DKA vs HHS: When the Glucose Won't Drop
You will meet both of these on the floor, and they will look deceptively similar at first glance: an older or younger diabetic, glucose in the 400s or 600s, dehydrated, a little confused, maybe vomiting. Your first instinct is to reach for the insulin drip. That instinct is right, but it is also the place where people go wrong, because the glucose number is a red herring. In diabetic ketoacidosis (DKA) the emergency is the acidosis and the ketones. In hyperosmolar hyperglycemic state (HHS) the emergency is the osmolality and the dehydration. Treat the number and you can miss the crisis. This is a practical guide to telling the two apart and managing each one, from the first bag of saline to the moment you are confident enough to stop the drip.
Why the glucose is the wrong thing to fixate on
Both DKA and HHS are hyperglycemic crises, and both start with the same basic problem: not enough insulin to keep glucose in check. Where they diverge is what that insulin shortage does to the rest of the body.
In DKA, insulin is so low that the body can no longer suppress the breakdown of fat. Free fatty acids flood the liver, get converted into ketone bodies (mostly beta-hydroxybutyrate and acetoacetate), and those ketones are acids. The bicarbonate buffer gets overwhelmed, the pH falls, and you get an anion-gap metabolic acidosis. The glucose is often elevated, but it is the ketones and the falling pH that make this dangerous and that drive the deep, gasping Kussmaul breathing you see in severe cases.
In HHS, there is just enough insulin left to stop the fat breakdown (so there is little or no ketosis) but not enough to control the glucose. The glucose climbs very high, often above 600 mg/dL, and that drags water and electrolytes out through the kidneys in an osmotic diuresis. The patient becomes profoundly dehydrated, the blood becomes concentrated, and the serum osmolality shoots past 320 mOsm/kg. It is that hyperosmolarity, not the acidosis, that drives the altered mental status, the seizures, and the thrombotic complications.
So the two crises are really two different emergencies wearing the same coat. The overlap is real. A patient can have both at once, and that is common enough in type 2 diabetes that you should not force every patient into one box or the other.
Who gets what
DKA is the classic crisis of type 1 diabetes, and it tends to come on fast, over hours. It is also the presentation that first reveals diabetes in a lot of young people. Roughly 30 percent of children with type 1 diabetes are diagnosed after a DKA episode, and DKA can be the first sign in adults who never knew they had diabetes. In the United States there are on the order of 135,000 DKA hospital admissions a year, and the cost runs to about $2.4 billion. With prompt, adequate treatment the mortality is low, under 1 to 5 percent, but a DKA admission is still a meaningful marker of future risk.
HHS is the crisis of the older, type 2 patient. It builds slowly, over days to weeks, and the dehydration is much more severe. The average fluid deficit in DKA is about 6 liters. In HHS it is closer to 9 to 12 liters. And the mortality tells the story: HHS kills around 15 percent of the patients who get it, several times the rate of DKA. When you see an older diabetic with a glucose over 600 and a confused, drowsy exam, HHS is the diagnosis to rule in, not rule out.
Telling them apart
Here is the working distinction. DKA is defined by three things together: hyperglycemia (glucose above 250 mg/dL), ketones in the blood or urine, and a metabolic acidosis (pH below 7.3, bicarbonate below 18). HHS is defined by a much higher glucose (usually above 600 mg/dL), a serum osmolality above 320 mOsm/kg, and the absence of significant acidosis (pH above 7.30, bicarbonate above 15), with little or no ketosis.
The American Diabetes Association grades DKA by severity, and it is worth memorizing because it maps directly to how sick the patient looks:
- Mild: pH 7.25 to 7.30, bicarbonate 15 to 18, patient alert.
- Moderate: pH 7.00 to 7.25, bicarbonate 10 to 15, mild drowsiness.
- Severe: pH below 7.00, bicarbonate below 10, stupor or coma.
A few practical points. A venous blood gas is fine for the pH; you only need an arterial sample if oxygenation is in question. For ketones, a capillary blood beta-hydroxybutyrate level is better than a urine dipstick for acetoacetate, because it tracks the actual ketone burden and can shorten the stay. And remember euglycemic DKA, which shows up in about 10 percent of cases: a real DKA with a glucose that looks almost normal. The classic modern cause is an SGLT2 inhibitor, but pregnancy, starvation, and liver disease can do it too. A "normal" glucose does not clear the diagnosis.
The workup
Once you suspect either crisis, the workup is largely the same and it serves two purposes: confirm the diagnosis and find the trigger. You want electrolytes, a basic metabolic panel (the BUN and creatinine will often be elevated from dehydration), a glucose, a blood gas, and a ketone level. Then you hunt for what set this off, because treating the trigger is what prevents the next admission. That means a CBC, a CRP, blood cultures, a urinalysis, and usually a chest X-ray to look for infection, which is the single most common precipitant. Check amylase and lipase if the abdomen is tender, because DKA can mimic an acute abdomen and pancreatitis can mimic DKA. If the mental status is out of proportion or you are worried about cerebral edema (more a pediatric concern), a CT head is reasonable.
Treating DKA: fluids, insulin, potassium, and the trigger
There are four pillars, and the order matters.
