Scandinavian Journal of Infectious Diseases, 2013; 45: 219–226

ORIGINAL ARTICLE

Length of stay after reaching clinical stability drives hospital

costs associated with adult community-acquired pneumonia

PIETER-JAN CORTOOS1,2, CHRISTA GILISSEN3, GERT LAEKEMAN1,

WILLY E. PEETERMANS4, HILDE LEENAERS3, LUC VANDORPE3 &
STEVEN SIMOENS1
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

From the 1Research Centre for Pharmaceutical Care & Pharmaco-economics, University of Leuven (KU Leuven), Leuven,
2Hospital Pharmacy Department, University Hospital Brussels (UZ Brussel), Brussels, 3Hospital Pharmacy Department,

Ziekenhuis Oost-Limburg, Genk, and 4Department of General Internal Medicine and Infectious Diseases, University Hospitals
Leuven, Leuven, Belgium

Abstract
Background: Community-acquired pneumonia (CAP) has a considerable clinical and economic impact. The aim of this
study was to identify drivers of hospital costs associated with CAP in 2 Belgian hospitals. Specifically, the influence of
For personal use only.

patient characteristics, quality indicators, and other treatment aspects on hospital costs was explored. Methods: The follow-
ing were registered for patients admitted with a confirmed diagnosis of CAP in a large university hospital (Universitaire
Ziekenhuizen Leuven, UZL) and a medium-sized secondary care hospital (Ziekenhuis Oost-Limburg, ZOL) in Belgium:
the pneumonia severity index (PSI), time to clinical stability, length of stay, antibiotic therapy, outcomes, compliance with
validated quality indicators, and the different costs (pharmacy, laboratory, and radiology, and total). Regression analysis
was used to identify influential variables. Results: Between October 2007 and June 2010, 803 patients were included, with
a median total cost of €4794.57. The length of stay after clinical stability and time to clinical stability had the highest
influence on the total cost (⫹6.3% and ⫹4.9% per additional day, respectively; p ⬍ 0.0001). Other important drivers of
higher costs were total therapy duration, PSI score, age, and admission to intensive care. Patients treated with moxifloxacin
had significantly, but limited, lower costs. Quality indicator compliance, including guideline-compliant antibiotic treatment
and therapy streamlining, had little influence. Conclusions: The most important driver of hospital costs associated with CAP
was the time between clinical stability and actual hospital discharge. In order to substantially decrease the costs of CAP
treatment, this period should be rigorously evaluated for possible intervention targets that would allow costs in CAP treat-
ment to be decreased in a substantial manner.

Keywords: Quality indicator, pneumonia, guidelines, antibiotics, Belgium

Introduction
Several studies have examined the cost impact of
Community-acquired pneumonia (CAP) is a com- patient characteristics, guideline compliance, and
mon infectious disease with considerable clinical other aspects of antimicrobial therapy such as anti-
and economic impact. In North America, the inci- biotic streamlining. However, these are often not
dence is about 1% of the adult population annually, combined, while other aspects of antimicrobial ther-
with an estimated annual hospital cost in 2007 of apy are often left out, and the variation explained by
between US$8 billion and US$10 billion [1]. A a single factor is limited [3,4]. The aim of this study
2010 review showed that the total cost of pneumonia was to identify the drivers of hospital costs associ-
in Europe was approximately €10.1 billion, with ated with CAP in 2 Belgian hospitals. Specifically,
inpatient care accounting for €5.7 billion, and this the influence of quality indicators on hospital costs
is likely to increase with the aging population [2]. was explored.

Correspondence: P.-J. Cortoos, Research Centre for Pharmaceutical Care & Pharmaco-economics, University of Leuven (KU Leuven), O&N 2, Herestraat
49, PO Box 521, B-3000 Leuven, Belgium. Tel: ⫹32 16 3 23465; Fax: ⫹32 16 3 23468. E-mail: [email protected]

(Received 6 March 2012 ; accepted 24 August 2012 )

ISSN 0036-5548 print/ISSN 1651-1980 online © 2013 Informa Healthcare
DOI: 10.3109/00365548.2012.726737
 220 P.-J. Cortoos et al.
Methods composed of demographics, vital signs, various
physical and laboratory parameters, and specific
Hospitals
co-morbidities (heart failure, renal, liver, cerebrovas-
Data were collected from a larger prospective survey cular and/or neoplastic disease), allows the quantifi-
on the use of antimicrobial guidelines in 2 Belgian cation of the morbidity and mortality risk for
hospitals: a large 1600-bed tertiary care hospital CAP patients, and is used to guide treatment. Other
(Universitaire Ziekenhuizen Leuven, UZL) and a pathologies that are not included in the PSI score
medium-sized 800-bed secondary care hospital (diabetes, obstructive/restrictive lung pathology, and
with expertise in pulmonary diseases (Ziekenhuis dementia) and whether the patient lived in a nursing
Oost-Limburg, ZOL) (ClinicalTrials.gov identifier home or other long-term care facility were also reg-
NCT00512772). istered. For each day in the hospital, antibiotic ther-
apy was registered and the patient was assessed for
clinical stability. The time to clinical stability was
Patient sample set at noon on the day when the following criteria
were fulfilled: temperature ⬍ 37.8°C with no fever
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

