Clinical Nutrition 31 (2012) 728e734

Contents lists available at SciVerse ScienceDirect

Clinical Nutrition
journal homepage: http://www.elsevier.com/locate/clnu

Original article

Is there a difference in bloodstream infections in critically ill patients

associated with ready-to-use versus compounded parenteral nutrition?
Alessandro Pontes-Arruda a, Gary Zaloga b, Paul Wischmeyer c, Robin Turpin d, e, *,
Frank X. Liu d, Catherine Mercaldi f
a
Department of Nutrition and Intensive Care, Fernandes Távora Hospital, Fortaleza, Ceará, Brazil
b
Global Medical Affairs, Baxter Healthcare, Deerfield, IL, USA
c
Department of Anesthesiology, University of Colorado, Denver, CO, USA
d
Global Health Economics, Baxter Healthcare Corporation, Deerfield, IL, USA
e
Public Policy Department, Thomas Jefferson Hospital, Philadelphia, PA, USA
f
United Biosource Corporation, Lexington, MA, USA

a r t i c l e i n f o s u m m a r y

Article history: Background & aims: Parenteral nutrition is widely used in critically ill patients receiving nutritional
Received 11 July 2011 support. Several previous studies associated the use of parenteral nutrition with the development of
Accepted 17 March 2012 bloodstream infections. This study compared bloodstream infections in critical care patients receiving
parenteral nutrition (PN) prepared via conventional compounding versus premixed multichamber bags.
Keywords: Methods: Records in the Premier PerspectiveÔ database for all in patients
18 years of age, with
Parenteral nutrition
a minimum 3-day intensive care unit stay, who received PN between 2005 and 2007 were analyzed
Infection
(n ¼ 15,328). Statistical analysis of data, grouped according to preparation method, compared differences
Bloodstream infection
Sepsis
in both observed bloodstream infection rates and adjusted rates, using logistic regression to examine the
Critical illness impact of hospital and patient baseline characteristics.
Outcomes Results: Patients receiving compounded parenteral nutrition had longer intensive care unit stays (11.3 vs.
9.1 days) and longer hospital stays (22.6 vs. 19.4 days); both P < .001. After adjusting for baseline
differences, the probability for bloodstream infections was 19% higher when using compounded
parenteral nutrition vs. multichamber bags (29.6 vs. 24.9%; odd ratio ¼ 1.29; 95% confidence
interval ¼ 1.06e1.59).
Conclusion: In this retrospective review of a large patient database the adjusted probability of blood-
stream infection was significantly lower in patients receiving multichamber bags than compounded
parenteral nutrition. These findings need to be investigated further in high quality observational studies
and prospective clinical trials.
Ó 2012 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.

1. Introduction
Central venous catheters are commonly used in the critical care
setting for administration of fluids, drugs, blood products and also
parenteral nutrition (PN).1 It is estimated that in the United States
(U.S.) about 3 million central venous catheters (CVC) are used
annually.2 The main complication from a CVC is the development of
bloodstream infections (BSI).3 The in-hospital mortality for adult
* Corresponding author. Global Health Economics, Baxter Healthcare, One Baxter
Parkway, Deerfield, IL 60015, USA. Tel.: þ1 847 948 3689; fax: þ1 847 948 2962.
E-mail addresses: pontes-arruda@secrel.com.br (A. Pontes-Arruda), gary_
zaloga@baxter.com (G. Zaloga), Paul.Wischmeyer@ucdenver.edu (P. Wischmeyer),
robin_turpin@baxter.com (R. Turpin), xiaoqing_liu@baxter.com (F.X. Liu),
Katie.Mercaldi@unitedbiosource.com (C. Mercaldi).
critically ill patients with catheter-related BSI varies from 11% in
France,4,5 12e15% in Germany,6,7 to 17.1% in Italy.8
BSI also imposes an important financial burden, as there are
additional charges associated with prolonged stays in the hospital
and intensive care unit (ICU), drug treatment, medical/surgical
procedures, and substantial indirect costs9 leading to important
economic costs to the healthcare system,10 which are particularly
burdensome among critically ill patients.11 Depending on the
causative microorganism, the increase in hospitalization associated
with nosocomial laboratory-confirmed BSI ranges from 7 to 30
days.4,12,13 with added costs from $3061-$40,000.14
Anywhere from 12% to 71% of critically ill patients receiving
nutritional support are reported to receive PN.15,16 Many studies
since the 1970’s associate the use of PN with the development of
BSI.17e19 In a recent study of ICU patients, Casear found that early

