The European Journal of Heart Failure 5 (2003) 201–209

Factors influencing the length of hospital stay of patients with heart

failure
S.P. Wrighta,*, D. Verouhisb, G. Gamblea, K. Swedbergb, N. Sharpea, R.N. Doughtya
a
Department of Medicine, Faculty of Medical and Health Sciences, University of Auckland, Private Bag 92019, Auckland, New Zealand
b ¨
Department of Medicine, Sahlgrenska University HospitalyOstra, ¨
Goteborg University, Goteborg, Sweden

Received 6 November 2001; received in revised form 9 July 2002; accepted 4 October 2002

Abstract

Introduction: Heart failure (HF) is characterised by frequent hospital admissions and prolonged length of hospital stay.
Admissions for HF have increased over the last decade while length of stay has decreased; the reasons for this change in length
of stay are uncertain. This study investigates the effect of patient-related variables, in-hospital progress and complications on
length of stay. Methods: Patients admitted to Auckland Hospital general medical service and randomised into the Auckland Heart
Failure Management Programme were included in this study. Results: One hundred and ninety-seven patients were included in
this study. Mean age 73 years, mean left ventricular ejection fraction 32%; 52% had one or more previous HF admissions and
75% were New York Heart Association class IV at admission. Median length of hospital stay was 6 days (IQR 4, 9) which is
comparable to the national average from New Zealand admission databases. Longer than average length of stay, defined as )6
days, was associated with the presence of peripheral congestion, duration of treatment with intravenous diuretic, the development
of renal impairment, other acute medical problems at admission, iatrogenic complications during hospital stay, and social problems
requiring intervention. Factors independently associated with length of stay in the top quartile ()10 days) on logistic regression
included the presence of oedema at admission (OR 10.5), change in weight during stay (OR 1.3), duration of treatment with iv
diuretic (OR 7.5), the development of renal impairment (OR 9.8), concurrent respiratory problems requiring specific treatment
(OR 3.8), and social problems requiring intervention (OR 6.8). Conclusions: Peripheral congestion, concomitant acute medical
problems requiring specific treatment, the development of renal impairment and the presence of social problems were related to a
longer than average length of hospital stay. Multivariate models only partly explained variance in hospital stay, suggesting the
importance of pre-admission and post-discharge factors, including the healthcare environment, the availability of primary and
secondary care resources, and the threshold for hospital admission.
䊚 2002 European Society of Cardiology. Published by Elsevier Science B.V. All rights reserved.

Keywords: Heart failure; Length of stay; Healthcare use; Hospitalisation; Inpatient management

1. Introduction
Heart failure (HF) is characterised by frequent hos-
pital admissions w1x and prolonged length of stay w2,3x.
These hospital admissions contribute to the significant
and increasing resource utilisation associated with HF,
costing approximately 1.5% of the annual healthcare
budget of most Western countries w2–5x. While hospital
admissions for HF are increasing w3,6–9x, the average
length of hospital stay has decreased in many countries
*Corresponding author. Tel.: q64-9-307-4949x7654; fax: q64-9-
302-2101.
E-mail address: sp.wright@auckland.ac.nz (S.P. Wright).
over the last decade, including Scotland w3x and the
Netherlands w9x. For example, the average length of stay
in Scotland has decreased from up to 3 weeks in 1985
to approximately 1 week in 1995 w3x. The shortest
hospital stay has been reported from Oregon, USA,
where the average length of stay for HF decreased from
5 days in 1991 to 4 days in 1995 w9x. The local
healthcare environment may have an important role in
determining the threshold for admission and subsequent
length of hospital stay.
Possible determinants of length of hospital stay for
patients with HF include socio-demographic variables
w9,10x, medical comorbidity w11,12x, disease severity