Patients were enrolled from 16 October 2007 until for at least 12 h, pulse ⬍ 100/min, respiratory rate
1 July 2010 (UZL) and from 1 December 2007 until ⬍24/min, systolic blood pressure ⬎90 mmHg, oxy-
1 June 2010 (ZOL). Eligible patients were identified gen saturation ⱖ 95% with supplementary oxygen or
in the emergency department (ED) and met the ⱖ 90% without supplement, able to eat and drink,
following inclusion criteria: adult (⬎ 18 y) patients and stable mental status [6]. This moment can be
with a suspected or confirmed diagnosis of (lower) used as a theoretical reference point for intravenous-
respiratory tract infection on admission and con- to-oral switch and hospital discharge. Clinical data
firmed diagnosis of CAP on discharge by the attend- were extracted by PC (UZL) or HL (ZOL) from the
ing physician(s), with radiological confirmation of a nursing and medical files and the electronic patient
new pulmonary infiltrate, admitted through the ED, management systems, and entered into an electronic
For personal use only.

with at least 2 of the following symptoms: fever database (MS Access for Windows). Financial data
(⬎ 38.5°C) or hypothermia (⬍ 36°C), chills, exces- were obtained from the pharmacy database (ZOL)
sive sweating, new cough or changed sputa in the or accounting department (UZL). Costs reflected
case of chronic cough, dyspnoea, and abnormal lung official national list prices, including third-party
auscultation. Exclusion criteria were: failure to payer reimbursement and patient co-payment, and
obtain written informed consent, other major non- were divided into pharmacy costs, technical costs
pulmonary infections, pneumonia not present on (laboratory, radiology, and physician consultations),
admission, patients with a recent history (30 days) of hotel costs (costs inherent in hospital operation),
hospitalization, transfer from another hospital, death and total hospital costs. Cost data did not include
within 24 h of admission, (suspected) aspiration implants such as pacemakers and were not corrected
pneumonia, immunocompromised patients (active for inflation due to the short time horizon of the
haematological malignancies, asplenia, any dose of study, the absence of important changes in official
systemic glucocorticoids or other immunomodulat- drug prices or clinical practice at both hospitals, and
ing drugs, cytotoxic treatment within the last limitations of the provided data.
3 months, and HIV infection), pregnancy, patients
with cystic fibrosis, admission for social reasons,
known colonization with methicillin-resistant Staph- Measurement
ylococcus aureus (MRSA) or other multi-resistant
The variables used for our cost evaluation are pre-
microorganism, and patients in a palliative setting.
sented in Table I. Registered patient variables were
This protocol was reviewed and approved by the
the PSI score, presence of other co-morbidities (dia-
ethics committees of both institutions. In accordance
betes, chronic obstructive/restrictive lung pathology,
with their recommendations and national legislation,
and dementia, with each co-morbidity separately
written informed consent was obtained from all
dummy-coded and entered into the regression anal-
patients or their representatives before definitive
ysis), the hospital of admission, residency in nursing
inclusion.
home/long-term care, re-admission within 30 days
after discharge, mortality during admission or within
30 days after discharge, time to clinical stability, and
Data collection
length of stay after the patient had become clinically
For each patient, the disease severity was assessed stable. These last 2 variables were used instead of
at the time of admission using the pneumonia sever- total length of stay, as separately they give a better
ity index (PSI) score [5]. This severity score is representation of the initial patient recovery in view
 Drivers of pneumonia hospital costs evaluated 221
Table I. Overview patient characteristics and quality indicators used in regression analysis for UZL and
ZOL.a

Variables UZL (n ⫽ 477) ZOL (n ⫽ 326)

Patient characteristics and PSI components

Age, y 71.6 (55.3–80.6) 67.6 (53.8–77.4)
Female 47.8 (228) 35.3 (115)
PSI score 90 (67–120) 84 (55–103)
CAP classb
I and II 28.9 (138) 38.4 (125)
III 21.8 (104) 24.5 (80)
IV 34.2 (163) 27.9 (91)
V 15.1 (72) 9.2 (30)
Admission to intensive care ⬍ 48 h after admission 13.6 (65) 6.4 (21)
Mortality during hospitalization 1.9 (9) 2.5 (8)
Mortality within 30 days after discharge 1.0 (5) 2.1 (7)
Readmission within 30 days after discharge 10.9 (52) 12.9 (42)
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

