Journal of Critical Care 30 (2015) 438.e1438.

e5

Contents lists available at ScienceDirect

Journal of Critical Care

journal homepage: www.jccjournal.org

Reevaluation of the utilization of arterial blood gas analysis in the

Intensive Care Unit: Effects on patient safety and patient outcome
Franziska E. Blum, MD a,, Elisa Takalo Lund, MS a, Heather A. Hall, MD b, Allan D. Tachauer, MD a,
Edgar G. Chedrawy, MD, MSc c,d, Jeffrey Zilberstein, MD a,e
a

Department of Internal Medicine, Weiss Memorial Hospital, afliate of the University of Illinois, Chicago, IL
Department of Vascular Surgery, Weiss Memorial Hospital, Chicago, IL
Department of Cardiovascular and Thoracic Surgery, Weiss Memorial Hospital, Chicago, IL
d
Department of Cardiothoracic Surgery, University of IllinoisChicago at Chicago, Chicago, IL
e
Department of Medicine, Section of Pulmonary, Critical Care, Sleep and Allergy, University of IllinoisChicago at Chicago, Chicago, IL
b
c

a r t i c l e

i n f o

a b s t r a c t

Keywords:
Arterial blood gas
Ventilator days
Routine testing
Unnecessary testing
Intensive care unit

Purpose: Arterial blood gas (ABG) analysis is a useful tool to evaluate hypercapnia in the context of conditions and
diseases affecting the lungs. Oftentimes, indications for ABG analysis are broad and nonspecic and lead to
frequent testing without test results inuencing patient management.
Materials and methods: Electronic charts of 300 intensive care unit (ICU) patients at a single institution were
reviewed retrospectively. Reassessment of indications for ABGs led to a decrease of the number of ABGs in the
ICU between March and November 2012. Data relating to ventilator days, length of stay, number of reintubations,
mortality, complications after arterial puncture, demographics, and medications in 159 ICU patients
between December 2011 and February 2012 (group 1) were compared with 141 ICU patients between
December 2012 and February 2013 (group 2). Subgroup analysis in ventilated patients was performed.
Results: A decrease of number of ABGs per patient (6.12 5.9, group 1 vs 2.03 1.66, group 2 in ventilated
patients; P = .007) was found along with a decrease in the number of ventilator days per patient (P = .004)
and a shorter length of stay for ventilated patients in group 2 compared with group 1 (P = .04).
Conclusion: A signicant decrease of ABGs obtained in the ICU does not negatively impact patient outcome
and safety. A decrease in the number of ABGs per patient allows cost-efcient patient care with a lower risk
for complications.
2014 Elsevier Inc. All rights reserved.

1. Introduction

that indications for ABG analysis should be based on the clinical assessment of the patient.
Arterial puncture for ABG analysis is an invasive procedure; and
potential complications include occlusion of the artery, digital
embolization leading to digital ischemia, sepsis, local infection,
pseudoaneurysm, hematoma, bleeding, and skin necrosis [5]. Arterial
blood gas samples are frequently obtained for reasons such as change in
ventilator settings, a respiratory or cardiac event, and as routine testing
[3]. In 2007, Melanson et al [3] determined the utilization of ABG analysis
in a tertiary care hospital by having physicians and nurses ll out a utilization survey inquiring about the level of training of the ordering clinician, reason for ordering ABG, and the effect of the results on patient
management. The study showed that 79% of ABG test results were expected; a change in patient management based on the ABG results occurred in
42% of cases; and ABG analysis was frequently performed on a routine
basis or to assess parameters, which can potentially be assessed clinically
or through other measures, such as capnometry [3].
Arterial blood gas analysis is a costly intervention and can lead to
serious complications for the patient [3,4]. The current literature does
not sufciently reect if a cost-efcient utilization of ABG analysis
through signicant reduction of the number of ABG samples affects

