The screening tool in this app is deliberately mechanical: enter the score, tick the criteria, get a pathway. That is exactly what you want at three in the morning, and exactly what will mislead you if you do not know what the criteria are standing in for. The tool follows the NICE guidelines only — NG253, NG254 and NG255. Where these chapters teach the UK Sepsis Trust’s red/amber flags and Sepsis Six, they are labelled as such: that is history and context, not the pathway the tool runs. These ten chapters cover what sepsis is under the current definitions, why the UK runs three different risk instruments for three different populations, what the fluid and timing trials actually showed as opposed to what they are usually quoted as showing, and — most importantly — the patients in whom a reassuring set of observations is not reassuring at all.
The 2016 international consensus definitions Sepsis-3 defined sepsis as “life-threatening organ dysfunction caused by a dysregulated host response to infection”. Two things in that sentence do real work. Organ dysfunction means an infection without organ dysfunction is not sepsis — it is an infection. Dysregulated host response means the damage under discussion is substantially the body's own, not the organism's. Sepsis-3 operationalised organ dysfunction as a rise in the SOFA score of 2 points or more, a threshold chosen because it carries an in-hospital mortality greater than 10%.
The same document defined septic shock as the subset of sepsis with circulatory, cellular and metabolic abnormalities severe enough to carry a substantially higher risk of death: clinically, a patient requiring a vasopressor to maintain a mean arterial pressure of 65 mmHg or above and a serum lactate above 2 mmol/L, in the absence of hypovolaemia. That combination carries a hospital mortality greater than 40%.
Sepsis-3 also deleted a term most of us were trained on. The task force concluded that “severe sepsis” was redundant — under the new definition sepsis is already, by construction, the severe thing. If you still write “severe sepsis” in a set of notes, you are using a category that formally no longer exists.
Sepsis-3 introduced quickSOFA (qSOFA) as a bedside prompt for patients outside the ICU: respiratory rate of 22/min or greater, altered mentation, or systolic blood pressure of 100 mmHg or less, with two or more counting as positive. It was derived to identify patients with suspected infection who were more likely to have poor outcomes typical of sepsis — that is, as a prognostic instrument.
It was then widely adopted as a screening instrument, which is not the same job, and it performed poorly at it. The Surviving Sepsis Campaign's 2021 international guidelines carry an explicit recommendation on this: “We recommend against using qSOFA compared to SIRS, NEWS, or MEWS as a single screening tool for sepsis or septic shock” — a strong recommendation on moderate-quality evidence SSC 2021. The guideline's own rationale notes that in the original derivation work only about a quarter of infected patients were qSOFA-positive, yet those patients accounted for the large majority of poor outcomes: excellent for telling you who is in trouble, useless for telling you who is not.
That asymmetry is the general lesson, and it applies to every instrument on this page. A tool built to identify the sickest is a tool with poor sensitivity by design. See the Sepsis-3 and SSC cards in Evidence.
NICE devotes a whole section of NG253 to interpreting findings, and it is essentially a list of ways the numbers mislead. It instructs that you should not rely on fever or hypothermia alone to rule sepsis either in or out, and names the groups who may not mount a temperature at all: people who are older or very frail, people having treatment for cancer, people severely ill with sepsis, and people with a spinal cord injury. Note the third item — being too sick to mount a fever is itself on the list.
The same section warns that older people with infection may not develop a tachycardia and may instead present with a new arrhythmia; that heart rate response is blunted by beta-blockers; and that normal blood pressure does not exclude sepsis in young people, whose compensation holds until it abruptly does not. Changes in cognition may be subtle, may present as altered behaviour or irritability in a person with a learning disability or dementia, and in older people may show up as an acute loss of functional ability rather than as confusion.
| Observation | Why it can be falsely reassuring | In whom |
|---|---|---|
| Temperature | May be normal or low despite established sepsis | Older or very frail; on cancer treatment; severely ill with sepsis; spinal cord injury |
| Heart rate | May not rise; may instead appear as new arrhythmia | Older people; anyone on rate-limiting drugs such as beta-blockers |
| Blood pressure | Normal BP does not exclude sepsis; compensation fails late and suddenly | Young people especially; interpret against the patient's own baseline |
| Mental state | Change may be subtle, or appear as behaviour change or lost function | Learning disability, dementia, older people |
| SpO2 | Difficulty obtaining a trace may itself indicate shock; readings can overestimate in dark skin | Poorly perfused patients; NICE cites the NHS England alert on probe placement |
| NEWS2 total | Less accurate where baseline physiology is abnormal | Spinal injury; chronic heart or lung disease |
Read that table as a single instruction: interpret every observation against the patient in front of you, not against the population. NICE says exactly this about NEWS2 — interpret the score within the context of the person's underlying physiology and comorbidities, and use it to support clinical decision making, not to replace it.
UK sepsis practice is not the Surviving Sepsis Campaign with a British accent. It is a distinct system, built on a national early warning score that most of the world does not use, an operational red/amber flag construct invented by a charity, and a six-item bundle with no exact international equivalent. Knowing the provenance tells you which parts are evidence and which parts are pragmatic policy.