0261-5614/$ e see front matter Ó 2012 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.
doi:10.1016/j.clnu.2012.03.004
 A. Pontes-Arruda et al. / Clinical Nutrition 31 (2012) 728e734
use of PN (in combination with EN) was associated with an
increased risk of infection, compared to the use of EN only.20
Beghetto et al found PN to be an independent risk factor for
catheter-related BSI in nonselected hospitalized adult patients
using a multivariate analysis.19 Published data evaluating the
impact of mutichamber bags (MCB) upon PN-associated BSI are not
available.
The objective of the present study was to perform a large
retrospective evaluation of the claims records associated with PN
prescriptions in critically ill patients in the U.S. in order to verify if
different methods of PN preparation affect the incidence of BSI.
Other outcomes in PN patients, such as time of hospitalization and
time in the ICU, were also analyzed to determine the economic
value of PN sourced via multichamber bags.
2. Materials and methods
2.1. Data source
Infectious complications associated with parenteral nutrition
were explored using claims data from a representative group of
hospitals in the U.S. managed by Premier Healthcare Alliance. The
Premier PerspectiveÔ database is the largest inpatient clinical and
financial data warehouse in the U.S., containing more than 45 million
discharges from acute care facilities. This database is publically
available and a copy of the database was obtained with the necessary
license to analyze the data with authorization to report the results
through scientific publications. Since this is an epidemiological
study involving only the collection and evaluation of publically
available data, this is allowable without approval by an ethics
committee. Data records include patients’ complete billing and
coding history, with procedures, medications, laboratory, diagnostic
and therapeutic services detailed at the individual level, but do not
include any patient identification information. Further details about
the database, overall population, hospital characteristics, and
methods of analysis have been published by Mercaldi et al.21
2.2. Study population
All patients included in this retrospective evaluation fulfilled the
following entry criteria: must be an inpatient with at least 18 years
of age, and who received parenteral nutrition between January 1,
2005 and December 31, 2007. A quality filter was added as previ-
ously suggested22 to be sure that only true critically ill patients
were included in this analysis. Therefore, all included patients
spent at least 3 days at the ICU. Excluded were all patients admitted
with any infection, hepatic dysfunction, hypoglycemia, acute
cholecystitis, cholestasis, renal failure, phlebitis or thrombophle-
bitis, since these diagnoses were expected to skew the data based
on higher mortality. However, patients who developed these
conditions during their admission were included in the analyses.
Patients received PN that was prepared via two-different
methods: premixed ready-to-use MCB or compounded admix-
tures. Compounded PN is prepared by mixing multiple source
solutions together to meet the prescribed nutrient formula. This
complex mixing process is typically performed by a pharmacy
following strict guidelines for handling sterile products. PN com-
pounding is often done in a hospital pharmacy, but may also be
done at a centralized location for multiple hospitals or be out-
sourced to another pharmacy. Premixed ready-to-use formulations
consisted of a dual-chamber bag containing glucose and amino
acids (MCB) that may have included additions of minerals, vita-
mins, electrolytes and/or lipid emulsion. Compounded admixtures
were prepared either in a hospital or off-site (outsourced) at
a centralized compounding facility. Outcomes were compared
729
across the two treatment groups (MCB vs. compounded PN) for all
patients in the database, with a separate subgroup analysis for
those patients where it was certain that they received PN com-
pounded at their hospital only.
2.3. Outcomes
The primary outcome of this analysis was the development of
bloodstream infection. Patients with infections were identified
using International Classification of the Diseases, Version 9 (ICD-9)
codes in accordance with the method previously described.23e25
BSIs were defined by the occurrence of a charge for any intrave-
nous (IV) antibiotic and any of the following ICD-9 codes: 038 x
(septicemia); 790.7 (bacteremia); 999.31 (infection due to central
venous catheter); or 996.62 (infection and inflammatory reaction
due to vascular device, implant and graft). The presence of a BSI was
confirmed in a sensitivity analysis by verifying that codes for IV
antibiotics needed to treat the BSI were also documented at the
same time.21
Secondary clinical outcomes included additional adverse events
potentially related to PN administration, including hypoglycemia,
hyperglycemia, hepatic dysfunction, acute cholecystitis, phlebitis or
thrombophlebitis of upper extremities, pulmonary embolism, and
in-hospital mortality (determined by reported discharge status).
The characteristics of patients who were administered parenteral
nutrition were also collected in terms of number of days using PN,
daily PN volume administered, additional use of nutrients (such as
additional amino acids, dextrose, lipids, minerals and electrolytes),
and total number of intravenous lines.
2.4. Analysis
Baseline demographics, comorbidities, and hospitalization
characteristics were reported descriptively for each treatment
group using means with standard deviations, using medians with
ranges for continuous variables, and by raw counts with percent-
ages for categorical measures. Since number of catheter days could
not be reliably tracked in the database, this differs from the typical
reporting of BSI incidence, usually in number of events per 1000
catheter days. All Patient Refined Diagnostic Reported Group (APR-
DRG) classifications were used to compare severity of illness since
neither APACHE, SOFA or SAP scores were available in the data set.
Differences were reported via chi-square for categorical measures
and t-tests for continuous variables. Differences for each level of
those categorical measures were reported via Z-test for two
proportions. Observed incidence rates for outcomes were reported
by treatment group. A P-value < 0.05 was considered statistically
significant. All data were analyzed using Statistical Analysis Soft-
ware (SAS Version 9.1, Cary, NC).
2.5. Infectious complications model
Multivariate logistic regression was used to create a model to
estimate the risk of infection for patients receiving MCB vs. com-
pounded PN. The model adjusted for risk factors and potential
confounders available in the database, including hospital variables
(size, geographic region, teaching status, and urban/rural location);
patient demographics (age, sex, ethnicity, payer), patient comor-
bidities and other indicators of acuity (malnutrition, intestinal
malabsorption, acute pancreatitis, peritonitis, gastrointestinal
fistula, Crohn’s disease, cancers, cirrhosis or chronic liver failure,
renal failure, diabetes, tuberculosis, emergency or urgent admis-
sion, surgical status, prior surgery or prior PN) as well as PN
treatment variables (length of PN treatment, PN volume, number of
