How much sliding scale insulin should I give?

The American Diabetes Association goals for inpatient management of diabetes can be found here. The goal fingerstick for non-critically ill patients is 140-180 mg/dL. Even though we don’t always do it (read: rarely do it), it is ideal to calculate a patient’s inpatient insulin regimen based on body weight. Sliding scales should not be the sole form of insulin administration as there is actually no evidence to support their use!

Sliding scale insulin is extra coverage and usually consists of short-acting insulin like lispro (Humalog) or insulin aspart (Novolog). These forms of insulin start working in 15-60 minues and peak in 1-3 hours.

Short-acting insulin vs. regular insulin for a sliding scale? Studies have found that there is no significant difference in outcomes, but generally regular insulin should be used for patients who are NPO, on TPN or continuous tube feeds.

To create an insulin sliding scale, calculate your patient’s “insulin sensitivity factor.”

ISF = 1700/total daily dose of insulin

For example: if someone takes 20 U insulin glargine (Lantus) and 2 U insulin lispro (Humalog) with meals, their total daily dose of insulin is 20 + 2*3= 26. 1700/26=65, so 1 U Humalog for sliding scale would be expected to lower their fingerstick glucose by 65.

The University of Pittsburgh put out a patient safety study with a preset protocol for low, moderate, and high-risk patient sliding scales that can be found here.

Important note: bedtime or “QHS” sliding scales should, generally speaking, be more gentle than mealtime sliding scales, as there is a risk for overnight hypoglycemia.


Testing for adrenal insufficiency: random cortisol level versus cort stim test

Let’s say a patient presents to the medical floor with hypotension and hyperkalemia, and generalized fatigue and weakness. Adrenal insufficiency might be on your differential. What’s the most accurate way to test for it? Can you get away with using doing a random cortisol level?

Screen Shot 2017-08-15 at 7.53.01 PM

A random serum cortisol level must always be in the context of what time of day it was taken. There are different laboratory cutoffs depending on what time of day the sample was taken. Typically, a morning cortisol level is most helpful. An AM cortisol level >15 mcg/dL is very reassuring that someone doesn’t have adrenal insufficiency. But any level <15 does not exclude adrenal insufficiency! As this case report shows.

A cort stim test (also called ACTH/cosyntropin stim test) involves measuring an AM serum cortisol, injecting 250 mcg of ACTH or cosyntropin, waiting 30 minutes, and then measuring serum cortisol again. An “adequate response” ruling out adrenal insufficiency is  ≥18 to 20 mcg/dL before or after ACTH injection. If there is an inadequate response, you may consider directly measuring ACTH levels or doing other tests to further evaluate for primary vs. secondary adrenal insufficiency.

Note 1: patients with higher levels of cortisol-binding globulin (like cirrhotics or those with nephrotic syndrome) may have lower levels of cortisol, and may be incorrectly diagnosed with adrenal insufficiency using normal cut-off ranges.

Note 2: this doesn’t apply to patients who are really, really sick in the ICU. As this review discusses, critically ill patients are probably relatively adrenally insufficient because they need tons of cortisol to maintain perfusion and create an inflammatory response.

One point that this review makes is:

Our belief is that adrenal insufficiency appears to be unlikely when a random cortisol measurement is greater than 34 μg per deciliter. Conversely, adrenal insufficiency is likely if the serum cortisol level is below 15 μg per deciliter during acute severe illness. For persons with cortisol levels between these two values, a poor response on a corticotropin test would indicate the possibility of adrenal insufficiency and a need for supplemental corticosteroids.

What’s the difference between D2 and D3 for supplementation? And when should you supplement?

I have to look this up myself every time:


One study, albeit small, looked at the effects of vitamin D3 versus D2 supplementation and found that D2 potency is less than 30% of that of D3 and that it has a markedly shorter duration of action. Because the 1,25 form is metabolized in the kidney, D2 is not recommended for patients with CKD or ESRD (D3 should be used). One paper even argues that vitamin D2 should not be sold as a supplement anymore.

Random fact: vitamin D supplementation is a USPSTF grade B recommendation for elderly adults for fall prevention.

Prescribing vitamin D for vitamin D-deficient patients is surprisingly controversial. There are gray areas like what truly counts as “low vitamin D,” racial differences in vitamin D levels (most discussions of vitamin D supplementation are based on evidence in Caucasians), and who should be screened in the first place.

However, here are the quick and dirty guidelines from UpToDate:

Generally, vitamin D deficiency is a serum 25 (OH)D level <20 ng/ml. A couple of specialty societies suggest that a level <30 ng/ml is cause for supplementation in pts >age 65. However people are usually not at risk for osteomalacia unless <10 ng/ml.

Normal adults do NOT need to be screened, but the elderly, those with poor sunlight exposure and malabsorptive disease, should be.