The split matters clinically, because the three guidelines do not risk-stratify the same way.
| Guideline | Population | Risk instrument | Role of lactate |
|---|---|---|---|
| NG253 NICE | 16 or over, not and not recently pregnant | NEWS2, in acute hospital, acute mental health and ambulance settings | Does not set the band. At medium risk, lactate over 2 mmol/L or AKI means treat as high risk (1.8.20) |
| NG254 NICE | Under 16s | High / moderate risk criteria grid — NEWS2 is not validated below 16 | Retained as a decision threshold: bands above 4, 2–4 and below 2 mmol/L |
| NG255 NICE | Pregnant or recently pregnant | Risk criteria grid; NICE says consider using an early warning score | Used within the criteria rather than as the driver |
So the single most-taught rule in adult UK sepsis — escalate on lactate — survives in adults only in a narrow form: a raised lactate no longer makes a patient high risk on its own, but it does raise a medium-risk patient (NEWS2 5 or 6) to high risk under NG253 1.8.20. In children NG254 keeps full lactate bands. That is covered properly in Chapter 6 and Chapter 8.
NG253 recommendation 1.6.2 sets four bands, and explicitly tells you to use clinical judgement in interpreting them:
| NEWS2 | Risk of severe illness or death from sepsis | Recalculate |
|---|---|---|
| 7 or more | High | Every 30 minutes |
| 5 or 6 | Medium | Every hour |
| 1 to 4 | Low | Every 4 to 6 hours |
| 0 | Very low | With standard observations, per local protocol |
Three riders on that table are where the safety lives, and all three are in the guideline itself:
The red/amber construct is a UK Sepsis Trust instrument, and this screening tool no longer uses it: since October 2026 adult mode grades from NEWS2 under NG253, and the paediatric and pregnancy modes use NG254’s and NG255’s own criteria. NG253 does not use the words “red flag”, “amber” or “Sepsis Six” anywhere in its 97 pages. The UK Sepsis Trust developed Red Flag Sepsis criteria in 2015; they were incorporated into the original NG51 in 2016 and have been revised since. The Trust is candid in its manual that these are “an operational solution to the complex process of sepsis risk stratification” UKST — a usable proxy, not a biological boundary. The Trust's own worked example is worth keeping: it would be illogical to think a patient with a heart rate of 131 needs a blue-light transfer while another at 129 can be left at home.
In the Trust’s 2024 manual the high-risk set is NEWS2 7 or above, or NEWS2 5–6 with any one of: a single parameter scoring 3, mottled or ashen skin, non-blanching rash, cyanosis, deterioration since last assessment, deterioration since a recent intervention, or lactate above 2 mmol/L or known AKI. The 2024 revision also added the patient who looks extremely unwell to a health professional, or who is actively deteriorating — an explicit licence for clinical judgement to create a red flag on its own.
Where that differs from NG253, NG253 is what this tool applies. Under NG253 a single parameter scoring 3 prompts an FY2-or-above review for a definite decision (1.6.2); it raises the band only when the clinician judges the 3 points are caused by the current infection (1.8.18 at NEWS2 5–6, 1.8.24 at 3–4). Mottling, rash, cyanosis and deterioration are things to consider when deciding whether risk is higher than NEWS2 suggests (1.6.3, 1.6.4) — the tool prompts on them but does not re-band automatically. Lactate over 2 or AKI raises medium risk to high (1.8.20), which matches the Trust’s rule in effect.
The Sepsis Six is a UK Sepsis Trust construct, designed as six tasks deliverable by junior staff working as a team, all within one hour for red flag patients. It is not part of NG253, and this tool’s adult panel lists NG253’s own actions instead. The current version in the 7th edition of the manual is:
How each step lines up with NG253 for a high-risk adult, and where the two disagree:
| Sepsis Six (UK Sepsis Trust) | NG253 (what this tool shows) | Agree? |
|---|---|---|
| 1. Senior clinician attends (ST3+) | FY2-or-above urgent assessment, thinking about alternative diagnoses; refer to the senior clinical decision maker as soon as possible (1.8.2). The senior decision maker attends in person only if there is no response within 1 hour of any intervention (1.8.15) | Differs — grade and timing of senior input |
| 2. Oxygen if required | Target 94–98%, or 88–92% if at risk of hypercapnic respiratory failure, aged 18 or over (1.10.1); aged 16–18, give for shock or SpO2 below 92% in air (1.10.2) | Broadly |
| 3. IV access and bloods, including cultures and lactate | Samples before antimicrobials (1.8.1); venous gas with glucose and lactate, culture, FBC, CRP and more (1.8.2); consider procalcitonin (1.8.2) | Yes |
| 4. IV antibiotics, consider source control | Within 1 hour of calculating the NEWS2 score on initial assessment or ward deterioration (1.8.3) — for high risk only; 3 and 6 hours are the maximum at medium and low risk | Yes for high risk; the clock starts at the NEWS2, not “recognition” |
| 5. IV fluids, up to 20 mL/kg in divided boluses | 250 mL, ideally over 10–15 minutes (1.8.6); further 250 mL boluses up to 1,000 mL total including fluid already given (1.8.7); reassess after each (1.8.8); not improved after 1,000 mL → senior advice (1.8.9) | Conflicts — 20 mL/kg is 1,400 mL at 70 kg |
| 6. Monitor: NEWS2, urine output, lactate | Recalculate NEWS2 every 30 minutes at high risk (1.6.5) and on any deterioration (1.6.6) | Broadly |
In the UK Sepsis Trust construct (not the NICE pathway this tool uses), amber is where most of the thinking happens, and it is easy to treat it as “not red, therefore wait”. The Trust's position is more specific: a single amber flag in hospital should prompt a full set of bloods including a venous gas with lactate, and strong consideration of discussion with a senior decision-maker. The key decision is whether antimicrobials are needed — and if they are needed, they should be given as soon as possible but always within three hours. If they are not needed, the reasons should be documented.
