The Association between Serum Potassium Level and 30-day Mortality for Patients with Cardiac Arrhythmias in the Emergency Department: Secondary Analysis of a Cohort Study

Background: Clinical evidence shows that mortality in relation to the potassium ranges was U-shaped in patients with hypertension, heart failure and cardiovascular disease. However, evidence is lacking for patients with cardiac arrhythmias in the emergency department. Present study aimed to clear the relationship between serum potassium level and mortality for patients with cardiac arrhythmias in the emergency department. Methods: Using data from a retrospective cohort study, we identied 7532 visits involving 3882 patients with cardiac arrhythmias referred for any reason to the emergency department (ED) between January 1, 2010, and December 31, 2015. All-cause mortality within 30-days after the date of visit to the ED. We estimated the risk of mortality using multivariable logistic-regression model. Results: There were 448 deaths (5.9%). The probability of mortality rose rapidly when potassium higher than the turning point (potassium ≥ 4.3 mmol/L) with a adjusted OR of 1.13 (95% CI:1.09, 1.16) for every 0.1 mmol/L increment of potassium. Moreover, the probability of mortality decreased when potassium lower than the turning point (potassium < 4.3 mmol/L) with a adjusted OR of 0.95 (95% CI:0.91, 0.98). Conclusions: Higher serum potassium, even at a low level, was associated with a higher risk of 30-day mortality for patients with cardiac arrhythmias in the emergency department. The probability of mortality rose rapidly when potassium higher than the turning point (may at 4.3 mmol/L).


Background
Heart rhythm disorders are responsible for considerable morbidity and mortality, particularly in developed nations. Atrial brillation (AF) is the leading cause of stroke and a common contributor to overall morbidity and mortality [1] .
Clinical evidence shows that mortality in relation to the potassium ranges was U-shaped in patients with hypertension [2,3] , chronic heart failure (HF) [4,5] , acute HF and cardiovascular disease (CVD) [2] , and even high and low potassium levels within the accepted normal range of potassium were associated with increased mortality suggesting a narrower safe level for these patients.
However, robust evidence is lacking to guide the emergency management of patients with cardiac arrhythmias [6] . Serum potassium levels of > 6.0 or > 6.5 mmol/L [7] or electrocardiogram (ECG) manifestations of hyperkalemia (regardless of serum potassium level) have been recommended as thresholds for initiation of emergency therapy due to the risk of acute life-threatening cardiac rhythm disorders [6] .
Exploring the threshold potassium level where risk of death signi cantly increases is a high priority in patients with cardiac arrhythmias. We used a emergency department (ED) database from a retrospective cohort study to investigate association between serum potassium levels and 30-day mortality. In our observational study, we hypothesized that in patients with cardiac arrhythmias even high and low potassium is associated with a higher risk of all-cause mortality within 30-days after the date of visit to the ED.

Data Source
Data from a retrospective cohort study were used for the present study [8] . All relevant data are available in the Dryad Digital Repository and this website permitted users to freely download the raw data. According to Dryad Terms of Service, we cited Dryad data package in the present study. Complete information regarding data collection is publicly available from the original paper [8] . Brie y, data were extracted from the Diakonhjemmet Hospital's Department of Medical Biochemistry database and the patient administrative system in Oslo, Norway.

Study population
Tazmini, Kiarash et al. [8] completed the entire study. Brie y, the original cohort is an unselected population in an emergency department. All patients ≥ 18 years referred for any reason to the ED between January 1, 2010, and December 31, 2015, who had measured blood electrolytes were included. Who were diagnosed with cardiac arrhythmias (see ICD-10 codes in Supplementary material online, S1) were included. The following exclusion criteria were applied: (i) dehydration, sepsis or cancer; (ii) missing serum potassium; (iii) with an extreme high potassium measurement (> 7 mmol/L). In total, 7532 visits involving 3882 patients with cardiac arrhythmias were included. The study ow chart is presented in Fig. 1.
In the previously published article [8] , Tazmini, Kiarash et al. Has clearly stated that: the study was approved by the institutional review board (The Research Committee, Diakonhjemmet Hospital). The data are anonymous, and the requirement for informed consent was therefore waived.

Variables
Measurement of serum levels of potassium, sodium, calcium, albumin and glucose are performed routinely in all medical patients in the ED (serum levels of calcium, albumin and glucose are not measured in surgical patients). For every visit to the ED, age, sex, serum-electrolyte values and serumalbumin and glucose levels were registered. The time at which the patients have a serum potassium measurement represents the baseline of our study.
Serum-sodium levels were corrected for serum-glucose by lowering the sodium concentration by 2.4 mmol/L for every 5.5 mmol/L increase in glucose to account for the diluting effect of hyperglycemia [9] .

Concomitant diseases, and conditions
We identi ed clinical relevant comorbidities, and conditions used as covariates in the analysis. Hospital discharge diagnoses (primary or secondary) were classi ed by the International Classi cation of Diseases, 10th revision (ICD-10). Patients with hypertension, diabetes, pneumonia, coronary heart disease, heart failure, atrial brillation/ utter, chronic obstructive pulmonary disease (COPD), kidney failure, dehydration, sepsis or cancer were identi ed from the Danish National Patient Registry based on ICD-10 (see ICD-10 codes in Supplementary material online, S1).

