Nephrology Topic

AKI is a definition, not a diagnosis.

Creatinine is a lagging indicator. Eight cards on what AKI is, what it costs, and when the circuit helps: the definition, the three questions, the pressure gradient, dialysis indications and timing, CRRT modes and prescription, and recovery.

Reviewed October 2026 · verify against current guidelines

Renal · 1 of 8

AKI is a definition, not a diagnosis.

Two numbers meet it: a creatinine that has risen, or a urine output that has fallen. Both describe filtration, which is a function. Neither names the injury, and neither tells you when it started.

Any one of three meets it — KDIGO, and the windows are part of the criterion

≥ 0.3	A creatinine rise in mg/dL, or 26.5 µmol/L, within 48 hours.
≥ 1.5 ×	Creatinine against baseline, known or presumed, within 7 days.
< 0.5	Urine in mL/kg/h, sustained for 6 hours.

Stage on the worse of the two — whichever criterion is further along

Stage	Creatinine	Urine output
1	1.5 to 1.9 × baseline, or a rise of 0.3 mg/dL	< 0.5 mL/kg/h for 6 to 12 hours
2	2.0 to 2.9 × baseline	< 0.5 mL/kg/h for 12 hours or more
3	3.0 × baseline, or a creatinine of 4.0 mg/dL, or dialysis started at any level	< 0.3 mL/kg/h for 24 hours, or anuria for 12

Four reasons the creatinine reads better than the kidney is

It measures function, not damage. Tubules can be injured while filtration is still compensated, so the number moves after the harm.
Volume dilutes it. A resuscitated patient can hold a flat creatinine on a falling filtration rate.
Production falls. Sepsis, immobility and low muscle mass all cut generation, blunting the rise.
The baseline is often a guess. No prior value means the fold-change criterion is being applied to an assumption.

Two traps — a good urine output does not exclude AKI, and roughly two thirds of cases never become oliguric at all — the creatinine criterion stands on its own. And stage 3 is reached the moment dialysis starts, whatever the numbers were: the stage records the decision, so it cannot be used to justify it.

Renal · 2 of 8

Prerenal is a hypothesis.

It is a label you can only apply afterwards, once function has recovered with perfusion restored. Before that it is a guess, so ask the three questions instead and treat what each one turns up.

Three questions, asked together — not in sequence; two can be true at once

Is flow getting in?	Pressure, output, venous congestion, abdominal pressure. Also the drugs that uncouple flow from filtration — NSAIDs, ACE inhibitors, ARBs.
Is urine getting out?	Scan the bladder and the kidneys. A blocked catheter is the commonest obstruction on the unit, and it is the fastest thing on this card to fix.
Is the tissue injured?	Sepsis, nephrotoxins, contrast, myoglobin, crystals, an immune glomerular or interstitial process. The urine is where you look.

What the urine shows — a fresh sample, looked at under a microscope

Finding	Points to	Next step
Muddy brown granular casts	Tubular injury	Stop the toxin, support flow
Dysmorphic red cells or casts	Glomerulonephritis	Nephrology, serology, biopsy
White cell casts, pyuria	Interstitial nephritis	Stop any drug new in 2 weeks
Blood on dipstick, no cells	Myoglobin or hemoglobin	Creatine kinase, then volume

Why the sodium fraction disappoints on the unit — a fractional excreted sodium under 1% means the tubule is still avidly reabsorbing. That is a statement about tubular behavior, not about the cause — and in sepsis, cirrhosis, heart failure, contrast and early obstruction the tubule behaves that way while injured. Any diuretic in the last day invalidates it, and the urea fraction under 35% performs no better. Both hold up in one narrow question — oliguric prerenal against tubular injury — and add little outside it.

Two traps — a normal ultrasound does not exclude obstruction — hydronephrosis takes time to appear and may never appear in retroperitoneal disease. And urine eosinophils are not a test for interstitial nephritis: they are neither sensitive nor specific, and a negative result should not talk you out of stopping the drug.

Renal · 3 of 8

The kidney is a pressure gradient.

Blood has to cross the organ, so perfusion is the arterial pressure minus whatever presses on the other side. Raise the venous pressure and the gradient narrows just as surely as if the mean arterial pressure had fallen.

Same mean arterial pressure, two different kidneys — MAP 65, unchanged

	Venous pressure	Gradient left to perfuse the kidney
Uncongested	8	65 − 8 = 57
Congested	20	65 − 20 = 45

Three pressures to know — two of them are not the blood pressure

65 — mean arterial, mmHg	The default target. SEPSISPAM found a higher target cut dialysis only in chronic hypertensives, at the cost of more atrial fibrillation.
> 12 — central venous, mmHg	Above this the association with AKI strengthens step by step. It is the cheapest gradient to widen.
> 12 — intra-abdominal, mmHg	Above 20 with a new organ failure is compartment syndrome, and the kidney usually fails first.