intravenous lines). There were very few missing values, with the
730 A. Pontes-Arruda et al. / Clinical Nutrition 31 (2012) 728e734
largest category hospital disposition (1.6%); all other missing values
were less than 0.5%.
The initial model contained all independent variables listed
above. Variable-selection procedures were aimed at selecting
a reduced set of independent variables providing the best fit of the
model. The final model was determined using backward elimina-
tion variable selection, with a significance level of 0.10 for the Wald
chi-square statistic to specify the covariates that would remain in
each step. Model fit was assessed using the C-statistic, which
corresponds to the area under a receiver operating characteristic
(ROC) curve. The ROC curve indicated the probability that randomly
selected patients with the event have a higher predicted probability
of the event than randomly selected patients without the event.
Risk of infection was reported as adjusted probability of the event
and odds ratio with 95% confidence level.
BSI-attributable LOS was estimated using multivariate ordinary
least squares regression models constructed to measure the impact
of infection events on LOS after controlling for important baseline
and clinical patient characteristics.
3. Results
3.1. Patients and baseline characteristics
A total of 11,193,887 in patients claims records were evaluated in
the Premier PerspectiveÔ Database during the defined 3-year
period, out of which 112,845 records were positively associated
with the use of PN during hospitalization. There were 37,390
records of PN use in pediatric and neonatal patients eliminated
from the data set. Of the remaining adult records, only 15,328 adult
patients stayed in the ICU for at least 3 days and were included in
the analysis.
The baseline characteristics of the patients can be found in
Table 1. Patients receiving MCB formulations were older than
patients who received compounded PN. The use of MCB was more
commonly observed in smaller sized hospitals (50-249 beds) than
compounded PN (60.3% vs. 10.8%, P < .0001), in non-teaching
hospitals (80.6% vs. 55.2%, P < .0001), and in hospitals located in
the rural areas (34.7% vs. 8.5%, P < .0001). The majority of the
patients in the compounded PN group (77.9%) were receiving care
at mid-size (250-499 beds) to large-size hospitals (500-749 beds).
No differences between the two groups were observed in terms of
in-hospital mortality (19.4% for MCB vs. 18.1% for compounded PN,
P > .1). More patients in the compounded PN group were discharged
to home as compared to those receiving MCB (42.0% vs. 33.3%,
P < .0001). No differences in home discharge were found between
the MCB and hospital compounded subgroup (33.3% vs. 38.5%, P >.1).
More patients in the MCB group were transferred to other institu-
tions when compared to either the compounded PN or hospital
compounded groups (14.8% vs. 10.6% or 4.9%, P < .05 for both).
The incidence of diabetes was higher in the MCB group (26.7%
vs. 22.6% for compounded PN, P < .05). No other differences were
observed for most of the major comorbidities evaluated, whether
comparing patients receiving MCB with those in the compounded
PN group or hospital compounded PN subgroup, except for GI
fistula which was higher in patients receiving hospital com-
pounded PN (1.4% vs. 0.1% for MCB, P < .05), but was not different in
patients receiving MCB vs. compounded PN (0.1% vs. 0.4%, P > .05).
3.2. Characteristics of the parenteral nutrition
The characteristics of the PN are depicted in Table 2. Patients in
the MCB group used PN for fewer days when compared with either
the compounded PN group (5.9  5.2.0 vs. 9.0  9.3 days, P < .0001)
or the hospital compounding subgroup (9.9  10.6 days, P < .0001).
The mean daily volume of PN was higher in the MCB group when
compared with the compounded PN group (2241.6  966 vs.
2098  1445 mL, P < .05), but the lowest observed mean daily
volume was found in the hospital compounded PN subgroup
(1152  333 mL, P < .0001 when compared with the MCB).
The exception was the additional use of dextrose, which was
higher in the compounded PN group (80.1% vs. 57.5% for the MCB,
P < .0001). The total number of intravenous lines including both
central and peripheral catheters was higher in the compounded
group (5.0  7.8 vs. 3.3  5.8 for the MCB, P < .0001).
3.3. Incidence of bloodstream infections
The incidence of any infection associated with an antibiotic
billing charge and ICD-9 code was higher in the compounded PN
group when compared to the MCB group (n ¼ 8699/60.0% vs.
n ¼ 444/54.3%, P ¼ .001); the incidence of infection was highest in
the hospital compounded PN subgroup (n ¼ 185/64.7%, P < .002).
After adjustment, the incidence of all post-admission infections
remained significantly lower in the MCB group when compared
with the compounded PN group (55.9% vs. 60.0%; OR ¼ 1.21, 95%
CI ¼ 1.03e1.44), but no different between the MCB group and
hospital compounded PN subgroup (56.5% vs. 59.1%; OR ¼ 1.13, 95%
CI ¼ 0.81e1.59).
The observed and adjusted results for the primary outcome, the
incidence of bloodstream infection, can be found in Fig. 1. When
considering the observed BSI rates defined by the ICD-9 code plus
IV antibiotic charges, the incidence was lower in the MCB group
compared to either the compounded PN group or hospital com-
pounded PN group (21.9% MCB group [n ¼ 179] vs. 29.2% [n ¼ 4235],
or 38.8% [n ¼ 111]; P < .0001 for both comparisons) (Fig. 1a). After
adjustment using multivariate logistic regression, the probability of
BSI remained higher for the compounded PN group when
compared with the MCB group (29.6% vs. 24.9%; OR ¼ 1.29, 95%
CI ¼ 1.06e1.59) (Fig. 1b). The same effect was found when
comparing the probability of BSI in the hospital compounded PN
subgroup versus the MCB group (33.1% vs. 23.4%; OR ¼ 1.69, 95%
CI ¼ 1.20e2.38) (Fig. 1c). The within group differences in BSI rates
between adjusted and non-adjusted results are accounted for by
the different variables used in the logistic regression that accounted
for risk factors and potential confounders.
3.4. Hospital days and ICU-days
The hospital compounded PN subgroup had the highest mean
duration of hospitalization (Fig. 2). The median hospital stay was
also significantly higher for the hospital compounded PN subgroup
(23 days; range-4e316 days, P < .0001 compared to MCB), as well as
for the compounded PN group (18 days; range ¼ 3e366 days,
P < .0001 compared to MCB) when compared with the MCB group
(16 days; range ¼ 3e142 days).
The hospital compounded subgroup also had the longest mean
ICU stay (Fig. 3). The median observed ICU stay was 8 days
(range ¼ 3e250 days) for all compounded PN, 11 days
(range ¼ 3e100 days) for the hospital compounded PN subgroup,
and 6 days (range ¼ 3e61 days) for the MCB group (P < .0001 for
hospital compounded PN vs. MCB and for compounded PN vs. MCB).
A separate evaluation was performed to evaluate the number of
hospitalization days and ICU-days after BSI. The highest hospitali-
zation and ICU stay was found in the hospital compounded PN
subgroup, followed by the compounded PN group using either the
mean or the median values. The length of stay attributable to BSI
was also calculated and represents an additional 8.1 days in the
hospital (P < .0001).
 A. Pontes-Arruda et al. / Clinical Nutrition 31 (2012) 728e734 731

