Anaesth Intensive Care 2013; 41: 231-241

Eosinopenia as a predictor of unexpected re-admission and

mortality after intensive care unit discharge
C. B. Yip*, K. M. Ho†
Department of Intensive Care Medicine, Royal Perth Hospital and School of Population Health of University of Western Australia,
Perth Western Australia, Australia

Summary
Predicting unexpected intensive care unit (ICU) re-admission and mortality after critical illness is difficult.
This study assessed the associations between eosinopenia on the day of ICU discharge and outcomes after
critical illness. This retrospective cohort study involved a total of 1446 critically ill patients who survived their
first ICU admission between January 2009 and March 2010 in a multidisciplinary ICU in Western Australia.
Eosinopenia was defined as eosinophil count <0.01×109/l and the date of censor for survival was 31 October
2011. Of the 1446 patients included in the study, 106 patients (7.3%) were re-admitted to the ICU during
the same hospitalisation and 178 patients died (12.3%) after ICU discharge. Eosinopenia at ICU discharge
occurred in 130 patients (9.7%) and was more common among those who were subsequently re-admitted (18.6
vs 8.6%) or died after ICU discharge (22.5 vs 7.5%). Eosinopenia remained associated with ICU re-admission
(odds ratio 2.50, 95% confidence interval 1.38–4.50; P=0.002) and post-ICU mortality (hazard ratio 2.65, 95%
confidence interval 1.77–3.98; P=0.001) after adjusting for age, gender, nocturnal discharge, neutrophil count at
ICU discharge, elective surgical admission, Sequential Organ Failure Assessment scores, Acute Physiology and
Chronic Health Evaluation II predicted mortality and chronic medical diseases. Eosinopenia at ICU discharge
explained about 8.4% of the variability and was the third most important factor in explaining the variability in
survival after ICU discharge. In summary, eosinopenia at ICU discharge was associated with an increased risk
of unexpected ICU re-admission and post-ICU mortality.
Key Words: infection, quality indicators, re-admission, sensitivity and specificity, survival