Resident in elderly home/long-term care facility 7.3 (35) 3.1 (10)

Neoplastic disease present 4.2 (20) 3.7 (12)
History of liver failure 1.3 (6) 1.2 (4)
History of congestive heart failure 23.9 (114) 23.9 (78)
History of cerebrovascular disease 11.3 (54) 5.2 (17)
History of renal failure 18.2 (87) 9.2 (30)
Confusion/altered mental status on admission 14.7 (70) 8.6 (28)
Pleural effusion 28.5 (136) 18.4 (60)
Other co-morbidities and therapy characteristics
History of COPD or other chronic pulmonary disease 28.7 (137) 36.5 (119)
Diabetes type I/II 17.6 (84) 15.0 (49)
Dementia or other mental illness 6.9 (33) 4.0 (13)
For personal use only.

Only moxifloxacin given 12.6 (60) 13.8 (45)

Moxifloxacin and other AB given 21.2 (101) 7.7 (25)
Urine culture taken 62.1 (296) 17.2 (56)
Length of stayb 8.0 (5.2–13.9) 7.0 (4.9–10.0)
Remaining duration of hospitalization after clinical stability 4.29 (2.2–8.9) 4.97 (3.1–7.1)
Time to clinical stability 2.86 (1.77–4.98) 1.75 (0.95–3.10)
Quality indicators
Guideline compliant therapy on first dose 56.0 (267) 66.0 (215)
Guideline compliant therapy 24–48 h after admission 58.3 (278) 55.8 (182)
First AB dose ⬍ 4 h after admission 67.5 (322) 57.7 (188)
Ratio intravenous/total administrations 0.70 (0.48–0.88) 0.79 (0.62–1.00)
Therapy streamlined from broad-spectrum to narrower 25.9 (22/85)c 33.3 (10/30)c
spectrum in accordance with antibiogram
Sputum culture taken after first dose of AB 24.3 (116) 35.0 (114)
(vs no sputum culture)
Sputum culture taken before first dose of AB 28.9 (138) 4.9 (16)
(vs no sputum culture)
Blood culture taken after first dose of AB 17.4 (83) 47.5 (155)
(vs no blood culture)
Blood culture taken before first dose of AB 70.6 (337) 23.6 (77)
(vs no blood culture)
Urinary Legionella antigen test taken 32.5 (155) 9.5 (31)

AB, antibiotic; CAP, community-acquired pneumonia; COPD, chronic obstructive pulmonary disease;
PSI, pneumonia severity index; UZL, Universitaire Ziekenhuizen Leuven (tertiary care hospital); ZOL,
Ziekenhuis Oost-Limburg (secondary care hospital).
aContinuous variables are presented as median (interquartile range); other variables are presented as

percentage (total number).

bNot entered in the regression analysis.
cPercentage (proportion) calculated only for patients with positive blood and/or sputum cultures, or

positive Legionella test.

of those factors that may interfere with hospital dis- separately to determine its specific influence. We
charge, such as additional examinations or problems also included antibiotic choices, primarily the use of
with transfer to home care. Although patient age is moxifloxacin, which has been shown to be a cost-
already part of the PSI, this variable was assessed effective treatment of CAP [7,8]. This variable was
 222 P.-J. Cortoos et al.
dummy-coded as ‘No moxifloxacin given vs only regression analysis. For each remaining variable, the
moxifloxacin given’ and ‘Both moxifloxacin and other relative weight (RW) as a percentage of R2 was calcu-
antibiotics given vs only moxifloxacin given’. Finally, lated using the lmg metric in the R package ‘relaimpo’.
in order to assess the impact of different aspects This metric allows quantification of the individual vari-
of CAP management on costs, validated quality indi- able’s contribution to the variation in CAP treatment
cators were applied using electronic algorithms on expenditure and potential for future intervention [11].
each individual treatment (Table I) [9]. These quality Potential interaction terms were calculated between the
indicators are selected from key guideline recom- following dichotomous variables: location, ICU admis-
mendations, rated to have relevance to health benefit, sion, history of chronic lung disease, antibiotic choice,
antimicrobial resistance, or cost-effectiveness, repre- and taking of Legionella urinary antigen, and between
senting good clinical practice independent of changes the 2 major continuous variables: time to clinical stabil-
in antibiotic choices, and have previously been used ity and post-stability length of stay. Each interaction
to monitor antimicrobial practice [10]. The indicator was separately added to and compared with the initially
‘Prescribe empirical (antibiotic) therapy adherent found model. The indicator ‘Streamlining therapy in
to national guidelines’ was evaluated using an elec- accordance with antibiogram’ was not included in this
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

