Red cell distribution width as a novel predictor of mortality in ICU patients

Abstract

Background. The red cell distribution width (RDW) in ICU patients has never been investigated. Methods. A total of 602 consecutive patients were prospectively enrolled. We collected each patient's base-line characteristics including the Acute Physiology and Chronic Health Evaluation II (APACHE-II) score, RDW, and C-reactive protein (CRP). The primary outcome for this analysis was ICU mortality, and secondary outcome was the total length of stay in hospital (TLSH). Potential predictors were analyzed for possible association with outcomes. Results. There was a significantly graded increase in APACHE-II score (tertile I = 10.7 ± 6.5 versus tertile II = 13.3 ± 6.2 versus tertile III = 15.8 ± 7.2; all P < 0.001) and ICU mortality rate (tertile I = 11.2% versus tertile II = 18.8% versus tertile III = 33.8%; all P < 0.001) across increasing of RDW tertile. As compared with APACHE-II score, combination of RDW and APACHE-II score increased the area under the curve (AUC) for predicting ICU mortality from 0.832 ± 0.020 to 0.885 ± 0.017 (P < 0.05). Multivariate analysis demonstrated that RDW, APACHE-II score, and CRP were independent predictors of ICU mortality (P < 0.05). RDW was also independently associated with TLSH in patients alive (P < 0.05). Conclusion. We found a graded independent relation between higher RDW and adverse outcomes in ICU patients. RDW has the potentially clinical utility to predict outcome in ICU patients.

Key words::

• APACHE-II score
• ICU
• mortality
• red cell distribution width

Abbreviations
APACHE-II score	=	Acute Physiology and Chronic Health Evaluation II score
AUC	=	area under the curve
CRP	=	C-reactive protein
eGFR	=	estimated glomerular filtration rate
MCV	=	mean corpuscular volume
MDRD	=	Modification of Diet in Renal Disease
RDW	=	red blood cell distribution width
ROC curve	=	receiver operating characteristic curve
SCr	=	serum creatinine
TLSH	=	total length of stay in hospital

Key messages

• There was a graded increase in APACHE-II score and ICU mortality rate across increasing of RDW levels in ICU patients.

• Combining RDW and APACHE-II score added to the ability to discriminate mortality risk, and RDW was additive with APACHE-II score for predicting ICU mortality.

• RDW was an independent predictor of ICU mortality and the total length of stay in hospital in ICU patients.

Introduction

Red blood cell distribution width (RDW) is a quantitative measurement of the variability in size of circulating erythrocytes (1). RDW is widely available to clinicians because it is routinely reported as part of the complete blood count. During past decades, it has been used to differentiate the causes of anemia (2). However, considerable attention has been paid to the association between RDW and clinical outcomes in multiple patient populations in the past few years (3–12). A series of studies have demonstrated that RDW can serve as a novel, independent predictor of prognosis in patients with cardiovascular diseases (e.g. heart failure (3–5), stable coronary diseases (6), acute myocardial infarction (7), strokes (8), and pulmonary hypertension (9)). Elevated RDW values were also shown to be associated with increased risk of mortality in the general population (10–12). To our knowledge, the RDW in ICU patients has never been investigated. We hypothesized that higher RDW would be independently associated with worse clinical outcomes in ICU patients. Therefore, we conducted a prospective study to evaluate the prognostic value of RDW in ICU patients.

Material and methods

Participants

We conducted a prospective study involving adult patients admitted to the ICU of Xin Hua Hospital affiliated to Shanghai Jiaotong University School of Medicine. Eligible patients were those needing to be hospitalized to ICU who transferred from the emergency department or other departments of our hospital, including medical and trauma patients (no surgical patients were included). The decision to transfer the patients into ICU was made by at least one critical care expert and one medical expert or trauma expert. Again, the decision to discharge patients or to transfer patients to general wards was also made by these experts. We decided a priori to exclude patients with the following criteria: 1) age < 18 years; 2) pregnant; 3) known hematologic disease such as leukemia, myelodysplastic syndrome, neoplastic metastases to marrow; 4) history of recent blood transfusion (less than 2 weeks); and 5) patients who died or were discharged from the ICU within 4 h of admission were excluded because data collection was difficult for these patients. Participants were recruited consecutively from January 2009 to March 2010.