Table 1
Characteristics of PN patients, comorbid conditions, facilities, and disposition.

Characteristic MCB (n ¼ 818) Compounded Hospital compounded P-value for MCB vs.
PN (n ¼ 14,510) PN (n ¼ 286)
Compounded PN Hospital
compounded PN
Age, years
Mean  SD 68.3  15.7 64.5  17.3 59.3  17.2 <0.0001 <0.0001
Median (range) 71.5 (18.0e89.0) 68.0 (18.0e89.0) 62.0 (18.0e89.0)
Age categorized in years, n (%)
18e34 28 (3.4) 1071 (7.4) 32 (11.2) <0.0001 <0.0001
35e54 124 (15.2) 2664 (18.4) 65 (22.7) 0.0209 0.0034
55e74 318 (38.9) 5765 (39.7) 133 (46.5) 0.6264 0.0239
þ75 348 (42.5) 5010 (34.5) 56 (19.6) <0.0001 <0.0001
Gender, n (%)
Male 378 (46.2) 7136 (49.2) 159 (55.6) 0.0983 0.0063
Female 440 (53.8) 7374 (50.8) 127 (44.4) 0.0983 0.0063
Hospital Size, n (%)
50e249 beds 493 (60.3) 1570 (10.8) 5 (1.7) <0.0001 <0.0001
250e499 beds 286 (35.0) 6607 (45.5) 280 (97.9) <0.0001 <0.0001
500e749 beds 3 (0.4) 4708 (32.4) 1 (0.3) <0.0001 0.967
750e999 beds 23 (2.8) 1430 (9.9) 0 (0.0) <0.0001 e
þ1000 beds 0 (0.0) 161 (1.1) 0 (0.0) e e
Unknown 13 (1.6) 34 (0.2) 0 (0.0) <0.0001 e
Teaching hospital, n (%)
No 659 (80.6) 8004 (55.2) 23 (8.0) <0.0001 <0.0001
Yes 159 (19.4) 6506 (44.8) 263 (92.0) <0.0001 <0.0001
Hospital location, n (%)
Urban 534 (65.3) 13,274 (91.5) 24 (8.4) <0.0001 <0.0001
Rural 284 (34.7) 1236 (8.5) 262 (91.6) <0.0001 <0.0001
Hospital disposition, n (%)
Died 159 (19.4) 2620 (18.1) 61 (21.3) 0.3185 0.4907
Home 272 (33.3) 6093 (42.0) 110 (38.5) <0.0001 0.1109
Rehab/SNF 241 (29.5) 3926 (27.1) 100 (35.0) 0.1325 0.083
Transferred 121 (14.8) 1543 (10.6) 14 (4.9) 0.0002 <0.0001
Other/Unknown 25 (3.1) 328 (2.3) 1 (0.3) 0.1399 0.0094
Surgical patients, n (%)
No 79 (9.7) 709 (4.9) 29 (10.1) <0.0001 0.8132
Yes 739 (90.3) 13,801 (95.1) 257 (89.9) <0.0001 0.8132
Prior use of PN, n (%)
No 784 (95.8) 14,178 (97.7) 280 (97.9) 0.0007 0.1088
Yes 34 (4.2) 332 (2.3) 6 (2.1) 0.0007 0.1088
APR-DRG Severity Class
Minor, n (%) 13 (1.6) 143 (1.0) 6 (2.1) 0.0942 0.5691
Moderate, n (%) 79 (9.7) 1043 (7.2) 14 (4.9) 0.0083 0.0126
Major, n (%) 285 (34.8) 4320 (29.8) 53 (18.5) 0.0021 <0.0001
Extreme, n (%) 441 (53.9) 9003 (62.0) 213 (74.5) <0.0001 <0.0001
Comorbidities, n (%)
Undernutrition 238 (29.1) 4566 (31.5) 94 (32.9) 0.1547 0.2312
Intestinal malabsortion 8 (1.0) 139 (1.0) 2 (0.7) 0.9544 0.6685
Acute pancreatitis 47 (5.7) 754 (5.2) 12 (4.2) 0.4922 0.3158
Peritonitis 83 (10.1) 1638 (11.3) 35 (12.2) 0.3141 0.3245
GI Fistula 1 (0.1) 52 (0.4) 4 (1.4) 0.263 0.0057
Crohns disease 6 (0.7) 210 (1.4) 2 (0.7) 0.092 0.9532
Cancer 289 (35.3) 5034 (34.7) 119 (41.6) 0.7098 0.0583
Diabetes 218 (26.7) 3274 (22.6) 49 (17.1) 0.0067 0.0012
Tuberculosis 0 (0.0) 22 (0.2) 1 (0.3) 0.2651 0.0907

