Accepted Manuscript

Hypokalemia in acute medical patients: risk factors and prognosis

Helene Kildegaard Jensen, BSc.Med. Mikkel Brabrand, MD, PhD Pernille Just
Vinholt, MD Jesper Hallas, MD, Professor, DMSci Annmarie Touborg Lassen, MD,
Professor, DMSci
PII: S0002-9343(14)00660-3
DOI: 10.1016/j.amjmed.2014.07.022
Reference: AJM 12623

To appear in: The American Journal of Medicine

Received Date: 23 February 2014

Revised Date: 10 June 2014
Accepted Date: 18 July 2014

Please cite this article as: Jensen HK, Brabrand M, Vinholt PJ, Hallas J, Lassen AT, Hypokalemia
in acute medical patients: risk factors and prognosis, The American Journal of Medicine (2014), doi:
10.1016/j.amjmed.2014.07.022.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Hypokalemia in acute medical patients: risk factors and

prognosis
Helene Kildegaard Jensen, BSc.Med.a*; Mikkel Brabrand, MD, PhDb; Pernille Just Vinholt, MDc;

Jesper Hallas, MD, Professor, DMScid; Annmarie Touborg Lassen, MD, Professor, DMScia

PT
a: Department of Emergency Medicine, Odense University Hospital, Denmark

RI
b: Department of Medicine, Sydvestjysk Sygehus Esbjerg, Denmark

c: Department of Clinical Biochemistry and Pharmacology, Odense University Hospital, Denmark

SC
d: Clinical Pharmacology, Institute of Public Health, University of Southern Denmark, Denmark

U
*Corresponding author information: AN
Helene Kildegaard Jensen, clinical research assistant, Department of Emergency Medicine, Odense University Hospital, Sdr.
M
Boulevard 29, entrance 130, 1st floor, DK5000 Odense C, Denmark. Helene.Jensen@rsyd.dk; telephone: +45 29 27 01 72;
fax +45 65 41 15 71
D

Funding: The study was funded by an independent grant from The Research Foundation of Odense University Hospital. The
Research Foundation had no role in the conception, design or conduct of the study; management, analysis, or interpretation
TE

of the data; or the preparation, review, or approval of the manuscript.

Conflicts of interest / financial disclosures: HKJ, MB and PJV declare no conflicts of interest. ATL is supported by an
unrestricted grant from the private philanthropic fund TrygFonden given to the University of Southern Denmark. JH has
EP

participated in research projects funded by Novartis, Pfizer, Menarini, MSD, Nycomed, and Astellas with grants paid to the
institution where he was employed. He has personally received fees for teaching or consulting from the Danish Association
of Pharmaceutical Manufacturers and from Nycomed, Pfizer, Novartis, Astra Zeneca, Lundbeck, Menarini, Leo
C

Pharmaceuticals and Ferring.

AC

Author contributions: HKJ conceived and designed the study, analyzed, and interpreted the data, and wrote the report. ATL,
MB, JH and PJV conceived and designed the study and assisted with analysis, interpretation of the data and writing of the
report. All authors have had access to the data and have approved the final version of the manuscript.

Article type: Original clinical research article

Key words: hypokalemia; potassium; electrolytes; mortality; hospitalization; comorbidity; diuretics; medications

Manuscript word count: 2828

1
 ACCEPTED MANUSCRIPT

Abstract

Background: Hypokalemia is one of the most common electrolyte disorders in hospitalized

patients. It is associated with a high mortality rate among patients with cardiovascular disease.

PT
Whether hypokalemia confers a similar risk in an unselected hospitalized population is not well

established.

RI
SC
Methods: We conducted a prospective cohort study involving all first time admissions (n=11988)

to the Acute Medical Department at Odense University Hospital linking potassium level at

U
admission with registry data on patient characteristics, laboratory data, redeemed prescriptions

AN
and time of death for the period from August 2009 to August 2011. We estimated hazard ratios

for all cause mortality within 0-7 days and 8-30 days after admission, comparing patients with
M
hypokalemia at admission (plasma [K+] level < 3.4 mmol/L) with patients with eukalemia at

admission ([K+] level of 3.4-3.8 mmol/l).

D
TE

Results: Hypokalemia occurred in 16.8% of first time admissions (n=2011). It was associated with

an adjusted hazard ratio [HR] of 1.34 (95% confidence interval [CI], 0.98-1.85) for 7-day mortality
EP

and 1.56 (CI, 1.18-3.06) for 8-30 day mortality. Among patients with more severe hypokalemia
C

(plasma [K+]<2.9 mmol/l) the adjusted HR was 2.17 (CI, 1.34-3.49) for 7-day mortality and 1.90
AC

(CI,1.18-3.06) for 8-30 day mortality. Prognostic factors for both 7-day and 8-30 day mortality

among hypokalemic patients were increasing age and Charlson comobidity index whereas there

was no prognostic effect of current diuretic or betagonist use.

2
 ACCEPTED MANUSCRIPT

Conclusions: In a mixed population of hospitalized medical patients, hypokalemia is common and

plasma [K+]<2.9 mmol/l is associated with increased 7-day and 8-30 day mortality.

PT
ABSTRACT WORD COUNT: 249

RI
U SC
AN
M
D
TE
C EP
AC

3
 ACCEPTED MANUSCRIPT

Introduction

Hypokalemia is one the most common electrolyte disorders encountered in hospitalized patients

with a reported frequency of approximately 20%(1). Most frequently it is asymptomatic and

PT
identified only on routine screening. However, severe hypokalemia is associated with potential

life threatening complications such as cardiac dysrhythmias, paralysis, rhabdomyolysis and

RI
diaphragmatic weakness (2-4). Where many studies have addressed the causes and symptoms

SC
associated with hypokalemia, risk factors and prognosis have almost exclusively been investigated

in patients with cardiovascular disease or chronic kidney disease (5-9). In these studies, even mild

U
to moderate hypokalemia was found to increase morbidity and mortality up to 50%. However,
AN
whether hypokalemia confers a similar risk in other patient populations is not well established.