1388-9842/03/$ - see front matter 䊚 2002 European Society of Cardiology. Published by Elsevier Science B.V. All rights reserved.
PII: S 1 3 8 8 - 9 8 4 2 Ž 0 2 . 0 0 2 0 1 - 5
202 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209
w13x, clinical presentation, in-patient treatment, in-hos-
pital progress w14x and the development of iatrogenic
complications w16x. Several studies exhibit the impor-
tance of concurrent medical diagnoses, including studies
from the US w10x and Scotland w11x. The latter study
illustrates the particular importance of concurrent stroke,
renal failure, atrial fibrillation, chronic lung disease and
ischaemic heart disease in prolonging length of hospital
stay for HF w12x.
Other studies of length of hospital stay for HF focus
on inter-institutional variation w14,16x and differences in
patient insurance status w9x using large hospital discharge
databases. Alternatively, some studies have assessed the
effect of specific clinical parameters. Such studies show,
for example, that low left ventricular ejection fraction
(LVEF) w13x, severe renal impairment w17x, or specific
HF aetiology w15x are associated with longer hospital
stay. However, these factors are usually associated with
more severe or end-stage HF, when hospital stay maybe
prolonged for many reasons.
There are little prospective data on determinants of
length of stay in patients with HF, and the reasons for
the decrease in length of stay documented in many
countries are not well understood. Previous studies
performed when length of hospital stay was longer for
HF may no longer be relevant. This study examines the
determinants of length of hospital stay in a cohort of
prospectively-identified HF patients in New Zealand.
2. Methods
Socio-demographic and clinical characteristics, treat-
ment-related factors and in-hospital progress were exam-
ined in relation to length of stay in patients with HF in
a single institution in New Zealand.
2.1. Patient population
Patients admitted to the acute general medical wards
of Auckland Hospital, New Zealand with either a first
diagnosis or an exacerbation of pre-existing HF between
1996 and 1997 were prospectively identified from hos-
pital admission registers. Acute medical care in New
Zealand is provided predominantly by state-funded pub-
lic hospitals. Primary care is provided by general prac-
titioners who are funded by a fee-for-service from
patients with a per-patient government-funded subsidy.
Auckland Hospital has a capacity of 500 beds and serves
a population of approximately 350 000. Measurement of
hospital inpatient volumes and the relative resource
consumption of different diagnostic groups in New
Zealand is performed using the Weighted Inlier Equiv-
alent Separations (WEIS) system. Length of hospital
stay in different diagnostic groups are classified as
prolonged based on the average length of stay w19x.
Patients were identified for inclusion in a randomised,
controlled single-centre study of an integrated HF out-
patient management programme, the Auckland Heart
Failure Management Study w18x. Patients were random-
ised either to attend a hospital-based HF management
clinic or to receive usual care (mainly based in primary
care). This intervention did not commence until after
discharge from hospital, and thus did not influence
length of stay of the index hospital admission.
Detailed data were recorded prospectively from the
hospital records of each index hospital admission. Data
included social and demographic characteristics; clinical
signs and symptoms at admission and discharge; labor-
atory and cardiac imaging parameters; in-patient social
worker consults; and complications (such as acute renal
failure secondary to treatment) or concomitant acute
medical problems (such as angina or pneumonia) that
occurred during the hospital stay and required specific
treatment or intervention.
2.2. Statistical analysis
In univariate analyses, Spearman’s rank coefficient
was used to determine correlation between variables.
For continuous variables, non-parametric analyses (Wil-
coxon rank sum tests) were utilised to examine differ-
ences in patients stratified above or below the median
length of hospital stay. For categorical variables, Fisher’s
exact tests were utilised. All tests were 2-tailed and
significant at the 5% level. Length of hospital stay was
stratified above and below the median for two reasons.
Firstly, length of hospital stay was not normally distrib-
uted; secondly, in-patient funding weights in New Zea-
land hospitals are calculated on whether length of
hospital stay is above or below the average for that
diagnostic group w19x. Analyses using a variety of
iterative multivariate linear regression methods (step-
wise, forward and backward selection) were performed
with length of hospital stay as a continuous variable.
Logistic regression was performed with length of hos-
pital stay stratified above and below the upper quartile.
Parsimony and biological plausibility were added to the
standard goodness of fit statistics to choose between
competing models.
Doses of loop diuretics and angiotensin-converting-
enzyme inhibitor medications are expressed in frusemide
and enalapril equivalents respectively. The need for
social intervention during in-hospital stay was coded if
documented referral to social work services was record-
ed in the hospital records.
3. Results
3.1. Patient characteristics
One hundred and ninety-seven patients were included
in this study (Table 1). The mean age was 73 years
 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209 203

Table 1
Patient characteristics

Demographicyclinical n (%) Echoylab investigations Mean (95% CI)

