What does hyperkalemia, low hemoglobin and artificial intelligence have in common?

They were all discussed in today’s journal article!

There is growing recognition that the causes of hyperkalemia are multi-factorial. This has increased the encouragement for clinicians to consider all potential causes before initiating (or doubling down on) low potassium diets.

There are also a growing number of studies starting to use artificial intelligence to find predictors of hyperkalemia causes. So, today I am summarizing:

• Chang HH, Chiang JH, Tsai CC, Chiu PF. Predicting hyperkalemia in patients with advanced chronic kidney disease using the XGBoost model. BMC nephrology. 2023 Jun 12;24(1):169.

With a special shout-out to this article, referenced by Chang et al: Mansoor F, Bai P, Kaur N, Sultan S, Sharma S, Dilip A, Kammawal Y, Shahid S, Rizwan A. Evaluation of serum electrolyte levels in patients with anemia. Cureus. 2021 Oct 1;13(10).

What did they do?

The authors of this study used an artificial intelligence model to help identify predictors of hyperkalemia for adults living with advanced kidney disease.  They provided a data set containing information about 1,965 adults into the model. Finally, they compared the ability of nephrologists or the computer to better predict who would have hyperkalemia.  The variables they included were:

• Age
• Gender
• Lab data
• Comorbidities
• Medications

The outcome? The ability of either the human or the computer to predict whether or not the next serum potassium value would be at or above 5.5mmol/L.

What did they find?

The top four predictors of hyperkalemia were:

1. The preceding potassium value – with a high potassium in the preceding values being the strongest predictor of a high potassium value on the next check.
2. Angiotensin-receptor blocker (ARB) use – such that those on ARBs where more likely to experience hyperkalemia (likely related to reduced serum aldosterone levels and decreased renal blood flow)
3. Low Hemoglobin levels – likely related to iron-deficiency anemia, sickle cell anemia or GI bleeding.
4. Calcium polystyrene sulfonate use – predicted recurrent hyperkalemia likely because this medication is being used in those who are predisposed to higher baseline serum potassium values.

And who did better at predicting hyperkalemia, the clinicians or the machine? The machine!

Why does this study matter?

I choose to talk about this study for two reasons.  First, it is good for dietitians to be aware of things like artificial intelligence and the potential for it to inform our practice and work.  Artificial intelligence is of keen interest in the global community. We are likely to see more interest in it and be encouraged to use it to help inform our practice.

My second reason for choosing this article was the predictors that they found.  Did you know that low hemoglobin was associated with hyperkalemia? I didn’t!  According to the study referenced in the discussion of the Chang et al article (Mansoor et al), the cause of this has to do with the Sodium-Potassium pumps found on the cell membranes of red blood cells. Changes in the red blood cell from either iron-deficiency anemia or sickle-cell anemia can affect these sodium-potassium pumps and cause hyponatremia and hyperkalemia. 

Take Aways

Next time you get a high potassium level on your desk, take a quick look at the hemoglobin and sodium.  If the sodium and hemoglobin are low, consider that this could be a potential contributor to hyperkalemia.