The aim of the present study was (i) to characterize the population of hypokalemic patients
M
in a large unselected group of medical patients who were acutely admitted at an urban university
D

hospital; (ii) to determine the relationship between hypokalemia and mortality within 0-7 and 8-30
TE

days of admission as well as prognostic factors for mortality among hypokalemic patients, and (iii)

to determine whether patients without a pharmacological cause of hypokalemia had a particularly

EP

poor prognosis.
C

Methods
AC

Study design and setting

The present study is an observational cohort study of all consecutively admitted medical patients

(age >15 years) to the Acute Medical Department at Odense University Hospital (OUH), Denmark,

from 1 August 2009 to 31 August 2011. The hospital is a 1300-bed level 1 trauma centre and

4
 ACCEPTED MANUSCRIPT

university teaching hospital with a contingency population of 290,000. The Acute Medical

Department has approximately 9,000 admissions annually and serves as a medical admission unit

for the following medical specialities: general internal medicine, infectious diseases, geriatric

PT
medicine, rheumatology, endocrinology and respiratory medicine. The Acute Medical Department

receives acute medical patients referred from either primary care, out-patient clinics or from the

RI
Emergency Department. Patients transferred directly from the Emergency Department to the ICU

were not included in the cohort. Only the patient’s first admission to the Acute Medical

SC
Department within the inclusion period was considered for analysis.

U
Data sources

AN
Data were retrieved from six sources: a research database of all acute admissions at OUH, the

hospital’s laboratory database, the Danish National Patient Register, the Odense University
M
Pharmacoepidemiological Database, the Danish National Alcohol and Drug Treatment Register and
D

the Civil Registration System in Denmark. From the laboratory database, we extracted plasma
TE

potassium values, plasma creatinine, arterial pH and bicarbonate values, if measured. Information

regarding previous discharge diagnoses was retrieved from the Danish National Patient Registry
EP

(10, 11). Data on the use of thiazid and loop diuretics, inhaled betaagonists and disulfiram was

obtained from the Odense University Pharmacoepidemiological Database, which systematically

C

records person-identifiable data on all reimbursed drugs (12-14). Finally data regarding death or
AC

migration within 30 days of admission was obtained from the Civil Registration System in Denmark

(15). All linkages were performed by use of a mutual person identifier; the unique and permanent

10-digit CPR-number assigned to all Danish residents.

5
 ACCEPTED MANUSCRIPT

Definition of hypokalemia

As the plasma [K+]-mortality association is likely to be U-shaped, we chose to define eukalemia as

the range with the lowest observed mortality (8, 16). Using a restricted cubic spline with four

PT
knots (17), we ascertained the crude relationship between admission plasma [K+], defined as the

first obtained plasma [K+] value within 24 hours of admission, and 7-day mortality. A U-shaped

RI
relationship was observed with a plasma [K+] value of 3.7 mmol/l associated with the lowest risk of

SC
mortality (Figure 1). Using this value as reference, we evaluated the crude odds ratio for 7-day

mortality for each 0.1 mmol/l change in plasma [K+] and found that 7-day mortality was

U
significantly associated with a plasma [K+] level of less than 3.4 mmol/l and greater than 3.8

AN
mmol/l. Consequently, the reference range for eukalemia was defined as 3.4-3.8 mmol/l in this

study, and hypokalemia was defined as a plasma [K+] level below 3.4 mmol/l.
M
Covariates and outcomes of interest
D

The following covariates were considered: age, sex, Charlson Comorbidity Index (0, 1-2 and
TE

>2) (18-20), history of alcohol abuse, use of drugs known to cause hypokalemia (diuretics or

betaagonists) and plasma creatinine. Patients were classified as having an alcohol abuse if they
EP

had ≥ 2 admissions with a discharge diagnosis of an acute alcohol episode, at least one admission
C

with a discharge diagnosis of a chronic alcohol related diagnosis, a redeemed prescription of

AC

disulfiram from 2007 and onwards, or a registration in The National Register of Alcohol Abuse

Treatment (21). We defined current use of loop diuretics, thiazid diuretics and betaagonists by the

redeeming of a prescription within 90 days before admission. Plasma creatinine was categorized

into three groups (normal, below and above reference range) according to the hospital’s reference

values (45-84 µmol/l for adult women, 59-104 µmol/l for adult men).

6
 ACCEPTED MANUSCRIPT

The primary outcome was all-cause mortality within 7 days of admission. In order to

ascertain the effect of hypokalemia on mortality beyond the acute setting, we also evaluated

mortality within 8 to 30 days of admission as a secondary outcome.

PT
Statistical methods

RI
Continuous variables are described as medians with interquartile ranges. Categorical variables are

expressed as proportions. Comparisons of baseline characteristics were performed using Wilcoxon

SC
rank sum test and Pearson’s chi-square test, as appropriate. In addition, age, gender,

comorbidities implemented in the Charlson Comorbidity Index, history of alcohol abuse and use of

U
diuretics and betaagonists prior to admission were evaluated as potential risk factors for
AN
hospitalization with hypokalemia in a multivariable logistic regression analysis.

The independent effect of hypokalemia on mortality was evaluated in a multivariable Cox

M
regression analysis controlling for age, sex, Charlson comorbidity index and admission plasma
D

creatinine. The scale of age was assessed using fractional polynomials, and we found no evidence
TE

against linearity in the log hazards. The remaining variables were included as categorical variables.