Critically, the amber pathway has a defined escalation trigger: if the bloods show an acute kidney injury, or a lactate above 2 mmol/L, the patient moves immediately onto the red flag pathway and the Sepsis Six should be delivered.
NG253 has the same two rules in its own terms, and these are what this tool applies. At medium risk (NEWS2 5 or 6), evidence of hypoperfusion — “for example, lactate over 2 mmol/litre or evidence of acute kidney injury” — means treat as if at high risk (1.8.20). In adult mode a lactate over 2 entered in the sidebar, or a ticked AKI, re-bands a medium-risk patient to high automatically. Two boundaries differ from the folk version: the rule applies only at medium risk, and NICE says over 2, so a lactate of exactly 2.0 does not upgrade. On antibiotics, an FY2-or-above clinician may consider deferring them for up to 3 hours from the first NEWS2 score at medium risk (1.8.17), and NICE’s committee calls the 1-, 3- and 6-hour limits a maximum, not an aim.
It is tempting to skip the mechanism. Do not: almost every counterintuitive thing about treating sepsis — why the fluid runs out of the circulation, why a patient can be warm and vasodilated yet hypoperfused, why lactate stays up after the blood pressure is fixed — follows directly from it.
The defining move of Sepsis-3 was to place the pathology in the host response. Infection triggers innate immune activation, and in sepsis that activation becomes both excessive and, simultaneously, dysfunctional. Pro-inflammatory and anti-inflammatory processes run at once rather than in sequence. This is why “the immune system overreacting” is only half the story: many patients with sepsis are concurrently immunosuppressed, which is part of why secondary and late infections are common.
The practical consequence is that killing the organism does not stop the syndrome. Antibiotics remove the stimulus; they do not switch off the response already under way. A patient can deteriorate for many hours after entirely appropriate antimicrobials, and that deterioration is not evidence that the antibiotic was wrong.
The vascular endothelium is a continuous organ with an enormous surface area, and in sepsis it is both an actor and a casualty. Activated endothelium becomes pro-coagulant, upregulates adhesion molecules that tether leucocytes, and — most importantly at the bedside — becomes leaky. Intercellular junctions loosen and protein-rich fluid escapes into the interstitium.
Coating the luminal surface is the glycocalyx, a gel-like layer of proteoglycans and glycosaminoglycans that governs permeability, keeps cells from adhering, and senses shear stress. In sepsis it is degraded and shed. Once the glycocalyx is damaged, the barrier properties assumed by classical fluid physiology no longer hold: the compartment you are infusing into does not retain what you put in it.
Shock in sepsis is usually distributive, but rarely purely so. Separating the components tells you what each intervention can and cannot fix:
| Component | Mechanism | What fixes it |
|---|---|---|
| Vasoplegia | Loss of vascular tone — nitric oxide and other mediators; impaired vasopressin response | Vasopressors. Fluid alone cannot restore tone |
| Absolute hypovolaemia | Reduced intake, losses, insensible loss from fever | Fluid — genuinely replacing a deficit |
| Relative hypovolaemia / capillary leak | Venodilation and endothelial leak move volume out of the effective circulation | Partly fluid, but the leak means returns diminish quickly |
| Septic cardiomyopathy | Reversible myocardial depression, often with a normal or high cardiac output masking it | Recognition; inotropy only on specialist assessment |
The UK Sepsis Trust's rationale for fluid in the Sepsis Six names the first three: hypovolaemia, both absolute and relative, contributes to shock in sepsis, and restoring volume can help correct it. Note the honest verb — help correct, not correct.
The clinically important failure in sepsis happens below the resolution of anything you can measure at the bedside. Capillary flow becomes heterogeneous: some capillaries are plugged by leucocytes, platelets and microthrombi, others are shunted through, and the net effect is that oxygen delivery and oxygen demand become uncoupled within the same tissue. Adjacent capillary units can be hyperperfused and ischaemic.
This decoupling is why a patient can have a mean arterial pressure you are pleased with and still be accumulating an oxygen debt — and it is the single best explanation for why treating the numbers has repeatedly failed to beat treating the patient. It is also the reason a lactate can remain elevated after the blood pressure has normalised, and why capillary refill time has held up surprisingly well as a resuscitation target in trial conditions (ANDROMEDA-SHOCK).
The reflex equation — lactate equals anaerobic metabolism equals tissue hypoxia — is at best incomplete. The UK Sepsis Trust manual makes the point directly: lactate does not diagnose sepsis, though it is a good indicator of how compromised the circulation and tissue oxygenation are. In sepsis, raised lactate arises from tissue hypoxia, but also from adrenergic stimulation driving aerobic glycolysis, from impaired hepatic clearance, and from the effects of drugs. All of this is why a raised lactate in sepsis is a prognostic signal of considerable power and a diagnostic signal of very little. The clinical handling of that distinction is in Chapter 7.
Antibiotics treat bacteraemia and tissue infection. They do not drain an abscess, remove an infected line, decompress an obstructed ureter, or debride necrotising fascia. For a patient whose sepsis is driven by an undrained collection or an obstructed system, no antimicrobial regimen and no volume of fluid will substitute for the procedure, and the deterioration that follows is frequently attributed to “the wrong antibiotic” when the real problem is an uncontrolled source.
