection that might go systemic
  • Any infant (< 6 months) sick enough to need admission for infectious cause

Lactate is not the “golden-bullet” for sepsis, but it is better than nothing, and cheap as chips.  The idea is to identify patients who were off to the ward for “usual care” who might benefit rfom a more aggressive strategy and closer monitoring.

Let me know what you think. More sepsis stuff this week
Check out the post on Shock: beyond the BP if you are interested


Add a Comment

Your email address will not be published. Required fields are marked *