MCB: ready-to-use, multichamber bag; PN: parenteral nutrition; SD: standard deviation; GI: gastrointestinal.

4. Discussion use the gastrointestinal tract or intolerance of enteral nutrients. The

current critical care guidelines of the European Society for Clinical
PN represents an important alternative or supplement to EN Nutrition and Metabolism (ESPEN)26 recommend that PN, when
when there is a deficiency in meeting the nutritional requirements indicated, should be initiated within 24e48 h of ICU admission. The
in critically ill patients via the enteral route as a result of inability to current ASPEN/SCCM critical care guidelines27 recommend that PN

Table 2
Characteristics of parenteral nutrition and lines.

Characteristic MCB (n ¼ 818) Compounded PN (n ¼ 14,510) Hospital compounded PN (n ¼ 286) P-value for MCB vs.

Compounded PN Hospital compounded PN

Days of PN
Mean  SD 5.9  5.2 9.0  9.3 9.9  10.6 <0.0001 <0.0001
Median (range) 4.0 (1.0e58.0) 7.0 (1.0e171.0) 7.0 (1.0e74.0)
PN Volume, mL per day, Mean  SD 2241  966 2098  1445 1152  333 0.0061 <0.0001
Total number of lines, Mean  SD 3.3  5.8 5.0  7.8 3.2  6.1 <0.0001 0.7619

MCB: Ready to use, multichamber bag; PN: parenteral nutrition; SD: standard deviation; GI: gastrointestinal.
732 A. Pontes-Arruda et al. / Clinical Nutrition 31 (2012) 728e734