Clinically plausible interaction terms were tested and none were included. Analyses were
EP

conducted using plasma potassium as a dichotomous variable; in decrements of 0.5 mmol/l below

3.4 mmol/l and secondarily as a continuous variable below 3.4 mmol/l.

C

All statistical tests were two-sided, with P-values below 0.05 considered statistically
AC

significant. All analyses and plots were generated using Stata software (version 13.0, Stata Corp

LP, Texas, USA).

7
 ACCEPTED MANUSCRIPT

The study was approved by the Danish Data Protection Agency (No. 2008-58-0035) and the

Danish National Board of Health (No. 3-3013-35). According to Danish legislation, informed

consent or review by an ethics board is not required for register-based studies.

PT
Results

RI
During the inclusion period, there were 11998 first time admissions. Reasons for exclusion

SC
included age<15 years (n=2), residency outside of Denmark at the time of admission (n=8),

absence of a plasma [K+] value within the first 24 hours of admission (n=472) and an admission

U
plasma [K+] level of greater than 3.8 mmol/l (n=4973). Thus, 6543 first time admissions were

included in the final analysis.

AN
M
Baseline characteristics

Hypokalemia was observed in 16.8% (n=2011) of all first time admissions; 13.5% of patients
D

(n=1615) presented with a plasma [K+] level of 2.9-3.3 mmol/l and 3.3% (n=396) had a plasma [K+]
TE

level<2.9 mmol/l. Table 1 summarizes the characteristics of hypokalemic and eukalemic patients.

Hypokalemic patients were older and had a significantly higher comorbidity score. Furthermore,
EP

hypokalemic patients were more likely to be female, have a history of alcohol abuse, have
C

redeemed a prescription on loop or thiazid diuretics within 90 days prior to admission and they
AC

were slightly more alkalotic at admission. When evaluated in a multivariable logistic regression

model; age, female sex, history of alcohol abuse, known liver disease or malignant disease as well

as redeeming a prescription on diuretics within 90 days prior to admission were all independent

risk factors for hospitalization with hypokalemia (Table 1).

8
 ACCEPTED MANUSCRIPT

Hypokalemia and 7-day mortality

Death within 7 days of admission occurred in 63 (3.1%) and 96 (2.1%) of patients with

hypokalemia and eukalemia respectively. Hypokalemia was associated with an adjusted hazard

PT
ratio [HR] of 1.34 (95% confidence interval [CI], 0.98-1.85) for 7-day mortality. When analyzed by

strata of plasma [K+] values, a plasma [K+] level of 2.9-3.3 mmol/l was not associated with an

RI
increased mortality rate (adjusted HR, 1.13; 95% CI, 0.79-1.62), whereas more severe hypokalemia

SC
(plasma [K+]<2.9 mmol/l) was independently associated with an increased risk of mortality

(adjusted HR, 2.17; 95% CI 1.34-3.49). When evaluated as a continuous variable, the adjusted

U
mortality was increased by 11% (adjusted HR 1.11; 95% CI 1.04-1.18) with each 0.1 mmol/l decline

in plasma [K+] below 3.4 mmol/l. AN

Hypokalemia and 8-30 day mortality
M

Among the 6384 patients that survived the first 7 days after admission, there were 84 and 117
D

deaths within the hypokalemic and eukalemic group respectively during the following 23 days.
TE

Compared with patients admitted with eukalemia, patients admitted with hypokalemia had an

increased 8-30 day mortality (4.3% vs. 2.6%), and hypokalemia was independently associated with
EP

a higher mortality within 8-30 days (adjusted HR, 1.56, 95% CI 1.18-2.07) (Table 2). Even a plasma

[K+] level of 2.9-3.3 mmol/l was associated with higher 8-30 day mortality (adjusted HR, 1.48; 95%
C
AC

CI 1.09-2.01), and the risk of death increased as hypokalemia worsened. Each 0.1 mmol/l decline

in plasma [K+] below 3.4 mmol/l was associated with a 4% increased risk of death (adjusted HR,

1.04; 95% CI 0.97-1.12).

9
 ACCEPTED MANUSCRIPT

Prognostic factors

Increasing age and Charlson comorbidity index were associated with an increased risk of mortality

within the first 30 days after admission (Table 3). Gender had no significant association with

PT
mortality; however, there was a tendency towards a prognostic value of male sex when assessed

in day 8-30 after admission (Table 3). When evaluating the prognostic value of whether patients

RI
had redeemed prescriptions on medications with potential of inducing hypokalemia prior to

SC
admission or had no such pharmacological disposition to developing hypokalemia, we found that

there was no prognostic value of a medical history with current use of diuretics or betaagonists in

U
the first 30 days after admission (Table 3).

Subgroup analyses
AN
M
The impact of hypokalemia on mortality was explored in a range of demographic and clinical

subgroups (Figure 2). The adjusted risk of 7- and 8-30 day mortality associated with hypokalemia
D

was most prominent in first time admissions involving patients younger than 80 years, whereas
TE

hypokalemia had no independent association with mortality among patients older than 80 years.

Furthermore, the association between hypokalemia and mortality was present primarily among
EP

women, as well as among patients with a high comorbidity index and patients with no current use
C

of diuretics or betaagonists (Figure 2).

AC

Sensitivity analysis

We performed sensitivity analyses using the conventional reference range for plasma potassium

of 3.5-4.4 mmol/l (22). As in the primary analyses, only plasma [K+] levels below 2.9 mmol/l was

associated with an increased 7-day mortality (adjusted HR, 1.98; 95% CI 1.25-3.12), whereas all

10
 ACCEPTED MANUSCRIPT

degrees of hypokalemia were associated with 8-30 day mortality. Effect estimates were similar to

those obtained in the primary analyses, although inclusion of patients with increased mortality

due to high [K+] values generally lessened the HR estimates by 5-10% (eTable 2).