Socio-demographic Chest radiography
Gender CTR, % 61.7 (60, 63)
Male 119 (60)
ECG
Ethnicity Abnormal rhythm 69 (35)
Caucasian 157 (79) Q waves 60 (30)
MaoriyPacific Island 39 (20) L bundle branch block 59 (30)
Other 3 (1) LV hypertrophy 57 (30)
Living alone 69 (35) Echocardiographic
LVEF, % 32.1 (30, 34)
Aetiology LVEDD, mm 64.1 (62, 66)
Ischaemic 106 (54) LVESD, mm 51.3 (49, 53)
Non-ischaemic 93 (47) LVEDV, ml 178.5 (168, 189)
LVESV, ml 126.7 (117, 137)
Serum biochemistry
Comorbidities and medical history Serum Naq, mmolyl 138.6 (138.0, 139.2)
Previous myocardial infarct 90 (45) Serum Kq, mmolyl 4.2 (4.1, 4.3)
Previous coronary bypass 22 (11) Creatinine, mmolyl 0.131 (0.124, 0.139)
Prior hypertension 103 (52) Peak creatinine, mmolyl 0.157 (0.14, 0.167)
Current angina 51 (25)
Diabetes 57 (29)
Chronic atrial fibrillation 64 (33)
Previous stroke 41 (21)
Chronic airways disease 37 (19)
Symptomsysigns at admission
Chest pain 68 (34)
Dyspnoea on exertion 193 (98)
Dyspnoea at rest 144 (73)
Paroxysmal nocturnal dyspnoea 107 (54)
Orthopnea 141 (71)
NYHA class III or IV 165 (84)
Third heart sound 51 (26)
Hepatomegaly 59 (30)
Oedema 132 (67)
Rales 178 (90)
Laboratory values are from measurements at hospital admission. Normal laboratory values: serum sodium (Naq): 135–146 mmolyl; serum
potassium (Kq): 3.5–5.0 mmolyl; serum creatinine 0.05–0.12 mmolyl. NYHA, New York Heart Association; CTR, cardio-thoracic ratio; LVEF,
left ventricular ejection fraction; LVEDD, left ventricular diastolic diameter; LVESD, left ventricular systolic diameter; LVEDV, left ventricular
diastolic volume; LVESV, left ventricular systolic volume.

(S.D. 10.8) and 60% were male. Forty-five percent of
patients had a documented history of prior myocardial
infarction, 52% had prior hypertension, 29% diabetes,
19% obstructive airways disease, and 21% a prior stroke.
Half the patients had a prior admission for HF and 17%
had 3 or more previous HF admissions. Three quarters
of the patients were classified as being New York Heart
Association (NYHA) functional class IV on admission.
The median duration of symptoms prior to admission
was 7 days (IQR 2, 21). Patients were on a median of
5 medications at admission (IQR 3, 12) and 6 medica-
tions at discharge (IQR 5, 8).
The average LVEF was 32% (S.D. 13). The mean
serum sodium, potassium, albumin and haemoglobin
values at admission were within the normal range for
our laboratory (Table 1). However, renal function was
impaired: mean serum creatinine was 0.13 mmolyl
(normal range 0.05–0.12 mmolyl); the average creati-
nine clearance for the cohort was 48.9 mlymin (S.D.
24, normal range 90–140 mlymin).
3.2. Length of hospital stay
The median length of stay was 6 days (IQR 4, 9).
Thirty-four patients (17%) had a hospital stay in the
upper quartile (9 days or longer), and 10 patients above
the 95th centile (21 days or longer). There was no
difference in length of hospital stay between patients
admitted with HF for the first time and those with one
or more previous admissions. All patients were managed
on general medical wards and none required transfer to
the intensive care unit (ICU). Patients were treated with
intravenous diuretic for a median of 1 day (IQR 0, 3),
with the maximum at 33 days. During their hospital
204 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209

Table 2
Changes of clinical characteristics from admission to discharge

Variable Admission value Discharge value Mean change P-value

mean (95% CI) mean (95% CI) mean (95% CI)
Heart rate 93 (90.0, 96.1) 80 (77.9, 81.5) y13 (y16.7, y10.4) 0.0001
Systolic BP, mmHg 145 (140.8, 149.2) 125 (121.9, 127.9) y20.0 (y23.7, y16.2) 0.0001
Diastolic BP, mmHg 86 (83.2, 88.3) 71 (69.7, 73.1) y14.9 (y16.8, y11.5) 0.0001
JVP, cm 6.2 (5.1, 7.4) 2.2 (1.01, 3.34) y4 (y4.5, y3.3) 0.0001
Weight, kg 76.5 (73.8, 79.2) 73.7 (71.2, 76.2) y2.8 (y3.16, y2.04) 0.001
Serum creatinine, mmolyl 0.13 (0.10, 0.15) 0.15 (0.13, 0.16) 0.01 (0.007, 0.017) 0.0001
Number of medicationsa 5 (3, 7)a 6 (5, 8)a 1 (0, 3) 0.0001b
Dose of frusemide, mg 67.2 (55.5, 78.8) 111.2 (99.3, 123.1) 44 (35, 53) 0.0001
Dose of ACE-I, mg 7.7 (6.4, 8.9) 10 (8.9, 11.1) 2.2 (1.2, 3.3) 0.0001
Values are means and 95% confidence intervals unless otherwise specified. BP, blood pressure; JVP, jugular venous pressure; ACE-I, angiotensin
converting enzyme inhibitor (dose expressed as enalapril equivalents).
a
Median (IQR).
b
McNemar’s test used to determine significance.