Laboratory methods

The RDW, hemoglobin level, and mean corpuscular volume (MCV) were determined using the Beckman Coulter LH-750 Hematology Analyzer (Beckman Coulter, Inc., Fullerton, California), as one part of a complete blood cell count. The normal reference range for RDW in the laboratory of our hospital is 11.6%–15.0%. In order to assess the reproducibility of RDW measurement, the RDW of 60 patients randomly chosen were measured twice. The coefficient of variability was 2.25% (0.32%/14.25% × 100%). Serum creatinine (SCr) and albumin were measured by Hitachi 7600-120 (Hitachi, Tokyo, Japan) analyzer. We calculated the estimated glomerular filtration rate (eGFR) using the abbreviated Modification of Diet in Renal Disease (MDRD) study equation: eGFR (expressed in mL/min/1.73 m²) = 186 × (SCr) – 1.154 × (age) – 0.203 × 0.742 (if female), where SCr is serum creatinine in mg/dL (13). Serum C-reactive protein (CRP) levels were measured using Quick Read CRP test kit (Orion Corporation, Orion Diagnostica, Espoo, Finland).

Study outcomes

At base-line, demographic and clinical characteristics, including the Acute Physiology and Chronic Health Evaluation II (APACHE-II) score (which can range from 0 to 71, with higher scores indicating more severe illness), were collected. Then the patients were followed up during the hospitalization. The primary outcome for this analysis was the ICU mortality. We also recorded the total length of stay in the hospital (TLSH) as a second outcome for analysis.

Statistical analysis

All continuous variables were presented as mean value ± SD, and categorical data were summarized as percentages. CRP and eGFR values were logarithmically normalized (presented as log-CRP and log-eGFR, respectively) for statistical calculations. We divided RDW into tertiles and compared demographics, clinical characteristics, laboratory test results, and APACHE-II score, with analysis of variance or Kruskal-Wallis tests for continuous variables and chi-square or Fisher's exact tests for categorical variables. Then, Bonferroni's post-hoc test was performed to evaluate the differences of APACHE-II score and TLSH among patients in different tertiles. Univariate logistic regression analysis was utilized to examine the association between the mortality and each of the predictors separately. We also conducted a multivariate stepwise logistic regression to determine the independent predictors of ICU mortality. Stepwise multivariate linear regression was also performed to determine the factors independently associated with TLSH. A criterion of P < 0.05 for entry and a P ≥ 0.10 for removal was imposed in this procedure. Odds ratios (ORs) for continuous variables were described using standardized ORs, which was associated with a 1-SD change in the variable. The receiver operating characteristic (ROC) curve was used to examine the performance of APACHE-II score and RDW to predict ICU mortality. The curve represented a plot of sensitivity versus 1–specificity. The area under the curve (AUC) was derived from the ROC curve. A statistically derived value, based on the Youden index, maximizing the sum of the sensitivity and specificity was used to define the optimal cut-off value (14). A ROC curve was also constructed for the combination of APACHE-II score and RDW for predicting ICU mortality according to the Mackinnon and Mulligan's weighted sum rule (15). Weighted sum formula: logit (case) = 0.198 × APACHE-II score + 0.202 × RDW – 7.322, where logit (case) is the logarithm of the odds of a subject dying in ICU. The differences between AUC were tested by Hanley-McNeil methods (16). Correlation between two continuous variables was assessed by Pearson correlation analysis or Spearman correlation analysis. A two-sided P value of less than 0.05 was considered to indicate statistical significance. All analyses were performed with SPSS 13.0 software.

Results

Base-line characteristics and base-line factors related with RDW

A total of 602 participants (58.1% male; mean age 70.39 ± 16.73 years) were eligible for this study. RDW ranged from 11.2% to 26.9% (median 13.9%; mean 14.5 ± 20.1%), and the mean APACHE-II score was 13.2 ± 7.0. Differences in clinical and laboratory characteristics among the three tertiles of RDW are listed in Table I. Patients with in the higher tertile of RDW tended to be older, were more likely to have pulmonary disease and accompanying infection, were less likely to have poisoning, had lower levels of hemoglobin, eGFR, and serum albumin, and had higher CRP values. The APACHE-II scores in tertile I, tertile II, and tertile III were 10.7 ± 6.5, 13.3 ± 6.2, and 15.8 ± 7.2, respectively. There was a stepwise increase in APACHE-II score with increasing tertile of RDW (all P < 0.001) (Figure 1A). The sex ratio and MCV were not different among the three tertiles of RDW.

