Escolar Documentos
Profissional Documentos
Cultura Documentos
of pages: 6; 4C:
Clinical Biochemistry xxx (2015) xxxxxx
Clinical Biochemistry
journal homepage: www.elsevier.com/locate/clinbiochem
a r t i c l e
i n f o
Article history:
Received 27 February 2015
Received in revised form 31 March 2015
Accepted 2 April 2015
Available online xxxx
Keywords:
APACHE
RDW
ICU scoring systems
a b s t r a c t
Background: The aim of this study was to evaluate whether the addition of red blood cell distribution width
(RDW) improves the prognostic value of current intensive care unit (ICU) scoring systems, namely APACHE III.
Design and Methods: All patients admitted to a mixed ICU in Brisbane between June 2013 and July 2014 for
whom RDW was available were included in the study. Analyses included descriptive statistics, linear regression
correlation, and receiver operating characteristic (ROC) curves.
Results: The study included 708 patients for whom both ICU mortality prediction and RDW were available. In
univariate analysis higher RDW values were associated with increased hospital mortality. Adding RDW to
APACHE III increased the area under the ROC marginally (from 0.9586 to 0.9613). RDW was not correlated
with C-reactive protein, white cell count, or patient's length of stay in ICU.
Conclusion: RDW was an independent predictor of mortality. The addition of RDW to APACHE III improved
its mortality prediction marginally. The underlying mechanism of RDW elevation warrants further investigation.
2015 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Background
ICU scoring systems use physiological parameters to predict patient
mortality. A small number of studies have validated the usefulness of including RDW in ICU scoring systems owing to its prognostic value. A summary of the key issues and ndings of current studies is provided below.
ICU scoring systems
ICU scoring systems are a means of quantifying the severity of illness
in critical patients [1]. Scoring systems, though not typically used clinically in the decision to withhold or withdraw treatment, are essential
in improvement of clinical decisions [2,3]. Their uses include: predicting
patient outcome; monitoring treatment efcacy; optimizing the use of
resources; and comparing performance within clinical trials and across
different centres [3,4]. The Acute Physiologic and Chronic Health Evaluation (APACHE) is one of the most commonly used ICU scoring systems,
predicting mortality based upon data collected on the day of ICU admission [4].
APACHE was the rst scoring system proposed to predict mortality
[5]. Developed in the 1980s with data from two American ICUs, it
included 33 physiological variables in addition to an assessment of
chronically ill health [5,6]. Subsequent revisions have been made since
its inception to improve APACHE's mortality prediction [7]:
Corresponding author at: Suite 201, level 2, 8 Herbert Street, St Leonards, Sydney
Australia 2065.
E-mail address: tony.badrick@rcpaqap.com.au (T. Badrick).
http://dx.doi.org/10.1016/j.clinbiochem.2015.04.002
0009-9120/ 2015 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Please cite this article as: Loveday S, et al, Does the addition of RDW improve current ICU scoring systems?, Clin Biochem (2015), http://
dx.doi.org/10.1016/j.clinbiochem.2015.04.002
Table 3
Statistical signicance of RDW in predicting mortality.
Study
Statistical signicance
[13]
[8]
Cigarette smoking
Nutritional deciencies
Obesity
Oxidative stress
Physical inactivity/sedentary lifestyle
Increased circulating immature red blood cells
Impaired iron metabolism
Increased blood viscosity and other thrombotic mechanisms
Suppressed bone marrow function
Haemolysis
Abnormal haemoglobin variants
Neurohormonal and endocrine system activation
Diminished capacity for systemic repair, recovery and defense
Tissue hypoxia/hypoxemia
Physiologic impairments associated with ageing
Chronic hyperglycemia
Shorter telomere length
[4]
[15]
[9]
[14]
Zhang et al. [14] and Hunziker et al. [15] whereas Wang et al. [4]
excluded patients if they had a history of recent blood transfusion
(less than 2 weeks); Meynaar et al. [9] excluded patients who had
received a blood transfusion in the preceding three months; and Bazick
et al. [8] included patients who had received transfusion 48 h before
critical care initiation. Exclusion of patients with recent transfusion is
an important consideration when investigating whether the addition
of RDW improves ICU scoring systems, as this parameter is not typically
included in the calculation. RDW may be increased by transfusions [9],
which may subsequently skew the relationship between RDW and mortality if blood transfusion is not an independent predictor of mortality
[17], or strengthen it if patients who receive transfusions have an
increased mortality rate [18,19]. The investigation of populations
stratied by transfusion history is warranted to validate whether such
a phenomenon exists.