stay, patients’ weight, heart rate and blood pressures
decreased significantly (Table 2). The mean decrease in
weight was 2.6 kg (S.D. 3.8). The mean increase in
serum creatinine was 0.013 mmolyl (S.D. 0.036). Doses
of frusemide and ACE-inhibitor were significantly
increased from admission to discharge.
3.3. Concomitant acute medical conditions and in-
hospital complications
Many patients had a complication or concomitant
acute medical condition requiring specific treatment
during their hospital stay (Table 3). Approximately 1 in
10 patients had concomitant renal problems, such as the
development of acute renal failure or the exacerbation
of pre-existing renal impairment. Cardiac conditions
were common (38%), particularly angina, myocardial
infarction and arrhythmias. Approximately 20% of the
cohort had concomitant acute respiratory diagnoses,
mostly pneumonia or exacerbation of chronic obstructive
airways disease. Other concurrent medical problems
included infections such as sepsis and cellulitis; gastro-
intestinal complaints including abdominal pain, nausea,
vomiting, and acute peptic ulceration; neurological prob-
lems including delirium, headache and seizures; diabetes
needing monitoring, intervention and education; mus-
culoskeletal complaints including back pain and acute
gout; anemia requiring transfusion; and metabolic abnor-
malities requiring further diagnostic tests. Eighteen
patients (9.1%) had a longer than average hospital stay

Table 3
Medical and social problems

n (%) n (%)
Renal 21 (10.6) Respiratory 38 (19.3)
Renal impairment 15 Respiratory infection 36
Other renal conditions 6 Other respiratory 2
Neurological 7 (3.5) Gastrointestinal 14 (7)
Cardiac 74 (37.6) Infection 12 (6.1)
Anginayacute MI 26
Atrial fibrillationyarrythmia 33 Other 87 (44.2)
Other cardiac conditions 15 Diabetes 23
Musculoskeletal (excl gout) 15
Social 32 (16.2) Gout 9
Refuses discharge 2 Haematologicytransfused 13
Mobilityyindependence 12 Other endocrinological 3
Home situation 14 Awaiting investigations 18
Languageycompliance 4 Metabolic 6

Iatrogenic 12 (6.1)
Drug side effects 10
Procedural complications 2
MI, myocardial infarction.
 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209 205

Table 4
Univariate analysis

LOS-6 days LOS06 days P-value

(ns89) (ns106)
Symptomsysigns at admission
NYHA class II or III 24 (27) 7 (7) 0.09
IV 58 (65) 79 (74) –
01 prior HF admission 43 (48) 59 (56) 0.84
Symptom duration, days 14 (8.5, 19.5) 18.2 (13.7, 22.7) 0.71
Orthopnea 61 (68) 80 (75) 0.30
Paroxysmal nocturnal dyspnoea 49 (55) 58 (54) 0.76
Fatigue 79 (89) 104 (98) 0.012
Oedema 51 (57) 81 (76) 0.01
Chest pain 33 (37) 35 (33) 0.05
Presence of third heart sound 25 (28) 26 (24) 0.05
Presence of rales 80 (90) 98 (92) 0.8
Presence of hepatomegaly 21 (23) 38 (35) 0.15
Heart rate, rateymin 92.7 (88.5, 96.8)a 93.4 (88.9, 97.9)a 0.80
Systolic blood pressure, mmHg 150 (142.5, 155.8)a 141.5 (136.1, 147.0)a 0.07
Diastolic blood pressure, mmHg 87.6 (83.7, 91.5)a 84.1 (80.9, 87.4)a 0.17
Jugular venous pressure, cm 5 (3, 7)b 5 (3, 10)b 0.04
Treatment
Dose loop diuretic at admission, mg 49.4 (35.9, 62.9)a 81.7 (64, 99.5)a 0.06
Dose loop diuretic at discharge, mg 84.6 (71.5, 97.7)a 133.1 (114.9, 155)a 0.001
Change in loop diuretic dose, mg q35.1 (25.8, 44.5)a q51.3 (36.6, 66)a 0.07
Dose ACE-I at admission, mg 7.2 (5.5, 8.8)a 8.1 (6.4, 9.9)a 0.92
Dose ACE-I at discharge, mg 9.4 (7.8, 10.9)a 10.5 (8.9, 12.0)a 0.31
Change in ACE-I dose, mg q2.2 (0.9, 3.5)a q2.3 (0.7, 4.0)a 0.92
Days on intravenous diuretic 1.2 (1, 1.4)a 3.3 (2.5, 4.2)a 0.0001
Number of medications at admission 5 (3, 6)b 5 (4, 7)b 0.04
Number of medications at discharge 6 (4, 7)b 7 (5, 8)b 0.008
Change in weight, kg y1.2 (0.9, 3.5)a y3.6 (2.7, 4.4)a 0.001
Characteristics of patients stratified for length of hospital stay (LOS), Values are n (%) unless specified. LOS, length of stay; ACE-I, angiotensin
converting enzyme inhibitor (dose expressed as enalapril equivalents).
a
Mean (95% CI).
b
Median (IQR).