Respiratory depression is a potentially lethal condition and has

received much attention in the literature [1]. Hypercapnia is a direct
indicator of respiratory depression, and arterial blood gas (ABG)
analysis is an accurate and reliable tool to evaluate hypercapnia in the
context of respiratory diseases and conditions affecting the lungs [2].
Most ABG samples are obtained in the Intensive Care Unit (ICU). Ideally,
an ABG sample should be obtained, when the results are highly likely to
inuence patient management [3]. Common indications for ABG sample
are the need to evaluate the adequacy of patient ventilation, the need to
quantify the response to therapeutic or diagnostic interventions, monitoring of severity and progression of documented disease process, and
the assessment of acid base status [4]. The current literature suggests

Conict of interest disclosure: The authors declare that they have no nancial and
other conict of interest related to the submitted manuscript.
Corresponding author at: Department of Internal Medicine, Weiss Memorial Hospital,
4646 North Marine Drive, Chicago, IL 60640. Tel.: +1 7735647400; fax: +1 7735645226.
E-mail address: fblum@weisshospital.com (F.E. Blum).
http://dx.doi.org/10.1016/j.jcrc.2014.10.025
0883-9441/ 2014 Elsevier Inc. All rights reserved.

438.e2

F.E. Blum et al. / Journal of Critical Care 30 (2015) 438.e1438.e5

patient outcome and patient safety. This study determines the effect of
reconsideration of the indications for ABG analysis, on patient outcome
and safety.
2. Methods
2.1. Data collection
The study was conducted at Weiss Memorial Hospital, an academic
teaching hospital and afliate of the University of Illinois at Chicago,
with a 16-bed multidisciplinary ICU. A total of 300 patients were included in this retrospective data review. With the goal to provide excellent
yet cost-efcient patient care, the indications to obtain an ABG sample
in the ICU (including, for example, change in ventilator settings,
respiratory or cardiac event, routine testing, metabolic event,
postintubation and postextubation as well as preintubation and
preextubation, follow-up on abnormal test results, unreliable pulse
oximetry data, and altered mental status) were reevaluated based on
an evidence-based review of the literature between March and November 2012. This change in the ICU model included intensivist-led team
discussion between attending physicians, resident physicians, and nursing staff during rounds, assessing the indication to obtain an ABG for
each individual patient and individual clinical situations based on the
question if the results from an ABG analysis would lead to a change in
patient management. The decision to obtain an ABG sample was made
based on the assessment of the patient rather than routine daily ABG
sampling, which included physical examination; ventilator parameters;
and the awakening, breathing, coordination, delirium screening, and exercise/early mobility assessment [6]. Before this change in practice was
introduced in daily patient care, ABG analysis was commonly ordered
by single health care providers with various levels of experience as a
matter of routine and without an intensivist-led team assessment of
the indication for ABG analysis based on the question if test results are
likely to lead to change in patient management. To determine the effect
of this measure on patient outcome and safety, we conducted a retrospective data review for the period between December 2011 and February 2012 (group 1) and between December 2012 and February 2013
(group 2). We included the number of ABG samples obtained in the
ICU; number of ventilator days; number of reintubations; length of
stay (LOS) in the ICU; 30-day mortality after admission to the ICU;

