Diagnosis needs all three components, not two of three.

• Glucose above 11.0 mmol/L, or known diabetes
• Ketones above 3.0 mmol/L — the ketone measured is 3-beta-hydroxybutyrate — or ketonuria of 2+ or more
• Bicarbonate below 15.0 mmol/L and/or venous pH below 7.3

Known diabetes satisfies the glucose criterion, so the glucose does not have to be raised at all for the diagnosis to stand. And the pH is venous, not arterial — there is no need for an arterial sample to diagnose or monitor DKA.

Severe DKA in adults is any one of the following, and one is enough:

• Ketones above 6.0 · bicarbonate below 5.0 · pH below 7.0 · anion gap above 16
• Admission potassium below 3.5
• GCS below 12 or an abnormal AVPU score
• Oxygen saturations below 92% on air, in someone with normal baseline respiratory function
• Systolic blood pressure below 90 · pulse above 100 or below 60

Severity is not just a label — it is an indication that the patient is likely to need a higher level of care. A patient who still meets severity criteria after initial resuscitation, or is failing to improve, or cannot be monitored adequately where they are, needs Level 2 care with critical care input and review by a consultant physician. A drowsy patient with DKA needs critical care assessment, and consideration of a nasogastric tube with airway protection because of the aspiration risk. Problems with intravenous access mean requesting critical care support immediately, not persevering — and concentrated potassium may itself require Level 2 or 3.

Yes, and it is the reason the guideline moved away from glucose-centric care. Euglycaemic DKA is DKA in someone with known diabetes whose glucose is normal or not particularly raised. It is most often seen with the SGLT2 inhibitors — dapagliflozin, empagliflozin, canagliflozin, ertugliflozin and sotagliflozin.

The mechanism is entirely renal, and it is worth being precise about what it is not. Blocking SGLT2 in the proximal tubule stops filtered glucose being reabsorbed, so it is excreted in the urine. This has nothing to do with insulin-driven uptake into cells: the plasma glucose is held down by renal loss while the insulin deficiency goes on driving ketogenesis unchecked. The reading looks reassuring and the acidosis is severe.

It is treated in exactly the same way as hyperglycaemic DKA, with three modifications:

• Start 10% glucose immediately at 125 mL/hour, rather than waiting for the glucose to fall below 14, because it is already below 14
• Begin fixed rate insulin at the usual 0.1 units/kg/hour
• If the glucose falls despite the glucose infusion, reduce to 0.05 units/kg/hour

If DKA occurs on an SGLT2 inhibitor, stop the drug and complete a Yellow Card. Restarting it is a decision for the diabetes team, not one to make in the Emergency Department.

A second assumption worth dropping at the same time: DKA is not exclusive to type 1. Ketosis-prone type 2 diabetes accounts for around a third of DKA admissions, is commonest in people of Afro-Caribbean or Hispanic descent, and the initial treatment is identical.

Ketones and a low pH are not the same as DKA. The diagnosis needs all three criteria, and several conditions produce some of them without producing all of them.

• Alcoholic ketoacidosis — a normal glucose is the key difference, but the histories overlap with euglycaemic DKA and a careful one is needed to separate them. Measure beta-hydroxybutyrate, not urine ketones: acetoacetate production is suppressed, and the dipstick measures acetoacetate, so the ketones read falsely low.
• Starvation ketosis — ketones can exceed 6 mmol/L, which looks alarming. Because it develops over days, renal compensation means the acid–base and electrolyte disturbance is often minimal. High ketones with a preserved pH is the pattern.
• The hyperosmolar hyperglycaemic state — hyperglycaemic and profoundly dry, but without significant ketonaemia or acidosis. It has a separate guideline and a different insulin strategy, and the overlap with DKA is common enough to be worth knowing precisely. See below.

Beyond those, a raised anion gap acidosis has a short standard differential worth running mentally before you commit to the DKA pathway: lactic acidosis, renal failure, toxic alcohols such as methanol and ethylene glycol, and salicylate. None of these is in the DKA guidelines — the point is that ketones on a gas do not finish the thinking.

The hyperosmolar hyperglycaemic state has its own guideline, JBDS 06, and its own diagnostic indicators. A precise definition has not been agreed, so what follows are characteristic features rather than diagnostic criteria:

• Marked hypovolaemia
• Measured or calculated osmolality usually 320 mOsm/kg or above
• Marked hyperglycaemia, 30 mmol/L or above
• Without significant hyperketonaemia — ketones 3.0 mmol/L or below
• Without significant acidosis — pH 7.3 or above and bicarbonate 15.0 or above

Osmolality is (2 × Na⁺) + glucose + urea. HHS should not be diagnosed on biochemistry alone.