D3 (cholecalciferol) is thought to be more efficacious than D2 (ergocalciferol). Although you will often see someone prescribed 50,000 U weekly followed by 600-800 U daily, there is no evidence behind the 50,000–so you might as well just start them on 600-800 U daily. Vitamin D levels should be monitored every 3-4 months until the target level is met. If someone has malabsorptive disease or isn’t responding to initial treatment, they may require increase of their dose.

How does the cosyntropin test work?

The cosyntropin test is used when adrenal insufficiency is suspected.

At T0, a baseline cortisol level is drawn.

250 mcg of cosyntropin (an ACTH derivative) is administered.

30 minutes later, a repeat cortisol level is drawn.

If the cortisol response is inappropriately low, this suggests adrenal insufficiency.

Further reading: a brief review of the pharmacokinetics of cosyntropin and adrenal insufficiency 


How do insulin pumps work?

Insulin pump: a game-changer in the management of type I diabetes, allowing for more natural insulin administration, tighter control and patient empowerment; a cause of dread for residents who have never had to manage one before.

Screen Shot 2016-07-17 at 9.13.53 AM.png
Parts of a pump: the pump itself, the infusion set (the tubing), the cannula (the needle that goes into the skin). While many people do the traditional “check your fingerstick four times a day” thing, some also use continuous glucose monitoring (CGM) with their pumps. 


It’s worth noting that hospital policy (at the hospital I work at, at least) is that patients should be allowed to manage their own insulin pump as long as they have capacity. There are tales passed down about patients who refuse to give up their insulin pump and send themselves into repeated episodes of hypoglycemia…but there is more going on there than just diabetes.

The pump requires a lot of education, and are costly. But they are also pretty cool. Every patient has basal infusions of insulin pre-programmed into their pump.ONLY rapid-acting insulin is used (U100) but it is infusing 24 hours a day.The program is usually reviewed every few months with their endocrinologist based on their sugar logs and adjusted if the blood glucoses are too high or low.

Patients still have to count their carbs, do four times a day fingersticks, and use a correction factor, but they can calculate their own mealtime bolus and adjust accordingly. Here is a general page about diabetes management for patients on the Joslin Center website that includes information about insulin pumps.

Calculating a bolus: 

There’s a patient-version handy guide from Medtronic, which has this example:

Screen Shot 2016-07-17 at 9.50.11 AM.png


  • hypoglycemia: has the patient been exercising more than usual? Are they drinking alcohol and not eating as much as they should? Is their correction factor too aggressive?
  • hyperglycemia: is the infusion site properly connected? Is the pump itself malfunctioning, and do they need new supplies? Is the infusion site scarred over and not absorbing insulin as well? Is it not calculating the right amount of correction insulin?



Inpatient insulin management and adjustment

Insulin management, in truth, deserves a whole chapter of a textbook, but here are a few quick-and-dirty methods I’ve learned for type 2 diabetics in the hospital:

Selecting a type of insulin:

  • I often try to use lantus over NPH if I have a choice, because lantus is longer-acting and doesn’t have a peak-decrease response
  • I also avoid 70/30 mixed insulin unless a patient uses that at home because 70/30 is more prone to giving people labile blood sugars
  • Short acting insulins, like aspart (lispro) and humalog (humulin) I think of as more interchangeable

Brand Name

(length of time before
insulin reaches bloodstream)
(time period when
insulin is most effective)
(how long insulin
works for)


10 – 30 minutes 30 minutes – 3 hours 3 – 5 hours
Short-acting Regular (R) 30 minutes – 1 hour 2 – 5 hours Up to 12 hours
NPH (N) 1.5 – 4 hours 4 – 12 hours Up to 24 hours
Long-acting Lantus
0.8 – 4 hours Minimal peak Up to 24 hours

(Table credit: The Joslin Center)

Calculating an initial dose of long acting (basal) insulin: Body weight (kg) x 0.2 if insulin naive; and x 0.5 if already on insulin.

Calculating the appropriate sliding scale instead of using your hospital’s predetermined one: 5% of total daily insulin requirement. For example, someone who requires 20 units of basal insulin should start with a sliding scale of 1 unit when glucose 150-200, 3 units when 201-249, etc.

Adjusting insulin on a daily basis: Add up the sliding scale units they received over the last 24 hours. Assign 50% to long-acting insulin and 50% to mealtime insulin. So, if someone is getting 10 units long-acting insulin and 2 units mealtime, and required an extra 12 units yesterday, they should now get 13 units long-acting insulin and 4 units at each mealtime.



Landmark papers in critical care: Insulin

When diabetics get sick, their blood sugars shoot up to ungodly levels. Intuitively, it makes sense to keep their blood sugars within normal range because we like lab values that are normal. But is that really the best thing to do?

The NICE-SUGAR trial set out to examine this question. It enrolled over 6,000 patients in ICUs, almost all of whom were mechanically ventilated. Patients were divided into blood sugar goal groups of 81-108 or <180. Ultimately, the study found 90-day mortality was significantly higher in the 81-108 group, by about 3%, and that severe hypoglycemia was about 7% more common.