medications including anesthetics and opiate-derived analgesics (alprazolam, clonazepam, chlordiazepoxide, diazepam, hydromorphone,
lorazepam, morphine, methadone, oxycodone, tramadol, fentanyl, midazolam, propofol, and remifentanil); readmissions to the ICU within the
periods mentioned above; complications from arterial puncture; and
demographic data including age, sex, Body Mass Index (BMI) as well
as cardiac and pulmonary comorbidities. Ventilator days and LOS in
the ICU were dened as primary outcome factors. Number of
reintubations, 30-day mortality, and complications after arterial puncture were secondary outcome measures. Subgroup analysis was performed in ventilated patients only (66 vs 60 patients in group 1 and
group 2, respectively). Data were extracted from Horizon Physician Portal (McKesson Corporation, Chicago, IL) and MIDAS (version 8.1.4;
MidasPlus Inc, Tuscon, AZ).
2.2. Statistical analysis
Statistical analysis was performed using SPSS version 21 (IBM Corp,
Armonk, NY) and Microsoft Excel 2010 (Redmond, WA). After assessment of the normality of distribution of data collected with the
Kolmogorov-Smirnov test, Mann-Whitney U test was applied to analyze
the differences of ventilator days, LOS in the ICU, reintubation rates, and
medications. The t test was used to analyze patient age and BMI; and
Fisher exact test was applied to assess sex, 30-day mortality, cardiac
and pulmonary comorbidities, and regression analysis; and Pearson
product correlation was performed. Data are presented as mean SD.
This study was approved by the Institutional Review Board at
Vanguard Health Chicago Institutional Review Board/Tenet Health
Care. Waiver of consent was obtained.
3. Results
A total of 300 patients were included in the study (159 in group 1 and
141 in group 2). Sixty-six patients in group 1 and 60 patients in group 2
were ventilated in the ICU. The number of ABG samples obtained per patient was lower in group 2 (all patients, 3.7 3.7; ventilated patients, 2.03
1.66) (Figs. 1 and 2) compared with group 1 (all patients, 5.5 4.7;
ventilated patients, 6.12 5.9) (all patients, P b .001; ventilated patients,
P b .001) (Figs. 2 and 3) (Table 1). A decrease of ABGs of more than 60%
per patient was observed for ventilated patients.

Fig. 1. Count of the number of ABG analysis per patient for all patients in group 1 and group 2.

F.E. Blum et al. / Journal of Critical Care 30 (2015) 438.e1438.e5

438.e3

Fig. 2. Count of the number of ABG analysis per patient in group 1 and group 2, in ventilated patients only.

The number of ventilator days per patient differed signicantly

between groups (group 1, 6.46 5.04 vs group 2, 3.7 2.61; P =
.004) (Mann-Whitney U test) (Fig. 3) (Table 1). Patients had a significantly shorter LOS in the ICU in group 2 compared with group 1
(ventilated patients: group 1, 8 5.6 vs group 2, 6 5.1; P =
.036) (all patients: group 1, 6 4.7 vs group 2, 5 4.7 P = .02)
(Mann-Whitney U test), whereas the LOS for all patients did not differ between groups (group 1, 6 4.7 vs group 2, 5 4.7; P = .09)
(Mann-Whitney U test) (Table 1) (Fig. 3). The total dose of propofol
was signicantly higher in group 2 compared with group 1 in all patients as well as in ventilated patients only (group 1 ventilated patients, 102.78 303.63 vs group 2, 235.02 449.64; P = .012)
(group 1 all patients, 48.75 210.56 vs group 2, 31.04 148.47;
P = .013) (Mann-Whitney U test). The total dose of lorazepam was
signicantly lower in group 2 compared with group 1 in all patients
and in ventilated patients only (group 1 ventilated patients, 81.59
440.38 vs group 2, 5.70 30.29; P = .008) (group 1 all patients,
33.56 281.78 vs group 2, 0.212 1.69; P = .022) (Mann-Whitney

U test). Pearson product correlation reported a positive correlation

between the LOS and ventilator days (Correlation Coefcient (CC)
0.79, P b .001) as well as LOS and ABGs per patient (CC 0.532, P b
.001). Furthermore, a positive correlation was reported between
ventilator days and ABGs per patient (CC, 0.522 P b .001). Both
groups had similar demographics (Table 2), reintubation rates, complications from arterial puncture, and mortality at 30 days (Fisher
exact test). Four patients in group 1 and 1 patient in group 2 were
readmitted to the ICU between December 2011 and February 2012
and between December 2012 and February 2013, respectively.
Body mass index was signicantly higher in group 2 compared
with group 1, and a signicant difference in BMI was found as well
in a subgroup analysis of ventilated patients between group 1 and
group 2 (group 1, 27 8.5 vs group 2, 29 8.9; P = .044) (all patients: group 1, 25 7.1 vs group 2, 30 8.5 in ventilated patients;
P = .007) (t test) (Table 2).
The total dose of alprazolam, clonazepam, chlordiazepoxide, diazepam, hydromorphone, morphine, methadone, oxycodone,
tramadol, fentanyl, midazolam, and remifentanil did not differ between groups and subgroups of ventilated patients (P N .05 each)
(Mann-Whitney U test).
4. Discussion
Our data indicate that improving costs by virtue of decreasing ABG
utilization, based on the reassessment of indications to obtain ABG
Table 1
Data primary outcome
Variable

