On the same track as my previous post about artificial intelligence, I am continuing in the trend by reviewing an article that looked that predictors of something bad happening to someone after having hyperkalemia.
As the authors very nicely explain, this work is important as it relates to diet because we want to avoid being overly restrictive with dietary potassium intake, especially if the risk of someone having a bad outcome related to their hyperkalemia is relatively low.
What did they do?
The authors used different Artificial Intelligence models to look at the risk of:
- All-cause mortality
- Starting renal replacement therapy (dialysis or transplant)
- Hospitalization for heart failure
- Cardiovascular events (such as MI, arrhythmia, cardiac arrest or a stroke)
They were looking for the predictors of these events in a data set of 24,949 adults who had a previous hyperkalemic event (defined as a K at or above 5.1mmol/L).
What did they find?
| Outcome | Predictor |
| Heart Failure | History of ER visit High Brain Natriuretic Peptide (BNP) Older Age |
| Starting Renal Replacement Therapy | Lower eGFR/advanced CKD stage History of AKI Younger Age |
| Cardiovascular Events | History of CVD, A.Fib or MI |
| Death | Older age History of Chronic Pulmonary Disease Sepsis ER visit |
Take Aways
Ok, well I don’t think any of these identified predictors are that awe-inspiring for me. Mostly, it just looks like the sicker our patients are, the more likely they are to experience an adverse event after hyperkalemia. That makes good intuitive sense.
But that being said, I still like the concept of the article and I am 100% on board with the goal of developing personalized risk evaluations for our patients to ensure that we match the right patient with the right diet advice. Clearly an area for us to keep watching!