Why more fluid so often makes it worse — the kidney sits inside a fibrous capsule, so it cannot swell outwards. Congestion raises the pressure within it, the gradient narrows, and filtration falls — while the chart shows a normal blood pressure throughout. Fluid responsiveness predicts that stroke volume will rise. It does not predict that the kidney will. Once the shock has resolved, taking fluid back is the intervention.

Two traps — oliguria in a congested patient is not a signal to give volume — it is often the consequence of the volume already given. And no drug protects the kidney: neither low-dose dopamine, diuretics, nor mannitol change the outcome. Diuretics buy room for fluid and time for lungs, not renal recovery.

Renal · 4 of 8

When to start dialysis.

Either there is an indication or there is not, and the creatinine is never one. Every line below means refractory — you have already given the thing that usually works, and it has not worked.

Five indications — any one of them is enough on its own

Indication	What makes it one	Already tried
Potassium	Rising past 6.5 mmol/L, or any level with ECG changes	Calcium, insulin with dextrose, a beta agonist, a binder
Acidosis	A pH near 7.15 that will not lift, usually with a rising lactate	The cause, ventilation, and bicarbonate if there is room for it
Overload	Oxygenation failing on a lung you cannot dry out	A diuretic challenge, and stopping what is still going in
Uremia	Encephalopathy, pericarditis, or bleeding — never a number alone	Nothing. These are the indication, not the warning before it
Toxin	Lithium, salicylate, metformin, methanol, ethylene glycol, valproate	The antidote and supportive care, where one exists

What a circuit commits you to

A catheter. A large vein, held for days, with the infection and thrombosis that follow it.
An anticoagulant. Running continuously, and needing its own monitoring and its own rescue.
A clearance. Every drug now has to be dosed around a kidney that is partly a machine.

Watch the trajectory

Ask what the kidney has failed to do, and where that failure is heading. In AKIKI, half the patients randomized to wait never needed dialysis at all — the indication simply never arrived.

Two traps — a urea or creatinine threshold is not an indication, and no trial has supported starting on the number alone. And in the poisoned patient the logic inverts: the toxin is the indication, the kidney may be working perfectly, and the decision belongs to toxicology rather than to this card.

Renal · 5 of 8

What the timing trials found.

Five trials asked whether starting before an indication appears helps. One said yes, three said no, and the fifth showed that waiting past an indication does harm. The window has two edges: before an indication — no benefit, and more left dependent; when one appears — start here, and not before; past an indication — waiting itself now costs lives.

Starting early, against waiting — all randomized, all in critical illness

Trial	Patients	Verdict on starting early
ELAIN	231	Lower mortality — but a single surgical center, and the only trial to favor it
AKIKI	620	No difference, and half the delayed arm never needed dialysis at all
IDEAL-ICU	488	No difference in septic shock; stopped early for futility
STARRT-AKI	2927	No difference at 90 days, and more survivors were still dialysis-dependent
AKIKI 2	278	Delaying further was worse — waiting has a floor as well as a ceiling

What the five of them settled together — waiting is safe, and it is not free: nothing is gained by pre-empting an indication, and AKIKI 2 showed that waiting through one costs lives. Watchful waiting means watching. Early starting has its own cost: more STARRT-AKI survivors randomized to start early were still dialysis-dependent at 90 days — a circuit begun sooner is not a circuit ended sooner.

Two traps — ELAIN is the trial people quote, and it is the smallest, from one surgical unit, against 4300 patients pointing the other way — weigh it accordingly. And “delayed” in these trials meant waiting for an indication, not waiting and hoping: every delayed arm was monitored to a protocol, which is the arm you are copying at the bedside.

Renal · 6 of 8

Convection, diffusion, or both.

The letters on the machine describe how solute crosses the membrane, not how sick the patient is. There are only two mechanisms, and every mode is one of them, the other, or the pair.

Two ways across the same membrane

Diffusion — solute crosses, the water stays	Counter-current dialysate holds the far concentration low, so small molecules keep crossing. Urea and potassium clear fast.
Convection — water crosses, the solute rides along	Pressure drives plasma water across, dragging solute along. Slower per liter, but larger molecules come too.

Reading the mode — the H is convection, the D is dialysate

Mode	Clears by	What you hang, and what it is for
SCUF	Minimal	Nothing. Water comes off, solute barely follows — for overload alone
CVVH	Convection	Replacement fluid, before or after the filter
CVVHD	Diffusion	Dialysate, counter-current. Efficient for urea and potassium
CVVHDF	Both	Dialysate and replacement. What most units run by default

Why convection has a ceiling, and where the fluid goes in — pulling water off concentrates the blood inside the filter. Past roughly a fifth to a quarter of the plasma flow, it clots — so convective clearance is capped by rheology, not by the prescription. Replacement fluid before the filter dilutes the blood and buys filter life, at the cost of clearance per liter. After the filter is the reverse: better clearance, shorter filter.