tronic algorithm based on the local guidelines and at regression due to the low number of valid cases, and
2 different time points – at the first dose and after was analyzed using a Mann–Whitney test and the asso-
24 h – in order to allow early modifications, which ciated effect size r for each hospital separately due to
are very common in a teaching hospital. The indica- the potential impact of hospital of admission. Analysis
tor ‘Streamlining therapy (from broad-spectrum to was done using SPSS 16.0 for Windows and R.
narrow spectrum antibiotic) in accordance with anti-
biogram’ was analyzed using a similar method. Only
blood or sputum cultures taken ⬍ 3 days after admis- Results
sion, without divergent results or interfering positive Of the 5339 patients with symptoms of respiratory
urine cultures, and positive Legionella tests were tract infection who were screened in UZL, 4258
For personal use only.

assessed. Compliance or not with the indicators patients had no confirmed diagnosis of CAP, 20
above was dummy-coded and entered into the anal- patients had a known history of multidrug-resistant
ysis. The indicator ‘Intravenous–per os switch accord- organisms, 11 patients died before informed con-
ing to existing recommendations’ was replaced by the sent could be obtained, 61 patients refused to par-
intravenous/total number of administrations ratio, ticipate in the study, and 170 patients left the
because for patients on combination therapy, an hospital before formal inclusion. Five patients had
unequivocal time of switch could often not be calcu- incomplete files and were also excluded from the
lated. ‘Stop antibiotic therapy 3 days after deferves- analysis. In ZOL, 11 ICU patients could also not
cence’ was replaced by the total therapy duration. be included due to local interference with other
The indicators ‘Perform culture of a blood sample’ studies. In total, 803 eligible patients gave their con-
and ‘Perform culture of a blood sample before em- sent. One patient was identified as an outlier and
pirical therapy’ were dummy-coded as ‘No sample vs was excluded from further analysis in most models
blood sample before empirical therapy’ and ‘No except pharmacy costs. Patient characteristics are
sample vs blood sample after empirical therapy’. Per- presented in Table I. The mean total cost for UZL
forming a sputum sample and its timing were coded was €8117.78, while for ZOL the mean cost for
similarly. Dose adaptation to renal function, therapy CAP was €4970.11 (Table II).
change if there is no improvement within 72 h, and Our final regression model was able to explain
presence of hospital-acquired complications were not 83.9% (p ⬍ 0.001) of variation in total costs. The
included as indicators due to incomplete data. most influential significant predictors were length of
stay after clinical stability (⫹6.3% per day) and time
to clinical stability (⫹4.9% per day), followed by
Statistical analysis
total therapy duration (⫹0.7% per day), initial PSI
Influential variables were selected using a stepwise score (⫹0.2% per unit), age on admission (⫹0.2%
multiple regression analysis based on Akaike’s informa- per y), and admission to intensive care (⫹18.0%)
tion criterion. Given that costs were not normally dis- (Table III). Similar influence was also present in the
tributed (as determined by visual analysis), original models for pharmacy costs (R2 ⫽ 0.595), technical
cost data were natural log transformed. Each model costs (R2 ⫽ 0.727), and hotel costs (R2 ⫽ 0.822)
was checked for multicollinearity, with a variance infla- (Supplementary Tables I–III, to be found online
tion factor equal to 5 as the upper limit, and for pos- at http://www.informahealthcare.com/doi/abs/10.31
sible outliers (using Cook’s distance criterion). All 09/00365548.2012.726737). Regarding antibiotic
variables mentioned above were initially entered in the choice, patients showed higher total expenses when
 Drivers of pneumonia hospital costs evaluated 223
treated without moxifloxacin (⫹13.5% on total

4794.57; (6853.98; 3419.82–7491.07)

3535.74 (5061.54; 2425.57–5744.90)
913.38 (1350.84; 672.31–1394.20)
costs) or when treated with a combination of moxi-

278.33 (391.62; 233.81–372.55)

floxacin and other antibiotics (⫹17.8%), but without
Aggregate costs (n ⫽ 797)
median (mean; IQR)
important impact on cost variation. Patients who
had sputum cultures taken showed higher total costs
than patients without sputum culture, whether
cultures were taken in a timely manner (⫹9.1%) or
not (⫹5.4%). Pharmacy, technical, and hotel costs
showed similar results. Finally, streamlining antibiotic
therapies resulted in rather similar or even higher costs
for one of the hospitals, although with only a small to
moderate effect size (Supplementary Table IV, to be
IQR, interquartile range; UZL, Universitaire Ziekenhuizen Leuven (tertiary care hospital); ZOL, Ziekenhuis Oost-Limburg (secondary care hospital).