Following apparent successful intensive care unit
(ICU) discharge, 5–10% of the patients are re-
admitted to the ICU1,2 or die in the hospital ward
unexpectedly3,4, increasing the overall morbidity,
mortality and resource requirements of the critically
ill5. The clinical problems of unexpected ICU re-
admission and post-ICU mortality may arise from
incomplete resolution of the primary critical illness
or from the development of new complications
such as venous thromboembolism and nosocomial
infection1,4–7. Numerous clinical risk factors for adverse
outcomes after ICU discharge have been identified,
including age, chronic medical illness, frailty, severity
* MBBS, FANZCA, Senior Anaesthetic Registrar, Department of
Anaesthesia.
† MB, BS, Postgrad Dip (Echo), MPH, PhD, FRCP(Glasg), FCICM,
FANZCA, Consultant Intensivist; and School of Population Health,
University of Western Australia.
Address for correspondence: Dr. K. M. Ho, Intensive Care Unit, Royal
Perth Hospital, Wellington Street, Perth, Western Australia, Australia 6000.
Email: kwok.ho@health.wa.gov.au
Accepted for publication on January 9, 2013
Anaesthesia and Intensive Care, Vol. 41, No. 1, March 2013
of acute illness at ICU admission, unstable vital
signs or high levels of inflammatory markers at ICU
discharge, nocturnal discharge, and insufficient level
of nursing care on the discharge ward5,8–11. Although
prognostic scoring systems at ICU discharge have
been developed to reduce inappropriate ICU dis-
charges12–14, their clinical utility for decision-making at
the bedside may be limited by their complexity.
Eosinophils normally only account for a small
proportion of the peripheral white blood cells (<3%)
and are regulated by a number of mechanisms
including cytokines of acute inflammation15. As
such, eosinopenia is expected in a number of clinical
conditions including bloodstream infections, viral
infections, therapy with corticosteroids and catecho-
lamines and physiological stress16–18. Recently, eosin-
openia has been reported as a sensitive biomarker
of sepsis and its associated mortality17,19,20. We
hypothesised that, as a marker of latent inflammation
and sepsis, eosinopenia may be useful in predicting
adverse outcomes after ICU discharge and we con-
ducted a retrospective cohort study to assess the
prognostic value of eosinopenia at ICU discharge.
232 C. B. Yip, K. M. Ho
MATERIALS AND METHODS
After obtaining approval from the Ethics Com-
mittee of Royal Perth Hospital (107/2012), the ICU
data of all the patients admitted to the ICU of Royal
Perth Hospital between January 2009 and March 2010
were linked to the laboratory databases and death
registry. The ICU database contained comprehensive
clinical data on severity of acute illness including
Acute Physiology and Chronic Health Evaluation
(APACHE) II score and its derived predicted mort-
ality, daily Sequential Organ Failure Assessment
scores21,22, principal and secondary admission diag-
noses coded according to the APACHE model
and International Classification of Diseases codes,
comorbidity, re-admission status, and outcomes
after ICU discharge. Royal Perth Hospital is an
800-bed tertiary teaching hospital and the 23-bed
multidisciplinary ICU admits about 1500 critically
ill adult patients per annum from all specialties,
including multiple trauma, burns, cardiothoracic
surgery and heart-lung transplantation. All results
analysed were performed for clinical purposes and
were subsequently retrieved for this study.
The clinical predictors analysed in this study
included age, sex, nocturnal discharge (6.00 pm–
8.00 am), neutrophil and eosinophil counts on the
day of ICU discharge, C-reactive protein (CRP)
concentrations on the day of ICU discharge, elective
surgical admission, chronic medical diseases as de-
fined by the APACHE prognostic model, APACHE
II score and its derived predicted mortality, and
maximum and discharge Sequential Organ Failure
Assessment scores21,22. Serum CRP concentrations
were measured by an immunoenzyme analyser
(Hitachi 917, Tokyo, Japan) and eosinophil counts
were determined by an automated method (Cell-Dyn
Sapphire, Abbott Diagnostics, IL, USA) with minimal
measurable eosinophil counts of 0.01×109/l (10 cells/
mm3)18,23. The slope, y-intercept and correlation
between eosinophil counts measured by this device
and manual differential study are 1.000 (95%
confidence interval [CI] 0.920–1.040), 0.100 (95%
CI 0.100–0.200) and 0.950, respectively24. Eosinophil
counts <0.01×109/l were not measurable by this
automated method and were defined as eosinopenia
in this study. For patients who were re-admitted
to the ICU during the same hospitalisation, only
data from the first ICU admission and at first ICU
discharge were used. Unexpected ICU re-admission
was defined as re-admission of those patients who
were discharged from ICU without a plan to be re-
admitted after elective surgery. We included all
unexpected ICU re-admissions after the first ICU
discharge to increase the power of the study. As for
survival after hospital discharge, the date of censor
for survival was 31 October 2011 and post-ICU
mortality was defined by death at any time between
ICU discharge and date of censor, independent of
hospital stay. No patients were lost to follow-up or
had missing mortality data and the median follow-
up time for survival was 26 months (interquartile
range 22–30). The venous thromboembolism data in
relation to presence of thrombocytosis of this cohort
of patients were described in our recent publication25.
Statistical analysis
The associations between unplanned ICU
re-admission during the same hospitalisation or
mortality after ICU discharge and eosinopenia were
tested with univariate analyses followed by multi-
variate analysis. Categorical variables and con-
tinuous variables with skewed distributions were
analysed using chi-square and Mann-Whitney tests,
respectively. Logistic regression was used to assess
the relationship between eosinopenia and risk of ICU
re-admission during the same hospitalisation, after
adjusting for known risk factors for unplanned ICU
re-admission5. Cox proportional hazards regression
was used to assess whether eosinopenia at ICU
discharge was associated with an increased risk of
subsequent unexpected mortality, after confirming
the proportionality assumption of eosinopenia for
mortality.
Multicollinearity between continuous predictors
were excluded, by Pearson’s correlation coefficient
<0.8, before these predictors were entered into the
multivariate Cox and logistic models. During the
modelling process, no predictors were removed.
Because CRP concentrations within 24 hours of ICU
discharge were available only in 755 (52%) patients,
CRP was only added to the final multivariate model
as a sensitivity analysis to assess whether CRP would
change the direction and magnitude of the associations
between eosinopenia and adverse outcomes after
ICU discharge.
In a separate logistic regression analysis, we also
modelled the relationship between eosinophil counts
as a continuous variable, and risk of a composite
outcome of either unplanned ICU re-admission or
post-ICU mortality using a three-knot restricted
cubic spline function26. This was performed to assess
whether eosinophil counts at ICU discharge had a
non-linear relationship with either of the adverse
outcomes after discharge. The relative importance of
each predictor in explaining the variability in survival
after ICU discharge was assessed using the chi-square
statistic minus the degrees of freedom25. In this study
a P value <0.05 was taken as significant. All tests were
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
 Eosinopenia and outcomes after ICU discharge 233

Patients admitted to the ICU between January 2009 and

March 2010 (n=1599, a total of 1731 ICU admissions)

153 patients died in first

ICU admission (9.6%)

1446 patients (90.4%) survived their first ICU admission to ICU discharge.
Eosinophil count was available on the day of ICU discharge in 1394
patients (96.4%) and eosinopenia occurred in 130 patients (9.7%).

1. Patients with ICU re-admission during the same hospitalisation (n=106 [6.6%] constituted a total of 132 ICU re-admissions)
2. Patients died in the hospital ward or ICU after first ICU discharge during the same hospitalisation (n=103, 6.4%)
3. Patients died after hospital discharge censored of 31 October 2011 (n=75, 4.7%)
4. Total number of patients died after ICU discharge as of 31 October 2011 (n=178, 11.1%)

Figure 1: Inclusion of patients in the study. ICU=intensive care unit

two-tailed and performed by the SPSS for Windows 600

2011 (Version 19.0, IBM, WA, USA) and S-PLUS

(version 8.0, 2007, Insightful Corp, Seattle, WA,
500
USA). Mean eosinophil count=0.15, median=0.08,
interquartile range=0.02–0.22; number of patients
with eosinopenia at ICU discharge=130 (9.7%)
RESULTS 400

Study population and outcomes

Frequency

300
Of the 1599 patients admitted to the ICU during the
study period, 1446 (90.4%) survived to their first ICU
discharge, 106 (6.6%) patients were re-admitted to 200

the ICU after discharge during the same hospitali-

sation and 103 (6.4%) died unexpectedly after their
100
first ICU discharge. Another 75 patients (4.7%)
died after hospital discharge as of 31 October 2011
(Figure 1). A total of 253 patients (17.5%) had either 0
unexpected re-admission during the same hospitali- 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 1.50

Eosinophil count (x109/l)

sation or post-ICU mortality before 31 October 2011.
Eosinophil count was available on the day of ICU
discharge in 1394 patients (96.4%) and eosinopenia
occurred in 130 patients (9.7%) (Figure 2).
Figure 2: Distribution of eosinophil count of all patients at the
time of intensive care unit (ICU) discharge (n=1394). Eosinophil
counts are measured in 109/l and eosinopenia is defined as count
<0.01×109/l.

