The prevalence and association with transfusion, intensive care unit stay and mortality of pre-operative anaemia in a cohort of cardiac surgery patients


  • Presented in part at the Association of Cardiothoracic Anaesthetists Annual Autumn Meeting, Brighton, UK, November 2010

Dr Andrew Klein


Anaemia is increasingly prevalent in the United Kingdom. Despite recommendations to the contrary, many patients undergo cardiac surgery with undiagnosed and untreated anaemia. According to the World Health Organization definition, 1463/2688 (54.4%) patients undergoing cardiac surgery between 2008 and 2009 in our institution were anaemic. Compared with non-anaemic patients, anaemia was significantly associated with transfusion (791 (54.1%) vs 275 (22.4%), p < 0.001, OR (95% CI) 3.4 (2.8–4.1)), death (45 (3.1%) vs 13 (1.1%), p = 0.0005, OR 2.4 (1.2–4.5)), and prolonged ICU stay (287 (19.6%) vs 168 (13.7%) p < 0.001, OR 1.3 (1.0–1.6)). The prevalence of anaemia in this cohort is much greater than that previously reported. The cause of this excess is not clear. Pre-operative anaemia is a strong predictor of increased transfusion requirement, risk of ICU stay and death during cardiac surgery. The effect of increasing haemoglobin concentration therapeutically is not yet clear.

The World Health Organization defines anaemia as haemoglobin concentration (Hb) less than 13 g.dl−1 for men and 12 g.dl−1 for women [1]. Anaemia is becoming increasingly prevalent in the ageing population in the UK and in developed countries in general, with an estimated prevalence of 17% among people aged 65 years and over, and up to 40% in people admitted to hospitals [2]. Consequently, anaemia is frequently encountered during the pre-operative assessment of patients for surgery. Anaemia has been related to a variety of medical illnesses. Among older people, approximately one third of cases of anaemia are due to nutritional deficiencies of iron, folate and/or vitamin B12, another third are due to chronic disease and/or renal insufficiency, and the remaining third are unexplained [3].

A UK observational study reported that severe pre-operative anaemia (Hb ≤ 10 g.dl−1) was a marker of comorbidity that had a major effect on survival after coronary artery bypass surgery [4]. However, the study did not provide data on the prevalence of pre-operative anaemia using the World Health Organization definition, or its impact on postoperative outcomes. Several non-UK studies have also demonstrated an association between pre-operative anaemia and increased morbidity and mortality after cardiac surgery [5–7], however, the population of patients included in these studies may differ markedly from those in the UK. Understanding the effects of a mild to moderate degree of anaemia has a particular clinical relevance in the modern era, given that it is common clinical practice for anaemic cardiac surgical patients to undergo scheduled surgery provided they have a value of Hb > 10 g.dl−1.

In addition, the Health Service Circular ‘Better Blood Transfusion’, published in 2007 [8], recommended that all efforts should be made to diagnose and correct anaemia in advance of planned surgery. Despite this, there is a paucity of published UK data on this important public health issue in cardiac surgery. Therefore, we designed a study to estimate the prevalence of pre-operative anaemia in a cohort of UK patients presenting for elective cardiac surgery, using the World Health Organization definition of anaemia. We also assessed the relationship between the presence of anaemia and three outcomes in these patients: requirement for transfusion; prolonged stay in ICU; and death in hospital.


The study was conducted according to the STROBE guidelines for cohort studies [].With the approval of the Cambridgeshire Research Ethics Committee, we studied all consecutive patients presenting for elective cardiac surgery at a single cardiothoracic surgical centre over a 2-year period (January 2008–December 2009). These included surgery with and without cardiopulmonary bypass. Patients undergoing urgent or emergency procedures were not studied. No formal sample size calculations were conducted. All data were collected prospectively and stored in a surgical database, and the requirement for patient consent was waived.

In this cohort study the ‘exposure’ of primary interest was presence of anaemia, defined as Hb < 12 g.dl−1 for women and < 13 g.dl−1 for men, as measured by the laboratory on the day before surgery. A limited number of variables that may affect outcomes were recorded and adjusted for in the analysis; these were surgical procedure, patients’ age and logistic Euroscore. Surgical procedures were categorised as coronary artery bypass graft (CABG) alone, single-valve procedure, double-valve procedure, combined CABG and valve, or other. Other operations included re-do sternotomy, congenital heart defect procedures and operations on the aorta.