Different patient populations also exist between studies because of
the patient types within the different hospital settings. Hunziker et al.
[15] included all adult patients admitted to ICU including medical, cardiac, surgical, and cardiothoracic; Meynaar et al. [9] did not include cardiac surgery and neurosurgery patients; and Wang et al. [4] did not
include any surgery patients at all. This is an important consideration
as heterogeneity may lower the discriminating power of RDW, hence
further studies are warranted to investigate whether analysis restricted
to certain subgroups can improve the diagnostic performance of RDW
[14]. Future studies in the improvement of ICU scoring systems by the
addition of RDW must include all types of ICU patients to reect the
utility of current scoring systems.
Differences also exist in the type of data recorded and the adjustments made in each study. Whilst all studies recorded age, sex and
Table 2
Current study ndings linking RDW to mortality.
Table 4
Adjustments in multivariate analysis.
Study Adjustments
[8]
[15]
[9]
[14]
DeyoCharlson index, age, sex, race, coronary artery bypass grafting, myocardial
infarction, haematocrit, WBC, MCV, BUN, sepsis, creatinine
SAPS, age, gender, haematocrit, different comorbidities
APACHE II score, age, admission type, mechanical ventilation
DeyoCharlson Index
Table 5
Patient inclusion and admittance data.
Study
[13]
[8]
[4]
[15]
[9]
[14]
702
51,413
602
17,922
2915
1539
Please cite this article as: Loveday S, et al, Does the addition of RDW improve current ICU scoring systems?, Clin Biochem (2015), http://
dx.doi.org/10.1016/j.clinbiochem.2015.04.002
Table 6
Population characteristics.
Summary statistics
Total population
Survived ICU
Died in ICU
Gender
1058
Patients
650
Male
408
Female
133
Ward
855
OT/Recovery
42
Accident & Emergency
28
Other Hospital
275
Cardiovascular
154
Gastrointestinal
27
Gynaecological
5
Haematological
17
Metabolic
60
Musculoskeletal/skin
209
Neurological
6
Other medical disorder
42
Renal/Genitourinary
200
Respiratory
57
Sepsis
5
Trauma
1
Undened/Unknown
66.1 (15.5)
2.4 (3.5)
13.3 (5.6)
0.1450 (0.1600)
45.0 (18.6)
25.2 (12.3)
0.1070 (0.1483)
14.5 (2.0)
1031
Patients
635
Male
396
Female
115
Ward
850
OT/Recovery
40
Accident & Emergency
26
Other Hospital
269
Cardiovascular
154
Gastrointestinal
27
Gynaecological
3
Haematological
17
Metabolic
60
Musculoskeletal/Skin
207
Neurological
6
Other medical disorder
41
Renal/Genitourinary
198
Respiratory
44
Sepsis
4
Trauma
1
Undened/Unknown
66.0 (15.4)
2.3 (3.4)
12.9 (5.1)
0.1320 (0.1368)#
43.6 (16.1)
24.4 4.4 (11.0)
0.0943 (0.1200)
14.5 (2.0)##
27
15
12
18
5
2
2
6
2
2
1
2
13
1
Source
ICU Diagnosis
Age
Length of ICU stay
APACHE II
APACHE II ROD
APACHE III
SAPS II
SAPS II ROD
RDW
Patients
Male
Female
Ward
OT/Recovery
Accident & Emergency
Other Hospital
Cardiovascular
Haematological
Neurological
Renal/Genitourinary
Respiratory
Sepsis
Trauma
71.4 (17.8)
4.3 (4.1)
27.1 (7.1)
0.6014 (0.2367)^
99.7 (26.7)
57.1 (16.7)
0.5913 (0.2718)
16.7 (2.2)^^
907 patients.