Characteristics of the patients who were re-admitted to
ICU and risk factors for ICU re-admission
The median time between the first ICU discharge
and re-admission was seven days (interquartile range
3–26). Those who were subsequently re-admitted to
ICU were slightly older (53 vs 50 years old), more
likely to be initially admitted to the ICU because of
sepsis or multiple trauma, and had a more severe
acute illness (APACHE II predicted mortality 20
vs 17%, P=0.02) than those who did not have ICU
re-admission. Eosinopenia at ICU discharge was
significantly more common among those who were
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
re-admitted to the ICU (18.6 vs 8.6%; P=0.002) than
those who did not have ICU re-admission (Table 1).
In the multivariate logistic regression analysis, only
eosinopenia (odds ratio 2.50, 95% CI 1.38–4.50;
P=0.002) and diabetes mellitus requiring insulin
(odds ratio 2.97, 95% CI 1.33–6.65; P=0.008) were
significantly associated with an increased risk of ICU
re-admission (Table 2). Including CRP concentrations
at ICU discharge in the final multivariate model also
did not change the direction and magnitude of the
association between eosinopenia and risk of ICU
re-admission (OR 2.99, 95% CI 1.47–6.07; P=0.002).
234 C. B. Yip, K. M. Ho

Table 1
Difference in characteristics between those who were re-admitted to ICU during the same hospitalisation and those who did not

Variables Re-admitted No re-admission P value

(n=106) (n=1340)
Age, years, (IQR) 53 (34–67) 50 (32–65) 0.049*
Female, number (%) 31 (29.2) 485 (36.2) 0.171
Elective surgery, number (%) 32 (30.1) 373 (27.8) 0.653
ICU major admission diagnosis, number (%) 0.009
Sepsis 11 (10.4) 71 (5.3)
Trauma 23 (21.7) 222 (16.6)
Obstructive airway disease 3 (2.8) 35 (2.6)
Cardiac or respiratory arrest 6 (5.7) 38 (2.8)
Pneumonia 8 (7.6) 79 (5.9)
Drug overdose 3 (2.8) 135 (10.1)
Diabetic ketoacidosis 0 (0) 5 (0.4)
Heart failure or myocardial infarction 3 (2.8) 9 (0.7)
Major cardiac or vascular surgery 21 (19.8) 333 (24.9)
Other 28 (26.4) 413 (30.8)
APACHE II scores (IQR) 20 (16–24) 19 (14–24) 0.080*
APACHE II predicted mortality, % (IQR) 20 (11–42) 17 (8–38) 0.020*
Admission SOFA score (IQR) 6 (4–9) 5 (3–8) 0.083*
SOFA score at ICU discharge (IQR) 2 (1–5) 2 (1–4) 0.243*
Maximum SOFA during ICU stay (IQR) 7 (4–10) 6 (4–9) 0.063*
ICU length-of-stay, days (IQR) 2.3 (1.1–6.7) 1.9 (1.0–5.0) 0.019*
Hospital length-of-stay, days (IQR) 28 (16–48) 12 (6–24) 0.001*
Nocturnal ICU discharge (6 pm–8 am), number (%) 25 (23.5) 404 (30.1) 0.185
Pre-existing chronic diseases
Immunosuppressive disease 1 (0.9) 7 (0.5) 0.457
Immunosuppressive treatment 3 (2.8) 33 (2.5) 0.744
HIV infection 0 (0) 3 (0.2) 0.999
Leukaemia/myeloma 1 (0.9) 15 (1.1) 0.999
Metastatic cancer 2 (1.9) 14 (1.0) 0.330
Lymphoma 0 (0) 9 (0.7) 0.999
History of hepatic failure 0 (0) 4 (0.3) 0.999
Cirrhosis 3 (2.8) 28 (2.1) 0.492
Chronic respiratory disease† 6 (5.7) 61 (4.6) 0.628
Chronic cardiovascular disease‡ 18 (17.0) 156 (11.6) 0.119
Chronic renal failure requiring dialysis 6 (5.7) 54 (4.0) 0.442
Diabetes mellitus requiring insulin 9 (8.5) 39 (2.9) 0.006

Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013

 Eosinopenia and outcomes after ICU discharge 235

Table 1
Difference in characteristics between those who were re-admitted to ICU during the same hospitalisation and those who did not (continued)

Variables Re-admitted No re-admission P value

(n=106) (n=1340)
Eosinopenia (<0.01×109/l), number (%) 19 (18.6) 111 (8.6) 0.002
Mean neutrophil counts at discharge (×10 /l) (IQR)
9
9 (6–12) 9 (6–11) 0.558
Mean CRP concentrations at discharge§ (IQR), mg/l 102 (32–140) 105 (37–150) 0.820
Mortality in the ward, number (%) 26 (24.5) 77 (5.7) 0.001
Mortality either in the ward or after hospital discharge, number (%) 31 (29.3) 147 (11.0) 0.001
Data of all continuous variables are presented as median with IQR unless stated otherwise. * Median and IQR of the variables are
presented and P values are generated by Mann-Whitney test. † Presence of pulmonary hypertension or requiring home oxygen therapy. ‡
New York Heart Association Class IV functional status. § Only 755 patients had CRP results available within 24 hours of ICU discharge.
IQR=interquartile range, APACHE=Acute Physiology and Chronic Health Evaluation, CRP=C-reactive protein, ICU=intensive care
unit, SOFA=Sequential Organ Failure Assessment.