All patients routinely received 2000 mg tranexamic acid after induction of general anaesthesia. For procedures with high risk of bleeding, tranexamic acid infusion, 500 mg.h−1, was also administered until the end of surgery. During the study period, Papworth Hospital used a strict transfusion protocol. Haemoglobin levels were maintained ≥ 8.0 g.dl−1 intra- and postoperatively, whilst the transfusion trigger was < 7.0 g.dl−1 during cardiopulmonary bypass.

The primary outcome measure was the requirement for peri-operative transfusion of packed red blood cells. Secondary outcome measures were in-hospital death from all causes, length of stay in ICU and a crude assessment of the costs of transfusion, given as the number of units of blood products multiplied by the unit cost for these products. The distribution of ICU stay was skewed with a very long tail; therefore this was grouped into (1) 0–2 days, and (2) > 2 days. Patients were not followed up beyond discharge from hospital and there was no loss to follow up.

In exploratory analyses, outcomes for patients with and without anaemia were compared using Pearson’s chi-squared for contingency tables or Wilcoxon signed-rank test as appropriate. Logistic regression models were developed to assess the relationship between Hb and the outcome variables. Initially the confounders logistic Euroscore, patient age and operation type were included in the model, and then anaemia was included to assess the additional predictive value. Significance values from the likelihood ratio tests comparing models with and without anaemia are presented. For completeness, a propensity score analysis was undertaken; the propensity score was based on age, logistic Euroscore and operation type. Since this did not substantially alter the estimates, it is not reported. Both the mean cost per patient (as recommended in health economic studies) and the median cost per transfusion were calculated and the groups were compared using the Wilcoxon signed-rank test.


During the study period, a total of 2688 patients underwent elective cardiac surgery at Papworth Hospital and were included in the study. Isolated CABG accounted for 44% of the procedures, 22% were single-valve surgery, 2% were combined-valve surgery, 15% were combined CABG and valve surgery, and 17% were other type of surgery (Table 1).

Table 1.  Characteristics of patients undergoing cardiac surgery. Anaemia defined as haemoglobin concentration < 12 g.dl−1 in women or < 13 g.dl−1 in men. Values are mean (SD) or number (proportion).
 Non-anaemic (n = 1225)Anaemic (n = 1463)
  1. CAGB, coronary artery bypass graft surgery.

Age; years64.5 (11.9)70.8 (10.3)
Women292 (22.8%)469 (32.1%)
Body mass index; kg.m−228.6 (4.6)27.5 (4.7)
Logistic Euroscore4.6 (5.7)6.9 (7.4)
Pre-operative haemoglobin; g.dl−113.7 (0.9)11.4 (1.1)
 CABG572 (46.7%)603 (41.2%)
 Single-valve282 (23.0%)308 (21.1%)
 Double-valve22 (1.8%)34 (2.3%)
 CABG + valve148 (12.1%)261 (17.8%)
 Other201 (16.4%)257 (17.6%)

The prevalence of pre-operative anaemia was 54.4% (1463/2688), according to the World Health Organization definition. The mean (SD) Hb in anaemic patients was 11.4 (1.1) g.dl−1 vs 13.7 (0.9) g.dl−1 in non-anaemic patients. Severe anaemia, defined as Hb < 10 g.dl−1, was present in 149 patients (5.5%). Patients with anaemia were significantly older, more likely to be women, had lower body mass index, were more likely to undergo complex surgery and were at higher risk of postoperative death as assessed by logistic Euroscore (Table 1). Overall, 1066 patients had a transfusion (40%), 58 patients died (2.2%), and 455 patients (17%) stayed in ICU for more than 2 days. There were significant differences between the non-anaemic and anaemic groups in the incidence of transfusion, death and prolonged length of ICU stay (Table 2 and Fig. 1). In addition, anaemic patients were transfused significantly more units of red blood cells and their postoperative Hb was significantly lower. In the cohort of patients who were not transfused (1622 in total), there was no significant difference in mortality or prolonged ICU stay between those patients who were and were not anaemic.

Table 2.  Outcome data of patients undergoing cardiac surgery. Anaemia defined as haemoglobin concentration < 12 g.dl−1 in women or < 13 g.dl−1 in men. Values are number (proportion), median (IQR [range]) or mean (SD).
 Non-anaemic (n = 1225)Anaemic (n = 1463)p value*
  1. RBC, red blood cells. *p value refers to the Wilcoxon signed-rank test or Pearson’s chi-squared test for contingency tables.