Fig. 3. Intensive care unit LOS in days, number of ABG analysis, and ventilator days for ventilated patients between December 2011 and February 2013.

No. of ABG analysis

Ventilated patients
All patients
ICU LOS
Ventilated patients
All patients
Vent days
Data are presented as mean SD.

Group 1

Group 2

6.12 5.9
5.5 4.7

2.03 1.66
3.7 3.7

b.001
b.001

8 5.6
6 4.7
6.46 5.04

6 5.1
5 4.7
3.7 2.61

.04
.09
.004

438.e4

F.E. Blum et al. / Journal of Critical Care 30 (2015) 438.e1438.e5

Table 2
Demographic characteristics
Variable

Group 1

Group 2

No. of
patients (n)
Sex (male/female)
All
Ventilated

159 (66 patients

on ventilator)

141 (60 patients

on ventilator)

N/A

77/86
30/36

82/60
33/27

.08
1.00

64 16.9
64 14.2

61 17.9
65 13.8

.082
.903

BMI (kilograms per square meter)

All
27 8.5
Ventilated
25 7.1

29 8.9
30 8.5

.044
.007

93
44

76
33

.565
.264

Pulmonary comorbidities (n)

All
60
Ventilated
30

51
27

.905
1.000

Age
All
Ventilated

Cardiac comorbidities (n)

All
Ventilated

Data presented as mean SD.

samples is safe and does not negatively impact patient outcome. A decrease of ABG utilization was correlated with a decrease in ventilator
days and LOS in the ICU, allowing further improvement of costs for patient care as well as improvement of safety and quality.

4.1. Reduction of unnecessary ABG analysis in the ICU

A signicant reduction of the number of ABGs of up to more than 60%
per patient in the ICU does not negatively impact patient outcome or patient safety. The utilization of unnecessary testing has been addressed in
the literature and most recently in the Choosing Wisely campaign [7,8].
However, the impact of a signicant reduction of ABG analysis on patient
safety and outcome in a closed ICU model has not been addressed yet.
In 1997, Pilon et al [9] suggested indications to obtain ABG samples
including, for example, a change in ventilator settings.
Since this study has been published, no further evaluation of indications for ABG analysis has been clearly reported. Furthermore, the validity of these indications is questionable in current practice, as a change in
ventilator setting, for example, is not considered an absolute indication
for ABG analysis and is controversial in the current literature as well
as in routine clinical practice. Furthermore, the change of Positive EndExpiratory Pressure was found to have no signicant inuence of ABG
analysis results [10].
Other studies published also raise concern about commonly used indications for ABG analysis.
Pawson and DePriest [11], for example, found that ABG measurement
does not necessarily need to precede extubation after a clinically successful spontaneous breathing trial. This nding was further supported by a
study published by Salam et al [12] in 2003, showing that ABG analysis
results do not change extubation decisions in more than 90% of cases.
Furthermore, the role of pulse oximetry and capnometry as an alternative for ABG analysis has to be considered [13]. Expiratory carbon dioxide, for example, was found to correlate well with PaCO2 [13]; the
expiratory carbon dioxidePaCO2 gradient in patients without compromise in gas exchange, such as an underlying pulmonary pathology, for
example, is reported as 5 mm Hg [13]. More research is needed to establish specic guidelines, in which cases, pulse oximetry and capnometry
can be used to replace conventional ABG analysis. Despite previous
data suggesting that indications for ABG analysis should be reevaluated,
indications to obtain ABG samples should be based on clinical assessment and only if they are likely to change patient management [7,9].
Melanson et al [3] have shown in 2007 that clinical routine is still a