Table 2
Association between eosinopenia (<0.01×109/l) at the time of ICU discharge and unplanned
ICU re-admission during the same hospitalisation, after adjusting for other covariates by
logistic regression analysis

Variables in the model Adjusted odds ratio P value

(95% CI)
Eosinopenia at ICU discharge 2.50 (1.38–4.50) 0.002
Neutrophil count at ICU discharge 0.99 (0.95–1.04) 0.732
APACHE II predicted mortality 1.04 (0.94–1.17) 0.436
(per 10% increment)
Age 1.00 (0.99–1.02) 0.494
(per year increment)
Gender (female) 0.75 (0.47–1.18) 0.230
SOFA at discharge 1.01 (0.92–1.11) 0.831
(per score increment)
Maximum SOFA score during ICU stay 0.99 (0.95–1.04) 0.769
(per score increment)
Nocturnal discharge (6 pm–8 am) 0.69 (0.42–1.12) 0.135
Immunosuppressive disease 2.32 (0.25–21.5) 0.460
Immunosuppressive treatment 0.82 (0.23–3.00) 0.766
HIV infection 0.02 (0.01–10.0) 0.999
Leukaemia/myeloma 0.56 (0.06–4.90) 0.601
Metastatic cancer 2.87 (0.59–13.9) 0.189
Lymphoma 0.02 (0.01–9.9) 0.999
History of hepatic failure 0.02 (0.01–10.0) 0.999
Cirrhosis 1.36 (0.38–4.86) 0.641
Chronic respiratory disease* 1.03 (0.42–2.54) 0.946
Chronic cardiovascular disease† 1.23 (0.67–2.29) 0.503
Chronic renal failure requiring dialysis 0.89 (0.33–2.42) 0.823
Diabetes mellitus requiring insulin 2.97 (1.33–6.65) 0.008
Elective surgery 1.10 (0.66–1.83) 0.710
Unadjusted odds ratio of eosinopenia on risk of unplanned ICU re-admission during the same
hospitalisation was 2.44 (95% CI 1.43–4.16; P=0.001). * Presence of pulmonary hypertension
or requiring home oxygen therapy. † New York Heart Association class IV functional status.
Hosmer-Lemeshow chi-square of the model was 8.6 (P=0.377). CI=confidence interval,
ICU=Intensive care unit, APACHE=Acute Physiology and Chronic Health Evaluation,
SOFA=Sequential Organ Failure Assessment, HIV=human immunodeficiency virus.
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
236 C. B. Yip, K. M. Ho

Table 3
Difference in characteristics between those who died after ICU discharge and those who survived until 31 October 2011

Variables Died (n=178) Survived (n=1268) P value

Age, years (IQR) 65 (52–76) 48 (30–63) 0.001*
Female, number (%) 66 (37.1) 450 (35.5) 0.677
Elective surgery, number (%) 41 (23.0) 364 (28.7) 0.130
ICU major admission diagnosis, number (%) 0.001
Sepsis 31 (17.4) 51 (4.0)
Trauma 8 (4.5) 237 (18.7)
Obstructive airway disease 5 (2.8) 33 (2.6)
Cardiac or respiratory arrest 10 (5.6) 34 (2.7)
Pneumonia 14 (7.9) 73 (5.8)
Drug overdose 5 (2.8) 133 (10.5)
Diabetic ketoacidosis 0 (0) 5 (0.4)
Heart failure or myocardial infarction 7 (3.9) 5 (0.4)
Major cardiac or vascular surgery 27 (15.2) 327 (25.8)
Others 71 (39.9) 370 (29.2)
APACHE II scores (IQR) 18 (14–23) 15 (12–20) 0.001*
APACHE II predicted mortality, % (IQR) 41 (19–62) 16 (7–34) 0.001*
Admission SOFA score (IQR) 7 (4–9) 5 (3–8) 0.001*
SOFA score at ICU discharge (IQR) 4 (1–6) 2 (1–4) 0.001*
Maximum SOFA during ICU stay (IQR) 8 (5–10) 6 (4–9) 0.001*
ICU length-of-stay, days (IQR) 3.6 (1.4–6.4) 1.9 (0.9–4.8) 0.001*
Hospital length-of-stay, days (IQR) 21 (11–38) 12 (6–25) 0.001*
Nocturnal ICU discharge (6 pm–8 am), number (%) 59 (33.2) 370 (29.2) 0.293
Pre-existing chronic diseases
Immunosuppressive disease 3 (1.7) 5 (0.4) 0.064
Immunosuppressive treatment 7 (3.9) 29 (2.3) 0.196
HIV infection 1 (0.6) 2 (0.2) 0.326
Leukaemia/myeloma 8 (4.5) 8 (0.6) 0.001
Metastatic cancer 7 (3.9) 9 (0.7) 0.002
Lymphoma 5 (2.8) 4 (0.3) 0.002
History of hepatic failure 1 (0.6) 3 (0.2) 0.409
Cirrhosis 10 (5.6) 21 (1.7) 0.003
Chronic respiratory disease† 20 (11.2) 47 (3.7) 0.001
Chronic cardiovascular disease‡ 34 (17.0) 140 (11.0) 0.003
Chronic renal failure requiring dialysis 21 (11.8) 39 (3.1) 0.001
Diabetes mellitus requiring insulin 21 (11.8) 27 (2.1) 0.001
Eosinopenia (<0.01×109/mm3), number (%) 38 (22.5) 92 (7.5) 0.001
Mean neutrophil counts at discharge (×109/l) (IQR) 9 (5–11) 9 (6–11) 0.735
Mean CRP concentrations at discharge§ (IQR), mg/l 119 (46–168) 101 (36–150) 0.037
Data of all continuous variables are presented as median with IQR unless stated otherwise. * Median and IQR of the variables are
presented and P values are generated by Mann-Whitney test. † Presence of pulmonary hypertension or requiring home oxygen ther-
apy. ‡ New York Heart Association Class IV functional status. § Only 755 patients had C-reactive protein results available within
24 hours of ICU discharge. IQR=interquartile range, ICU=Intensive care unit, APACHE=Acute Physiology and Chronic Health
Evaluation, SOFA=Sequential Organ Failure Assessment, HIV=human immunodeficiency virus, CRP=C-reactive protein.
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
 Eosinopenia and outcomes after ICU discharge 237