due to the need to wait for inpatient hospital consulta-
tions and investigations for various co-existing
conditions.
Twelve patients had iatrogenic complications prolong-
ing their stay (Table 3): 5 had drug side effects and 2
had complications from procedures. Thirty-two patients
(16%) had social problems requiring intervention,
including issues of poor mobility and independence, the
need for home help or other home-based social inter-
ventions, or problems with language needing medical
translators.
3.4. Predictors of length of hospital stay: univariate
analysis
3.4.1. Clinical factors
Length of stay greater than 6 days was associated
with the presence at hospital admission of the symptoms
of peripheral oedema, chest pain, or fatigue; the clinical
findings of elevated jugular venous pressure or a third
heart sound; and weight increase during hospital stay
(Table 4). Longer than average length of hospital stay
was also associated with the number of patients’ medi-
cations at admission and at discharge, diuretic dose at
admission and at discharge, change in diuretic dose
during hospital stay, number of days treated with intra-
venous diuretic, and the development of complications
or other acute medical problems. Factors not associated
with length of hospital stay greater than 6 days included
age, gender, ethnicity, the number of previous admis-
sions with HF, duration of symptoms prior to admission,
NYHA functional class at admission, the presence at
admission of rales, orthopnea or paroxysmal nocturnal
dyspnoea.
There was no difference in length of hospital stay
between incident and prevalent cases; both had a median
length of stay of 6 days (Ps0.24). Thus incident and
prevalent cases were combined for analysis.
3.4.2. Laboratory parameters and indices of LV function
Serum albumin, serum sodium at admission and peak
creatinine were also associated with length of stay
greater than 6 days. Radiographic cardio-thoracic ratio,
echocardiographic LVEF and left ventricular end-dia-
206 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209