708 patients.
#
882 patients.
##
599 patients.
^
25 patients.
^^
19 patients.
conclusions can be drawn. For example, Hunziker et al. [15] noted that
their patient-history lacked transfusion information hence they were
unable to correct for this factor. Potential bias due to lack of information
can be overcome, in part, by using multivariate analysis, as in Zhang
et al. [14], but ultimately gaps in patient data limit the accuracy and
application of the models developed.
Missing information
Retrospective observational studies utilize previously recorded data.
Consequently, data necessary to understand the pathophysiology
underlying RDW differences (e.g. reticulocyte count, iron studies, transfusion history) may be lacking for all or a proportion of patients [9,15].
Lack of information limits what adjustments can be made and what
Table 7
RDW as a predictor of mortality in all ICU patients.
Table 8
RDW as a predictor of mortality in ICU patients with normal MCVs.
Parameter
Number
Hospital mortality
Parameter
Number
Hospital mortality
RDW b13.1
RDW 13.113.9
RDW 13.915.3
RDW 15.3
161
164
188
195
0 (0.0%)
1 (0.6%)
3 (1.6%)
15 (7.7%)
RDW b13.1
RDW 13.113.9
RDW 13.914.9
RDW 14.9
120
122
110
151
0 (0.0%)
1 (0.8%)
2 (1.8%)
10 (6.6%)
Please cite this article as: Loveday S, et al, Does the addition of RDW improve current ICU scoring systems?, Clin Biochem (2015), http://
dx.doi.org/10.1016/j.clinbiochem.2015.04.002
Brisbane, would improve its mortality prediction. Deidentied laboratory and ICU data was gathered for all patients admitted to ICU from July
1st 2013 to June 30th 2014. The ICU provided data for all 1058 patients
admitted in this time period with measurements taken within the rst
24 h after patients were admitted to ICU. A single laboratory provided
data for 989 of these patients, including measurements within the rst
24 h after patients were admitted to ICU and measurements taken up
until the patients left ICU. The datasets were merged using date of
birth and gender as common identiers, as all ICU patients were determined to have unique dates of birth with respect to their ICU number,
and patients were excluded if mortality prediction or RDW was not
available. This resulted in a total of 708 patients for whom both ICU
mortality prediction and RDW were available. A summary of the patient
population is provided in Table 6.
Whilst ICU calculates APACHE II, APACHE III and SAPS II scores, only
the most current scoring system, APACHE III, is investigated in this
study.
To adjust for patients who may have an elevated RDW due to recent
transfusion, nutritional deciencies, haemoglobinopathies, etc., a population subset was created to minimize confounding effects. This subset
included only those patients with a normal mean cell volume (MCV),
between 85 and 95 inclusive to exclude those patients with an elevated
RDW due to B12/folate deciency or iron deciency anaemia.
Univariate analysis, using quartiles, was undertaken to assess the
prognostic power of RDW and RWD + APACHE III score. Descriptive
statistics were calculated online (http://www.calculatorsoup.com/
calculators/ statistics/descriptivestatistics.php) with counts determined
using Microsoft Excel 2010.
To assess the accuracy of the APACHE III score in predicting mortality, the AUROC was calculated. Similarly, the accuracy of RDW to predict
mortality and the accuracy of RDW + APACHE III score to predict
mortality was assessed by calculating the AUROC. ROCs were determined by calculating the sensitivity and specicity at every threshold
(708 data points for the entire population, 503 data points for the normal population); AUROCs were calculated using the trapezoidal rule.
(Calculations performed in Microsoft Ofce 2010 from rst principles.)
