everything in the department, ranked by how fast it kills
| In the hospital | What it actually is | What you already call it |
|---|---|---|
| Sensitivity | Of everyone who has the disease, the fraction the test catches | recall / TPR |
| Specificity | Of everyone who doesn't, the fraction the test clears | TNR (1 − FPR) |
| Positive predictive value | Of everyone who tested positive, the fraction who really have it | precision |
| Pre-test probability | What you believed before the test came back | the prior |
| Likelihood ratio | How far one result is allowed to move you | Bayes factor |
| Post-test probability | Where you land after updating | the posterior |
| Protocol (ACLS, ATLS) | A fixed response to a class of failure, run without re-deriving it | runbook · state machine |
| Triage | Service order under permanent scarcity, with preemption | priority queue |
| Compensation | Reserve absorbing an error so the outputs still look normal | headroom before saturation |
| Decompensation | Reserve exhausted; the curve stops being a line | falling off the cliff |
| Homeostasis | Several loops holding a variable near a setpoint | closed-loop control |
| The H's and T's | The enumerated reversible causes, worked through out loud | a known-failure-mode checklist |
| Red flags | Cheap questions that catch the catastrophic misses | asserts on the dangerous invariants |
| A before B before C | Work the failures in order of how fast each one kills | ordering by blast radius |
Move the prior and watch what happens. The test characteristics never change — the same sensitivity and specificity — and yet the answer changes completely, because a result is evidence, not a verdict. Run FAST on a patient you were already sure about and a positive confirms what you knew. Run it on someone you thought was fine and a positive is mostly a false alarm, because the false positives are drawn from a much larger pool.
Now look at D-dimer: 41% specificity. Six of every ten healthy people test positive. No engineer would ship that classifier. It is a genuinely excellent test, and the reason is cost asymmetry — a missed pulmonary embolism kills the patient, a false positive costs a CT scan. So the test is tuned all the way toward recall, deliberately, and used for exactly one job: ruling out, in patients whose prior was already low.
That is also the discipline it demands. A test built to rule out tells you almost nothing when it comes back positive, and using it that way is the single commonest reasoning error in the department. Which test you reach for is a statement about which mistake you are least willing to make.
RSI is rapid sequence intubation: sedate and paralyse in one go, then place the tube, with no bag-mask ventilation in between. The point of the sequence is to keep air out of the stomach and stomach contents out of the lungs.
Almost the whole clock runs before anyone picks up a blade. The intubation itself is step six. When it goes badly on screen, it is nearly always because a step before it was skipped — the pressure was never optimised, or the preoxygenation was cut short.
Sux (succinylcholine) works in about 45 seconds and wears off in eight minutes — which is why it is the choice when you want the option of a patient who starts breathing again. It is off the table in hyperkalaemia, crush injury, or burns more than a day old. Roc (rocuronium) has the same onset and lasts three quarters of an hour.
A bougie is the thin flexible introducer you pass first when you can only see part of the cords; the tube railroads over it. A cric — cricothyrotomy — is the neck incision you make when you can neither intubate nor oxygenate. It is the last box on the algorithm, and reaching it is a failure of everything above, not a flourish.
Listen carefully to that word, though. Cric pressure in the quote above is not the surgical airway at all — it is cricoid pressure, a thumb pressed on the cartilage to flatten the oesophagus behind it. Same syllable, opposite stakes; only the sentence around it tells you which one is being asked for.
The loop is two minutes long and it does not stop for conversation. Compress, check the rhythm, shock it if it is shockable, resume within seconds, repeat. ROSC — return of spontaneous circulation — is the goal, and the moment everyone stops moving.
Only two rhythms get shocked: VF (ventricular fibrillation) and pulseless VT. The other two — PEA (organised electrical activity with no pulse) and asystole — get compressions, adrenaline, and a hunt through the H's and T's, the reversible causes. Shocking asystole does nothing, which is why the defibrillator scene in most other shows is wrong.
Watch the ETCO₂ column. End-tidal CO₂ measures carbon dioxide coming back out of the lungs, which requires blood to have carried it there — so it is a live readout of whether the compressions are actually perfusing. Under 10 means they are not. A sudden jump toward normal is often the first sign of ROSC, arriving before anyone has found a pulse.
It is the most honest number in the room, because unlike a rhythm on a screen it cannot show you electrical activity that is moving no blood.
A tension pneumothorax is air entering the chest through a one-way flap and never leaving. The pressure collapses the lung, then keeps climbing until it shoves the heart and great vessels across the midline and kinks the veins returning blood to the heart. Death here is not from the lung failing. It is from the circulation being squeezed shut.
That is why it is one of the very few diagnoses made on examination and treated before imaging: hypotension, absent breath sounds on one side, distended neck veins. Tracheal deviation is the classic sign and also a late one — if you can see it, you have waited too long.
Needle decompression converts a tension pneumothorax into a simple one by giving the air a way out. It is a temporising move, minutes of grace, and it can fail or kink. Tube thoracostomy — the chest tube — is the actual treatment.
Placement is the triangle of safety, in the fifth intercostal space at the anterior axillary line, and always over the top of the rib: the artery, vein and nerve run in a groove along the rib's underside. Trauma guidance moved the needle site here from the old second-space landmark, because the chest wall is thinner laterally and the older site missed often in adults.