be initiated in malnourished patients in whom EN is not feasible as

soon as possible following admission and adequate resuscitation.
These recommendations are based on evidence that the use of PN
does not increase mortality when appropriately prescribed in
comparison to EN. Nevertheless, studies have reported that the use
of PN was significantly associated with the occurrence of BSI.
Beghetto et al conducted a concurrent cohort study to compare BSI
rates among patients using central venous catheters with and
without PN and found PN to be independently associated with the
incidence of BSI.19 Casaer et al found that critical care patients who
received early PN in combination with EN experienced a 7.5% rate
for BSI, compared to the 6.1% rate of the EN-only group (p < .05).20
Moreover, a recent meta-analysis supports the recommendation for
PN use in patients in whom EN could not be initiated with a grade
Bþ evidence-based rating, even when considering that PN was
associated with a statistically significant increase in infectious
complications (OR ¼ 1.66).28 With a possible increase in PN usage
due to new guidelines and the association of this nutrition therapy
with BSI, more evidence is needed to evaluate possible strategies to
reduce PN-associated BSI. This study observed that the rate of BSI is
lower when using MCB compared to compounded PN. Confirma-
tion of these results will require a prospective randomized clinical
trial.
An association between PN preparation and BSI does not
establish a cause and effect relationship, but instead is hypothesis-
generating for further research studies. In this study, the evaluation
of the BSI rates associated with two PN preparation methods
indicated that the incidence of BSI was significantly higher in
patients receiving compounded PN when compared with those
individuals receiving MCB. Although the rates of BSI reported here
are higher than expected, it is important to keep in mind that these
patients are not only critically ill but are also receiving PN. A cohort
study in 19 Canadian hospitals found that patients who received PN
had 4 times as many BSIs as those who did not receive any PN.29
The reason that hospital compounded PN was associated with
higher BSI than the entire compounded PN group is unclear and
represents an interesting opportunity for further research. One
possible reason may be contamination with either bacteria and/or
toxins (ie. endotoxins) during preparation of PN in the pharmacy. A
simulation study of PN compounding in a hospital pharmacy sug-
gested a contamination rate of 5.2% during preparation.30 Other
explanations for the differences may include: storage time after
mixing, frequency of catheter access, use of appropriate barrier
precautions and bundles, introduction of non-sterile room air
during preparation or administration, minimal number of additives
to MCB, terminal sterilization of MCB preparations, and underlying
severity of illness. Better aseptic techniques and hand-washing has
been suggested to help reduce bacterial infections.31 Future studies
will need to be designed to determine potential etiologies. It is also
possible that differences in the patient populations that were not
controlled in these data (i.e. length of time in the ICU) resulted in
the observed differences. Future trials should be designed to
confirm the findings from this study and to evaluate possible
etiologies for the differences in BSI rates.
Although this retrospective database analysis has several limi-
tations, its greatest value is the ability to reflect the real-world use

Fig. 1. a: Observed incidence of bloodstream infections associated with different