Table 5 Mulivariate logistic regression was performed in order

Multivariate analysis of length of hospital stay to investigate independent predictors of hospital stay
Variable Partial R 2 P-value longer than the top quartile of 10 days. Odds ratios are
a
shown in Table 6. Admission variables associated with
(a) Hospital admission variables
Living alone 0.04 0.07
length of hospital stay of more than 10 days included
Oedema at admission 0.14 0.002 the presence of oedema at admission. In-hospital pro-
Frusemide dose at admission 0.03 0.1 gress variables associated with length of stay )10 days
Cardio-thoracic ratio 0.03 0.09 included duration of treatment with intravenous diuretic;
(b) In-hospital progressb change in weight during stay; the development of renal
Days treated with intravenous diuretic 0.32 0.001 impairment; concurrent acute respiratory conditions
Development of renal impairment 0.11 0.001 requiring medical treatment (such as pneumonia or
Social problems 0.004 0.0005
Iatrogenic complications 0.04 0.0004
exacerbation of existing airways disease); and social
Change in weight 0.02 0.004 problems requiring in-hospital assessment.
Concomitant respiratory problems 0.02 0.01
a
Length of stay is modeled as a continuous variable (stepwise 3.6. Charlson comorbidity index
selection). Variables in model: age, race, gender, whether living alone,
duration of symptoms, NYHA class at admission, number of previous
admissions, number of medications, symptoms at admission, signs at A simple model for predicting length of stay from
admission, any in-hospital complication, comorbid medical problems admission and in-hospital progress data would be useful
(diabetes, chronic obstructive airways disease, cerebrovascular dis- for administrative purposes, including casemix analysis
ease), clinical variables at admission (chest pain, orthopnea, PND,
fatigue, peripheral oedema, heart rate, blood pressure, jugular venous and cost calculations. The Charlson comorbidity index
pressure, third heart sound), dose of ACE-inhibitor and frusemide at w20x is one such score, using patient age, ICD coding at
admission, LVEDD, LVESV, radiologic cardio-thoracic ratio, labora- discharge and weighting for specific conditions includ-
tory parameters at admission (serum sodium, serum potassium, serum ing ventricular arrhythmias, shock, malignancies, and
albumin).
b
Length of stay is modeled as a continuous variable (stepwise
the need for care in the ICU. However, the patients in
selection). Variables in model: age, race, gender, whether living alone, this study were managed on general medical wards and
number of previous HF admissions, days treated with intravenous diu- none required ICU care. In this cohort of elderly patients
retic, change in weight, change in frusemide dose, change in ACE- with exacerbations of chronic HF who survived to
inhibitor dose, change in creatinine, social problems requiring hospital discharge, shock did not occur, metastatic
intervention, concomitant diagnoses (respiratory, cardiovascular, neu-
rological, gastrointestinal, infections), development of iatrogenic malignancy was an exclusion criterion for the study, and
problems, development of renal impairment. ventricular arrhythmias were uncommon (in contrast to
atrial arrhythmias which were common). Hence, patient
stolic and end-systolic volumes were not associated with age was the main component of Charlson scores in this
longer than average length of stay. study. The mean Charlson score for this cohort was 2.4
(S.D. 1.32; range 0.5–9). The Charlson index was not
3.5. Predictors of length of stay: multivariate analysis predictive of length of stay in this cohort and correlated
poorly to length of stay with a r value of 0.08 (Ps
0.26).
Variables relating to the clinical status at admission
and in-hospital progress were modeled separately using
logistic regression methods. Admission variables asso- Table 6
ciated with longer than average hospital stay (06 days) Independent predictors of hospital stay of 10 days or more
included the presence of peripheral oedema, living alone
at home, oral diuretic dose at admission, and cardio- OR (95% CI) P
thoracic ratio (Table 5). This model explained 31% of (a) Hospital admission variables
the variance in length of hospital stay. Oedema at admission 10.5 (2.2, 50.3) 0.003
In-hospital progress variables associated with longer (b) In-hospital progress
than average length of stay (06 days) included duration Development of renal impairment 9.8 (2.5, 38.6) 0.001
of treatment with intravenous diuretic, the development Social problems 6.8 (1.7, 26.3) 0.006
of renal or non-renal iatrogenic complications, change Change in weighta 1.3 (1.1, 1.4) 0.0002
Days on intravenous diuretica 7.5 (1.2, 44.3) 0.03
in weight, and comorbid respiratory problems requiring Respiratory conditions 3.8 (1.3, 11.3) 0.01
specific treatment. The number of days of treatment
with intravenous diuretic was strongly associated with The model uses logistic regression with length of stay stratified
above and below the top quartile ()10 days). Variables in the model
length of hospital stay, partial r 2 0.32 (Ps0.001). This are the same as Table 5a and b respectively.
model explained 55% of the variance in length of a
Odds ratios are for 1 day on iv loop diuretic; and 1 kg weight
hospital stay. change, respectively.
 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209
Fig. 1. Conceptual model of factors influencing length of hospital stay.
4. Discussion
This study investigates the determinants of length of