Fluid comes first, and lots of it. These patients are profoundly dry — losses are 100 to 220 mL/kg — and they may not look as dehydrated as they are, because hypertonicity preserves intravascular volume. Give a litre of 0.9% saline over the first hour, then 0.5 to 1 litre per hour, aiming for a positive balance of 2 to 3 litres by 6 hours and 3 to 6 litres by 12 hours. Be more cautious, at 0.25 mL/kg/hr, in someone under 50 kg or with heart or renal disease.

Insulin waits. It is not started until the glucose has plateaued on fluid alone, and then at 0.05 units/kg/hour, because starting an intravenous insulin infusion too early could result in circulatory collapse. If ketones are between 1.0 and 3.0 and the patient is not acidotic, 0.05 units/kg/hour from the outset is appropriate.

Everything is corrected slowly. Glucose should fall by no more than 5.0 mmol/L per hour, osmolality by 3 to 8 mOsm/kg per hour, and sodium by no more than 10 mmol/L in 24 hours. A rising sodium as the glucose falls is expected — a 5.5 mmol/L fall in glucose raises the sodium by 2.4 — and is not itself a reason to switch to hypotonic fluid. Only move to 0.45% saline if the osmolality is not declining despite an adequate positive balance. Everyone gets prophylactic low molecular weight heparin for the whole admission unless contraindicated, because the venous thromboembolism risk is greater than in DKA.

The overlap is what gets tested. Mixed DKA and HHS occurs relatively frequently: marked hypovolaemia and hyperosmolality, but with a pH below 7.3 and ketones above 3.0. That patient goes down the DKA pathway at 0.1 units/kg/hour, not the HHS one. Resolution of HHS is osmolality below 300, hypovolaemia corrected with a urine output of at least 0.5 mL/kg/hr, cognition back to baseline, and glucose below 15 — a normal glucose or sodium alone is not enough.

As an obstetric emergency, and it must be admitted to delivery suite or a high dependency unit. This section was new in the March 2023 revision, and two points from it are worth repeating.

The first: DKA in pregnancy may manifest as abdominal pain, and should always be considered as a possible alternative to pre-term or term labour. The second: it can occur with only very modest elevation of glucose, below 13.9 mmol/L — so a reading that looks unalarming does not exclude it.

UK incidence is 6.3 per 100,000. There were no maternal deaths in the reported series, but fetal mortality was 16%, because ketones are toxic to the fetus. The commonest precipitants are infection, vomiting, steroids and medication errors. Use her present weight for the insulin calculation, and call for immediate senior obstetric help. A pregnancy test is on the initial investigation list for women of childbearing age with DKA, which is where some of these are picked up.

Adults from 18 follow JBDS 02. Children up to 16 follow BSPED. The 16 to 18 year olds fall into either, depending on which team is admitting them — adult team, adult guideline; paediatric team, paediatric guideline.

The reason is safety rather than physiology: harm happens when a team works from a treatment plan it does not know well.

Fluid first, then insulin. The order is specified, not conventional, and an option that gives the insulin before the fluid is wrong on the order alone.

If the systolic is below 90: 500 mL of 0.9% sodium chloride over 10 to 15 minutes, repeated as needed — most patients need 500 to 1000 mL. If there is no improvement, look for another cause of the hypotension and consider critical care. The guideline calls for senior assessment at this point; in the exam, and in resus, that senior is you — what matters is recognising it as the moment a junior should be calling you. Once the systolic is above 90, give a litre over the next hour; potassium is likely to be needed in that second litre.

If the systolic is 90 or above: The illustrative regimen for a previously well 70 kg adult is one litre over one hour, then over two hours, two, four, four and six. All but the first litre carry potassium chloride, and cardiovascular reassessment at 12 hours is mandatory. The typical water deficit is 100 mL/kg — a 70 kg adult may be up to seven litres down.

The fluid is 0.9% sodium chloride with pre-mixed potassium, chosen partly because adding potassium on a general ward is unsafe. Two randomised trials comparing it with Hartmann's showed no difference between them, and critical care units that prefer a balanced crystalloid are explicitly accommodated. The stated downside of saline is hyperchloraemic metabolic acidosis, which matters later when you are trying to judge resolution.