Figure 1. Comparisons of Acute Physiology and Chronic Health Evaluation (APACHE) II score (A), ICU mortality rate (B) and the total length of stay in the hospital (TLSH) (C) among patients with different red blood cell distribution width (RDW) tertiles.

Table I. Base-line clinical and laboratory characteristics by tertile of RDW.

	Tertile I	Tertile II	Tertile III	P
RDW (%)	12.9 ± 0.4	14.0 ± 0.4	16.6 ± 2.2	<0.001
	(11.2–13.4)	(13.5–14.6)	(14.7–26.9)
n	205	202	195	–
Male (%)	58.5	57.9	58.0	1.000
Age (years)	61.8 ± 19.3	73.7 ± 13.0	76.0 ± 13.4	<0.001
Principal diagnosis leading to ICU admission (%)
ICU admission (%)
Pulmonary disease	25.3	29.7	37.9	0.022
Cardiovascular disease	36.1	39.6	30.3	0.145
Neurological disease	10.2	8.9	4.1	0.056
Digestive disease	8.3	4.0	5.6	0.178
Infectious disease/sepsis	2.4	5.9	7.7	0.057
Renal insufficiency	1.0	2.5	4.1	0.141
Poisoning	8.3	4.5	2.1	0.015
Trauma	2.9	1.0	1.5	0.323
Other	5.4	4.0	6.7	0.485
Accompanying infection (%)	32.2%	46.0%	56.4%	<0.001
Hemoglobin (g/L)	124.9 ± 19.5	119.0 ± 23.9	109.5 ± 25.8	<0.001
MCV (μm^3)	92.9 ± 4.3	92.6 ± 5.1	92.0 ± 9.5	1.000
APACHE-II score (points)	10.7 ± 6.5	13.3 ± 6.2	15.8 ± 7.2	<0.001
eGFR (mL/min/1.73 m^2)	85.3 ± 42.5	70.5 ± 42.8	60.3 ± 41.8	<0.001
Albumin (g/L)	35.1 ± 4.8	34.0 ± 4.5	31.7 ± 5.1	<0.001
CRP (mg/L)	55.7 ± 57.3	59.7 ± 57.4	69.5 ± 61.6	0.027

APACHE-II score = Acute Physiology and Chronic Health Evaluation II score; CRP = C-reactive protein; eGFR = estimated glomerular filtration rate; MCV = mean corpuscular volume; RDW = red cell distribution width.

Association of RDW with ICU mortality

There was a significantly graded increase in ICU mortality rate across increasing RDW tertile (tertile I = 11.2% versus tertile II = 18.8% versus tertile III = 33.8%; all P < 0.001) (Figure 1B). Univariate logistic regression analysis demonstrated that those older, with higher RDW, higher APACHE-II score, higher CRP, lower eGFR, lower hemoglobin level, and lower albumin concentration had significantly greater death hazard (Table II). To evaluate the value for RDW and APACHE-II score to predict ICU mortality, a ROC curve was drawn (Figure 2). The AUC was calculated as 0.832 ± 0.020 (P < 0.001) for APACHE-II score and 0.672 ± 0.027 (P < 0.001) for RDW. The optimal cut-off value of APACHE-II score for predicting death was ≥ 15 points, which gave a sensitivity of 78.7% and a specificity of 73.9%. The optimal cut-off value of RDW (≥ 14.8%) provided sensitivity of 51.2% and specificity of 74.7%. To further clarify whether RDW had an additive power with APACHE-II score for ICU mortality, we combined RDW and APACHE-II score to draw a third ROC curve in Figure 2. As compared with APACHE-II score, combination of RDW and APACHE-II score increased AUC for predicting ICU mortality from 0.832 ± 0.020 to 0.885 ± 0.017 (P < 0.05) (Figure 2), suggesting that combining RDW and APACHE-II added to the ability to discriminate mortality risk. To further understand the additive prognostic value of RDW, we divided each RDW and APACHE-II score into two (low and high) groups according to the cut-off value for predicting ICU mortality and examined the odds ratio for ICU mortality in each group. Using low RDW (< 14.8%) and APACHE-II score (< 15 points) as the reference group, we found a trend toward increasing odds ratio in other groups (Figure 3). Among them, high RDW (≥ 14.8%) and high APACHE-II score group (≥ 15 points) had the highest risk of ICU death (OR 23.33; 95% CI 12.17–44.82; P < 0.001), as compared with the reference group (low RDW and low APACHE-II group). The multivariate logistic regression analysis showed that only log-CRP, RDW, and APACHE-II score can predict primary outcome (Table III).