Table 4
Association between eosinopenia (<0.01×109/l) at the time of ICU discharge and survival time after discharge, after
adjusting for other covariates by Cox proportional hazards regression analysis

Variables in the model Adjusted hazard ratio (95% CI) P value

Eosinopenia at ICU discharge 2.65 (1.77–3.98) 0.001
Neutrophil count at ICU discharge 0.99 (0.96–1.03) 0.677
APACHE II predicted mortality 1.17 (1.09–1.26) (per 10% increment) 0.001
Age 1.03 (1.02–1.04) (per year increment) 0.001
Gender (female) 1.02 (0.74–1.40) 0.901
SOFA at discharge 1.04 (0.99–1.10) (per score increment) 0.200
Maximum SOFA score during ICU stay 0.99 (0.98–1.02) (per score increment) 0.819
Nocturnal discharge (6 pm–8 am) 1.04 (0.74–1.44) 0.840
Immunosuppressive disease 2.34 (0.68–8.1) 0.180
Immunosuppressive treatment 0.70 (0.30–1.64) 0.410
HIV infection 4.04 (0.54–30.3) 0.175
Leukaemia/myeloma 2.96 (1.27–6.88) 0.012
Metastatic cancer 5.49 (2.30–13.1) 0.001
Lymphoma 3.95 (1.53–10.2) 0.004
History of hepatic failure 0.99 (0.12–7.87) 0.991
Cirrhosis 2.09 (1.03–4.23) 0.041
Chronic respiratory disease* 1.94 (1.18–3.20) 0.009
Chronic cardiovascular disease† 0.95 (0.60–1.51) 0.830
Chronic renal failure requiring dialysis 1.34 (0.75–2.40) 0.323
Diabetes mellitus requiring insulin 3.51 (2.09–5.91) 0.001
Elective surgery 0.78 (0.51–1.19) 0.251
Unadjusted hazard ratio of eosinopenia on survival time after ICU discharge was 3.23 (95% CI 2.25–4.64;
P=0.001). * Presence of pulmonary hypertension or requiring home oxygen therapy. † New York Heart Association
Class IV functional status. ICU=intensive care unit, CI=confidence interval, APACHE=Acute Physiology and
Chronic Health Evaluation, SOFA=Sequential Organ Failure Assessment, HIV=human immunodeficiency virus.

1.0
Risk factors for mortality after first ICU discharge
Those who died after ICU discharge were signif-
icantly older (65 vs 48 years old), more likely to be
0.9 initially admitted to ICU because of sepsis (17.4
vs 4.0%), had a more severe acute illness and also
Cumulative survival

more co-existing medical illnesses, including diabetes

0.8
mellitus requiring insulin, cirrhosis, chronic renal
failure requiring dialysis, metastatic cancer and
0.7 Adjusted hazard ratio=2.65 (95% CI 1.77–3.98; P=0.001) hematological malignancies (Table 3). Eosinopenia
(22.5 vs 7.5%, P=0.001) and a higher CRP concen-
tration (119 vs 101 mg/l, P=0.037) were more common
0.6 among those who died after ICU discharge than those
Eosinopenia
No
who survived until the date of censor for survival.
Yes In the multivariate Cox regression analysis,
0.5 eosinopenia (hazard ratio 2.65, 95% CI 1.77–3.98;
P=0.001) remained associated with an increased
0 200 400 600 800 1000
Days since ICU discharge risk of unexpected mortality after adjusting for other
covariates (Table 4, Figure 3). Other important factors
Figure 3: Survival after ICU discharge for patients with and
without eosinopenia at ICU discharge after adjusting for other in determining the medium-term survival were age,
covariates. ICU=intensive care unit, CI=confidence interval. severity of acute illness as measured by the APACHE
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
 238 C. B. Yip, K. M. Ho

Age
Diabetes mellitus requiring insulin
Eosinopenia
APAHCE II predicted mortality
Metastasis
Chronic respiratory disease
Lymphoma
Leuokemia/myeloma
Cirrhosis
Elective
HIV
SOFA at discharge
Immunosuppressed disease
Chronic renal failure
Immunosuppressive treatment
SOFA maximum
Nocturnal discharge
Chronic cardiovascular disease
Gender
Hepatic failure
Neutrophil count