Outcomes
The outcome of the study was all-cause mortality within 30 days after the date of visit to the Emergency department. In supplementary analyses, we also analyzed 60-days mortality.

Statistical analysis
Continuous variables are described as means ± SD and categorical data are presented as number and percentage. The difference according to the tertiles of potassium was compared using one-way analysis of variance (ANOVA) for continuous data and Chi-squared tests for categorical variables.
We applied a generalized additive model (GAM) to investigate dose-response relationships between potassium and 30-day mortality (Fig. 1). We applied logistic-regression model to estimate the association between potassium and 30-day mortality. The results were presented as odds ratios (ORs) with their 95% con dence intervals (95% CIs). Crude regression estimates are presented, as well as estimates adjusted for covariates. We selected these confounders on the basis of their associations with the outcomes of interest or a change in effect estimate of more than 10%. [11] Adjusted for the following potential confounders: age (as a continuous variable), sex, glucose-corrected sodium (as a continuous variable), albumin-corrected calcium (as a continuous variable), glucose (as a continuous variable), albumin (as a continuous variable) and comorbid conditions (no, yes): hypertension, diabetes, pneumonia, coronary heart disease, heart failure, atrial brillation/ utte, COPD and kidney failure.
We further applied two-piece-wise linear regression model to examine the threshold effect of potassium on mortality ( Table 2). The turning point of potassium was determined using "exploratory" analyses, which is to move the trial turning point along the pre-de ned interval and pick up the one which gave maximum model likelihood. We also conducted log likelihood ratio test comparing one-line linear regression model with two-piece-wise linear model. As described in previous analyses [12,13] .
To examine the robustness of our results, we conducted strati ed analyses according to covariates.
Dummy variables were used to indicate missing covariate values. The multiple visits per patient were analyzed with generalized estimating equation (GEE) model [14] . The two-sided alpha level was set at 0.05.
All the statistical analysis was performed using the Empower Stats (www.empowerstats.com, X&Y solutions, Inc. Boston MA) and R software version 3.6.1 (http://www.r-project.org).

Baseline characteristics
A total of 7532 visits involving 3882 patients with cardiac arrhythmias were included in the study. The median age of all visits was 76.2 years (IQR 68-87 years). 3788 visits (50.3%) were female. Around 92% had a atrial brillation/ utter. No signi cant statistical difference in coronary heart disease was detected across the tertiles of potassium (see Table 1).

Identi cation of non-linear relationship
We observed a nonlinear dose-response relationship between potassium and mortality ( Fig. 2 and Table 2). The probability of mortality rose rapidly when potassium higher than the turning point (potassium ≥ 4.3 mmol/L) with a adjusted OR of 1.13 (95% CI:1.09, 1.16) for every 0.1mmol/L increment of potassium. Moreover, the probability of mortality decreased when potassium lower than the turning point (potassium < 4.3 mmol/L) with a adjusted OR of 0.95 (95% CI:0.91, 0.98) for every 0.1 mmol/L increment of potassium.
Using the generalized additive model, the non-linear association between potassium and 30-day mortality was detected ( Table 2). The linear regression model and a two-piece-wise linear regression model were compared, and the P value for the log-likelihood ratio test is < 0.001. This result demonstrates that the two-piece-wise linear regression model should be used to t the model.

Additional analyses
We performed ve sensitivity analyses to test the robustness of our results.
First, we analyzed 60-days mortality. These analyses showed that different outcome did not change the overall results. See Supplementary material online, S2 for these analyses.