found online at http://www.informahealthcare.com/

doi/abs/10.3109/00365548.2012.726737).
Various interactions were present amongst the
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

p-Valueb

variables found.The most important one was between

⬍ 0.001
⬍ 0.001
⬍ 0.001
⬍ 0.001

ICU admission and length of stay after stability,

although additional explained variability was limited
(Table III and Supplementary Tables I–III to be
found online at http://www.informahealthcare.com/
doi/abs/10.3109/00365548.2012.726737).
A post-hoc analysis of the factors that had the
3118.59 (3560.73; 2079.06–4158.12)
4136.59 (4970.11; 3079.72–5603.27)
848.66 (1026.00; 651.05–1086.90)

highest impact on length of stay after clinical stability,

238.28 (383.38; 165.57–379.73)

showed that PSI score had the strongest correlation

median (mean; IQR)

with this time period in both hospitals (Spearman’s

Table II. Overview of different cost types (pharmacy, technical, and total) in Euro (€) for UZL and ZOL.a
For personal use only.

ZOL (n ⫽ 320)

rho⫽ 0.487; p ⬍ 0.001). After correcting for PSI, age

(Spearman’s rho⫽ 0.165; p ⬍ 0.001), total therapy
duration (Spearman’s rho⫽ 0.252; p ⬍ 0.001), read-
mission at ⬍ 30 days (Spearman’s rho⫽ 0.129;
p ⬍ 0.001), blood culture taken after the first dose
of antibiotic (Spearman’s rho⫽ 0.088; p ⫽ 0.015),
and location (Spearman’s rho⫽ ⫺0.090; p ⫽ 0.013)
were associated with the length of stay after clinical
stability.

Discussion
(6151.88; 2679.50–7133.28)
(8117.78; 3811.31–9454.51)
(1568.76; 697.12–1664.59)

Our aim was to explore whether variations in the

(397.14; 252.54–371.30)

bMann–Whitney test for difference between UZL and ZOL.

costs of CAP could be explained by patient charac-

median (mean; IQR)

teristics in combination with existing quality indica-

UZL (n ⫽ 477)

tors that reflect different aspects in the treatment of

CAP. Length of stay after clinical stability, time to
clinical stability, total therapy duration, admission to
intensive care, and disease severity explained more
aCosts do not include medical implants.

than half of the cost variations in all models. Gener-

291.88
1017.57
4163.84
5489.58

ally speaking, the influence of the used quality

indicators on treatment costs is very limited.
The length of stay after clinical stability was the
most important single driver of total costs, followed
by the time to clinical stability, consistent with
the observation that final length of stay will mainly
determine treatment costs [12]. The influence of the
Pharmacy costs
Technical costs

PSI score on this period was rather straightforward,

Hotel costs
Total costs

with more seriously ill patients or those with more

co-morbidities taking longer to recover, for whom the
factor of clinical stability may not be sensitive enough.
 224 P.-J. Cortoos et al.
Table III. Regression analysis of total costs (natural log transformed) with separate contribution to R2 (n ⫽ 708).

Relative
Unstandardized contribution
Variables coefficients (SE) t to R2 (%) Adjusted R2

Constant 7.476 (0.060) 124.121d 0.839d

LOS after clinical stability (days) 0.061 (0.002) 32.590d 42.82
Time to clinical stability (days) 0.048 (0.003) 17.435d 15.29
Total duration AB therapy (days) 0.007 (0.002) 3.398d 13.08
PSI score 0.0017 (0.0005) 3.592d 10.15
Patient age 0.002 (0.001) 2.666c 6.00
Admission to intensive care ⬍ 48 h after admission 0.166 (0.039) 4.267d 4.75
Urinary culture taken 0.035 (0.023) 1.503 2.39
Hospital (UZL vs ZOL) 0.068 (0.0258) 2.699c 2.13
Sputum culture taken after first dose of AB (vs no sputum culture) 0.052 (0.024) 2.194b 1.01
Urinary Legionella antigen test taken 0.051 (0.027) 1.906a 0.84
Moxifloxacin and other AB given (vs only moxifloxacin given) 0.164 (0.043) 3.860d 0.59
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