NICE builds the search into the first assessment: NG253 1.1.5 directs you to assess anyone with suspected infection to identify the possible source, the factors increasing risk, and any indications of clinical concern. The UK Sepsis Trust’s Sepsis Six (not part of NG253) pairs antibiotics with source control in the same step for the same reason.
| Source | Why it is missed | What actually finds it |
|---|---|---|
| Intra-abdominal collection | Examination unreliable in the elderly, obese, immunosuppressed and post-operative abdomen | Imaging with a low threshold; surgical opinion rather than serial observation |
| Obstructed infected urinary tract | Treated as “urosepsis” on antibiotics alone; obstruction not sought | Imaging of the renal tract — obstruction makes this a drainage emergency |
| Line or device infection | The line is part of the furniture; exit site never uncovered | Inspect every line and device, including under dressings; consider removal |
| Necrotising soft tissue infection | Early appearance is bland; pain out of proportion is dismissed | Pain disproportionate to findings, rapid progression — a surgical, not radiological, decision |
| Spinal or deep collection | Back pain in a septic patient is attributed to lying still | Fever plus new spinal pain, or any neurology, needs imaging |
| Meningococcal disease | Rash appears late or is not looked for on dark skin | Undressed full-skin examination; mottling, non-blanching rash and cyanosis all raise risk above NEWS2 alone |
| Post-partum / post-procedural | Symptoms attributed to normal recovery | Explicit obstetric or surgical review — see Chapter 9 |
Blood cultures come before antimicrobials — NG253 1.8.1, and step 3 of the UK Sepsis Trust’s Sepsis Six — because they are the route to targeted therapy and to stopping the broad-spectrum agent later. The sequencing intent is that they are taken before antimicrobials, since yield falls once antibiotics are in. The competing priority is obvious, and the UK Sepsis Trust manual is explicit about it in the meningococcal context: where sampling is likely to cause delay, the antibiotic goes first. Taking cultures must not be the reason a high-risk patient waits for antibiotics.
NG253 sets out the venous blood panel for high-risk patients: blood gas including glucose and lactate, blood culture, full blood count and C-reactive protein, among others. Note that the gas is listed for the glucose and lactate, and is obtainable in seconds — there is rarely a defensible reason not to have one early.
Few numbers in medicine are quoted as confidently, or as loosely, as the sepsis antibiotic clock. It is worth knowing precisely what the underlying studies measured, because the answer changes what you should do at 55 minutes.
Kumar and colleagues (2006) Retrospective cohort studied patients with septic shock who received effective antimicrobials only after the onset of recurrent or persistent hypotension — 2,154 patients. Administration within the first hour of documented hypotension was associated with a survival to hospital discharge of 79.9%, and each hour of delay over the ensuing six hours was associated with an average decrease in survival of 7.6%. The adjusted odds ratio for death was 1.119 per hour of delay (95% CI 1.103–1.136). Time to effective antimicrobial therapy was the single strongest predictor of outcome in multivariate analysis.
Two details are routinely dropped when this study is quoted. First, the population is septic shock with documented hypotension — not undifferentiated suspected sepsis. Second, it is a retrospective association, and the patients who got antibiotics late differ systematically from those who got them early, most obviously in how clearly they declared themselves.
Seymour and colleagues (2017) Cohort examined 49,331 patients at 149 hospitals under New York's mandated sepsis regulations. Longer time to completing the three-hour bundle was associated with higher risk-adjusted in-hospital mortality (OR 1.04 per hour, 95% CI 1.02–1.05), as was longer time to antibiotics (OR 1.04 per hour, 95% CI 1.03–1.06). Critically, longer time to completion of the initial fluid bolus was not associated with mortality (OR 1.01 per hour, 95% CI 0.99–1.02, P=0.21).
NG253 recommendation 1.8.3 is specific in a way the folk version is not: give people aged 16 or over who are at high risk of severe illness or death from sepsis broad-spectrum intravenous antibiotics within 1 hour of calculating the person's NEWS2 score on initial assessment in the emergency department or on ward deterioration.
Three things follow. The rule is banded — it applies to high risk, with longer timeframes at lower risk levels. The clock starts from calculating the NEWS2 score, not from arrival, triage, or a vague notion of “recognition”. And NICE states the purpose of the tiering openly: deferring antibiotic delivery in lower-risk patients is not intended to delay treatment, but to allow time to gather information for a more specific diagnosis and more targeted therapy.
NG253’s lower bands carry their own bounds. At medium risk an FY2-or-above clinician should consider deferring antibiotics for up to 3 hours after the first NEWS2 score, using the time to reach a more specific diagnosis and discussing with a senior clinical decision maker (1.8.17); at low risk, up to 6 hours (1.8.23). Once the decision to give antibiotics is made, there is to be no further delay (1.8.18, 1.8.24). The committee is explicit that these limits are a maximum, not an aim. This is what the tool applies.
The UK Sepsis Trust’s amber pathway (not the NICE pathway this tool uses) lands in a similar place: if antimicrobials are required, they should be given as soon as possible but always within three hours, and if they are not required the reasons should be documented.
NICE's position on choice is deliberately structural rather than a named drug list: where the source is unknown, give broad-spectrum treatment within the timeframe for the patient's risk category; once the source is confirmed, use source-specific antibiotics instead. For people aged 18 or over NG253 names no empirical drug at all and defers to the agreed local formulary (1.9.7). The UK Sepsis Trust’s Sepsis Six specifies maximum-dose broad-spectrum therapy, taking account of local policy, allergies and the possible need for antivirals.
In practice the binding constraint is your local antimicrobial policy, which encodes local resistance patterns that no national guideline can. The empirical suggestions in this app's action panel are a prompt, not a policy. For allergy handling, organism-specific choice and the reasoning behind agent selection, the ResusDoc Antibiotic Guide covers this in depth.
