Study population
This observational cohort study was performed in two medical centers during a 3-year period. The institutions involved were Seoul National University Hospital (Seoul, Korea) and Seoul National University Boramae Medical Center (Seoul, Korea), which are tertiary referral hospitals with 1,600 and 800 beds, respectively, and an academic affiliation with Seoul National University College of Medicine. Using the electronic medical record system, we identified the population of hospitalized patients at these centers between August 2007 and July 2010; we enrolled patients at or over the age of 19 years who had at least one severe hyperkalemic event, with serum K+ levels of at least 6.5 mEq/L. In patients who had several of these events, the first event was used for analysis. All cases of severe hyperkalemia were diagnosed either at the time of admission to the hospital or during the period of hospitalization. This study was approved by the institutional review boards of both hospitals; the need for informed consent was waived because of the study's retrospective design. All clinical investigations were conducted in accordance with the guidelines of the 2008 Declaration of Helsinki.
Data collection
Detailed evaluations of hospitalizations, prescriptions, and laboratory findings were performed for all identified patients by using the electronic medical record systems of the institutions. Data, including patients' medical histories, comorbid diseases, medications, coexisting medical conditions, ECG findings, and hyperkalemia management strategies, were abstracted from admission records, progress records, nursing records, discharge summaries, and the records of the emergency department and the intensive care unit (ICU). The type of admission and the timing of the onset of hyperkalemia were reviewed; in patients with hospital-acquired hyperkalemia, the period from admission to diagnosis and the hospital location at diagnosis were also reviewed. The symptoms associated with hyperkalemia and the occurrence of multi-organ failure (MOF) or cardiac arrest (or both) at the time of severe hyperkalemia diagnosis were examined closely.
Chronic kidney disease (CKD) was classified into five groups on the basis of the estimated glomerular filtration rate (eGFR): normal renal function and stage I CKD, eGFR of at least 90 mL/minute per 1.73 m2; stage II, eGFR of 60 to 89 mL/minute per 1.73 m2; stage III, eGFR of 30 to 59 mL/minute per 1.73 m2; stage IV, eGFR of 15 to 29 mL/minute per 1.73 m2; and stage V, eGFR of less than 15 mL/minute per 1.73 m2 or requiring renal replacement therapy (RRT). Patients with a previous diagnosis of severe hyperkalemia were considered to have recurrent severe hyperkalemia. Hypertension was defined as a systolic blood pressure of greater than 140 mm Hg or a diastolic pressure of greater than 90 mm Hg or by the use of antihypertensive drugs. Diabetes mellitus was diagnosed in patients with a random blood glucose concentration of greater than 200 mg/dL, a fasting plasma glucose level of greater than 126 mg/dL on at least two separate occasions, or a glycated hemoglobin of greater than 7.0% or by the use of oral hypoglycemic agents or insulin. Cirrhosis was defined by computed tomography or sonography, and congestive heart failure was defined as a New York Heart Association functional class III or IV. Coronary artery disease was defined by a prior diagnosis of ischemic heart disease and positive ECG findings; pulmonary diseases included tuberculosis, chronic obstructive pulmonary disease, and asthma.
We identified prescriptions for the most common medications that are capable of increasing serum K+ levels: angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, potassium-sparing diuretics, beta blockers, non-steroidal anti-inflammatory drugs, digoxin, and potassium supplements. These medications were considered potentially contributory if administered within 24 to 36 hours of the onset of hyperkalemia; if patients were already on dialysis, these medications were considered noncontributory.
Coexisting medical conditions affecting the occurrence of severe hyperkalemia were categorized into one of three groups: those causing renal impairment, those causing K+ shift across cell membranes, and others. The initial categorization was performed by electronic medical record review and confirmed on the basis of the clinical judgment of the researchers. Acute kidney injury (AKI) was defined by the Acute Kidney Injury Network criteria and consisted of an absolute increase in serum creatinine of at least 0.3 mg/dL, a percentage increase in the serum creatinine of at least 50%, and/or a reduction in urine output, defined as an output of less than 0.5 mL/kg per hour for greater than 6 hours; these changes were required to occur over a rapid time course (< 48 hours) to meet the definition of AKI [20, 21]. The diagnosis of infection required not only at least two signs of systemic inflammatory response syndrome but also clinical evidence of infection. Volume depletion was defined as a clinical situation resulting from decreased effective circulating volume and total extracellular fluid volume or as decreased effective circulating volume with increased total extracellular fluid volume. Bleeding was defined as class II or higher hemorrhage on the basis of Advanced Trauma Life Support guidelines, or grade 2 or higher hemorrhage on the basis of World Health Organization guidelines, with definite clinical signs and symptoms. Metabolic acidosis was defined as an arterial pH of less than 7.35. Poor compliance with K+-lowering agents was determined by review of the attending physician's records, including the patient's medical history, and the reviewer's judgment.