No moxifloxacin given (vs only moxifloxacin given) 0.127 (0.038) 3.384d 0.47
Sputum culture taken before first dose of AB (vs no sputum culture) 0.087 (0.028) 3.156c 0.30
Ratio intravenous/total administrations ⫺0.145 (0.046) ⫺3.250c 0.17
Admission to intensive care * LOS after clinical stability ⫺0.029 (0.004 ⫺7.476d ⫺ 0.851d
Admission to intensive care * time to clinical stability ⫺0.022 (0.005) ⫺4.507d ⫺ 0.844d
Hospital * LOS after clinical stability ⫺0.011 (0.004) ⫺2.968c ⫺ 0.841d
Hospital * time to clinical stability ⫺0.031 (0.007) ⫺4.312d ⫺ 0.843d
Moxifloxacin and other AB given * LOS after clinical stability 0.014 (0.004) 3.399c ⫺ 0.842d
Moxifloxacin and other AB given * time to clinical stability 0.019 (0.006) 2.980c ⫺ 0.841d
No moxifloxacin given * LOS after clinical stability ⫺0.018 (0.004) ⫺4.779d ⫺ 0.844d
No moxifloxacin given * time to clinical stability ⫺0.021 (0.006) ⫺3.349d ⫺ 0.842d
For personal use only.

AB, antibiotic; LOS, length of stay; PSI, pneumonia severity index; SE, standard error; UZL, Universitaire Ziekenhuizen Leuven (tertiary
care hospital); ZOL, Ziekenhuis Oost-Limburg (secondary care hospital).
Significance levels: a0.10 ⬎ p ⬎ 0.05; bp ⬍ 0.05; cp ⬍ 0.01; dp ⬍ 0.001.

More important was the independent influence of age, length of stay, drug expenses are fully reimbursed. In
readmission, and therapy duration, whereas the impact some of the more severe cases, this may lead to inten-
of location was almost negligible. This suggests that tionally delaying discharge in order to fully recuperate
elderly patients are confronted with specific issues costs. Unfortunately, the precise extent of this practice
such as an increased number of drug therapies, mul- is not known nor its impact on our data.
tiple co-morbidities, dormant social problems, and Another observation is the influence of microbiol-
reduced autonomy, which may delay discharge and ogy sampling. Our results showed both timely and late
result in frequent readmission [13]. Equally, the anti- sputum cultures to be associated with an important
biotic therapy itself appears to be a cause for delayed increase in various costs. However, for sputum samples
discharge. Possible solutions may be improvements in that were taken in a timely manner, the increase in costs
the switch from intravenous to oral antibiotics or an had a minimal impact. This is in accordance with the
increased use of outpatient parenteral antibiotic ther- recommendation of taking sputum cultures preferen-
apy for eligible patients [14]. Still, these factors can tially before the first dose of antibiotics, in order to
only partly explain the impact of post-stability length allow therapy de-escalation and the use of less expen-
of stay. Because this latter period represents the part sive antibiotic agents [1,9]. Important to note is that in
of a hospital stay where further hospitalization is often both hospitals, only a third of the therapies were stream-
debatable, it should be rigorously evaluated for other lined after results became available. In most cases, this
possible intervention targets, such as an increased is only done in more severe cases and because the
shift of non-CAP related procedures towards ambula- identified microorganism has proved to be resistant to
tory care, which may allow the treatment costs to be the current therapy, urging the need for more broad-
decreased in a substantial manner. A final determining spectrum and thus more expensive antibiotics. In com-
factor may be the Belgian reimbursement system by bination with a delayed discharge this may also explain
which each hospital gets an individually fixed amount our results regarding antibiotic streamlining, in par-
for each admission based on the estimated cost for a ticular at UZL. Such practice could nullify the cost-
particular diagnosis and the local case-mix, in addi- saving potential of correctly performed cultures and also
tion to 25% of the reimbursable drug expenses. How- explains why, apart from the diagnostic costs, these still
ever, for patients who are deemed outliers in terms of generate higher costs as compared to no bacteriological
 Drivers of pneumonia hospital costs evaluated 225
sampling. Similar observations have been made for Our study has some limitations. Firstly, Table I
positive blood cultures elsewhere [15]. This also adds points to differences in patient characteristics,
to the current evidence that sputum cultures usually co-morbidities, therapy characteristics, and quality
have low yields and that a stricter selection of patients indicators between our 2 hospitals. Nevertheless, sep-
eligible for sputum sampling may be necessary [1,16]. arate regression analyses for each hospital (available
Overall, moxifloxacin use proved to be an inde- from the authors upon request) point to the same
pendent driver of lower costs when used in mono- drivers of hospital costs associated with CAP, as iden-
therapy. However, the final impact on total costs was tified in our aggregate analysis reported in this manu-
limited and should not be a reason to completely script. Also, the concordance between the 2 data
abandon β-lactam-based therapies in its favour. extractors was not tested, which may have further
β-Lactam monotherapy proved to be equally effica- added to the variation between the hospitals involved.
cious as moxifloxacin in patients with a low mortality Secondly, our study population may not be entirely
risk score [17]. However, in higher risk groups, mox- representative, as only 14–16% of initially screened
ifloxacin monotherapy proved to be more beneficial patients were included, being those patients with a
in terms of mortality and treatment success. Interest- clear diagnosis and well-defined therapy options and
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