Two dosing points that matter more in sepsis than elsewhere: the first dose should be at the top of the range, because volume of distribution is increased and underdosing the first dose is common; and renal impairment is a reason to adjust subsequent doses, not to reduce the first one.
The history of fluid in sepsis is a useful corrective to confidence. A single-centre trial produced a protocol that became a global standard of care and a regulatory requirement; three large multicentre trials then found the protocol added nothing; and the subsequent attempt to settle how much fluid to give has produced two large neutral trials.
Rivers (2001) RCT randomised 263 patients to a six-hour protocol targeting central venous pressure, mean arterial pressure and central venous oxygen saturation. In-hospital mortality was 30.5% with EGDT versus 46.5% with standard therapy (P=0.009) — an effect size large enough to reshape practice worldwide.
Between 2014 and 2015, three trials tested it properly:
| Trial | n | Outcome | Result |
|---|---|---|---|
| ProCESS 2014 RCT | 1,341 | 60-day mortality | 21.0% protocol-EGDT / 18.2% protocol-standard / 18.9% usual care. RR vs usual care 1.04 (0.82–1.31), P=0.83 |
| ARISE 2014 RCT | 1,600 | 90-day mortality | 18.6% EGDT vs 18.8% usual care. Absolute difference −0.3 points (−4.1 to 3.6), P=0.90 |
| ProMISe 2015 RCT | 1,260 | 90-day mortality | 29.5% EGDT vs 29.2% usual care. RR 1.01 (0.85–1.20), P=0.90 |
ProMISe, the UK trial, went further than neutrality: the EGDT group received more fluids, vasoactive drugs and transfusions, and had significantly worse organ-failure scores, more days of advanced cardiovascular support and longer ICU stays. EGDT also increased costs, with a probability of being cost-effective below 20%.
CLASSIC (2022) RCT randomised 1,554 ICU patients with septic shock to restrictive or standard fluid. The separation achieved was substantial — a median of 1,798 mL versus 3,811 mL in ICU. Mortality at 90 days was 42.3% versus 42.1% (adjusted absolute difference 0.1 points, 95% CI −4.7 to 4.9, P=0.96), with similar rates of serious adverse events.
CLOVERS (2023) RCT asked the same question earlier in the illness, in 1,563 patients with sepsis-induced hypotension. The restrictive group received a median of 2,134 mL less fluid, and used vasopressors earlier, more often and for longer. Death before discharge home by day 90 was 14.0% versus 14.9% (estimated difference −0.9 points, 95% CI −4.4 to 2.6, P=0.61).
The honest reading of two well-conducted neutral trials is not “fluid doesn't matter” and not “give as much as you like”. It is that within the ranges these trials tested, the precise volume was not the thing determining survival, and that a strategy of less fluid with earlier vasopressors is a legitimate alternative rather than a deviation. It also means the arguments at the bedside about 30 mL/kg are arguments about a number that has never been shown to be the right one.
NG253's 2025 fluid recommendations moved deliberately away from lactate-triggered resuscitation in adults. The committee reviewed the evidence on indicators of hypoperfusion and concluded that lactate, mottled skin and capillary refill time could not be used to guide treatment decisions in isolation — noting that lactate can be high for numerous reasons, and that assessment of mottling is subjective and needs caution in people with brown or black skin. Risk is therefore categorised by NEWS2.
Note the divergence you will have to hold in your head: the UK Sepsis Trust's Sepsis Six still frames fluid as up to 20 mL/kg in divided boluses, using lactate to help guide further therapy. Both documents are current UK sources. The reconciliation is that NICE is describing the initial stepped approach and the point at which you must escalate, and neither document licenses continuing to bolus a patient who is not responding. Where they differ, this tool follows NG253: 250 mL boluses to a total of 1,000 mL, then senior advice.
SOAP II (2010) RCT randomised 1,679 patients with shock of any cause to dopamine or noradrenaline as first-line vasopressor. Death at 28 days was 52.5% with dopamine versus 48.5% with noradrenaline — odds ratio 1.17 (95% CI 0.97–1.42, P=0.10), which is not a significant difference. The decisive finding was arrhythmia: 24.1% with dopamine versus 12.4% with noradrenaline (P<0.001).
A pre-specified subgroup analysis found increased death with dopamine in the 280 patients with cardiogenic shock (P=0.03), but not in the 1,044 with septic shock (P=0.19). So the honest statement is: in septic shock specifically, dopamine was not shown to increase mortality — it roughly doubled the arrhythmia rate for no demonstrated benefit. That is a sufficient reason to use noradrenaline, and it is a more accurate reason than the one usually given.
The threshold embedded in the Sepsis-3 definition of septic shock is a vasopressor requirement to maintain a mean arterial pressure of 65 mmHg or greater, and 65 mmHg remains the usual default target. It is a default, not a physiological constant: a chronically hypertensive patient may need a higher pressure to perfuse a kidney accustomed to a higher baseline, and the right target is a specialist decision rather than an automatic one. What 65 mmHg is genuinely good for is defining the point at which you stop assuming fluid alone will do. Note that the 65 mmHg figure comes from Sepsis-3 and the Surviving Sepsis Campaign, not NICE: NG253 names no vasopressor, no dose and no MAP target (1.8.11–1.8.13), so the adult tool shows none.