PT
Discussion

RI
In this cohort study of acutely admitted medical patients, we found that hypokalemia was present

SC
in 16.8% of patients; with plasma [K+] levels below 2.9 mmol/l occurring in 3.3% of patients.

Compared with previous studies who have reported in-hospital mortality rates of 24.9% (1) and

U
20.4% (23) among hypokalemic patients (serum [K+] level <3.0 mmol/l) in a general hospitalized

AN
population, our mortality rates were notably lower (Table 2). Presumably, this is attributable to

the fact that these studies did not discriminate between community-acquired and hospital-
M
acquired hypokalemia, and included patients from a variety of non-medical departments and
D

intensive care units. Thus, in a mixed population of acutely hospitalized medical patients, the

mortality rate associated with community-acquired hypokalemia is lower than previously

TE

reported.
EP

In our material, we found that only plasma [K+] levels below 2.9 mmol/l was associated

with an increased risk of 7-day mortality. Unfortunately, the number of fatal events was too small
C

to allow further stratification. Based on our material, it is therefore impossible, with adequately
AC

statistical strength, to ascertain whether the increased mortality is due primarily to the very

severe cases of hypokalemia, or whether more moderate hypokalemia is also associated with an

increased risk of death.

11
 ACCEPTED MANUSCRIPT

Interestingly, we found that not only was more severe hypokalemia associated with an

increased 7-day mortality, but when evaluated on day 8-30, all degrees of hypokalemia were

associated with increased risk of death. Whether this association is causal, or hypokalemia is

PT
simply a marker of the severity of the underlying condition(s) that led to its development or a

general marker of frailty, is however a key question. Evidence suggests that hypokalemia may

RI
enhance membrane excitability, increase cardiac automaticity and predispose patients to

reentrant arrhythmias (24, 25). Furthermore, data from animal and human models suggest that

SC
low potassium may cause diastolic dysfunction, endothelial dysfunction and increased rates of

U
thrombosis and platelet aggregation (26-29). Therefore it seems biologically plausible that

AN
hypokalemia is a direct contributor to adverse outcomes occurring both in the acute setting and

after discharge. On the other hand, some authors have suggested that hypokalemia in acute
M
conditions might be seen as an epiphenomenon, conferred by elevated sympathetic activity (30-

32). It is thus conceivable that hypokalemia is a marker of severe stress rather than a cause of
D

increased mortality in itself. Regardlessly, admission with a plasma [K+]<2.9 mmol/l seems an
TE

interesting prognostic marker for both short-term and longer-term mortality among acutely

hospitalized medical patients, although the association needs to be confirmed in a larger

EP

multicenter study.
C

Subgroup analyses showed that the association between hypokalemia and mortality was
AC

present primarily in women, patients younger than 80 years and patients not receiving any

diuretics prior to admission. That hypokalemia is more common in women has been reported by

other researchers, but our study is the first to report that the impact of hypokalemia on mortality

is present primarily in women (33-35). The etiology is unclear, and no plausible hypotheses exist in

the current literature. An accentuated impact of electrolyte abnormalities among patients

12
 ACCEPTED MANUSCRIPT

younger than 65 years has also been reported in a study investigating the association between

hospital-associated hyponatremia and mortality (16). That younger patients might be more prone

to adverse outcomes when experiencing electrolyte abnormalities requires further investigation,

PT
but differences in the pace of development of the electrolyte abnormality might contribute to the

difference in risk. Where the elderly are more likely to be in treatment with an inciting drug and to

RI
develop chronic hypokalemia, hypokalemia in younger patients might more often be an indicator

of an acute development and of an inability to maintain normal homeostasis. Therefore, our

SC
hypothesis prior to analyses was that hypokalemic patients without an obvious pharmacological

U
disposition to develop hypokalemia might have a worse prognosis than patients in treatment with

AN
diuretics or betaagonists. However, we found no evidence supporting that the redeeming of a

prescription on these drugs prior to admission could be used as a prognostic factor for mortality.
M
Our study has several strengths. To our knowledge, this is the first study to report
D

mortality rate ratios associated with hypokalemia in a large diverse medical cohort while adjusting

for potential demographic and clinical confounders. We collected data prospectively with plasma
TE

potassium values being available in the majority of patients. Due to the completeness of the
EP

Danish registries (10, 13), we had excellent data on comorbidity, reimbursed prescription drugs

and complete follow-up data. Our study also has limitations. Because of its observational design,
C

there is a possibility of residual confounding. For instance we did not have information neither on
AC

the condition that led to admission nor its severity, and we had no data on magnesium levels and

do not know to what extent the observed effect was caused by magnesium depletion. As this

study was conducted at a single hospital, the generalizability of the findings may be impaired, and

the findings cannot be extended to an outpatient, intensive care or surgical setting. Furthermore,

potassium was measured in heparinized plasma and not in serum as used in many countries. Due

13
 ACCEPTED MANUSCRIPT

to potassium release from platelets during the clotting process, serum potassium values can be

0.1-0.5 mmol/l greater than in plasma, resulting in higher upper reference limits in serum

(reference range typically 3.5-5.0 mmol/l) as compared to plasma (reference range 3.5-4.4

PT
mmol/l) (22, 36, 37). The differences between serum and plasma potassium are, however, of

negligible clinical importance in lower potassium ranges and study results thus transferrable to

RI
settings using serum values(38). Finally, it was beyond the focus of the present study to account

for changes in potassium over time, which may modify the association between potassium level

SC
and mortality. Further studies should investigate potential hospital aggravation of hypokalemia,

U
hospital-acquired hypokalemia and the risks associated herewith.