ing to note here is that adding a macrolide to a therefore limiting the generalization of our findings.
β-lactam antibiotic restores its efficacy. Unfortunately, Compared to other studies, patients with a higher
our study was not designed to investigate this. Proper mortality risk are underrepresented in our study,
identification of and restriction to patient subgroups mostly due to the need for informed consent [2,22].
where moxifloxacin has the highest cost-effectiveness As such, for these patients who also have the highest
is therefore advisable [18]. It has been demonstrated expenses, not all cost-determining factors may have
that treatment failure is also a predictor of increased been fully identified. Our data were not corrected for
treatment costs [19]. However, both hospitals have a inflation or other changes in pricing. On the other hand,
training function with supervisors frequently chang- our time-frame was sufficiently short to minimize the
ing residents’ initial antibiotic regimens, making impact of inflation. We used the combination of time
For personal use only.

treatment changes an unreliable factor in our case. to clinical stability and length of stay after clinical stabil-
Nevertheless, the fact that in-hospital mortality and ity as an alternative for total length of stay. Length of
readmission within 30 days are correlated with higher stay is inherently highly dependent on disease severity,
pharmacy costs is consistent with this observation. which may not be completely accounted for, amongst
The virtual absence of an influence of guideline- other factors. Length of stay after clinical stability is
compliant antibiotic treatment can be explained by the based on objective clinical data and as such gives a bet-
specific type of registered deviations. More than half ter representation of the factors that may interfere with
of the non-compliant therapies are due to incorrect hospital discharge. Unfortunately, measured variables
addition or omission of atypical coverage, treatment are unable to fully explain this post-stabilization hospi-
of CAP class IV patients with co-amoxiclav instead of tal stay. More detailed information on the number and
a third-generation cephalosporin as prescribed by type of interventions during this period, either medical
local guidelines, or initial treatment with intravenous or not, and in relation to their acuteness is needed.
moxifloxacin (normally allowed for β-lactam allergic In conclusion, although quality indicators for
patients only), but with rapid conversion to oral ther- CAP treatment are useful for the assessment of anti-
apy (unpublished results). These deviations have only biotic therapy, their potential impact on total hospi-
limited potential to cause therapy failure, major mor- tal costs is limited. The most important driver of
bidities, and higher total costs, limiting the cost impact CAP cost variation was found to be the duration of
of guideline compliance in our study. Also, possible hospitalization after the patient becomes clinically
higher antibiotic costs will only have a marginal impact stable. Although disease severity, patient age, and
on total costs. At the same time, these results do not therapy duration can partly explain this influence,
vindicate neglecting the existing guidelines [20]. other factors, not necessarily related to the CAP
A final observation was the large cost difference diagnosis, may play an important role in prolonging
between the 2 studied hospitals. This is probably the post-stabilization hospital stay and should be
mostly attributable to the different settings – UZL is evaluated for cost containment.
a tertiary care hospital with extended training func-
tion and a different case-mix – and to the accompany-
ing difference in reimbursement in Belgian hospitals.
Acknowledgements
Location itself was shown to have only a limited
impact, with an overall 7.0% total cost increase at The authors wish to acknowledge Ilse Vautmans
UZL, but explaining only about 2.1% of the total cost from ZOL, and Krista D’Hondt from UZL, for pro-
variation, similar to other observations [21]. viding us with the financial data.
 226 P.-J. Cortoos et al.
Declaration of interest: All the authors have no [10] Schouten JA, Hulscher ME, Trap-Liefers J, Akkermans RP,
potential conflicts of interest to declare. Kullberg BJ, Grol RP, et al. Tailored interventions to improve
antibiotic use for lower respiratory tract infections in hospi-
This study was supported by an unconditional tals: a cluster-randomized, controlled trial. Clin Infect Dis
grant from the Flemish Society of Hospital Pharma- 2007;44:931–41.
cists (VZA), Belgium. The study sponsor was not [11] Grömping U. Relative importance for linear regression in R:
involved in any part of this study or manuscript. the package relaimpo. J Stat Softw 2006;17:1.
[12] Bauer TT, Welte T, Ernen C, Schlosser BM, Thate-Waschke I,
de Zeeuw J, et al. Cost analyses of community-acquired
pneumonia from the hospital perspective. Chest 2005;128:
References
2238–46.
[1] Mandell LA, Wunderink RG, Anzueto A, Bartlett JG, [13] Wong A, Elderkamp-de Groot R, Polder J, van Exel J. Predic-
Campbell GD, Dean NC, et al. Infectious Diseases Society tors of long-term care utilization by Dutch hospital patients
of America/American Thoracic Society consensus guidelines aged 65⫹. BMC Health Serv Res 2010;10:110.
on the management of community-acquired pneumonia in [14] Paladino JA, Poretz D. Outpatient parenteral antimicrobial
adults. Clin Infect Dis 2007;44(Suppl 2):S27–72. therapy today. Clin Infect Dis 2010;51(Suppl 2):S198–208.
[2] Welte T, Torres A, Nathwani D. Clinical and economic burden [15] Ramanujam P, Rathlev NK. Blood cultures do not change
of community-acquired pneumonia among adults in Europe. management in hospitalized patients with community-
Scand J Infect Dis Downloaded from informahealthcare.com by Michigan University on 10/25/14