Then fixed rate insulin at 0.1 units/kg/hour, made up as 50 units of soluble human insulin to 50 mL with saline. There is no priming bolus; the only exception is intramuscular insulin at 0.1 unit/kg if setting up the infusion is delayed. Continue the usual long-acting basal insulin at the usual dose and time, and give a newly diagnosed patient 0.25 units/kg of basal insulin to prevent rebound ketosis when the infusion stops.

Potassium is banded, and the bands are worth knowing exactly:

No potassium goes in the initial resuscitation fluid, none if the potassium is above 5.5, and none unless the patient is passing urine. Monitoring is hourly glucose and hourly ketones, a venous gas at 60 minutes, at 2 hours and 2-hourly thereafter, four-hourly electrolytes, and a urine output of at least 0.5 mL/kg/hour.

Three rates and one range, checked every hour:

• Ketones falling by at least 0.5 mmol/L per hour
• Bicarbonate rising by at least 3.0 mmol/L per hour
• Glucose falling by at least 3.0 mmol/L per hour
• Potassium held between 4.0 and 5.5 — outside that, recheck hourly, and if it is still abnormal an hour later, get senior advice immediately

Ketones are the primary target. Bicarbonate and glucose are the fallbacks where ketone measurement is unavailable.

If the targets are not being met, there is a step before the rate change. Check the pump is working and connected and that the residual volume is correct — then increase the fixed rate insulin by 1.0 unit/hour, hourly, until the ketones fall at target. A stem describing a patient whose ketones have not moved in three hours may be testing whether you reach for the rate or for the line.

Once the glucose falls below 14 mmol/L, add 10% glucose at 125 mL/hour alongside the saline, through a Y connector, rather than instead of it — and consider reducing the fixed rate insulin to 0.05 units/kg/hour. The word in the guideline is “consider”. Neither the insulin nor the glucose stops until the patient is eating and drinking normally.

Resolution is ketones below 0.6 mmol/L and venous pH above 7.3. Two things that look like markers of resolution are not.

Bicarbonate is not a surrogate at this stage. The hyperchloraemic acidosis that comes with large volumes of 0.9% sodium chloride lowers the bicarbonate, and that makes it difficult to judge whether the ketosis has resolved. The same acidosis may cause renal vasoconstriction and oliguria. Urinary ketone clearance is not a marker either — urine ketones persist after the DKA has resolved.

By 24 hours the ketonaemia and acidosis should have resolved in most people. Failure to resolve biochemically by 24 hours is unusual and needs senior and specialist input.

Conversion has a sequence, and the sequence is the examinable part. When the ketones are below 0.6, the pH is above 7.3 and the patient is ready and able to eat: give the subcutaneous fast-acting insulin with a meal, and stop the intravenous insulin 30 to 60 minutes later. The subcutaneous insulin must be running before the intravenous stops. Never convert at bedtime. Twice-daily fixed-mix regimens are restarted before breakfast or before the evening meal only.

If the ketonaemia has cleared but the patient is not eating and drinking, the answer is a variable rate infusion, not a subcutaneous regimen.

The pH gives you the dehydration to assume. It is not possible to assess the degree of dehydration clinically with any accuracy, so the estimate is based on the initial blood pH. The percentage never comes from the examination.

Note that severe in children is pH below 7.1, less acidotic than the adult threshold of 7.0. The band does a second job as well: a pH below 7.1, or an age under 2, means one-to-one nursing on a high dependency unit or a general paediatric ward, and transfer to paediatric intensive care should be considered if that cannot be provided.

The calculation itself is two separate sums over two different timeframes, added together.

• Deficit = weight (kg) × % dehydration × 10 = mL, replaced over 48 hours
• Maintenance = full Holliday-Segar, over 24 hours: 100 mL/kg/day for the first 10 kg, 50 mL/kg/day for the next 10 kg, and 20 mL/kg/day for every kilogram above 20
• Starting rate = deficit rate + maintenance rate

Full Holliday-Segar is more permissive than the pre-2021 version, and that change is flagged as a significant one — anyone revising from older material will have a smaller figure in mind.

Whether the child is shocked decides whether the bolus comes off the deficit. It carries a red alert in the pathway: subtract only the 10 mL/kg bolus given over 30 minutes to non-shocked patients, and do not subtract rapid resuscitation boluses given to shocked patients. Shock here is the APLS definition — tachycardia, prolonged central capillary refill, poor peripheral pulses and hypotension as a late sign — and explicitly not just poor peripheral perfusion, because acidosis and hypocapnia both cause peripheral vasoconstriction.