Conclusion
AN
M
We found that plasma potassium values <2.9 mmol/l at admission to general medical wards were
D

associated with a 2-fold increase in 7-day mortality and a 90% increase in 8-30 day mortality.

Whether the association between hypokalemia and mortality is causal or associative, hypokalemia
TE

seems an interesting prognostic marker of fatal outcome. However, whether the increased
EP

mortality associated with hypokalemia can be remedied, e.g., by potassium supplement, requires

further investigation.
C
AC

14
 ACCEPTED MANUSCRIPT

References:

1. Paice BJ, Paterson KR, Onyanga-Omara F, Donnelly T, Gray JM, Lawson DH. Record linkage study

of hypokalaemia in hospitalized patients. Postgrad Med J. 1986;62(725):187-91.

PT
2. Gennari FJ. Disorders of potassium homeostasis. Hypokalemia and hyperkalemia. Crit Care Clin.

2002;18(2):273-88, vi.

RI
3. Rastegar A, Soleimani M. Hypokalaemia and hyperkalaemia. Postgrad Med J. 2001;77(914):759-

SC
64.

4. Cohn JN, Kowey PR, Whelton PK, Prisant LM. New guidelines for potassium replacement in

U
clinical practice: a contemporary review by the National Council on Potassium in Clinical Practice.

AN
Archives of internal medicine. 2000;160(16):2429-36.

5. Alper AB, Campbell RC, Anker SD, Bakris G, Wahle C, Love TE, et al. A propensity-matched study
M
of low serum potassium and mortality in older adults with chronic heart failure. Int J Cardiol.

2009;137(1):1-8.
D

6. Ahmed A, Zannad F, Love TE, Tallaj J, Gheorghiade M, Ekundayo OJ, et al. A propensity-matched
TE

study of the association of low serum potassium levels and mortality in chronic heart failure. Eur

Heart J. 2007;28(11):1334-43.
EP

7. Bowling CB, Pitt B, Ahmed MI, Aban IB, Sanders PW, Mujib M, et al. Hypokalemia and outcomes
C

in patients with chronic heart failure and chronic kidney disease: findings from propensity-
AC

matched studies. Circ Heart Fail. 2010;3(2):253-60.

8. Goyal A, Spertus JA, Gosch K, Venkitachalam L, Jones PG, Van den Berghe G, et al. Serum

potassium levels and mortality in acute myocardial infarction. Jama. 2012;307(2):157-64.

9. Korgaonkar S, Tilea A, Gillespie BW, Kiser M, Eisele G, Finkelstein F, et al. Serum potassium and

outcomes in CKD: insights from the RRI-CKD cohort study. Clin J Am Soc Nephrol. 2010;5(5):762-9.

15
 ACCEPTED MANUSCRIPT

10. Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health.

2011;39(7 Suppl):30-3.

11. Thygesen SK, Christiansen CF, Christensen S, Lash TL, Sorensen HT. The predictive value of ICD-

PT
10 diagnostic coding used to assess Charlson comorbidity index conditions in the population-based

Danish National Registry of Patients. BMC Med Res Methodol. 2011;11:83.

RI
12. Clausen T. Hormonal and pharmacological modification of plasma potassium homeostasis.

Fundamental & clinical pharmacology. 2010;24(5):595-605.

SC
13. Gaist D, Sorensen HT, Hallas J. The Danish prescription registries. Dan Med Bull.

U
1997;44(4):445-8.

AN
14. Salpeter SR, Ormiston TM, Salpeter EE. Cardiovascular effects of beta-agonists in patients with

asthma and COPD: a meta-analysis. Chest. 2004;125(6):2309-21.

M
15. Pedersen CB. The Danish Civil Registration System. Scand J Public Health. 2011;39(7 Suppl):22-

5.
D

16. Wald R, Jaber BL, Price LL, Upadhyay A, Madias NE. Impact of hospital-associated
TE

hyponatremia on selected outcomes. Arch Intern Med. 2010;170(3):294-302.

17. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med. 1989;8(5):551-
EP

61.
C

18. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic
AC

comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-83.

19. de Groot V, Beckerman H, Lankhorst GJ, Bouter LM. How to measure comorbidity. a critical

review of available methods. J Clin Epidemiol. 2003;56(3):221-9.

20. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM

administrative databases. J Clin Epidemiol. 1992;45(6):613-9.

16
 ACCEPTED MANUSCRIPT

21. Henriksen DP, Nielsen SL, Laursen CB, Hallas J, Pedersen C, Lassen AT. How well do discharge

diagnoses identify hospitalised patients with community-acquired infections? - a validation study.

(1932-6203 (Electronic)).

PT
22. Rustad P, Felding P Fau - Franzson L, Franzson L Fau - Kairisto V, Kairisto V Fau - Lahti A, Lahti A

Fau - Martensson A, Martensson A Fau - Hyltoft Petersen P, et al. The Nordic Reference Interval

RI
Project 2000: recommended reference intervals for 25 common biochemical properties. (0036-

5513 (Print)).

SC
23. Paltiel O, Salakhov E, Ronen I, Berg D, Israeli A. Management of severe hypokalemia in

U
hospitalized patients: a study of quality of care based on computerized databases. Arch Intern

Med. 2001;161(8):1089-95. AN
24. Coca SG, Perazella MA, Buller GK. The cardiovascular implications of hypokalemia. Am J Kidney
M
Dis. 2005;45(2):233-47.

25. Schulman M, Narins RG. Hypokalemia and cardiovascular disease. The American journal of
D

cardiology. 1990;65(10):4E-9E; discussion 22E-3E.