Lactate does three jobs at once, and conflating them is the commonest error. It is a risk marker of considerable power; a trend monitor of some value; and a diagnostic test for sepsis of almost none.
| Question | Is lactate useful? | Why |
|---|---|---|
| Does this patient have sepsis? | No | Non-specific; raised by seizures, salbutamol, metformin, liver failure, ischaemic bowel, exercise. UKST: lactate does not diagnose sepsis |
| How much trouble is this patient in? | Yes | Above 2 mmol/L carries significantly higher mortality in sepsis; it is part of the Sepsis-3 septic shock definition |
| Is my resuscitation working? | Partly | A falling lactate with resuscitation is associated with better outcome; clearance is a trend, not a target to chase |
| Should I give this adult fluid? | Not on its own | NG253 stratifies fluid decisions by NEWS2; at medium risk a lactate over 2 mmol/L means treat as high risk (1.8.20) |
| Should I give this child fluid? | Yes, partly | NG254 retains lactate bands above 4, 2–4 and below 2 mmol/L as decision thresholds — see Chapter 8 |
ANDROMEDA-SHOCK (2019) RCT randomised 424 patients with septic shock to resuscitation targeting normalisation of capillary refill time versus targeting serum lactate. Mortality at 28 days was 34.9% in the capillary refill group versus 43.4% in the lactate group — hazard ratio 0.75 (95% CI 0.55–1.02), P=0.06, which did not meet significance. The trial was negative on its primary outcome and must be reported as such. It did show less organ dysfunction at 72 hours in the capillary refill group (mean SOFA 5.6 versus 6.6; mean difference −1.00, 95% CI −1.97 to −0.02, P=.045).
What it does not support is abandoning lactate. What it does support is the claim that a free, immediate, repeatable bedside sign of peripheral perfusion is not inferior to a laboratory number, which in a UK emergency department at handover is a practically useful thing to know. NICE, for its part, cautions that capillary refill and mottling are subjective and need particular care in people with brown or black skin.
Paediatric sepsis is not adult sepsis in a smaller body, and since the NICE split it is not even governed by the same document. NG254 covers under 16s, and it differs from the adult guideline in ways that matter at the bedside.
NG254's fluid recommendations, carried over from NICE's guideline on intravenous fluid therapy in children and young people in hospital:
Two points are easy to miss. The 250 mL cap means that above about 25 kg the weight-based calculation stops being the binding constraint. And the escalation after the second bolus is specified as a consultant attending in person — not a phone call.
| Situation | NG254 action |
|---|---|
| Any high-risk criteria and lactate over 4 mmol/L | IV fluid bolus without delay (within 1 hour of identifying high-risk criteria) and refer to a critical care specialist or team to review management, including need for central venous access and inotropes or vasopressors |
| Aged 12–15, high-risk criteria and lactate over 4 mmol/L or systolic BP 90 mmHg or less | As above — fluid without delay plus critical care referral |
| Any high-risk criteria and lactate 2–4 mmol/L | IV fluid bolus without delay, within 1 hour of identifying high-risk criteria |
| Any high-risk criteria and lactate below 2 mmol/L | Consider giving an IV fluid bolus |
Note what the last row does not say. A lactate below 2 mmol/L in a child who meets high-risk criteria does not remove the high-risk criteria, does not stop the antibiotic clock, and does not make the child safe. It changes a “without delay” into a “consider” for one intervention only.
NG254 directs that a broad-spectrum antimicrobial be given without delay — within 1 hour of identifying that a person under 16 in secondary care meets any high-risk criteria. Doses are weight-based. The screening tool in this app shows NG254’s own paediatric regimens in its paediatric modes — ceftriaxone 80 mg/kg once daily (maximum 4 g) from 28 days to 15 years, 50 mg/kg in the newborn, benzylpenicillin with gentamicin in the first 72 hours of life, and added listeria cover in any baby under 3 months — and never the adult doses. Confirm every dose against the BNF for Children and your local paediatric guideline before giving it.
NG254's vulnerability list is broader than the clinical one most of us carry, and includes factors that are social rather than physiological: being under 1 year; being from an ethnic minority background; frailty, multimorbidity or severe chronic conditions; impaired immune function; surgery or invasive procedures in the past 6 weeks; indwelling catheters; repeated antibiotic use; breach of skin integrity; communication challenges including learning disability, cognitive impairment or needing an interpreter; drug or alcohol misuse; and social, economic or environmental factors including homelessness and living in a deprived area.
Those last items are not decoration. They are in the guideline because they are associated with sepsis not being identified promptly — which is a statement about how we assess people, not only about the people themselves.
Screening instruments are built on populations. The patients who die of missed sepsis are largely the ones whose physiology, presentation or pathway sits outside the population the instrument was built on. Each group below defeats screening by a different mechanism, and it is worth knowing which.
This group defeats screening physiologically. NG253 warns that older or very frail people may not develop a raised temperature; that older people with infection may not mount a tachycardia and may instead present with a new arrhythmia; that beta-blockers blunt the heart rate response; and that cognitive change in older people may present as an acute change in functional abilities rather than as recognisable confusion.
The practical translation: “off legs”, “not herself”, a new fall, new incontinence or a sudden inability to manage at home are presenting features of sepsis in an older person, and they will not score on any observation chart. NG253 also instructs you to pay particular attention to concerns expressed by family or carers about a change from usual behaviour — in this group that history is often the most sensitive test available.
This group defeats screening in three ways: the physiology of pregnancy shifts the baseline observations, NEWS2 is not the instrument (NG255 says consider using an early warning score), and the pathway is fragmented across obstetrics, the emergency department and primary care.