parenteral nutrition delivery systems. b: Adjusted incidence of BSI comparing MCB
versus compounding. OR ¼ 1.29, 95% CI ¼ 1.06e1.59 variables in the final model
included: age group, gender, geographic region, admission type, year of admission,
hospital size, surgery, days of PN, additional dextrose, additional mineral/electrolyte,
cancer, malnutrition, peritonitis. c: Adjusted incidence of BSI comparing MCB versus
hospital compounding subgroup. OR ¼ 1.69, 95% CI ¼ 1.20e2.38 variables in the final
model included: days of PN, transfer status, malnutrition, peritonitis.
 A. Pontes-Arruda et al. / Clinical Nutrition 31 (2012) 728e734
Fig. 2. Total hospital stay associated with different parenteral nutrition delivery
systems.
of PN in the U.S. The retrospective nature of the study, and subse-
quent inability to randomize, limits the capacity to determine
a causal relationship between PN type and infection rates.
Furthermore, the baseline variable adjustment provided by logistic
regression could only include variables measured in the database.
The models that were used in this study can adjust infection rates
for numerous variables that impact them, but this adjustment
cannot guarantee that all sources of variance were identified.
Although data validation was not conducted through a chart
review, we expect inaccuracies to be random. We did conduct
a series of sensitivity analyses, including propensity score
modeling, to improve the balance between groups and help
account for any channeling of patients with more severe diseases
into one group over another. These results have been reported
previously.21 Nonetheless, the large population size provides the
opportunity to raise these issues and guide further study into this
relationship.
Fig. 3. Total ICU stay associated with different parenteral nutrition delivery systems.
733
This analysis identified BSI via ICD-9 codes and use of IV anti-
biotics, but this was not validated with patient records. However, it
is likely that any error in identifying infections in such a large
population was random across both treatment groups. Because the
recording of charges for antibiotics was used, this may underesti-
mate the true infection rate. Furthermore, it is generally accepted
practice to report infection rates as BSI/catheter days. Since the
presence of IV catheters was not reliably identified, this study
cannot generate a BSI/catheter days report.
For most patients receiving compounded PN, it was not possible
to identify whether this was prepared at an outsource facility or in
the hospital. Therefore, the hospital compounded PN subgroup
represents seven hospitals (n ¼ 286 patients) that were known to
compound PN on-site, but it is likely that additional hospitals in the
larger group compounded some of their PN on-site as well. Given
the differences in BSI rates between the total compounded PN and
hospital compounded PN groups, it is possible that the hospital
group contributed disproportionately to the overall rate, but we are
unable to determine this impact.
Finally, our population was limited to patients in the U.S. who
received PN, so generalization is limited to this patient group. Differ-
ences between aseptic compounding standards and methods in the
U.S. compared to those used in the rest of the world may limit the
extrapolation of these findings to other regions of the world. The
nutritional status of patients and use of supplemental nutrition outside
of PN (ie. dual feeding) is not accounted for in these data. This may be
an area for further research to determine if the observed difference
may be related to nutritional needs that could not be met with MCB.
Conflict of interest
The authors acknowledge that Baxter Healthcare is the maker of
CLINIMIX, a two-chamber bag parenteral nutrition product mar-
keted and sold in the United States where this study was con-
ducted. Catherine Mercaldi, Alessandro Pontes-Arruda, and Paul
Wischmeyer are paid consultants of Baxter Healthcare Corporation.
Robin Turpin, Gary Zaloga, and Frank X Liu are employees and
shareholders of Baxter Healthcare Corporation.
Statement of authorship
All authors have made substantial contributions to all of the
following: (1) the conception and design of the study, or acquisition
of data, or analysis and interpretation of data, (2) drafting the article
or revising it critically for important intellectual content, (3) final
approval of the version submitted. The specific contribution of each
author was as follows:
- APA carried out interpretation of data, and revising it critically
for important intellectual content.
- GZ carried out analysis and interpretation of data, and critically
revising the manuscript for important intellectual content.
- PW carried out analysis and interpretation of data, and criti-
cally revising the manuscript for important intellectual
content.
- RT carried out conception, study design, acquisition of data,
interpretation of data, drafting and revising the manuscript.
- FL carried out conception, study design, acquisition of data,
drafting and revising the manuscript.
- CM carried out conception and design of the study, or acqui-
sition of data, or analysis and interpretation of data, and
revising it critically for important intellectual content.
B The IMPROVE Study Group was involved in interpretation of
data. This group included:Todd W. Canada, PharmD, BCNSP,
FASHP
734 A. Pontes-Arruda et al. / Clinical Nutrition 31 (2012) 728e734
B Gordon L. Jensen, MD, PhD
B Alessandro Pontes-Arrunda, MD, MSc, PhD, FCCM
B Victor D. Rosenthal, MD, CIC, MSc
B Paul Wischmeyer, MD
B Fouad Amer, MD, MPH
B Cyrus Banan, PhD
B Mary Hise, PhD, RD
B Monica Marlowe, MD, FAAP
B Carol Schermer, MD, MPH
B Shane Scott, PharmD, BCPS, BCOP
B Robin Turpin, PhD
B Gary Zaloga, MD, FACP, FCCP, FACN, FCCM
B Matthew Reynolds, PhD
B Catherine Mercaldi, MPH
B Diane Nitzki-George, PharmD
All contributors who do not meet the criteria for authorship as
defined above are listed in the Acknowledgements section.
Acknowledgements
This study was possible thanks to a funding from Baxter
Healthcare Corporation. We want to thank Diane Nitzki-George,
Pharm D for her medical writing assistance and the other
