ICU Glucose Management: Why the Protocol Beats Your Corrections
Walk through any ICU and pull up the glucose flowsheet on every patient. Almost every one of them is running high, and most of them were never diabetic. The intern's instinct is to treat each number as it appears: a correction here, a few units there, like swatting flies. That instinct is wrong, and in the ICU it is dangerous. Critical illness glucose management is a systems problem, and the system that solves it is the nurse-driven protocol, not your one-off orders. This post is about why, and about the traps that hurt people when you forget it.
Why everyone in the ICU is hyperglycemic
Stress hyperglycemia is the expected physiology of being critically ill. Catecholamines, cortisol, glucagon, and inflammatory cytokines all push glucose up and drive insulin resistance at the same time. Your septic patient is making glucose in the liver at full throttle while the muscle that should be taking it up has stopped answering the phone.
Then we make it worse on purpose. We give stress dose steroids for the septic shock. We hang norepinephrine and epinephrine, which are literally counterregulatory hormones on a pump. We run continuous tube feeds, a steady carbohydrate infusion the patient never got at home. Add it up and a patient with a normal pancreas will run high, and the patient with any baseline insulin resistance will run very high.
So the first mental shift: hyperglycemia in the ICU is not a surprise finding that needs a workup. It is a load that needs a matched, continuous response. The workup question is only worth asking when the glucose behavior does not fit the clinical picture, and I will come back to the one scenario where that really matters.
Why chasing single values fails
Here is what happens when you manage ICU glucose by individual corrections. The nurse calls you with a high value. You order a few units. It works, sort of, two hours later, at which point the steroids have kicked in, or the feeds were held for a CT, and now you are correcting a different patient than the one you dosed. You are always reacting to a number that is already history, with a dose that lands in a future you did not predict. Stack a few of those and you get the sawtooth flowsheet every experienced nurse recognizes: spike, overcorrection, dip, panic, juice, spike.
The protocol wins because it does the two things you cannot do from the workroom. It samples frequently, and it adjusts based on the trend, not the value. That is a control loop. Your scattered corrections are not.
The modern consensus, after years of the pendulum swinging, is moderate glycemic targets in critical illness rather than tight control. Tight control looked elegant on paper and bought hypoglycemia in practice, and hypoglycemia in a sedated, intubated patient is a silent event with real consequences. The exact target range is your institution's call, written into its protocol, and I am deliberately not giving you numbers: use the ones on your unit's protocol, because those are the ones your nurses are titrating to. Moderate targets, protocol titration, and a healthy fear of the low end.
Thinking about the insulin infusion
When do you start a drip? When the glucose is persistently above your unit's threshold despite the easy fixes, when the patient is on pressors or stress steroids and the trajectory is obviously up, or in DKA and HHS, where the infusion is the treatment. In a patient with rapidly changing hemodynamics, absorption of subcutaneous insulin is unreliable anyway. What goes in the vein is what the patient gets.
Once the drip is running, your job is not to micromanage the titration. The protocol and the nurse own that. Your job is to manage the inputs: know what carbohydrate is going in, know what counterregulatory push is coming from steroids and pressors, and anticipate transitions. Control loops fail at the boundaries, not in the middle.
The handoff to subcutaneous, done badly and done well
The most common failure is the transition off the drip. Done badly, it looks like this: the patient is improving, someone turns off the infusion at shift change, orders sliding scale correction only, and walks away. Intravenous insulin has a very short half-life. Within the hour the patient has essentially no insulin on board, and by the time the sliding scale catches up, the glucose has rebounded and, in the type 1 diabetic or the DKA patient, ketosis is rebuilding. You spent two days fixing a problem and undid it with one checkbox.
Done well, the transition is planned like an extubation. You use the recent stable drip rate to estimate the daily requirement, you give a scheduled basal dose before the drip stops so there is overlap while the subcutaneous depot absorbs, and you continue scheduled, not merely reactive, insulin afterward. Sliding scale alone is not a maintenance strategy for anyone who needed a drip. I have written out the full basal-bolus logic, with the arithmetic and example orders, in the guide to inpatient insulin management, and if you take one link from this post, take that one. The transition is where the ICU's work gets preserved or squandered.