Figure 2. ROC curve for Acute Physiology and Chronic Health Evaluation (APACHE) II score, red blood cell distribution width (RDW) and combination of both in predicting ICU mortality.

Figure 3. Odds ratio for ICU mortality by logistic regression analysis in four groups. Red blood cell distribution width (RDW) as an additive marker with Acute Physiology and Chronic Health Evaluation (APACHE) II score for predicting ICU mortality.

Table II. Univariate odds ratios of variables for predicting ICU mortality.

Variables	Odds ratio	95% CI	P
Male	1.007	0.677–1.497	0.974
Age (years)	1.796	1.382–2.337	<0.001
RDW (%)	1.784	1.475–2.158	<0.001
Hemoglobin (g/L)	0.653	0.543–0.786	<0.001
MCV (μm^3)	0.851	0.699–1.037	0.107
APACHE-II score (points)	4.200	3.152–5.594	<0.001
Log-eGFR (mL/min/1.73 m^2)	0.463	0.378–0.566	<0.001
Albumin (g/L)	0.505	0.406–0.626	<0.001
Log-CRP (mg/L)	1.779	1.431–2.211	<0.001

Odds ratios for continuous variables shown as standardized odds ratios (OR per 1 SD). Abbreviations as in Table I.

Table III. Independent predictors of ICU mortality by multivariate logistic regression analysis.

Predictor	Odds ratio	95% CI	P
RDW (%)	1.551	1.250–1.926	<0.001
APACHE-II score (points)	3.034	2.027–4.172	<0.001
Log-CRP (mg/L)	1.400	1.078–1.819	0.012

Independent variables included sex, age, log-eGFR (estimated glomerular filtration rate), log-CRP, hemoglobin, RDW, mean corpuscular volume, APACHE-II score, albumin, accompanying infection, pulmonary disease, and poisoning. Variables not listed in the table were removed from the stepwise analysis. Odds ratios for all variables shown as standardized odds ratios (OR per 1 SD). Abbreviations as in Table I.

Association of RDW with TLSH in patients alive

There were 443 patients discharged alive for hospice. Patients with RDW in the lowest tertile had shorter TLSH than those with RDW in tertiles II and III (13.7 ± 12.5 days versus 17.8 ± 14.8 and 18.7 ± 14.5; P < 0.05) (Figure 1C). Age, hemoglobin, APACHE-II score, log-CRP, and albumin were also associated with TLSH by univariate analysis (r = 0.165, –0.097, 0.173, 0.104, and –0.150, respectively; all P < 0.05), whereas gender, MCV, and log-eGFR were not correlated with TLSH (P > 0.05). On multivariate analysis, the independent predictors of TLSH were RDW, age, and log-CRP (Table IV).

Table IV. Independent predictors of TLSH by multivariate linear regression analysis.

Predictor	Unstandardized coefficients	Standardized coefficients	P
Age (years)	0.012	0.179	<0.001
RDW (%)	0.079	0.114	0.019
Log-CRP (mg/L)	0.129	0.097	0.039

Independent variables including sex, age, log-eGFR (estimated glomerular filtration rate), log-CRP, hemoglobin, RDW, mean corpuscular volume, Acute Physiology and Chronic Health Evaluation II (APACHE-II) score, albumin, accompanying infection, pulmonary disease, and poisoning. Variables not listed in the table were removed from the stepwise analysis.

TLSH = total length of stay in hospital; other abbreviations as in Table I.