NG255 places the whole group at high risk: anyone pregnant, or who has given birth, had a termination of pregnancy or a miscarriage in the past 6 weeks. Within that, it flags particularly: impaired immunity through illness or drugs; diabetes, gestational diabetes or other comorbidities; invasive procedures such as caesarean section, forceps delivery or removal of retained products of conception; prolonged rupture of membranes; having or having been in close contact with someone with group A streptococcal infection, for example scarlet fever; and continued vaginal bleeding or an offensive vaginal discharge.
This group defeats screening immunologically: the signs we screen for are largely signs of an inflammatory response, and these patients cannot mount one. Fever may be absent — NG253 names people having treatment for cancer among those who may not develop a raised temperature — and localising signs can be minimal because pus requires neutrophils.
NG253 directs you to suspect neutropenic sepsis in anyone who becomes unwell and is having or has had systemic anticancer treatment within the last 30 days, or is receiving or has received immunosuppressant treatment for reasons unrelated to cancer — using clinical judgement about whether past treatment may still be causing neutropenia. Such patients should be referred immediately for assessment in secondary or tertiary care, and treated in line with NICE's guideline on neutropenic sepsis in people with cancer. NICE's committee recorded that people with suspected neutropenic sepsis are at very high risk.
NG253 recommendation 1.1.4 asks you to assess people with extra care where there is difficulty taking a history — naming English as a second language, and communication challenges associated with neurodiversity, cognitive impairment, learning disabilities, severe mental health conditions or brain injury. The vulnerability lists across all three guidelines go further, including drug and alcohol misuse, homelessness and living in a deprived area.
These are listed as factors that increase the risk of sepsis or of sepsis not being identified promptly. That second clause describes a failure of the assessment, not of the patient. Where the history is hard to obtain, the examination and the observations have to carry more weight, and the threshold for investigation should fall rather than rise.
| Group | How screening fails | What to do instead |
|---|---|---|
| Older / frail | No fever, no tachycardia, confusion is subtle or looks like function loss | Treat “off legs” and carer concern as presenting features; compare with the patient's own baseline |
| Pregnant / ≤6 weeks post | Shifted baseline physiology; NEWS2 not the instrument; split pathways | Use NG255; ask about miscarriage and termination; ask about group A strep contact |
| Neutropenic / immunosuppressed | Cannot mount fever or localising signs; deteriorates abruptly | Suspect on the treatment history — SACT within 30 days; refer immediately; do not wait for a score |
| Young adults | Compensate well; normal BP until sudden decompensation | NG253: normal blood pressure does not exclude sepsis in young people |
| Chronic heart / lung disease, spinal injury | Abnormal baseline physiology makes NEWS2 less accurate | Interpret the score against their usual values, not the population's |
| Communication barriers | History is incomplete, so the clinical picture looks milder than it is | NG253 1.1.4 — assess with extra care; lower the threshold to investigate |
Sepsis outcomes are almost always reported as mortality at 28 or 90 days, and every trial in the Evidence tab uses that endpoint. It is a genuinely poor summary of what happens to the people who live. Survivors commonly leave hospital with new physical weakness, new or worsened cognitive impairment, and psychological sequelae including anxiety, depression and post-traumatic symptoms — and with an elevated risk of readmission and of further infection.
NICE takes this seriously enough to devote several sections of each guideline to information at discharge, and it distinguishes three different groups, which is a useful clinical frame:
A useful discharge conversation in suspected sepsis names the symptoms that should trigger return rather than offering a general instruction to come back if worse: new confusion or drowsiness, breathlessness or fast breathing, a rash that does not fade under pressure, not passing urine, mottled or discoloured skin, or simply feeling much worse. It should also be given to someone other than the patient where possible, because the patient who becomes confused is the one least able to act on it.
NICE also asks that families and carers be involved in discussions — which connects back to recommendation 1.1.3, where concern expressed by a relative or carer is itself something to pay particular attention to. The person who will notice deterioration first is usually not a clinician.
Patients are rarely warned that recovery from sepsis is slow and uneven, and many interpret their own fatigue, poor concentration and low mood as a personal failure or as evidence that something was missed. Saying plainly that these are recognised consequences of sepsis, that they usually improve, and that there is support available — the UK Sepsis Trust runs a support service for survivors and bereaved families — is a genuinely therapeutic intervention that costs a minute.
Every figure on this page was read from the study's own abstract on PubMed rather than from a review, a summary or memory. Where a trial is negative it is presented as negative — several of the most influential sepsis trials found no mortality difference, and the way they are quoted in teaching often reverses that. Each card gives the design, the numbers, and the limitation that matters at the bedside.
Not a trial: a Delphi and data-informed consensus convened by SCCM and ESICM, with peer review and endorsement by 31 societies. Its authority is definitional rather than experimental.
Limitation: SOFA needs laboratory data and is not a bedside screening instrument. The task force itself identified poor sensitivity and specificity of SIRS as a driver for change — but its replacement, qSOFA, was derived for prognosis, and the card below explains why that distinction matters.
The guideline's rationale notes that qSOFA is more specific but less sensitive than two of four SIRS criteria for identifying infection-induced organ dysfunction, and that in the original derivation only about a quarter of infected patients were qSOFA-positive though those patients accounted for the large majority of poor outcomes.
Why it matters in the UK: this is the international guideline agreeing, on different grounds, with what NICE does — NG253 mandates NEWS2, not qSOFA, for adults in acute hospital, acute mental health and ambulance settings.
Limitation: single centre, modest sample, and a control arm reflecting the standard care of 2001. An effect size of this magnitude from one site should always have prompted replication before global adoption — and when replication came, it did not reproduce.