The tube feed interruption: the classic hypoglycemia setup
If you want to predict the next hypoglycemic event on your unit, do not look at the insulin orders. Look at the feeds. The patient is on continuous tube feeds with an insulin infusion, or with scheduled subcutaneous insulin dosed to cover those feeds. Then the feeds stop: the tube clogs, the patient pulls the NG, someone makes the patient NPO for a midnight procedure. The carbohydrate infusion is gone. The insulin is still there, either dripping in or sitting in a subcutaneous depot that does not care about your NPO order.
This is the classic, recurring, preventable ICU hypoglycemia story. The fix is a standing reflex: any interruption of feeds triggers an immediate look at the insulin. Good protocols build this in, with dextrose infusions bridging planned interruptions and explicit instructions for unplanned ones. Your job on rounds is to ask the question out loud: if the feeds stop tonight, what happens to this patient's insulin? If nobody has an answer, write one. While you are there, check the potassium and phosphorus too, because insulin drives both into cells; the electrolyte replacement guide covers the how and how much.
Steroid days have a shape
Steroid induced hyperglycemia is not uniform across the day. A morning steroid dose pushes glucose hardest in the afternoon and evening, then the effect fades overnight. Cover that pattern with a flat insulin regimen and you get the worst of both worlds: undertreated afternoons and a hypoglycemic dawn. The same trap appears in slow motion when the steroids taper: the insulin requirement falls with the dose, and a regimen that was right at the peak becomes an overdose by the end of the taper. When the steroids change, the insulin must change with them, same day, not eventually. The endocrinology board review walks through this pattern in more detail.
Euglycemic DKA: the trap the ICU keeps meeting
Here is the scenario where the glucose does need a second look, precisely because it looks fine. A patient on an SGLT2 inhibitor, one of the flozins, comes in sick: postoperative, septic, or simply not eating. The glucose is unimpressive, maybe barely elevated. But the anion gap is open, the bicarbonate is low, and the patient is acidotic. This is euglycemic DKA. The drug keeps dumping glucose into the urine, so the sugar never climbs to the number your brain has filed under DKA, while ketogenesis proceeds underneath.
The trap is that every screening instinct you have is keyed to the glucose. A normal sugar quietly removes DKA from your differential, and the acidosis gets attributed to lactate or renal failure while the ketones go unmeasured. So build the reflex: unexplained anion gap acidosis in anyone on an SGLT2 inhibitor gets ketones checked. Treatment, per your institution's protocol, generally means insulin with dextrose running alongside, because the insulin is there to shut off ketogenesis, not to lower a glucose that was never high. Hold the SGLT2 inhibitor. The ICU keeps meeting this patient because the drugs are everywhere now and the presentation is built to slip past a glucose focused screen.
What to actually do on rounds
Pull the glucose trend, not the last value. Name the inputs out loud: feeds, steroids, pressors, dextrose. Ask the boundary questions: what happens if feeds stop, what happens when the steroids taper, what is the plan for coming off the drip. Trust the protocol for the titration and spend your attention on the transitions. That is the whole game, and if you are heading into the unit, the critical care rotation guide covers how this fits into the rest of your ICU day.
If you are studying this for boards rather than for tonight's shift, the glucose material here maps directly onto tested territory: transition dosing, steroid patterns, and euglycemic DKA are all reliable question fodder. The board review section has practice questions built around exactly these traps, and membership gets you the full question banks and the rest of the guides in one place. The ICU will keep serving you these patients either way. Better to have seen the pattern first on a screen than at the bedside.
Where HistoryandPhysical.net fits in
Glucose is one protocol you should stop overriding. The unit runs on several more.
- Free ICU guides. Inpatient insulin, the sepsis bundle, ICU delirium and bedside procedures, all complete and all free.
- Critical care questions when the exam comes. A critical care bank and a 120 question PCCM board review bank with its own timed exam.
- Evidence with citations. The Evidence tool answers clinical questions against PubMed, Europe PMC and published guidelines and shows you the sources, so you can check the claim rather than trust it.
- Written in 2026. Protocols move. The whole question catalog was written this year.
Membership is $10 a month or $59 a year at the founding rate, with a full refund available within 30 days of any payment.
This is not medical advice. Glycemic targets, insulin infusion titration, transition dosing, and DKA management are governed by your institution's protocols and current references. Verify doses and thresholds locally before acting on anything here.