Discussion

The present study is a prospective clinical investigation of the prognostic value of RDW in ICU patients. The primary finding of this study is that increasing RDW levels can serve as a strong independent predictor of greater mortality in ICU patients. This association remained significant even after adjustment for APACHE-II score. Combining RDW and APACHE-II score added to the ability to discriminate mortality risk. We also found that RDW was independently associated with the length of hospital stay in patients alive. CRP was also shown to be an independent predictor of ICU mortality and length of hospital stay. To our knowledge, this represents the first report of elevated RDW as a potential prognostic marker in ICU patients. A recent meta-analysis of seven studies with over 11,000 community-dwelling older adults showed that RDW was strongly associated with multiple causes of death and long-term mortality within major demographic and disease subpopulations (17). The current study extends this previous work by demonstrating the prognostic power of RDW among critical care patients, predicting ICU mortality as well as length of stay. Thus, the clinical utility of RDW may potentially span from primary care to acute care settings.

The way to predict clinical outcome of ICU patients is not only useful in allocating resources (18,19) but is also helpful in monitoring treatment progress, comparing therapeutic efficacy, and comparing performance (20,21) in different centers. The APACHE-II score system developed in 1985 has shown a positive correlation with hospital mortality and length of hospital stay in ICU and is one of the most common models to evaluate patients’ condition (22). In accordance with previous studies (22), the optimal cut-off value of APACHE-II score (≥ 15) was also demonstrated to have a strong power to predict ICU mortality (AUC 0.832 ± 0.020; P < 0.001; sensitivity 78.7%; specificity 73.9%) in the present study. For the first time, we reported that RDW can be used for predicting ICU mortality, although the prediction power of RDW was relatively low (AUC 0.672 ± 0.027; P < 0.001; sensitivity 51.2%; specificity 74.7%). ROC curves in Figure 2 show that combining RDW and APACHE-II score added to the ability to discriminate mortality risk, and RDW was additive with APACHE-II score for predicting ICU mortality. Patients with high APACHE-II score and high RDW levels carry the highest risk of mortality (Figure 3). We also found that RDW was independently associated with TLSH in patients alive.

The pathophysiologic mechanism underlying the association of higher RDW with worse outcomes in ICU patients has yet to be defined. RDW is frequently high in situations of ineffective red cell production (such as iron deficiency, B₁₂ or folate deficiency, and hemoglobinopathies), increased red cell destruction (such as hemolysis), or after blood transfusion (23). Patients with higher RDW values tended to be older, were more likely to have accompanying infection, had lower levels of hemoglobin, eGFR, and serum albumin, and had higher CRP values in the present study. These results indicated that elevated RDW may conceivably represent an integrative measure of multiple harmful pathologic processes simultaneously occurring in critical illness (e.g. anemia, renal dysfunction, malnutrition, aging, and inflammation). Particularly, elevated RDW may be indicative of rapid red blood cell demise because of underlying inflammation, which is associated with adverse outcome (24–26). However, this hypothesis cannot completely explain the association of RDW with clinical outcomes, because the association still persisted in analyses after adjusting for CRP and APACHE-II score. RDW may be a reflection of general membrane integrity, and a high RDW might be a surrogate for membrane instability, which might adversely affect the functions of many organ systems (27). Future studies are needed to define further the underlying mechanism.

CRP is an exquisitely sensitive objective marker of inflammation, tissue damage, and infection. Its value for predicting outcome in ICU patients was just evaluated in a few studies (24–26). In the present study, CRP was also an independent predictor of ICU mortality and length of hospital stay. This result was in line with previous studies (24–26).

There are several limitations to this study. Firstly, we did not investigate the causes of elevated RDW, such as iron or vitamin B₁₂ deficiency, which could confound the relationship between RDW and adverse outcome. However, several other potential confounders, such as hemoglobin, mean corpuscular volume, and APACHE-II score, had been adjusted for. Secondly, this was a single-center study, and participants do not include surgery patients. The value for RDW in prediction of adverse outcome would be a bit different if the population was different. As such, the findings need to be confirmed in multi-center and prospectively designed studies. And lastly, RDW was not dynamically observed. Whether RDW level was stepwise elevated when patient's condition was progressive deteriorated was unclear.

In conclusion, we found a graded independent relation between higher RDW and adverse outcome in ICU patients. Because RDW is widely available at no additional cost to the routinely performed complete blood cell count and is highly reproducible, RDW has the potentially clinical utility to predict outcome for patients in ICU. CRP can also be as an independent predictor of clinical outcomes in ICU patients.

Acknowledgements

The authors would like to thank Kai Hu MD (Department of Medicine, University of Würzburg, Würzburg, Germany) for English editing for this paper.

Declaration of interest: The authors state no conflict of interest and have received no payment in preparation of this manuscript.