0 5 10 15 20 25
Chi-square test minus degrees of freedom

Figure 4: Variability in survival time explained by different covariates in the final Cox
proportional hazards model. APACHE=Acute Physiology and Chronic Health Evaluation,
HIV=human immunodeficiency virus, SOFA=Sequential Organ Failure Assessment Score.
log odds of either unplanned readmission or mortality after ICU discharge

II predicted mortality, diabetes mellitus requiring

insulin, chronic respiratory disease, haematological
malignancies and metastatic cancer. Eosinopenia at
0.4

ICU discharge explained about 8.4% of the variability

and was the third most important factor—after age
(13%) and diabetes mellitus (9.9%)—in explaining
0.2

the variability in survival after ICU discharge (Figure

4).
Including CRP concentrations at ICU discharge
in the final multivariable Cox model also did not
0.0

change the direction and magnitude of the associa-

tion between eosinopenia and risk of mortality after
ICU discharge (hazard ratio 3.1, 95% CI 1.9–5.0;
-0.2

P=0.001). CRP at ICU discharge was, however, not

associated with an increased risk of mortality after
adjusting for all covariates in the final model (hazard
ratio 1.0, 95% CI 0.9–1.1; P=0.478).
-0.4

Eosinophil count as a continuous predictor for either

unplanned re-admission or mortality after ICU dis-
0.0 0.1 0.2 0.3 0.4 0.5 charge
Absolute eosinophil count at ICU discharge (x109/l)
When eosinophil count was modelled as a
Figure 5: Non-linear relationship between absolute eosinophil continuous predictor, the risk of either unexpected
count (×109/l) at ICU discharge and odds of having unplanned re-admission or post-ICU mortality only increased
ICU re-admission during the same hospitalisation or mortality
after ICU discharge. Dotted lines signify 95% confidence interval. significantly when the eosinophil counts were less
ICU=intensive care unit. than 0.1×109/l (Figure 5), suggesting that only
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
 Eosinopenia and outcomes after ICU discharge
Diagnostic ability of eosinopenia as a predictor of adverse outcomes after intensive care unit discharge
Sensitivity, %
Specificity, %
Positive predictive value, %
Negative predictive value, %
Positive likelihood ratio
Negative likelihood ratio
Values in brackets are 95% confidence intervals. ICU=intensive care unit.
eosinopenia, but not eosinophilia, was important in
predicting adverse outcomes after ICU discharge.
The sensitivity, specificity, positive predictive value
and negative predictive value of eosinopenia to
predict unexpected re-admission, mortality or either
of these adverse events are described in Table 5.
DISCUSSION
In this study we assessed the incidence of eosin-
openia during the recovery phase of critical illness
and showed that eosinopenia was not rare (9.7%) and
was associated with an increased risk of unexpected
ICU re-admission and mortality after ICU discharge.
This result has some clinical implications and requires
careful consideration.
First, unexpected ICU re-admission has a significant
attributable effect on morbidity and mortality of
critically ill patients5. Many strategies have been
suggested to reduce inappropriate ICU discharges
and, at the same time, avoid prolonging unnecessary
ICU stay. Although complicated prognostic models
to predict risk of unexpected ICU re-admission or
mortality are available12–14, their clinical applicability
for decision-making at the bedside is limited by their
complexity. Thus, it will be attractive to clinicians to
have some readily available tests to stratify their
patients’ subsequent risk of adverse outcomes at the
time of ICU discharge. Our results suggested that
the utility of eosinopenia lies in its high ‘negative
predictive value’. Although an absence of eosinopenia
might be reassuring to clinicians about a lower risk
of unexpected re-admission or mortality after ICU
discharge, the presence of eosinopenia was not as
useful due to its low positive predictive value. We also
noted that eosinophil counts did not have a linear
relationship with the risk of adverse outcomes after
ICU discharge. In fact, only eosinopenia, defined by
eosinophil count <0.01×109/l, was associated with
an increased risk of adverse outcomes after ICU
discharge. As such, the absolute eosinophil count at
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
239
Table 5
ICU re-admission Post-ICU mortality Re-admission or
post-ICU mortality
18 (12–26) 21 (16–28) 20 (15–27)
92 (90–93) 93 (91–94) 92 (91–94)
15 (10–22) 29 (22–38) 27 (20–35)
93 (92–95) 89 (88–91) 89 (87–91)
2.2 (1.4–3.4) 2.9 (2.1–4.2) 2.6 (1.8–3.7)
0.9 (0.8–1.0) 0.8 (0.7–0.9) 0.9 (0.8–1.0)
ICU discharge is not useful and only eosinopenia, as a
dichotomised variable, may be useful as an adjunct to
other clinical predictors to predict adverse outcomes
after ICU discharge.
Second, recent studies suggested that CRP at
the time of ICU discharge may be useful to stratify
a patient’s risk of adverse outcomes after ICU
discharge1,4,11,27,28. However, the association between a
high CRP and an increased risk of adverse outcomes
after ICU discharge is by no means consistent in all
published studies29,30. Although CRP concentrations at
ICU discharge were associated with a higher risk of
mortality in the univariate analysis in this study, this
association was no longer significant after eosinopenia