Finally, to investigate the cause of RDW elevation, RDW and white
cell count (WCC), and RDW and C-reactive protein (CRP), were correlated. All cell counts and derived parameters were performed on the same
analyser for the entire period (Sysmex XN-1000, Sysmex Corporation,
Japan). To assess the prognostic value of RDW on length of stay (LOS),
RDW and LOS were correlated. A further investigation in how RDW
changes over a patient's LOS was also included for those patients who
died in ICU, for whom serial RDW measurement was obtained, compared
to similarly matched patients who survived ICU. Graphical data generated
in Microsoft Excel 2010.
Results
Univariate analysis, using quartiles, indicated that as RDW increased
so too did mortality for all ICU patients (Table 7) and for those with
normal MCVs (Table 8). Similarly, as RDW + APACHE III score
increased, so too did mortality in both patient groups (Tables 9 and 10),
suggesting both RDW and RDW + APACHE III score are good predictors
of mortality.
The AUROC analysis revealed that RDW is a better prognostic predictor of mortality in patients with a normal MCV and that the addition of
Table 9
RDW + APACHE III score as a predictor of mortality in all ICU patients.
Parameter
Number
Hospital mortality
176
177
177
178
0 (0.0%)
1 (0.6%)
0 (0.0%)
18 (10.1%)
Table 10
RDW + APACHE III score as a predictor of mortality in ICU patients with normal MCVs.
Parameter
Number
Hospital mortality
125
124
130
124
0 (0.0%)
1 (0.8%)
0 (0.0%)
12 (9.7%)
Discussion
The results are consistent with previous studies that have shown
that mortality is greater when RDW is elevated. Univariate analysis
also revealed that mortality increases when RDW + APACHE III score
increases.
The addition of RDW to APACHE III score increased mortality prediction marginally as quantied by AUROC analysis. Interestingly, RDW
was found to be a better prognostic predictor of mortality in patients
with a normal MCV than in the entire population, whereas APACHE
III and RDW + APACHE III score were better predictors in the entire
population when compared to the normal population.
The lack of correlation between RDW and WCC, and RDW and CRP
is consistent with the ndings of Maynaar et al. [9], supporting the conclusion that RDW is not a result of inammation. It is important to note
that CRP was only available for 86 patients, all of whom were survivors,
and some of whom had multiple serial CRP/RDW paired scores included
to give the 142 data points analysed. A possible association between
RDW and sepsis warrants further investigation as it was found to be
the most prevalent diagnosis of patients who died in ICU.
The nding of Zhang et al. [14] that RDW and LOS are correlated was
not supported. Neither the entire population nor the subset of survivors
in the total population showed a correlation between RDW and LOS.
Zhang et al. [14] had also reported that repeated measurements of
RDW provide no additional prognostic value in critically ill patients.
Quantifying how RDW changes over a patient's LOS revealed the greatest
rates of change, as either an increase or decrease, occurred in survivors,
potentially providing additional prognostic value and warranting further
investigation.
The strengths of the study are that differences in parameter measurement, by other laboratories or point of care devices, is minimized
as data from only one laboratory was used and patients without RDW
recorded in the laboratory data were excluded. Results from the laboratory are valid as blood samples are processed immediately on site before
any signicant changes in MCV, and hence RDW, occur.
Table 11
Improvement in APACHE III ROC with addition of RDW.
Parameter
RDW
APACHE III
RDW + APACHE III
AUROC
All patients
Normal MCV
0.8114
0.9586
0.9613
0.8124
0.9443
0.9466
Please cite this article as: Loveday S, et al, Does the addition of RDW improve current ICU scoring systems?, Clin Biochem (2015), http://
dx.doi.org/10.1016/j.clinbiochem.2015.04.002
if/when an acute change occurs. Thus the usefulness of RDW could also
be extended further to predict ICU admission. Additionally, parameters
not utilized in the APACHE scoring system warrant further investigation
to understand the pathophysiology underlying RDW differences and to
identify other potential indicators of mortality.
Conclusion
RDW was found to be an independent predictor of mortality in ICU
patients. The addition of RDW to APACHE III improves its mortality
prediction marginally. No correlation was found between RDW and
WCC, CRP or LOS. Further investigations are warranted to understand
the cause of elevated RDW and how RDW changes over hospital and
ICU LOS.