common indication for ABG analysis, and 79% of ABG analysis results
were as expected. In keeping with the new Society of Critical Care Medicine Choosing Wisely campaign, furthermore, one of the efforts is
based in avoiding routine daily laboratory sampling [8].
These ndings strongly indicate the need to reassess commonly
used indications for ABG analysis and to dene solid clinical guidelines for indications for this test. Our data suggest that a signicant
reduction in ABG analysis is safe and does not negatively impact patient outcome, and it may reduce the likelihood of potential complications from arterial puncture. Our study was not primarily
designed to detect complications from arterial puncture and compare them between groups; however, there were no documented
complications for either group. Further research needs to be done
on this particular topic with a larger patient population and a
prospective approach on the assessment of complications of arterial
puncture, including minor complications such as vasospasm, which
is difcult to assess in retrospect. The number of pulmonary and
cardiac comorbidities did not differ among patients in group 1 and
group 2, indicating that a decrease of ABG analysis is safe and does
not impair outcome in this high-risk patient population [14].
The observed signicant decrease of the LOS in the ICU in ventilated
patients and in group 2 compared with group 1 as well as the signicant
decrease of ventilator days in group 2 compared with group 1 is most
likely a multifactorial nding. Our results show that there is a signicant
correlation between LOS in the ICU and ventilator days, but at the same
time, the number of ABGs per patient was positively correlated with
ventilator days and LOS over time. This correlation in particular may reect the indirect effect of the combination of measures taken to decrease the number of ABG samples obtained per patient as part of an
intensivist-led closed ICU model including changes of sedative medication use. For example, the use of propofol instead of benzodiazepines for
sedation has been shown to improve patient outcome by reducing the
LOS in the ICU and the number of ventilator days [15].
Wang et al [7] reported in a study published in 2002 that education
of physicians and nursing staff has been shown to reduce frequent testing, such as frequent ABG analysis. Our approach mainly falls into this
category, as the application of a close clinical assessment of the patient
in combination with team discussions about the indication to obtain
an ABG in each particular case resulted in a decrease sampling of ABGs
of up to more than 60% per patient, which in retrospect may indicate
much of it prior was unnecessary. Hence, teamwork and frequent
assessment of the patient at the bedside in an intensivist-led closed
ICU model seems to be the most effective way to reduce unnecessary
ABG analysis. Other techniques to reduce frequent tests described in
the literature are electronic alerts for redundant testing in the computer
ordering system and changing in funding for tests. Furthermore, it can
be suggested that a feedback on change in ordering ABGs might
positively enforce ordering behavior. Solomon et al [16] have shown
that a combination of interventions to reduce unnecessary testing is
more effective compared with a single measure.
Given that the presence of an arterial line has been described as the
most important predictor of the number of ABG samples obtained from
the patient [17], it is clear that attention must be paid to the utility of
testing rather than the ease with which a sample can be obtained. The
indication to obtain ABG analysis based on the clinical assessment of
the patient may vary depending on the health care provider's experience and level of training; therefore, more research is needed to develop
solid guidelines and training programs for physicians and nursing staff
to establish an equal standard of care for ABG analysis and to increase
the awareness of costs and potential consequences of unnecessary testing for the patient. Despite the large size of study population, limitations
of our assessment were conduction of the study at a single center and
retrospective data review. Future multicenter studies are needed to develop new solid clinical practice guidelines leading to appropriate ABG
sampling in the ICU to assure patient safety, further improve patient
outcome, and to simultaneously provide cost-efcient patient care.