1. Fluids first
The patient is dry, often profoundly so. Normal saline (0.9 percent) is the fluid of choice, and small trials of balanced crystalloids have not shown a meaningful difference. If the patient is in shock or has a depressed level of consciousness, give 1 liter (adults) or 10 mL/kg (children) quickly to restore circulating volume, then slow down and replace the rest based on the estimated deficit. As the glucose falls toward 250 to 300 mg/dL, add dextrose (for example D5NS) so you can keep the insulin running without sending the glucose into the floor. One special case: if the low blood pressure is from a failing heart rather than dehydration (cardiogenic shock), the patient needs the ICU, central line monitoring, and inotropes, not just more fluid.
2. Insulin, but not before potassium
This is the step that kills people, so it deserves its own section, and it applies to HHS exactly the same way. Insulin drives potassium into the cells, and these patients have already lost a lot of total body potassium through the osmotic diuresis. The serum level can look reassuring at the start and then crash once the insulin starts. So the rule is simple: check the potassium, and if it is below 3.3 mmol/L, do not start the insulin yet. Replete the potassium first, then start the drip.
Once the potassium is safe, start regular insulin at 0.1 units per kilogram per hour. Some protocols give a 0.1 units/kg IV bolus first; others skip the bolus and just start the infusion, and both are defensible. The infusion should bring the glucose down by roughly 50 to 75 mg/dL per hour. When the glucose reaches 250 to 300 mg/dL, reduce the insulin rate and add dextrose, because at that point you are no longer trying to lower the glucose, you are trying to clear the remaining ketones. Keep going until the ketones are gone.
3. Potassium, the running game
Think of potassium as a moving target you manage all the way through. If the level is below 3.3, hold the insulin and replace it. If it is between 3.3 and 5.3, add 20 to 30 mEq per liter to the fluids to keep it in the 4 to 5 range. If it is above 5.3, hold the potassium and recheck, then add it once it drifts down. Continuous cardiac monitoring and frequent potassium checks are non-negotiable, because the arrhythmias that come from a falling potassium are the ones that will end the admission badly.
4. Find and treat the trigger
Infection is the most common cause, so treat it if you find it. But the list is long: a missed insulin dose, a new SGLT2 inhibitor or a course of steroids, a heart attack, a stroke, an eating disorder with deliberate insulin restriction. Document it, treat it, and use it for the discharge plan, because the trigger is usually the reason the patient will come back.
Treating HHS: slower, bigger, and more careful
The pillars are the same, but the tempo is different. The dehydration is much larger, so the rehydration is a 24-hour project rather than an hour-and-a-half one. Start normal saline at about 1 liter per hour for the first few hours, then titrate to the response. The glucose should come down slowly, and you want to avoid a rapid fall in osmolality because that is what drives cerebral edema. Insulin is the same 0.1 units/kg per hour once the potassium is above 3.3, and the potassium replacement logic is identical. Given the high thrombotic risk from the hyperviscosity, low molecular weight heparin for VTE prophylaxis is often reasonable. The goal of resolution is a glucose under 200 mg/dL, a normalized osmolality, and a patient who is back to their neurologic baseline, which typically takes a few days.
When is it safe to stop the drip?
This is the question that determines whether the patient rebounds. The criteria for resolution in DKA are that the patient is tolerating oral intake, the pH is above 7.3, and the ketones have cleared (beta-hydroxybutyrate below 1 mmol/L or a negative urine ketone). Once those are met, you transition to subcutaneous insulin. The key is the overlap: give the first dose of the patient's basal (long-acting) subcutaneous insulin, and then stop the IV drip about an hour later. That hour of overlap is what prevents a gap in insulin coverage. If you stop the drip the moment you give the basal shot, or if you stop the drip before the basal has had time to absorb, the patient can bounce straight back into DKA. Keep checking glucose and ketones for several hours after the switch, and make sure the patient (or the family) understands the sick-day rules before they leave.
Pitfalls worth remembering
- The glucose is a red herring. Treat the acidosis in DKA and the osmolality in HHS. A normal-ish glucose does not exclude DKA.
- Always check potassium before insulin. It is the single most testable and most dangerous step in the whole algorithm.
- Do not stop the IV insulin early. The one-hour overlap with the subcutaneous basal prevents rebound DKA.
- SGLT2 inhibitor euglycemic DKA is a modern classic. Hold the drug in acute illness and before surgery, and keep a low threshold for checking ketones in patients on these agents.
- Ketosis-prone type 2 diabetes ("Flatbush diabetes") is DKA in a patient who turns out to have type 2 diabetes, more common in African and Hispanic patients. Checking GAD and islet cell antibodies helps you decide whether they need lifelong insulin or can go back to oral agents.
- HHS is the deadlier crisis. Older patient, glucose over 600, confused exam, and a 15 percent mortality should make you move fast on the fluids.
Bottom line
Both crises are managed with the same four tools: fluids, insulin, potassium, and treatment of the trigger. What changes is what you are actually rescuing the patient from. In DKA you are fixing an acidosis; in HHS you are fixing an osmolality. Keep the glucose in your peripheral vision, keep your eyes on the pH and the osmolality, check the potassium before you touch the insulin, and give the subcutaneous basal an hour of overlap before you pull the drip. Do that and the "glucose that won't drop" usually ends up dropping exactly when it is supposed to.