was included in the multivariate analysis. This could
be due to the fact that only 52% of our patients had
CRP measured within 24 hours of ICU discharge, or
perhaps, eosinophil count is a more rapid biomarker
than CRP in response to changes in inflammation or
infection. In an animal model of infection, eosinophil
counts reduced to 80% of the trough level within
six hours of initiation of infection and returned to a
normal level within 12 hours after commencement
of appropriate antibiotic31. The dynamic response
of eosinophil count to infection and resolution of
inflammation compares favourably to the relatively
long half-life of CRP (~19 hours)32, especially im-
portant when the ICU length-of-stay is shorter than a
few half-lives of the biomarker concentrations. Large
observational studies comparing predictive ability of
different inflammatory markers for adverse outcomes
after ICU discharge are, however, needed to confirm
whether eosinopenia is a better biomarker than CRP
or other inflammatory markers in differentiating
adverse events after ICU discharge by having a larger
area under the receiver operating characteristic curve,
and whether they can be used together to facilitate
ICU discharge.
Third, we could not confirm nocturnal discharge as
a risk factor for adverse outcomes after ICU discharge
 240
in this study. This may be due to the fact that the
study centre has a consultant intensivist working in
the ICU until late at night (8.00 am–11.00 pm) during
the weekdays and patients are often discharged to the
high dependency unit instead of the general hospital
wards at night.
The last consideration is the limitations of this
study. First, it has to be emphasised that eosinopenia
is a non-specific marker of inflammation or infection.
Similar to all non-specific biomarkers, the diagnostic
or prognostic ability of eosinopenia is expected to vary
substantially depending on the pre-test probability
of the event in the cohort of patients33. As such, we
would expect eosinopenia to perform better as a
predictor of adverse outcomes after ICU discharge
when such events are relatively uncommon (e.g.
<10%) (Figure 6). Second, this was a single-centre
study and hence validation of our results by other
centres is essential to confirm the external validity
of our findings. Third, eosinopenia can be induced
by a number of factors including sepsis, therapy with
corticosteroids and catecholamines, and physiological
stress16–18. Systemic corticosteroids are not routinely
used for most patients with septic shock in the study
centre. Whether eosinopenia is also useful to predict
adverse outcomes in ICUs where corticosteroids
are frequently used remains uncertain. Finally, we
did not have post-mortem results for most of the
deaths included in this study. Whether eosinopenia
at ICU discharge is only useful to predict deaths or
re-admissions due to occult infections1,4, but not other
causes such as venous thromboembolism, remains
uncertain7.
1.0
0.9
0.8
0.7
Post-test probability
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5
Prevalence
Figure 6: The relationship between post-test probability of having
either unexpected re-admission or mortality after intensive care
unit discharge in patients who had eosinopenia and the prevalence
of these events. The relative change in odds is greatest when the
prevalence of adverse outcome is low (e.g. <10%).
C. B. Yip, K. M. Ho
CONCLUSIONS
Eosinopenia was not rare during the recovery phase
of critical illness and was associated with an increased
risk of unexpected re-admission and mortality after
ICU discharge, suggesting that latent inflammation
or incomplete recovery from infection at ICU dis-
charge may be associated with worse outcomes after
critical illness. The clinical utility of eosinopenia
relies mainly on its high negative predictive value: an
absence of eosinopenia at ICU discharge is reassuring
for a reduced risk of adverse outcomes after ICU
discharge. Further studies on the prognostic value of
eosinopenia at ICU discharge are needed to confirm
the results of this preliminary study.
ACKNOWLEDGEMENT
This study was solely funded by the Department
of Intensive Care Medicine, Royal Perth Hospital,
Perth, Western Australia, Australia.
References
  1. Ho KM, Dobb GJ, Lee KY, Towler SC, Webb SA. C-reactive
protein concentration as a predictor of intensive care unit
readmission: a nested case-control study. J Crit Care 2006;
21:259-265.
  2. Rosenberg AL, Hofer TP, Hayward RA, Strachan C, Watts CM.
Who bounces back? Physiologic and other predictors of inten-
sive care unit readmission. Crit Care Med 2001; 29:511-518.
 3. Azoulay E, Alberti C, Legendre I, Buisson CB, Le Gall JR,
European Sepsis Group. Post-ICU mortality in critically ill
infected patients: an international study. Intensive Care Med
2005; 31:56-63.
 4. Ho KM, Lee KY, Dobb GJ, Webb SA. C-reactive protein
concentration as a predictor of in-hospital mortality after ICU
discharge: a prospective cohort study. Intensive Care Med
2008; 34:481-487.
 5. Rosenberg AL, Watts C. Patients readmitted to ICUs: a
systematic review of risk factors and outcomes. Chest 2000;
118:492-502.
 6. Wallis CB, Davies HT, Shearer AJ. Why do patients die on
general wards after discharge from intensive care units?
Anaesthesia 1997; 52:9-14.
  7. Ho KM, Burrell M, Rao S, Baker R. Incidence and risk factors