F.E. Blum et al. / Journal of Critical Care 30 (2015) 438.e1438.e5

5. Conclusion
Arterial blood gas analysis is oftentimes ordered as a routine, and
ABG results do not change patient management in most cases. This results in increased risk for the patient and increased costs for patient care.
A signicant decrease of the number of ABGs obtained in the ICU
does not negatively affect patient safety, reduces the risk for complications, and likely lowers the overall costs for patient care. A decrease of
the number of ABGs can be done safely in a high-risk patient population.
Further research is needed to develop guidelines for appropriate indications for ABG analysis.
Acknowledgment
The authors would like to thank Kevin Baxter, RN, BS (senior data
analyst, Weiss Memorial Hospital), and Elda Pourshahbaz (technical director, Vascular Lab, Vascular Institute of Chicago at Weiss Memorial
Hospital), for excellent assistance in data acquisition and assessment
of data availability.
References
[1] http://www.apsf.org/announcements.php?id=13.
[2] Kelly AM, Klim S. Agreement between arterial and venous pH and pCO2 in patients
undergoing non-invasive ventilation in the emergency department. Emerg Med
Australas 2013;25(3):2036.
[3] Melanson SE, Szymanski T, Rogers SO, Jarolim P, Frendl G, Rawn JD, et al. Utilization
of arterial blood gas measurements in a large tertiary care hospital. Am J Clin Pathol
2007;127(4):6049.

438.e5

[4] AARC clinical practice guideline. Sampling for arterial blood gas analysis. American
Association for Respiratory Care. Respir Care 1992;37(8):9137.
[5] Brzezinski M, Luisetti T, London MJ. Radial artery cannulation: a comprehensive review of recent anatomic and physiologic investigations. Anesth Analg 2009;109(6):
176381.
[6] Balas MC, Vasilevskis EE, Burke WJ, Boehm L, Pun BT, Olsen KM, et al. Critical care
nurses' role in implementing the "ABCDE bundle" into practice. Crit Care Nurse
2012;32(2):358 [40-7].
[7] Wang TJ, Mort EA, Nordberg P, Chang Y, Cadigan ME, Mylott L, et al. A utilization
management intervention to reduce unnecessary testing in the coronary care unit.
Arch Intern Med 2002;162(16):188590.
[8] www.sccm.org/News/Pages/Choosing-Wisely.aspx.
[9] Pilon CS, Leathley M, London R, McLean S, Phang PT, Priestley R, et al. Practice guideline for arterial blood gas measurement in the intensive care unit decreases numbers
and increases appropriateness of tests. Crit Care Med 1997;25(8):130813.
[10] Liou CM, Lin CH, Kang HM, Liu YC, Tso HS. The inuence of position and PEEP on
arterial blood gas during operation. Ma Zui Xue Za Zhi 1993;31(2):10312.
[11] Pawson SR, DePriest JL. Are blood gases necessary in mechanically ventilated
patients who have successfully completed a spontaneous breathing trial? Respir
Care 2004;49(11):13169.
[12] Salam A, Smina M, Gada P, Tilluckdharry L, Upadya A, Amoateng-Adjepong Y, et al.
The effect of arterial blood gas values on extubation decisions. Respir Care 2003;
48(11):10337.
[13] Thawley V, Waddell LS. Pulse oximetry and capnometry. Top Companion Anim Med
2013;28(3):1248.
[14] Qureshi A, Ballard RD. Obstructive sleep apnea. J Allergy Clin Immunol 2003;112(4):
64351.
[15] Fraser GL, Devlin JW, Worby CP, Alhazzani W, Barr J, Dasta JF, et al. Benzodiazepine
versus nonbenzodiazepine-based sedation for mechanically ventilated, critically ill
adults: a systematic review and meta-analysis of randomized trials. Crit Care Med
2013;41(9 Suppl. 1):S308.
[16] Solomon DH, Hashimoto H, Daltroy L, Liang MH. Techniques to improve physicians'
use of diagnostic tests: a new conceptual framework. JAMA 1998;280(23):20207.
[17] Muakkassa FF, Rutledge R, Fakhry SM, Meyer AA, Sheldon GF. ABGs and arterial
lines: the relationship to unnecessarily drawn arterial blood gas samples. J Trauma
1990;30(9):108793.