The honest interpretation: ProMISe's own conclusion specifies a population “identified early and received intravenous antibiotics and adequate fluid resuscitation”. These trials did not show that early resuscitation is useless. They showed that once patients are found early and treated promptly, the catheters, protocolised targets and transfusion triggers add nothing — and in ProMISe, added harm and cost.
Limitation, and it is a large one: retrospective and observational, so causation is not established — patients treated late differ systematically from those treated early. The population is septic shock with documented hypotension, not undifferentiated suspected sepsis, and the figures are routinely quoted as though they applied to the latter.
Why this is the most practically useful sepsis study of the last decade: it separates the bundle into its parts and finds the urgency belongs to the antimicrobial, not the fluid. Limitation: observational, within a regulatory system that may not generalise, and residual confounding by severity is likely — but the internal contrast between two elements measured the same way in the same patients is more robust than either association alone.
Interpretation: a genuinely neutral trial with a real separation in exposure — which makes the neutrality informative rather than a failure of protocol adherence. Limitation: ICU patients already resuscitated before randomisation, so it does not address the first hour in the emergency department. That is what CLOVERS attempted.
Interpretation: taken with CLASSIC, two large RCTs at different points in the illness found that the volume strategy did not determine survival within the ranges tested. Importantly, the restrictive-with-earlier-vasopressors approach was not harmful — which is what licenses starting a vasopressor rather than reaching for a fourth bolus. Limitation: both are neutral trials, and neutrality is not evidence of equivalence for every patient; neither tested the extremes.
How it is misquoted: the 8.5-point mortality difference is frequently cited as though the trial were positive. It was not — P=0.06, and the confidence interval crosses 1. What it does support: a free, immediate, repeatable bedside sign performed at least as well as a laboratory measurement as a resuscitation target. Limitation: capillary refill assessment is subjective, and NICE specifically cautions that it and mottling need careful interpretation in people with brown or black skin.
Precise reading: in septic shock specifically, dopamine was not shown to increase mortality. What it did was roughly double the arrhythmia rate with no demonstrated benefit — which is a sufficient and more honest reason to use noradrenaline first-line than the mortality claim often attributed to this trial.
Most of the landmark trials are negative. ProCESS, ARISE, ProMISe, CLASSIC, CLOVERS and ANDROMEDA-SHOCK all failed to demonstrate a mortality benefit for the intervention tested. A field whose teaching is built largely on negative trials needs to be careful: the correct inference is usually “this specific protocol did not help”, not “this does not matter”.
The strongest timing evidence is observational. Kumar and Seymour are cohorts, not trials. It is not ethical to randomise patients to delayed antibiotics in septic shock, so this is probably the best evidence that will ever exist — which means the one-hour rule rests on association, and should be applied with that understood rather than treated as proven causation.
Populations differ from the patient in front of you. The fluid trials studied ICU patients or those already hypotensive; the screening studies are mostly non-UK; the definitions come from consensus. None of them studied the frail 88-year-old with a NEWS2 of 4 and a story of being off her legs — who is a large fraction of UK sepsis.
Where this leaves UK practice. Find patients early; give antimicrobials promptly to those at high risk; give fluid in measured steps and stop to think at a litre; reach for a vasopressor rather than a fourth bolus; and find and control the source. Every element of that is supported above. The elaborate protocols are the part that did not survive testing.
The screening tool encodes the NICE sepsis guidelines — NG253 for people aged 16 or over, NG254 for under 16s and NG255 for people who are or have recently been pregnant. It does not use the UK Sepsis Trust red and amber flag criteria. The teaching and evidence layers were added in September 2026.
Each of the four populations — adults aged 16 and over, children aged 5–15, children under 5, and people who are pregnant or recently pregnant — is graded by its own NICE guideline, and the action panel shows that guideline’s actions for the resulting risk level, each with its recommendation number. Until October 2026 the tool used the UK Sepsis Trust red/amber construct and the Sepsis Six; none of that drives the tool now. It survives only in the Teaching tab, labelled as UK Sepsis Trust material, where it is compared with NICE. The paediatric high-risk checklist was the Trust’s Paediatric Sepsis 6 until 20 October 2026; it is now NG254’s own high-risk actions.
How each mode works:
The whole subdomain — the NICE-based screening tool, Teaching and Evidence — was clinically reviewed in October 2026 by Dr Asim Ijaz, Consultant in Emergency Medicine and ResusDoc’s clinical governance lead. The amendments he requested have been applied; his final sign-off of the amended version is pending.
The screening tool was clinically reviewed in July 2026, but that review covered a different instrument: the risk criteria were then taken from the UK Sepsis Trust red and amber flag construct. In October 2026 all three pathways were rebuilt directly on the NICE guidelines — NG253 for people aged 16 or over, NG254 for under 16s, and NG255 for people who are or have recently been pregnant. The UKST red/amber construct no longer grades risk or drives any action anywhere in the tool (the last UKST element, the paediatric high-risk checklist, was replaced by NG254’s own actions on 20 October 2026), so the July 2026 review no longer describes what is on this page and should not be read as sign-off for it.
Those rebuilds were audited internally against the full text of each guideline, read from the published PDF rather than from memory, and the divergences found are recorded in the project notes. The Teaching and Evidence layers were added in September 2026 and are covered by the October 2026 review above. Every trial figure in the Evidence tab was read from the study's own PubMed abstract, and every NICE guideline code was verified against the guidance page it resolves to.
Built and maintained by Dr Nirmalya Hore, NHS Emergency Physician. Corrections are welcome via ResusDoc.