TE

26. Lin H, Young DB. Interaction between plasma potassium and epinephrine in coronary

thrombosis in dogs. Circulation. 1994;89(1):331-8.

EP

27. McCabe RD, Bakarich MA, Srivastava K, Young DB. Potassium inhibits free radical formation.
C

Hypertension. 1994;24(1):77-82.
AC

28. McCabe RD, Young DB. Potassium inhibits cultured vascular smooth muscle cell proliferation.

American journal of hypertension. 1994;7(4 Pt 1):346-50.

29. Srivastava TN, Young DB. Impairment of cardiac function by moderate potassium depletion.

Journal of cardiac failure. 1995;1(3):195-200.

17
 ACCEPTED MANUSCRIPT

30. Brown MJ. Hypokalemia from beta 2-receptor stimulation by circulating epinephrine. The

American journal of cardiology. 1985;56(6):3D-9D.

31. Manhem P, Nilsson LH, Moberg AL, Wadstein J, Hokfelt B. Hypokalaemia in alcohol withdrawal

PT
caused by high circulating adrenaline levels. Lancet. 1984;1(8378):679.

32. Struthers AD, Reid JL. Adrenaline causes hypokalaemia in man by beta 2 adrenoceptor

RI
stimulation. Clinical endocrinology. 1984;20(4):409-14.

33. Hawkins RC. Gender and age as risk factors for hypokalemia and hyperkalemia in a multiethnic

SC
Asian population. Clinica chimica acta; international journal of clinical chemistry. 2003;331(1-

U
2):171-2.

AN
34. Kleinfeld M, Borra S, Gavani S, Corcoran A. Hypokalemia: are elderly females more vulnerable?

Journal of the National Medical Association. 1993;85(11):861-4.

M
35. Liamis G, Rodenburg EM, Hofman A, Zietse R, Stricker BH, Hoorn EJ. Electrolyte disorders in

community subjects: prevalence and risk factors. The American journal of medicine.
D

2013;126(3):256-63.
TE

36. Nijsten M, Dofferhoff AS. Pseudohyperkalemia and Platelet Counts. New England Journal of

Medicine. 1991;325(15):1107-.
EP

37. Nguyen MK, Kurtz I. An unusual case of pseudohyperkalaemia. Nephrology Dialysis

C

Transplantation. 2003;18(8):1657-9.
AC

38. Drogies T, Ittermann T, Lüdemann J, Klinke D, Kohlmann T, Lubenow N, et al. Potassium –

reference intervals for lithium-heparin plasma and serum from a population-based cohort / Kalium

– Referenzbereiche für Lithium-Heparin-Plasma und Serum aus einer bevölkerungsbezogenen

Studie. LaboratoriumsMedizin2010. p. 39.

18
 ACCEPTED MANUSCRIPT

Figures

Figure 1: Association between admission plasma potassium level and 7-day mortality

Legend: Restricted cubic spline model of the unadjusted relationship between admission

PT
plasma potassium concentrations and 7-day mortality. Dashed lines represent the 95%

confidence interval.

RI
SC
Figure 2: Subgroup analysis of the association between hypokalemia and mortality

Legend: Plot of the adjusted relationship between hypokalemia (admission plasma [K+]<3.4

U
mmol/l) and (A) 7-day mortality and (B) 8-30 day mortality in selected demographic and

AN
clinical subgroups. Adjusted effect estimates are reported only if at least 30 deaths

occurred in the subgroup of interest. Abbreviations: HR, hazard ratio; CI confidence

M
interval.
D
TE
C EP
AC

19
 ACCEPTED MANUSCRIPT

Table 1

Table 1. Baseline characteristics of Hospitalizations With and Without Hypokalemia

Baseline characteristics Reference group Hypokalemia Crude OR Adjusted ORa

(95% CI) (95% CI)
No. of patients 4532 2011

PT
Age, years, median [IQR] 60.6 [39.6;77.3] 65.1 [46.8;78.5] 65.1 [46.8;78.5] 65.1[46.8;78.5]
Age 15-64 years – no. (%) 2547 (56.2) 1002 (49.8) 1 (ref) 1 (ref)
Age 65-79 years – no. (%) 1072 (23.7) 568 (28.2) 1.3 (1.2;1.5) 1.3 (1.1;1.5)

RI
Age ≥ 80 years – no. (%) 913 (20.2) 441 (21.9) 1.2 (1.1;1.4) 1.1 (0.9:1.3)
Sex, male – no. (%) 1972 (43.5) 662 (32.9) 0.6 (0.6;0.7) 0.6 (0.6;0.7)
Charlson Index score – no(%)

SC
b
Comorbidity index lox, (0) 2431 (53.6) 982 (48.8) 1 (ref)
b
Comorbidity index, medium (1-2) 959 (21.2) 446 (22.2) 1.2 (1.01;1.3)
b
Comorbidity index, high (>2) 1142 (25.2) 583 (29.0) 1.3 (1.1;1.4)
History of – no.(%)

