Abstract: FR-PO308

Predicting Complicated Hyperkalemia in Hospitalized Patients

Session Information

Category: Fluid and Electrolytes

  • 902 Fluid and Electrolytes: Clinical


  • Macedo, Etienne, University of California San Diego Medical Center, La Jolla, California, United States
  • Awdishu, Linda, UCSD Skaggs School of Pharmacy and Pharmaceutical Sciences , La Jolla, California, United States
  • Lee, Euyhyun, University of California San Diego Medical Center, La Jolla, California, United States
  • Mehta, Ravindra L., University of California San Diego Medical Center, La Jolla, California, United States

Hyperkalemia (HighK) is common in hospitalized patients and can be life-threatening. Patient-related factors and process of care, including drugs, are known to be associated with the development of hyperkalemia. In order to prevent hyperkalemia complications, it is fundamental to identify the cause and provide early treatment. In this study, we evaluated risk factors for complicated hyperkalemia (CK) during hospitalization.


Data regarding patient location, comorbidities, medications in use before and after HighK detection and outcomes were extracted from the EMR of all adult patients at an academic medical center with at least 2 consecutive K > 5mEq/L during a hospital stay. CK was defined as: (1) maximum K level ≥ 6.5 mEq/L, (2) need for ICU admission within 24h of HighK diagnosis, (3) need for dialysis, if not ESRD, (4) death with K ≥ 5 mEq/L, (5) more than 3 drugs for hyperkalemia treatment initiated after hyperkalemia diagnosis, and (6) more than 8 days for potassium normalization. We divided the cohort into derivation (n= 1,165), and validation (n=985) cohorts, to build and evaluate the performance of the model. The Least absolute shrinkage and selection operator (LASSO) with 10-fold cross-validation was used to build the model.


From January 1, 2013, to November 30, 2015, 133,807 hospitalized patients had at least one K assessment. Of these patients, 13,748 (10.3%) had one K value higher than 5 mEq/L, and 2,150 (1.6%) had two consecutive values, within 7 days, more than 5 mEq/L. CK occurred in 822(38%) of patients; 488(42%) derivation and 334(34%) in the validation cohort. The majority of the patients, 1,193 (55%) had eGFR > 60ml/min, and 1310 (61%) fulfilled criteria for AKI during the admission, 192 (9%) were ESRD. Mortality in CK was significantly higher, 200(24.3%) vs. 71(5.3%) HighK group (p<0.001). Presence of hypotension, cardiopulmonary diagnosis, fluid overload, infection and ICU location were independently associated with CK. ESRD, myopathy and HIV diagnosis were protective factors. The model AUC was 0.70 in the derivation and 0.64 in the validation dataset.


HighK is common in hospitalized patients and associated with high mortality. A model based on clinical information and process of care may help identify patients with HighK who are at highest risk for complications and require surveillance.


  • Commercial Support