When available, ECGs corresponding to the time of severe hyperkalemia diagnosis or those nearest in time to the diagnosis were reviewed and compared with baseline ECGs. The existence of ECG findings typical of severe hyperkalemia, or an alteration in ECG findings compared with previous results, was considered a 'change in ECG findings'. The findings considered typical of severe hyperkalemia were tall T waves, shortening of QT intervals, prolonged PR intervals, reductions in the amplitude of P waves, 'sine-wave' ventricular rhythms with wide QRS complexes, and the occurrences of ventricular fibrillation and asystole. Decisions on ECG findings were based on formal readings, documented by the attending cardiologist, and adjudicated by reviewers and researchers on the basis of an extensive literature review [14, 22–24]. The period from the diagnosis of hyperkalemia to 'change in ECG findings' was also recorded.
Management of severe hyperkalemia was divided into a conservative management group and an aggressive management group; patients requiring RRT were in the aggressive management group. The choice of RRT modality was made by the attending physician after considering the clinical characteristics of each patient. The criteria for initiation of RRT in AKI included volume overload, oliguria, acidosis, refractory hyperkalemia, and uremic symptoms or documented uremia. We categorized management techniques into 'level of support I' and 'level of support II' categories. 'Level of support I' contained seven initial conservative management strategies, all given a weight of 1: drug cessation, intravenous (IV) calcium gluconate, dextrose fluid with insulin, IV or oral (PO) sodium bicarbonate, calcium polystyrene sulfonate enema, PO calcium polystyrene sulfonate, and IV or PO loop diuretics. The sum of weighted values was defined as the 'level of support I' value for each patient. 'Level of support II' contained nine initial conservative management strategies and RRT treatments: the previously named seven strategies were included with the addition of hemodialysis (weight 1) and continuous renal replacement therapy (CRRT) (weight 2). The sum of weighted values was defined as the 'level of support II' value for each patient.
Clinical parameters that could influence either the development of severe hyperkalemia or in-hospital mortality - serum creatinine, eGFR, total carbon dioxide, and arterial pH level - were documented. All laboratory data were collected from the time that the serum K+ level reached at least 6.5 mEq/L. Serum creatinine levels were measured by using an assay based on the Jaffe method, and eGFR was calculated by using the following abbreviated Modification of Diet in Renal Disease formula: GFR (in mL/minute per 1.73 m2) = 186 × (serum creatinine) 1.154 × (age in years)-0.203 × (0.742 if female).
Clinical outcomes
Clinical outcomes included ICU treatment (including reasons for ICU treatment), cardiopulmonary resuscitation (CPR), improvement in severe hyperkalemia, and in-hospital mortality (including reasons for in-hospital mortality). Improvement in hyperkalemia was defined as a serum K+ level of less than 5.5 mEq/L, independent of deterioration in clinical condition or in-hospital mortality. In-hospital mortality was defined as death during the period of hospitalization, independent of an improvement in severe hyperkalemia.
Statistical analysis
Categorical variables were described as the frequency and proportion of variables and were compared by using the chi-square test. Continuous variables were expressed as the mean ± standard deviation and compared by using a Student t test after normality testing. A simple logistic regression model was used to determine the unadjusted odds ratios (ORs) and 95% confidence intervals (CIs). A correlation analysis was conducted in order to avoid multi-collinearity; only one variable in highly correlated variable sets was selected for multiple logistic regression analysis. Statistically significant covariables from univariate analysis and clinically important covariables were included in the final multiple logistic regression model, conducted in a forward stepwise manner. The end results of multiple logistic regression analysis were demonstrated as a forest plot. A P value of less than 0.05 was considered statistically significant. Statistical analysis was performed with the Statistical Package for the Social Sciences, version 18.0K (SPSS, Inc., Chicago, IL, USA).