hospital stay in a cohort of patients representative of HF
patients admitted to the acute general medical service
in a general hospital setting. Commonly, these patients
are elderly, with multiple comorbidities and are receiving
multiple medications. Several clinical variables were
associated with longer than average hospital stay ()
median of 6 days). Hospital admission variables includ-
ed the presence of co-existing medical problems and
symptoms of peripheral congestion. Factors relating to
in-hospital progress included the duration of treatment
with intravenous diuretic; change in weight during
hospital stay; the development of renal and non-renal
iatrogenic complications; concomitant respiratory con-
ditions requiring specific treatment; and social problems
requiring in-patient assessment. Stratifying length of
hospital stay at the top quartile reinforced the importance
of these clinical variables, particularly concurrent social
and medical comorbidity. Many studies of length of
hospital stay predict duration of stay from laboratory
parameters or other quantifiable variables w13,15,17x. In
this study, laboratory and echocardiographic variables
were not associated with length of hospital stay.
Duration of treatment with intravenous diuretic is a
surrogate measure of disease severity and the degree of
peripheral congestion, reflecting the need for prolonged
hospitalisation and treatment. Patients with hospital stays
less than 6 days lost a mean of 1.2 kg compared with a
loss of 3.6 kg for those who stayed longer (Ps0.001),
and change of weight was independently associated with
length of hospital stay. Peripheral congestion is often
resistant to the initial in-patient treatment of HF. The
effect of peripheral congestion as opposed to pulmonary
congestion on the prolongation of length of hospital stay
in HF has not previously been documented.
The occurrence of concomitant acute respiratory prob-
lems requiring treatment, and the development of renal
failure or iatrogenic complications were also indepen-
dently associated with longer than average length of
hospital stay as shown in previous studies w8,12,15x.
Administrative scores used to calculate casemix and in-
207
hospital costs should recognise the importance of periph-
eral congestion at admission, social problems and
concurrent medical problems requiring specific treat-
ment.
Social ‘complications’ and living alone were also
independently associated with longer than average hos-
pital stay in this cohort of patients with HF. The burden
of social problems on healthcare services is not easily
documented as social problems may not be consistently
recorded in hospital case records or captured using ICD
codes and may have been under-estimated. The direct
effect of patients’ social environment on length of
hospital stay, while clinically self-evident and an obvious
feature of chronic disease has not been previously
reported. The importance of the social and medical
comorbidities in this cohort of patients with HF may
suggest that length of stay in HF may not be further
modifiable in our institution by adjusting in-patient
treatment strategies. It may be more appropriate for
future interventions to target social problems affecting
readiness for discharge such as targeted assessment of
the home situation, social support, patient mobility, and
independence issues.
The multivariate models in this study explained only
30–50% of the variance in hospital stay of patients with
HF. Other factors more difficult to measure are likely to
have important effects on hospital stay. Examples may
include compliance, mobility, delay in awaiting rest
home placement, the use of inappropriate medications
and factors relating to the healthcare environment in a
broader sense. It is also likely that the impact of social
needs of patients with HF and the role of comorbid
medical conditions on hospital stay were under-estimated
in this study. Length of hospital stay in HF is affected
by many factors, which in part reflect the clinical
spectrum of HF presenting to hospital, the admission
threshold, the importance of concomitant medical prob-
lems and the consequences of HF therapy (Fig. 1).
The functioning of the greater healthcare environment
is also likely to be important, including healthcare
funding, access to hospital and primary care, and the
socio-cultural environment of each patient. The severity
of HF symptoms in patients admitted to hospital depends
208 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209
on an institution’s admission ‘threshold’. In New Zea-
land, this is affected by the evaluating primary care
practitioner who has a gate-keeping role, and by hospital
bed availability. Lastly, the discharge environment will
impact on length of stay. The assessment of patient
readiness for discharge encompasses the presence of
community supports and access to primary care as well
as a patient’s social situation and medical status (Fig.
1). The availability of community-based social support
may offset any increased risk of early readmission if
patients with medical and social comorbidity are dis-
charged too early. In sum, variables influencing length
of hospital stay in HF include a broad range of societal,
economic and political factors outside the scope of this
study, which is confined to the impact of clinical and
patient factors.
The optimal duration of hospital stay for treatment of
an exacerbation of chronic HF remains unknown. The
median length of hospital stay in this study, although
comparable to other countries w3x, is still considerably