Acknowledgements
The authors would like to thank Dr Ranald Pasoce, Director of
Intensive Care, The Wesley Hospital, Brisbane.
Please cite this article as: Loveday S, et al, Does the addition of RDW improve current ICU scoring systems?, Clin Biochem (2015), http://
dx.doi.org/10.1016/j.clinbiochem.2015.04.002
References
[1] Baruch M, Messer B. Criteria for intensive care unit admission and severity of illness.
Surgery (Oxford) 2012;30:22531.
[2] Juffermans NP. Predicting mortality in the critically ill: a tricky enterprise. Neth J
Med 2013;71:447.
[3] Rapsang AG, Shyam DC. Scoring systems in the intensive care unit: a compendium.
Indian J Crit Care Med 2014;18:2208.
[4] Wang F, Pan W, Pan S, Ge J, Wang S, Chen M. Red cell distribution width as a novel
predictor of mortality in ICU patients. Ann Med 2011;43:406.
[5] Chatterji C, Raj A. Intensive care unit scoring systems. Apollo Med 2011;8:8993.
[6] Williams C, Wheeler D. Criteria for ICU admission and severity of illness scoring.
Surgery (Oxford) 2009;27(5):2016.
[7] Cerner Corporation. The APACHE IV equations: benchmarks for mortality and
resource use [white paper]. Missouri: Cerner Corporation; 2005.
[8] Bazick HS, Chang D, Mahadevappa K, Gibbons FK, Christopher KB. Red cell
distribution width and all-cause mortality in critically ill patients. Crit Care
Med 2011;39:191321.
[9] Meynaar IA, Knook AHM, Coolen S, Le H, Bos MMEM, Van Der Dijs F, et al. Red cell
distribution width as predictor for mortality in critically ill patients. Neth J Med
2013;71:48893.
[10] Yang X, Du B. Red cell distribution width: the crystal ball in the hands of intensivists?
Thorac Dis 2014;6:645.
[11] Hunziker S, Stevens J, Howell MD. Red cell distribution width and mortality in newly
hospitalized patients. Am J Med 2012;125:28391.
[12] Patel A, Brett SJ. Identifying future risk from routine tests? Crit Care Med 2014;42:
9991000.
[13] Topeli A, Pepedil F, Ozen G, Er E, Yildiz Y, Cakir B, et al. Does elevated red blood cell
distribution width predict hospital mortality in critically-ill patients? Intensive Care
Med 2010;36:S140.
[14] Zhang Z, Xu X, Ni H, Deng H. Red cell distribution width is associated with hospital
mortality in unselected critically ill patients. Thorac Dis 2013;5:7306.
[15] Hunziker S, Celi LA, Lee J, Howell MD. Red cell distribution width improves the simplied acute physiology score for risk prediction in unselected critically ill patients.
Crit Care 2012;16:R89.
[16] George H. Red blood cell distribution width (RDW) mortality and morbidity implications. Retrieved from http://insureintell.com/content/rdw-mortality-and-morbidityimplications; March 17 2014.
[17] Dixon B, Santamaria JD, Reid D, Collins M, Rechnitzer T, Newcomb AE, et al. The
association of blood transfusion with mortality after cardiac surgery: cause or
confounding? Transfusion 2013;53(1):1927.
[18] Robinson III WP, Ahn J, Stifer A, Rutherford EJ, Hurd H, Zarzaur BL, et al. Blood
transfusion is an independent predictor of increased mortality in nonoperatively
managed blunt hepatic and splenic injuries. Trauma Acute Care Surg 2005;58:
43745.
[19] Anthony C, Shaikh AA, Walters M, Huehl S, Pomerantz R. Blood transfusion is an
independent predictor of mortality after blunt trauma. Am Surg 2007;73:15.
[20] Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;49(12):
13739.
Please cite this article as: Loveday S, et al, Does the addition of RDW improve current ICU scoring systems?, Clin Biochem (2015), http://
dx.doi.org/10.1016/j.clinbiochem.2015.04.002