Discussion
This study showed that higher serum potassium, even at a low level, was associated with a higher risk of 30-day mortality among patients with cardiac arrhythmias in the emergency department. The major nding was that the probability of mortality rose rapidly when potassium higher than the turning point (potassium ≥ 4.3 mmol/L) with a adjusted OR of 1.13 (95% CI:1.09, 1.16) for every 0.1 mmol/L increment of potassium. Moreover, the probability of mortality decreased when potassium lower than the turning point (potassium < 4.3 mmol/L) with a adjusted OR of 0.95 (95% CI:0.91, 0.98). To our knowledge, this is the rst study to report the relation between potassium and 30-day mortality in an unselected adult ED patients with cardiac arrhythmias.
Previous studies have demonstrated that mortality in relation to the potassium ranges was U-shaped. In a Danish National Registries-based study of 44 799 hypertensive patients, potassium levels outside the interval of 4.1-4.7 mmol/L were associated with increased mortality risk [2] . Krogager et al. also demonstrated that potassium levels outside the interval 3.9-4.5 mmol/L were associated with a substantial risk of death in patients requiring diuretic treatment after an myocardial infarction [15] . Moreover, in HF patients, potassium levels of < 4 and ≥ 5 mEq/L have been shown to be associated with poor outcomes when compared with 4-5 mEq/L [5,16] . These ndings are in agreement with our study. We observed that higher serum potassium, even at a low level, was associated with a higher risk of 30-day mortality among patients with cardiac arrhythmias.
Hypokalemia has traditionally been de ned as serum potassium < 3.5 mEq/L, in patients with HF, and in some study de ned hypokalemia as potassium < 4 mEq/L [17] . Several retrospective analyses from the digitalis investigation group trial [17] , [18] all suggest that a potassium level below 4 mmol/L was associated with excess mortality and diuretic induced potassium depletion has been suggested as a potential cause of arrhythmic death in chronic heart failure patients [19] . Similarly, our study demonstrated that the probability of mortality decreased when potassium lower than the turning point (potassium < 4.3 mmol/L) with a adjusted OR of 0.95 (95% CI:0.91, 0.98) for every 0.1mmol/L increment of potassium. In the present study, patients with HF in about 20% of cases. We performed an additional sensitivity analysis (see Supplementary material online, S6) and these analyses both show the same trend as the main analysis.
Hyperkalemia is a potentially serious condition that can result in life-threatening cardiac arrhythmias and is associated with an increased mortality risk [20] . Hyperkalemia is a potentially life-threatening condition that is de ned as a serum potassium level above a reference range, usually greater than 5.0 mEq/L; severe hyperkalemia is often de ned as a level greater than 6.0 mEq/L [21] . In some retrospective analyses, moderate hyperkalemia de ned as serum potassium level ≥ 5.5 mg/dL and < 6.0 mg/dL [22] and severe hyperkalemia de ned as ≥ 6.0 mg/dL [23] . Hyperkalemia is encountered frequently in patients with established CVD who are taking antihypertensive drugs and is associated with increases in all-cause mortality [24] . If not treated rapidly, the mortality rate for patients with severe hyperkalemia can be over 30% [25] . Similarly, we observed that the probability of mortality rose rapidly when potassium higher than the turning point (potassium ≥ 4.3 mmol/L) with a adjusted OR of 1.13 (95% CI:1.09, 1.16) for every 0.1mmol/L increment of potassium. The probability of mortality rose rapidly when potassium higher than the turning point (may at 4.3 mmol/L).
The current study is based on a emergency department data, included both patients with abnormal and normal potassium levels, of high age, with comorbid conditions. Explored the threshold potassium level where risk of death signi cantly increases is a high priority in patients with cardiac arrhythmias.

Study limitations
One limitation of this study is inherent to the observational nature of the study design which lends itself subject to limitations that should be considered including confounding by indication.
In our analysis, we adjusted for likely confounders, including age, sex, sodium, calcium, glucose, albumin and comorbid conditions (hypertension, diabetes, coronary heart disease, heart failure, COPD, pneumonia, kidney failure, dehydration, diabetes and pneumonia). Despite this adjustment, it is still possible that some amount of unmeasured confounding remains. Additional limitations of our study include missing data for some variables. Nonetheless, we used contemporary methods to deal with missing data to minimize bias.
Another limitation relates to the fact that the diagnoses were based on the ICD-10 coding which the responsible physician found relevant, and we did not have information concerning causes of death. Since we are examining mortality over a short period after the date of visit to the ED, we did not nd it bene cial to distinguish between cardiovascular and non-cardiovascular death. Furthermore, we lacked information about interventions during the initial stabilization, which may have in uenced potassium levels and survival. It is noteworthy that the potential resulting from interventions would bias toward to the null and thus result in an underestimation of the association between potassium level and mortality.
Further, we did not have electrocardiogram data available. Consequently, we were not able to examine whether the arrhythmia patient had malignant ventricular arrhythmias such as torsades de pointes (TdP) ventricular tachycardia and other lethal ventricular arrhythmias. Nevertheless, we performed an additional analysis including visits with atrial brillation/ utter. The result is generally similar to our main results (see Supplementary material online, S5).
Moreover, we were unable to obtain the variables, including past medical history, reason for visiting and vital status in the ED, and history of chronic electrolyte imbalances and treatment. Although the routine at Diakonhjemmet Hospital is to take blood samples shortly after the patient has arrived in the ED, we cannot exclude that blood from some patients was collected after initiation of treatment. Finally, we also acknowledge that as our participants were patients referred for any reason to the emergency department, which limits the generalizability of the ndings to other population.

Conclusions
Using data from a retrospective cohort study, we identi ed 7532 visits involving 3882 patients with cardiac arrhythmias referred for any reason to the emergency department. This study identi es a nonlinear dose-response relationship between potassium and 30-day mortality. The probability of mortality rose rapidly when potassium higher than the turning point (may at 4.3 mmol/L). < Abbreviations AF: atrial brillation; ANOVA: one-way analysis of variance; COPD: chronic obstructive pulmonary disease; CVD: cardiovascular disease; ECG: electrocardiogram; ED: emergency department; GEE: generalized estimating equation; HF: heart failure; ICD-10: International Classi cation of Diseases, 10th revision; IQR: interquartile range; K: potassium; ORs: odds ratios; TdP: torsades de pointes; 95% CIs: 95% con dence intervals. Figure 1 Study ow chart. *Subjects meet combined conditions