Thorax 2012;67:71–9. acquired pneumonia. Acad Emerg Med 2006;13:740–5.

[3] Reyes S, Martinez R, Valles JM, Cases E, Menendez R. [16] Ewig S, Schlochtermeier M, Goke N, Niederman MS.
Determinants of hospital costs in community-acquired pneu- Applying sputum as a diagnostic tool in pneumonia: limited
monia. Eur Respir J 2008;31:1061–7. yield, minimal impact on treatment decisions. Chest 2002;
[4] Merchant S, Mullins CD, Shih YC. Factors associated with 121:1486–92.
hospitalization costs for patients with community acquired [17] Ewig S, Hecker H, Suttorp N, Marre R, Welte T; CAPNETZ
pneumonia. Clin Ther 2003;25:593–610. study group. Moxifloxacin monotherapy versus ss-lactam
[5] Fine MJ, Auble TE, Yealy DM, Hanusa BH, Weissfeld LA, mono- or combination therapy in hospitalized patients with
Singer DE, et al. A prediction rule to identify low-risk community-acquired pneumonia. J Infect 2011;62:218–25.
patients with community-acquired pneumonia. N Engl J [18] Simoens S. Evidence for moxifloxacin in community-
Med 1997;336:243–50. acquired pneumonia: the impact of pharmaco-economic
[6] Halm EA, Fine MJ, Kapoor WN, Singer DE, Marrie TJ, considerations on guidelines. Curr Med Res Opin 2009;25:
For personal use only.

Siu AL. Instability on hospital discharge and the risk of 2447–57.

adverse outcomes in patients with pneumonia. Arch Intern [19] Ott SR, Hauptmeier BM, Ernen C, Lepper PM, Nuesch E,
Med 2002;162:1278–84. Pletz MW, et al. Treatment failure in pneumonia: impact of
[7] Drummond MF, Becker DL, Hux M, Chancellor JV, antibiotic treatment and cost analysis. Eur Respir J 2012;
Duprat-Lomon I, Kubin R, et al. An economic evaluation of 39:611–8.
sequential i.v./po moxifloxacin therapy compared to i.v./po co- [20] Martinez R, Reyes S, Lorenzo MJ, Menendez R. Impact of
amoxiclav with or without clarithromycin in the treatment of guidelines on outcome: the evidence. Semin Respir Crit Care
community-acquired pneumonia. Chest 2003;124:526–35. Med 2009;30:172–8.
[8] Lloyd A, Holman A, Evers T. A cost-minimisation analysis [21] Kane RL, Bershadsky B, Weinert C, Huntington S,
comparing moxifloxacin with levofloxacin plus ceftriaxone Riley W, Bershadsky J, et al. Estimating the patient care costs
for the treatment of patients with community-acquired pneu- of teaching in a teaching hospital. Am J Med 2005;118:
monia in Germany: results from the MOTIV trial. Curr Med 767–72.
Res Opin 2008;24:1279–84. [22] Ewig S, Birkner N, Strauss R, Schaefer E, Pauletzki J,
[9] Schouten JA, Hulscher ME, Wollersheim H, Braspennning J, Bischoff H, et al. New perspectives on community-acquired
Kullberg BJ, van der Meer JW, et al. Quality of antibiotic use pneumonia in 388 406 patients. Results from a nationwide
for lower respiratory tract infections at hospitals: (how) can mandatory performance measurement programme in health-
we measure it? Clin Infect Dis 2005;41:450–60. care quality. Thorax 2009;64:1062–9.

Supplementary material available online

Supplementary Tables I–IV.