Two traps — the dose is the effluent, which sums dialysate, replacement and net ultrafiltration — and with pre-filter replacement the delivered clearance is lower than that number reads. And convection does clear middle molecules better, but that has never translated into survival in AKI: choose the mode for the fluid you have and the filter life you need, not for the cytokines.

Renal · 7 of 8

Running the circuit.

Modality does not change survival, so choose it for the hemodynamics. Dose does not change survival above a threshold, so there is nothing to gain by going higher. What is left to get right is the anticoagulation and the drugs.

Choose on tolerance, not outcome — no modality has shown a mortality advantage

Continuous	Slow, gentle, and the choice on pressors or with a raised intracranial pressure. Costs immobility and continuous anticoagulation.
Intermittent	Fast clearance in hours, so it is the answer for refractory potassium or a dialyzable poison. Poorly tolerated in shock.
Prolonged	Six to twelve hours, usually overnight. Most of the tolerance of continuous, with the days left free.

The prescription — written on actual body weight

20–25 — effluent, mL/kg/h	Higher doses were tested to 35 and 40 and gave nothing.
25–30 — what to prescribe	Downtime for clotting, imaging and the OR eats the difference.
< 2.5 — total to ionized calcium	Above this, citrate is accumulating faster than the liver clears it.
Right IJ — access, in order	Then femoral, then left IJ. Not subclavian — it stenoses the vein they may need later.

Why citrate is first line, and how it bites — it chelates calcium inside the circuit only, so the filter is anticoagulated and the patient is not. Filters last longer and bleeding is less than with heparin, which is why guidelines put it first even without a bleeding risk. The liver turns the returned citrate into bicarbonate. A failing liver cannot, so citrate accumulates: the ionized calcium falls while the total rises, and the patient develops an acidosis on a machine prescribed to correct one.

Two traps — the circuit clears antibiotics, so a standard renal-failure dose underdoses the patient on continuous therapy — ask pharmacy, and watch phosphate and magnesium fall too. And weaning is judged by the kidney, not the numbers: a urine output recovering past roughly 400 mL a day without diuretics is the sign that liberation will hold.

Renal · 8 of 8

Recovery is not a return to baseline.

The creatinine comes down, the catheter comes out, and the patient goes home. None of that resets the risk. An episode of AKI is the start of a course that runs for months, and the number that looks normal can sit on a kidney with fewer nephrons.

One continuum, three windows — days from the injury

AKI — the first 7 days	Injury, and what it costs.
AKD — day 7 to day 90	Recovered, persistent, or relapsing.
CKD — beyond day 90	A new baseline, and a new risk.

The stage sets the tail — adjusted hazard of new CKD, against no AKI

Stage 1 — 2.0 ×	Even the mildest bump, the one that gets no mention in the discharge summary.
Stage 2 — 3.3 ×	The risk of dialysis in the long run rises five-fold at this stage.
Stage 3 — 28 ×	Severe injury is a different disease in the follow-up clinic, not a worse version of stage 1.

Why a normal creatinine can hide a smaller kidney — nephrons lost in the injury do not come back. The survivors hyperfiltrate to cover the shortfall, so filtration — and the creatinine that reports it — can look restored on a kidney that has lost reserve. That compensation is itself a driver of later scarring. This is where the damage markers belong: NGAL and TIMP-2·IGFBP7 rise with tubular injury before filtration falls, and the coming KDIGO framework folds them in beside the creatinine. None has yet been shown to change an outcome, so they inform rather than decide.

At three months

Reassess everyone who had AKI: creatinine and eGFR, urine albumin, blood pressure, and a look at what was restarted. That visit is what turns an episode into a diagnosis, and it is the one most often skipped.

Two traps — discharging on the drug that caused it is common, because the hold was recorded on the ward and never revisited; the ACE inhibitor, the NSAID and the metformin all need a decision, not a default. And a survivor whose creatinine normalized is still a survivor of AKI — the risk sits with the episode, so the discharge summary has to name it rather than leave the next clinician to infer it from a graph.

Educational reference only — not medical advice. Verify against local protocol.

Sources

Verify against current guidelines and local protocol before acting.

KDIGO Clinical Practice Guideline for Acute Kidney Injury — definition and staging.
SEPSISPAM — higher versus lower blood-pressure target in septic shock.
ELAIN — early versus delayed renal replacement therapy (n = 231).
AKIKI — early versus delayed renal replacement therapy (n = 620).
IDEAL-ICU — timing of renal replacement therapy in septic shock (n = 488).
STARRT-AKI — accelerated versus standard renal replacement therapy (n = 2,927).
AKIKI 2 — delayed versus more-delayed renal replacement therapy (n = 278).

Downloads

Every card behind this topic, as published — view full size or download.

Carousel8 slides