U
Myocardial infarction 202 (4.5) 75 (3.7) 0.8 (0.6;1.1)c 0.8 (0.6;1.03)c
Heart failure 279 (6.2) 125 (6.2) 1.0 (0.8;1.3)c 0.8 (0.6;1.04)c
Cerebrovascular disease
Peripheral vascular disease
Chronic pulmonary disease
AN
546 (12.1)
227 (5.0)
693 (15.3)
268 (13.3)
117 (5.8)
299 (14.9)
1.1 (0.96;1.3)
1.2 (0.9;1.5)
1.0 (0.8;1.1)c
c

c
1.0 (0.9;1.2)c
1.1 (0.9;1.4)c
0.9 (0.7;1.1)c
c
Chronic kidney disease 83 (1.8) 42 (2.1) 1.1 (0.8;1.7) 1.0 (0.7;1.5)c
M
c
Diabetes mellitus 386 (8.5) 189 (9.4) 1.1 (0.9;1.3) 1.0 (0.8;1.2)c
c
Any type of malignancy 399 (8.8) 228 (11.3) 1.3 (1.1;1.6) 1.3 (1.05;1.5)c
Liver disease, non-alcohol related 42 (0.9) 27 (1.3) 1.4 (0.9;2.4)c 1.7 (1.01;2.7)c
D

c
Dementia 184 (4.1) 72 (3.6) 0.9 (0.7;1.2) 0.8 (0.6;1.1)c
c
Alcohol abuse 450 (9.9) 247 (12.3) 1.3 (1.1;1.5) 1.5 (1.3;1.8)c
Use of inciting drug in past 3 months – no(%)
TE

Loop or thiazid diuretics 834 (18.4) 565 (28.1) 1.7 (1.5;2.0)c 1.8 (1.5;2.0)c
Inhaled β-agonists 641 (14.1) 287 (14.3) 1.0 (0.9;1.2) c
0.9 (0.8;1.1)c
In-hospital measures P-value
EP

Admission laboratory values, median [IQR]

Creatinine µmol/l 71 [59;88] 69 [57;88] 0.26
pH 7.44 [7.41;7.48] 7.47 [7.43;7.51] <0.001
Bicarbonate mmol/l 24.0 [22;26.2] 24.9 [21.8;27.5] <0.001
C

Length of hospital stay, days, median [IQR] 3 [2;6] 4 [2;8] <0.001

Transfer to the ICU within 2 days of admission 107 (2.4) 72 (3.6) 0.005
AC

Abbreviations: OR, odds ratio; CI, confidence interval; IQR, interquantile range.
-
Conversion factors: To convert [creatinine] to mg/dL divide by 76.26. To convert [HCO3 ] to mEq/L divide by
1.0.
a
Multivariable logistic regression model adjusted for age, sex, individual comorbidities implemented in the
Charlson Comorbidity Index, history of alcohol abuse, current use of diuretics and current use of
betaagonists.
 ACCEPTED MANUSCRIPT

b
The Charlson comorbidity index was not included in the multivariate analysis due to high correlation with
the individual comorbid diseases.
c
The estimated odds ratio associated with the particular disease or drug compared to no occurrence of the
particular disease or drug.

PT
RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

Table 2
+
Table 2. Unadjusted and Adjusted Relationship Between Admission plasma [K ]
and Mortality
Admission [K+], N Dead Mortality Crude HR Adjusted HRa
mmol/L (95% CI) (95% CI) (95% CI)
0-7 day mortality

PT
3.4-3.8 4520 96 2.1 (1.7-2.6) 1.0 (ref.) 1.0 (ref)
< 3.4 2008 63 3.1 (2.5-4.0) 1.48 (1.08-2.04) 1.34 (0.98-1.85)
2.9-3.3 1612 42 2.6 (1.9-3.5) 1.2 (0.9-1.8) 1.13 (0.79-1.62)
<2.9 396 21 5.3 (3.5-8.0) 2.5 (1.6-4.1) 2.17 (1.34-3.49)

RI
8-30-day mortality
3.4-3.8 4424 117 2.6 (2.2-3.2) 1.0 (ref) 1.0 (ref)

SC
< 3.4 1945 84 4.3 (3.5-5.3) 1.65 (1.25-2.18) 1.56 (1.18-2.07)
2.9-3.3 1570 64 4.1 (3.2-5.2) 1.56 (1.15-2.11) 1.48 (1.09-2.01)
<2.9 375 20 5.3 (3.5-8.1) 2.05 (1.28-3.30) 1.90 (1.18-3.06)

U
Abbreviations: [K+], potassium concentration; CI, confidence interval, HR, hazard ratio
+ AN
Conversion factors: To convert [K ] to mEq/l divide by 1.0.
a
Cox proportional hazard model adjusted for age, sex, Charlson Comorbidity Index and admission plasma
M
creatinine level. Patients who died during the first 7 days after admission, were excluded from the analyses

of 8-30 day mortality.

D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

Table 3

Table 3. Prognostic factors for mortality among patients admitted with

hypokalemia (plasma [K+]<3.4 mmol/l)
0-7 day 8-30 day

PT
N Dead Crude HR Adjusted HRa Dead Crude HR Adjusted HRa
(95% CI) (95% CI) (95% CI) (95% CI)
Current use of
diuretics or
betaagonists

RI
No 1281 34 1.0 (ref) 1.0 (ref) 48 1.0 (ref) 1.0 (ref)
Yes 727 29 1.5 (0.9-2.5) 0.9 (0.5-1.5) 36 1.3 (0.9-2.1) 0.8 (0.5-1.2)
Sex

SC
Female 1348 44 1.0 (ref) 1.0 (ref) 49 1.0 (ref) 1.0 (ref)
Male 660 19 0.9 (0.5-1.5) 0.8 (0.5-1.4) 35 1.5 (0.9-2.3) 1.6 (1.0-2.5)
Age
2008 63 1.5 (1.2-1.7)b 1.3 (1.1-1.6)b 84 1.6 (1.4-1.8)b 1.6 (1.4-1.9)b

U
Charlson index
0 980 17 1.0 (ref) 1.0 (ref) 23 1.0 (ref) 1.0 (ref)
1-2
>2
446
582
9
37
1.2 (0-5-2.6)
3.7 (2.1-6.6)
AN
0.9 (0.4-2.1)
2.3 (1.3-4.3)
15
46
1.4 (0.8-2.8)
3.6 (2.2-6.0)
1.1 (0.6-2.2)
2.2 (1.3-3.8)
M
Abbreviations: CI, confidence interval, HR, hazard ratio .
a
Cox proportional hazard model adjusted for age, sex, Charlson Comorbidity Index and admission plasma
D

creatinine. Hypokalemic patients who died during the first 7 days after admission, were excluded from the
TE

analyses of 8-30 day mortality.

b
Increase in hazard ratio for each 10-year increase in age.
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

Clinical significance

• Hypokalemia is present in 17% of acute medical patients and is more common among the elderly,

women, patients with malignant disease, alcohol abuse and patients in diuretic treatment.