members of the IMPROVE Study group for their perspective and
guidance.
References
1. Raad I, Hanna H, Maki D. Intravascular catheter-related infections: advances in
diagnosis, prevention, and management. Lancet Infect Dis 2007;7:645e57.
2. Hockenhull JC, Dwan KM, Smith GW, Gamble CL, Boland A, Walley TJ, et al.
The clinical effectiveness and cost-effectiveness of central venous catheters
treated with anti-infective agents in preventing bloodstream infections:
a systematic review and economic evaluation. Health Technol Assess 2008;
12:1e154.
3. Edgeworth J. Intravascular catheter infections. J Hosp Infect 2009;73:323e30.
4. Soufir L, Timsit JF, Mahe C, Carlet J, Regnier B, Chevret S. Attributable mortality
and mortality of catheter-related septicemia in critically ill patients:
a matched, risk-adjusted, cohort study. Infect Control Hosp Epidemiol 1999;20:
396e401.
5. Renaud B, Brun-Buisson C. ICU-Bacteremia Study Group: outcomes of primary
and catheter-related bactermia. A cohort and case-control study in critically ill
patients. Am J Respir Crit Care Med 2001;163:1584e90.
6. Gastmeier P, Weist K, Ruden H. Catheter-associated primary bloodstream
infections: epidemiology and preventive methods. Infection 1999;27(Suppl.1):
S1e6.
7. Gastmeier P, Sohr D, Geffers C, Zuschneid I, Behnke M, Rüden H. Mortality in
German intensive care units: dying from or with a nosocomial infection?
Anasthesiol Intensivmed Notfallmed Schmerzther 2005;40:267e72.
8. Endimiani A, Tamborini A, Luzzaro F, Lombardi G, Toniolo A. A two-year analysis
of risk factors and outcome in patients with bloodstream infection. Jpn J Infec Dis
2003;56:1e7.
9. Blot SI, Depuydt P, Annemans L, Benoit D, Hoste E, De Waele JJ, et al. Clinical
and economic outcomes in critically ill patients with nosocomial catheter-
related bloodstream infections. Clin Infect Dis 2005;41:1591e8.
10. Orsi GB, Di Stefano L, Noah N. Hospital-acquired, laboratory-confirmed
bloodstream infection: increased hospital stay and direct costs. Infect Control
Hosp Epidemiol 2002;23:190e7.
11. Karchmer AW. Bloodstream infections: the problem and the challenge. Int J
Antimicrob Agents 2009;34S:S2e4.
12. Digiovine B, Chenoweth C, Watts C, Higgins M. The attributable mortality and
costs of primary nosocomial bloodstream infections in the intensive care unit.
Am J Respir Crit Care Med 1999;160:976e81.
13. Pittet D, Tarara D, Wenzel RP. Nosocomial bloodstream infection in critically ill
patients: excess length of stay, extra costs and attributable mortality. JAMA
1994;271:1598e601.
14. Jarvis WR. Selected aspects of the socialeconomic impact of nosocomial
infections: morbidity, mortality, cost and prevention. Infect Control Hosp Epi-
demiol 1996;17:552e7.
15. Hill SA, Nielsen MS, Lennard-Jones JE. Nutrition support in intensive care units
in England and Wales: a survey. Eur J Clin Nutr 1995;49:371e8.
16. Garvin CG, Brown RO. Nutritional support in the intensive care unit: are
patients receiving what is prescribed? Crit Care Med 2001;29:204e5.
17. Ryan Jr JA, Abel RM, Abbott WM, Hopkins CC, Chesney TM, Colley R, et al.
Catheter complications in total parenteral nutrition: a prospective study of 200
consecutive patients. N Engl J Med 1974;290:757e61.
18. Opilla M. Epidemiology of bloodstream infection associated with parenteral
nutrition. Am J Infect Control 2008;36:S173.e5e8.
19. Beghetto MG, Victorino J, Teixeira L, de Azevedo MJ. Parenteral nutrition as
a risk factor for central venous catheter-related infection. J Parenter Enteral
Nutr 2005;29:367e73.
20. Casaer MP, Mesotten D, Hermans G, Wouters PJ, Schetz M, Meyfroidt G, et al.
Early versus late parenteral nutrition in critically ill adults. N Engl J Med 2011
Aug 11;365(6):506e17.
21. Mercaldi CJ, Reynolds MW, Turpin RS. Methods to identify and compare
parenteral nutrition administered from hospital compounded and premixed
multi-chamber bags in a retrospective hospital claims database. J Parenter
Enteral Nutr 2011 Jul 12 [Epub ahead of print].
22. Doig GS, Simpson F, Delaney A. A review of the true methodological quality of
nutritional support trials conducted in the critically ill: time for improvement.
Anesth Analg 2005;100:527e33.
23. Patkar NM, Curtis JR, Teng GG, Allison JJ, Saag M, Martin C, et al. Administrative
codes combined with medical records based criteria accurately identified
bacterial infections among rheumathoid arthritis patients. J Clin Epidemiol
2009;62:321e7.
24. MacLaren R, Bond CA, Martin SJ, Fike D. Clinical and economic outcomes of
involving pharmacists in the direct care of critically ill patients with infections.
Crit Care Med 2008;36:3184e9.
25. Danai PA, Sinha S, Moss M, Haber MJ, Martin GS. Seasonal variation in the
epidemiology of sepsis. Crit Care Med 2007;35:410e5.
26. Singer P, Berger MM, Van den Berghe G, Biolo G, Calder P, Forbes A, et al. ESPEN
guidelines on parenteral nutrition: intensive care. Clin Nutr 2009;28:387e400.
27. McClave SA, Martindale RG, Vanek VW, McCarthy M, Roberts P, Taylor B, et al.
Guidelines for the provision and assessment of nutrition support therapy in the
adult crticially ill patient: society of Critical Care Medicine (SCCM) and
American Society for Parenteral and Enteral Nutrition (A.S.P.E.N. J Parenter
Enteral Nutr 2009;33:277e316.
28. Simpson F, Doig GS. Parenteral vs. Enteral nutrition in the critically ill patient:
a meta-analysis of trials using the intention to treat principle. Intensive Care
Med 2005;31:12e23.
29. Holton D, Paton S, Conly J, Embree J, Taylor G, Thompson W. Central venous
catheter-associated bloodstream infections occurring in Canadian intensive
care units: a six-month cohort study. Can J Infec Dis Med Microbiol 2006;
17:169e76.
30. Trissel LA, Gentempo JA, Anderson RW, Lajeunesse JD. Using a medium-fill
simulation to evaluate the microbial contamination rate for USP medium-
risk-level compounding. Am J Health Syst Pharm 2005;62:285e8.
31. Kumpf VJ, Mirtallo JM, Petersen C. Parenteral nutrition formulations: prepa-
ration and ordering. In: Merritt R, editor. The A.S.P.E.N. Nutrition support practice
Manual. 2nd ed. Silver Springs, MD: American Society for Parenteral and
Enteral Nutrition; 2005. p. 102.