longer than in Oregon, where the average length of stay
in HF is 4 days w9x. It is possible that patients admitted
in Oregon have milder HF or the threshold for hospital
admission is lower leading to shorter hospital stay.
This study is confined to the examination of length
of hospital stay of patients who survived until discharge
and were recruited into an outpatient HF management
programme. Patients who died during their hospital stay
were not included in this study. Other studies have
shown that patients who die in hospital with HF have a
longer mean stay than those who survive to discharge
w21x. The factors affecting the length of stay of patients
with HF who die in hospital may differ from factors
that have been shown to be important in this study. This
study also excluded patients who refused informed
consent (16% of patients approached) or who fulfilled
exclusion criteria specified in the Auckland Heart Fail-
ure Management Study (43% of patients approached).
The substantial burden of HF in terms of healthcare
resource consumption includes hospital readmissions
and primary care consultations, two important compo-
nents not investigated in this study. Future studies of the
impact of HF on healthcare systems must include both
primary and secondary care in order to assess the total
burden of HF. Likewise, strategies for reducing hospital
use in HF must be developed and assessed with the
entire framework of healthcare services in mind, includ-
ing the phenomenon of readmissions and the role of
primary care.
Acknowledgments
The Auckland Heart Failure Management Study was
funded by a Project Grant from the National Heart
Foundation of New Zealand and an unrestricted educa-
tional grant from Merck Sharp and Dohme (NZ) Ltd.
RND was the recipient of the NZ Heart Foundation
BNZ Senior Fellowship. We acknowledge the involve-
ment of participating Auckland general practitioners
w19x.
References
w1x Sharpe N, Doughty R. Epidemiology of heart failure and
ventricular dysfunction. Lancet 1998;352(Suppl):S3 –7.
w2x Doughty R, Yee T, Sharpe N. Hospital admissions and deaths
due to congestive heart failure in New Zealand 1988–1991.
NZ Med J 1995;1012:473 –5.
w3x Stewart S, MacIntyre K, MacLeod MM, Baily AE, Capewell
S, McMurray JJV. Trends in hospitalization for heart failure in
Scotland 1990–1996. An epidemic that has reached its peak.
Eur Heart J 2001;22:209 –17.
w4x McMurray JJV, Petrie M, Murdoch D, Davie A. Clinical
epidemiology of heart failure: public and private health burden.
Eur Heart J 1998;19(Suppl P):P9 –16.
w5x Tavazzi L. Epidemiological burden of heart failure. Heart
1998;79(Suppl 2):S6 –9.
w6x Roughead L, Gilbert A. Australian trends in hospitalisation
and mortality associated with chronic heart failure. Med J Aust
1998;168:256.
w7x Haldeman GA, Croft JB, Giles WH, Rashidee A. Hospitalisa-
tion of patients with heart failure. National hospital discharge
survey 1985 to 1995. Am Heart J 1999;137(2):352 –60.
w8x Reitsma J, Mosterd A, de Craen A, et al. Increase in hospital
admission rates for heart failure in the Netherlands, 1980–
1993. Heart 1996;76:388 –92.
w9x Ni H, Nauman D, Hershberger R. Analysis of trends in
hospitalisations for heart failure. J Cardiac Failure
1999;5(2):79 –84.
w10x Philbin E, DiSalvo T. Managed care for congestive heart
failure: influence of payer status on process of care, re-
source utilization, and short-term outcomes. Am Heart J
1998;136:553 –61.
w11x Jencks S, Williams D, Kay T. Assessing hospital-associated
deaths from discharge data: the role of length of stay and
comorbidities. J Am Med Assoc 1988;260(15):2240 –6.
w12x Brown A, Cleland J. Influence of concomitant disease on
patterns of hospitalization in patients with heart failure dis-
charged from Scottish hospitals in 1995. Eur Heart J
1998;19:1063 –9.
w13x Harjal K, Nunez E, Turgut T, et al. The independent effects of
left ventricular ejection fraction on short-term outcomes and
resource utilization following hospitalisation for heart failure.
Clin Cardiol 1999;22:184 –90.
w14x Burns L, Wholey D. The effects of patient, hospital and
physician characteristics and length of stay and mortality.
Medical Care 1991;29(3):251 –71.
w15x Rich M, Shah A, Vinson J, Freedland K, Kuru T, Sperry J.
Iatrogenic congestive heart failure in older adults: clinical
course and prognosis. J Am Geriatr Soc 1996;44(6):638 –43.
w16x Rosenthal G, Harper D, Quinn L, Cooper G. Severity-adjusted
mortality and length of stay in teaching and nonteaching
hospitals: results of a regional study. J Am Med Assoc
1997;278:485 –90.
w17x Weinfield M, Chertow G, Stevenson L. Aggravated renal
dysfunction during intensive therapy for advanced chronic
heart failure. Am Heart J 1999;138:285 –90.
w18x Doughty RN, Wright SP, Pearl A, et al. Randomized, controlled
trial of integrated heart failure management: the Auckland
 S.P. Wright et al. / The European Journal of Heart Failure 5 (2003) 201–209 209

Heart Failure Management Study. Eur Heart J 2002;23:139 –
46.
w19x Rains M, Hogan J, Ayto A, Thomson R. National Service
Framework Project: report of the costs weight project group.
Ministry of Health, Wellington, New Zealand, 23 March 2001.
Online report available from www.moh.govt.nz.
w20x Charlson M, Szatrowski TP, Peterson J, Gold J. Valida-
tion of a combined comorbidity index. J Clin Epidemiol
1994;47:1245 –51.
w21x Philbin EF, McCullogh PA, Dec GW, DiSalvo TG. Length of
stay and procedure utilization are the major determinants of
hospital charges for heart failure. Clin Cardiol 2001;24:56 –62.