• In acute medical patients not admitted due to severe heart or kidney disease, hypokalemia <2.9

PT
mmol/l at admission is associated with an increased 7 and 8-30 day mortality: however, the absolute

mortality associated with hypokalemia is substantially lower than previously reported.

RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
Supplementary Appendix

RI
Supplement to: Jensen HK, Brabrand M, Vinholt PJ, Hallas J, Lassen AT. Hypokalemia in acute
medical patients: risk factors and prognosis.

SC
Manuscript number: 14-279

U
eTable 1. Covariates included in baseline description with ICD-10 codes and ATC-codes

eTable 2. Sensitity analyses of primary effect estimates using the conventional reference
range for plasma potassium AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

eTable 1:
Covariates included in baseline description with ICD-10 codes and ATC codes

Covariate Codes
Medical historya

PT
ICD-10 codes
Myocardial infarction I21-22, I25.2
Heart failure I09.9, I11.0, I13.0, I13.2, I25.5, I42.0, I42.5-42.9,
I29.0, I43, I50

RI
Cerebrovascular disease G45-46, I60-69, H34.0
Peripheral vascular disease I70, I71, I73.1, I73.8, I73.9, I77.1, I79.0, I79.2,
K55.1, K55.8, K55.9, Z95.8, Z95.9

SC
Chronic pulmonary disease J40-47, J60-67, J68.4, J70.1, J70.3, I27.8, I27.9
Chronic kidney disease N03.2-N03.7, N05.2-N05.7, N18-19, N25.0, I12.0,
I13.1, Z49.0-Z49.2, Z94.0, Z99.2

U
Diabetes E10.0-E10.9, E11.0-E11.9, E12.0-E12.9, E13.0-
E13.9, E14.0-E14.9
Any type of malignancy
ANC00-09, C10-19, C20-26, C30-34, C37-39, C40,
C41, C43, C45-49, C50-58, C60-69, C70-79, C80-
85, C88, C90-97
Liver disease, non-alcohol related B18, I85.0, I85,9, I86.4, I98.2, K71.1, K71.3-K71.5,
M
K71.7, K72.1, K72.9, K73-74, K76.0, K76.2-K76.9,
Z94.4 AND no registration with alcohol abuse
Dementia F00-03, F05.1, G30, G31.1
D

Alcohol abuse Acute alcohol episode: F10.0-F10.1; T519

Chronic alcohol related diagnoses: F10.2-F10.9,
TE

E24.4, E52.9A, G31.2, G62.1, G72.1, I42.6, K29.2,

K70, K85.2, K86.0, L27.8A, O35.4, P04.3, T50.0A,
Z71.4, Z72.1, Z50.2
Prescription drugsb ATC-codes
EP

Loop diuretics C03C

Thiazid diuretics C03A
Inhaled β-agonists R03AC
Disulfiram N07BB01
C
AC

ICD: international classification of diseases; ATC: anatomical therapeutic chemical classification system.
a
Discharge diagnoses from the National Patient Register from 10 years prior to admission/index date to 7
days prior.
b
Data from the Odense University Pharmacoepidemiological Database.
 ACCEPTED MANUSCRIPT

eTable 2:
Sensitivity analyses of primary effect estimates using the conventional reference
range for plasma potassium

PT
eTable 1. Sensitivity Analysis of the Unadjusted and Adjusted Relationship
between Admission plasma [K+] and Mortality
Admission [K+], N Dead Mortality Crude HR Adjusted HRa

RI
mmol/L (95% CI) (95% CI) (95% CI)
0-7 day mortality
3.5-4.4 7445 194 2.6 (2.3-3.0) 1.0 (ref.) 1.0 (ref)

SC
< 3.4 2008 63 3.1 (2.5-4.0) 1.2 (0.91-1.60) 1.25 (0.93-1.66)
2.9-3.3 1612 42 2.6 (1.9-3.5) 1.00 (0.71-1.40) 1.05 (0.75-1.47)
<2.9 396 21 5.3 (3.5-8.0) 2.05 (1.31-3.22) 1.98 (1.25-3.12)

U
8-30-day mortality
3.5-4.4 7251 246
AN
3.4 (3.0-3.8) 1.0 (ref) 1.0 (ref)
< 3.4 1945 84 4.3 (3.5-5.3) 1.28 (1.00-1.64) 1.40 (1.09-1.80)
M
2.9-3.3 1570 64 4.1 (3.2-5.2) 1.21 (0.92-1.59) 1.33 (1.01-1.76)
<2.9 375 20 5.3 (3.5-8.1) 1.59 (1.01-2.50) 1.70 (1.07-2.70)
D

Abbreviations: [K+], potassium concentration; CI, confidence interval, HR, hazard ratio
TE

Conversion factors: To convert [K+] to mEq/l divide by 1.0.

a
Cox proportional hazard model adjusted for age, sex, Charlson Comorbidity Index and admission plasma
EP

creatinine level. Patients who died during the first 7 days after admission, were excluded from the analyses
of 8-30 day mortality.
C
AC