Use the actual admission weight, not an estimate of the likely weight after rehydration, with consideration given to a maximum of 75 kg or the centile weight for age, whichever is lower.

The errors named in the pathway are the distractors: subtracting the shocked patient's resuscitation boluses; confusing the two timeframes; using an estimated or post-rehydration weight; and fluid calculation error itself, listed as a cause when the pH is not improving. Fluid given in the Emergency Department and on the way to the ward must be documented carefully, because that is where most mistakes occur.

Children are not scaled-down adults, and the divergences are specific. The largest is insulin: in children it starts one to two hours after intravenous fluids have begun, on the stated basis that there is some evidence cerebral oedema is more likely if insulin is started early.

* Three potassium exceptions, and all of them matter:

• Withhold potassium if anuria is suspected, or if there are peaked T waves.
• If the potassium is high on arrival, add none until they have passed urine, or until it has fallen back into range — which it typically does after the 10 mL/kg bolus.
• If the potassium is below 3.0 at presentation, defer the insulin until it is above 3.0. If it falls below 3.0 during treatment, consider temporarily stopping the insulin and discuss urgently with critical care; anything above 40 mmol/L needs a central line. Hypokalaemia can occur up to 48 hours after treatment starts.

Two other paediatric details do not appear in the adult guideline at all. Insulin is never given as a bolus. And when glucose-containing fluid is running, do not stop the insulin infusion — insulin is what switches off ketone production.

Immediately, and without waiting for imaging. Any fall in conscious level is the trigger. Doses are published for children; none are published for adults.

• Hypertonic saline 2.7% or 3% — 2.5 to 5 mL/kg over 10 to 15 minutes, or
• Mannitol 20% — 0.5 to 1 g/kg over 10 to 15 minutes
• Then restrict fluids to half maintenance and replace the deficit over 72 hours rather than 48
• Do not intubate and ventilate until an experienced doctor is available

In adults, cerebral oedema is uncommon and occurs in the physically slight or younger. Asymptomatic cerebral oedema may be common and may exist before treatment starts, and any fall in GCS is treated urgently with mannitol or hypertonic saline without waiting for imaging — but no adult doses are published. If you quote a dose, it is the paediatric one, and it should be given as such.

Cerebral oedema remains the commonest cause of death in DKA, particularly in young children and adolescents. Adult deaths are mainly from severe hypokalaemia, ARDS and the comorbid state that precipitated the episode.

This is a consultant-level exam written on the assumption that you can already start DKA treatment. So the questions are rarely “what is the diagnosis” and rarely “what fluid”. They are about thresholds and steps: at what glucose the 10% goes up, what you check before you increase the insulin, what counts as resolution, which bolus comes off the deficit, and what changes when the patient is admitted under a different team.

That has a practical consequence for how you read a stem. The pH is not there to tell you the patient is unwell — in a child it is the dehydration percentage. Whether the child is shocked is not scene-setting — it decides an entire bolus. The admitting team is not background — it decides the guideline. Each of those details is a criterion, and if you know them cold you register them while reading rather than hunting for them once you have seen the options.

The distractors follow the same logic. The most convincing wrong answers are correct actions at the wrong point: stopping the insulin when the glucose falls rather than adding glucose alongside it; calling the DKA resolved on a bicarbonate that hyperchloraemia has suppressed; giving a shocked child credit for a bolus that should not be subtracted; starting insulin in HHS before the fluid has done its work. Each is defensible in isolation and wrong in the sequence.

If you want the method behind keeping guideline revision current, see how I rebuilt my guideline revision for the FRCEM Final and our list of the guidelines worth knowing for the FRCEM Final. For the paediatric end of the same territory, see paediatric cardiac arrest in special circumstances.

Go to the source

• The adult guideline — JBDS 02, The Management of Diabetic Ketoacidosis in Adults (revised March 2023). The severity list, the potassium bands and the resolution criteria are all worth reading in the original.
• The paediatric guideline — BSPED DKA guidelines, where you will also find the Integrated Care Pathway and the worked fluid examples.
• The HHS guideline — JBDS 06, The Management of the Hyperosmolar Hyperglycaemic State in Adults (February 2022). Short, and separate from JBDS 02 for a reason.

Doses and thresholds here are a revision guide — always confirm against current guidance, the BNF or BNFc, and your local protocol.