The FAST exam is four ultrasound windows, done at the bedside in under two minutes, while everything else is still happening. It is deliberately narrow: it asks whether there is free fluid where there should be none, and in a trauma patient free fluid is assumed to be blood.
Fluid is anechoic — it reflects nothing back — so it reads black on the screen. The windows are chosen as the places gravity takes blood when someone is lying flat: the sac around the heart, the recess between liver and right kidney (Morison's pouch, the most sensitive of the four), the equivalent space beside the spleen, and the pelvis.
What makes it worth the two minutes is that it changes the destination, not just the chart. A hypotensive patient with a positive FAST goes to the operating room. The same patient with a negative FAST can survive the trip to CT, which is a far better test but a corridor away and useless to someone who is crashing.
E-FAST adds two windows on the front of the chest, looking for the sliding of lung against chest wall. When the sliding is absent, there is air in between — a pneumothorax, found before the x-ray exists. And note the limits: FAST is poor at retroperitoneal and hollow-organ injury, so a negative scan rules very little out.
Two layers of pleura normally slide against each other with every breath. On ultrasound that movement is visible directly, and it is the single fastest way to look for a pneumothorax — seconds, at the bedside, with the probe already in your hand from the FAST.
The bat sign is the landmark: two ribs casting dark shadows, with the pleural line slung between them like a body between wings. Below it, the horizontal A-lines are not anatomy at all — they are the same reflection echoing back at even intervals.
M-mode resolves the argument. It plots one line of the image against time, so anything that moves becomes grainy and anything still stays a straight stripe. A sliding lung gives motionless chest wall above and granular texture below — the seashore sign, waves over sand. When sliding stops, the sand becomes stripes: the barcode sign.
Here is the part worth holding onto. Absent sliding is sensitive but not specific. A bleb, a pleural adhesion, severe ARDS, a breath being held — all abolish sliding with no pneumothorax anywhere. So does an endotracheal tube pushed past the carina into the right mainstem bronchus, which stops the left lung ventilating altogether. That is a finding about the tube, not the pleura, and the fix is to pull it back a few centimetres rather than to put a needle in a chest that did not need one.
The one finding that is specific is the lung point — the exact spot where sliding stops and starts again with each breath, the edge of the collapsed lung caught in the frame. Find it and the diagnosis is made.
Shock is not a blood pressure. It is tissue not getting enough oxygen to do its work — which is why lactate rises and why a patient can be in shock with a pressure that still reads normal. The pressure is a late and lagging symptom of it.
Every cause is a failure of the same three-part circuit: a tank of volume, a pump, and the pipes between them. Naming which part failed is the entire clinical move, because the four failures want four different things, and three of them are made worse by the treatment the fourth needs.
Pour fluid into a failed pump and you flood the lungs. Give a pressor for haemorrhage and you squeeze a circuit that is empty — a better-looking number on the monitor, no more oxygen delivered, and the real problem hidden for another ten minutes.
Obstructive shock is the one worth watching for on screen, because the fix is never pharmacological. Tamponade — blood in the sac around the heart, stopping it from filling — is drained. A tension pneumothorax is decompressed. Beck's triad (low pressure, distended neck veins, muffled heart sounds) is the classic description, and, like tracheal deviation, all three together is a late and unreliable gift.
Real patients are frequently mixed: the septic patient is also dry, the trauma patient is also cold and acidotic. The taxonomy is not a set of boxes so much as a list of questions to ask fast, in a room where the answer changes what goes into the line.
| Reading | What it actually measures | Where you want it | When they start moving | What it changes |
|---|---|---|---|---|
SpO₂ |
Percentage of haemoglobin carrying oxygen — a saturation, not a delivery rate | 94–98% | < 90% and falling | Escalate the airway. Unreliable in cold, poor perfusion, or carbon-monoxide poisoning, where it reads falsely high. |
ETCO₂ |
CO₂ exhaled per breath — requires blood to have carried it to the lungs | 35–45 mmHg | < 10 during CPR | Confirms tube placement, grades compression quality, and a sudden rise is often the first sign of ROSC. |
MAP |
Mean arterial pressure — the average driving pressure organs actually perfuse at | ≥ 65 mmHg | < 65 sustained | Fluids, blood, or a pressor. MAP is the number that matters here, not the systolic everyone quotes. |
Shock index |
Heart rate divided by systolic pressure — the ratio, not either number alone | < 0.7 | > 0.9 | Catches compensated shock while the blood pressure still looks normal. Rises before the pressure falls. |
GCS |
Glasgow coma scale, 3–15: eye, verbal and motor response scored separately | 15 | ≤ 8 | "GCS 8, intubate" is a prompt, not a rule — the real question is whether they can protect the airway. |
Lactate |
Product of anaerobic metabolism — tissue running without enough oxygen | < 2 mmol/L | > 4 mmol/L | Marks occult hypoperfusion and drives the sepsis bundle. Clearance over hours matters more than any single value. |
Temp |
Core temperature — and in trauma, a clotting variable rather than a comfort one | ≥ 36 °C | < 35 °C | Warm the room, the fluids and the blood. Hypothermia, acidosis and coagulopathy reinforce each other. |