for fatal pulmonary embolism after major trauma: a nested
cohort study. Br J Anaesth 2010; 105:596-602.
  8. Ho KM, Dobb GJ, Lee KY, Finn J, Knuiman M, Webb SA. The
effect of comorbidities on risk of intensive care readmission
during the same hospitalization: a linked data cohort study. J
Crit Care 2009; 24:101-107.
 9. Pilcher DV, Duke GJ, George C, Bailey MJ, Hart G. After-
hours discharge from intensive care increases the risk of re-
admission and death. Anaesth Intensive Care 2007; 35:477-485.
0.6 0.7 0.8 0.9 1.0 10. Beck DH, McQuillan P, Smith GB. Waiting for the break of
dawn? The effects of discharge time, discharge TISS scores
and discharge facility on hospital mortality after intensive care.
Intensive Care Med 2002; 28:1287-1293.
11. Grander W, Dünser M, Stollenwerk B, Siebert U, Dengg C,
Koller B et al. C-reactive protein levels and post-ICU mortality
in nonsurgical intensive care patients. Chest 2010; 138:856-862.
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
 Eosinopenia and outcomes after ICU discharge
12. Daly K, Beale R, Chang RW. Reduction in mortality after
inappropriate early discharge from intensive care unit: logistic
regression triage model. BMJ 2001; 322:1274-1276.
13. Smith L, Orts CM, O’Neil I, Batchelor AM, Gascoigne AD,
Baudouin SV. TISS and mortality after discharge from inten-
sive care. Intensive Care Med 1999; 25:1061-1065.
14. Ouanes I, Schwebel C, Français A, Bruel C, Philippart F, Vesin
A et al. A model to predict short-term death or readmission
after intensive care unit discharge. J Crit Care 2012; 27:422.e1-9.
15. Bass DA, Gonwa TA, Szejda P, Cousart MS, DeChatelet LR,
McCall CE. Eosinopenia of acute infection: Production of
eosinopenia by chemotactic factors of acute inflammation. J
Clin Invest 1980; 65:1265-1271.
16. Gil H, Magy N, Mauny F, Dupond JL. Value of eosinopenia in
inflammatory disorders: an “old” marker revisited. Rev Med
Interne 2003; 24:431-435.
17. Abidi K, Belayachi J, Derras Y, Khayari ME, Dendane T,
Madani N et al. Eosinopenia, an early marker of increased
mortality in critically ill medical patients. Intensive Care Med
2011; 37:1136-1142.
18. Wibrow BA, Ho KM, Flexman JP, Keil AD, Kohrs DL.
Eosinopenia as a diagnostic marker of bloodstream infection in
hospitalised paediatric and adult patients: a case-control study.
Anaesth Intensive Care 2011; 39:224-230.
19. Abidi K, Khoudri I, Belayachi J, Madani N, Zekraoui A,
Zeggwagh AA et al. Eosinopenia is a reliable marker of sepsis
on admission to medical intensive care units. Crit Care 2008;
12:R59.
20. Shaaban H, Daniel S, Sison R, Slim J, Perez G. Eosinopenia: Is
it a good marker of sepsis in comparison to procalcitonin and
C-reactive protein levels for patients admitted to a critical care
unit in an urban hospital? J Crit Care 2010; 25:570-575.
21. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonça
A, Bruining H et al. The SOFA (Sepsis-related Organ Failure
Assessment) score to describe organ dysfunction/failure. On
behalf of the Working Group on Sepsis-Related Problems of
the European Society of Intensive Care Medicine. Intensive
Care Med 1996; 22:707-710.
22. Ho KM. Combining sequential organ failure assessment
(SOFA) score with acute physiology and chronic health evalu-
ation (APACHE) II score to predict hospital mortality of criti-
cally ill patients. Anaesth Intensive Care 2007; 35:515-521.
Anaesthesia and Intensive Care, Vol. 41, No. 2, March 2013
241
23. Ho KM, Towler SC. A comparison of eosinopenia and
C-reactive protein as a marker of bloodstream infections in
critically ill patients: a case control study. Anaesth Intensive
Care 2009; 37:450-456.
24. Tan BT, Nava AJ, George TI. Evaluation of the Beckman
Coulter UniCel DxH 800 and Abbott Diagnostics Cell-Dyn
Sapphire hematology analyzers on pediatric and neonatal
specimens in a tertiary care hospital. Am J Clin Pathol 2011;
135:929-938.
25. Ho KM, Yip CB, Duff O. Reactive thrombocytosis and risk
of subsequent venous thromboembolism: a cohort study. J
Thromb Haemost 2012; 10:1768-1774.
26. Ho KM, Knuiman M, Finn J, Webb SA. Estimating long-term
survival of critically ill patients: the PREDICT model. PLoS
One 2008; 3:e3226.
27. Kaben A, Corrêa F, Reinhart K, Settmacher U, Gummert J,
Kalff R et al. Readmission to a surgical intensive care unit:
incidence, outcome and risk factors. Crit Care 2008; 12:R123.
28. Araújo I, Gonçalves-Pereira J, Teixeira S, Nazareth R, Silvestre
J, Mendes V et al. Assessment of risk factors for in-hospital
mortality after intensive care unit discharge. Biomarkers 2012;
17:180-185.
29. Silvestre J, Coelho L, Póvoa P. Should C-reactive protein con-
centration at ICU discharge be used as a prognostic marker?
BMC Anesthesiol 2010; 10:17.
30. Al-Subaie N, Reynolds T, Myers A, Sunderland R, Rhodes
A, Grounds RM et al. C-reactive protein as a predictor of
outcome after discharge from the intensive care: a prospective
observational study. Br J Anaesth 2010; 105:318-325.
31. Bass DA. Behavior of eosinophil leukocytes in acute inflamma-
tion. II. Eosinophil dynamics during acute inflammation. J Clin
Invest 1975; 56:870-879.
32. Ho KM, Lipman J. An update on C-reactive protein for inten-
sivists. Anaesth Intensive Care 2009; 37:234-241.
33. Ho KM. Ten commandments of interpreting and applying
results of biomarker research. Anaesth Intensive Care 2011;
39:799-801.