Transfusion275 (22.4%)791 (54.1%)< 0.0001
Total units of RBC transfused0 (0–1 [0–20])0 (0–2 [0–34])< 0.0001
Units of RBC if transfused2 (1–3 [1–20])2 (1–3 [1–24])< 0.0001
Transfusion > 6 units RBC25 (2%)102 (7%)0.098
Postoperative haemoglobin10.5 (3.1)9.2 (3.4)< 0.001
In-hospital deaths13 (1.1%)45 (3.1%)0.0005
ICU stay; days1 (0–2 [0–81])1 (0–2) [0–69]< 0.0001
ICU stay > 2 days168 (13.7%)287 (19.6%)< 0.0001
Transfusion cost per patient; £133 (0–410 [0–4205]362 (0–795 [0–4205]< 0.0001
Figure 1.

 Odds ratio for each outcome for anaemic patients relative to non-anaemic patients. The error bars represent the 95% confidence intervals.

Logistic regression demonstrated an increased risk of transfusion associated with increased age, logistic EuroSCORE and complexity of surgery (Table 3). After adjusting for these variables, pre-operative anaemia was still significantly associated with increased risk of transfusion.

Table 3.  Multiple-variable logistic regression analysis of the probability of transfusion, in-hospital mortality and ICU stay more than 2 days. Values are odds ratio (95% CI).
 Transfusionp value*Mortalityp value*Prolonged ICU stayp value*
  1. CABG, coronary artery bypass surgery. *p value from likelihood ratio test comparing model with all variables included and a model excluding each variable.

Age (per 5 years)1.1 (1.1–1.2)< 0.011.0 (0.9–1.2)0.821.1 (1.0–1.1)0.09
Logistic EuroSCORE (per point)1.1 (1.0–1.1)< 0.011.1 (1.0–1.1)< 0.011.0 (1.0–1.1)< 0.01
CABG aloneReference< 0.01Reference0.20Reference< 0.01
Single-valve1.0 (0.8–1.3)1.1 (0.5–2.6)0.7 (0.5–1.0)
Double-valve1.9 (1.0–3.4)4.2 (1.3–13.6)1.5 (0.8–3.0)
CABG + valve2.5 (1.9–3.3)1.8 (0.8–3.9)1.9 (1.4–2.6)
Other2.3 (1.8–3.1)1.7 (0.8–3.9)2.3 (1.7–3.1)
Non-anaemicReference< 0.01Reference< 0.01Reference0.04
Anaemic3.4 (2.8–4.1)2.4 (1.2–4.5)1.3 (1.0–1.6)

Only 58 patients died in hospital, so that the power to detect significant associations was low, and confidence intervals from the logistic regression models were wide. However, operative risk (as expected) and anaemia were significantly associated with in-hospital death (Table 3). Operative risk, complexity of procedure and anaemia were all significantly associated with prolonged ICU stay (Table 3). Of the patients who died, 55 (95%) were transfused. There was a very strong relationship between transfusion and hospital death in the single variable logistic regression (OR 29.4 (95% CI 9.2–94.2), p < 0.001); including anaemia in this extended model was not significant (OR 1.2 (95% CI 0.6–2.4), p = 0.54). Transfusion was also strongly associated with prolonged hospital stay (more than 2 days), OR 5.8 (95% CI 4.5–7.2), p < 0.001; again, the effect of anaemia was not significant (OR = 0.8 (95% CI 0.6–1.0), p = 0.08). In patients who were not transfused, there were no significant differences between anaemic and non-anaemic patients (Table 4).

Table 4.  Outcomes in 1622 patients who were not transfused blood in the peri-operative period, out of total cohort of 2688 patients. Anaemia is defined as haemoglobin concentration < 12 g.dl−1 in women or < 13 g.dl−1 in men. Values are number (proportion) or median (IQR [range]).
 Non-anaemic (n = 950)Anaemic (n = 672)p value
In-hospital death1 (0.1%)2 (0.3%)0.7630
ICU stay; days1 (0–1 [0–43])1 (0–1 [0–51])0.2938
ICU stay; > 2 days80 (8.4%)40 (6.0%)0.0759

A crude costing suggested that the anaemic group incurred an additional average transfusion cost of £228.60 per patient compared with the non-anaemic group (Table 2). For those who were transfused, the cost of transfusion did not differ between the two groups. Thus, any saving in costs resulted from the lower transfusion rate rather than a difference in cost per transfusion.


Our study has three main findings. First, pre-operative anaemia was a highly prevalent condition among adult patients presenting for elective cardiac surgery at our centre. More than half of the patients in our cohort had an Hb that fulfilled the World Health Organization’s definition of anaemia. This high prevalence of pre-operative anaemia was an unexpected finding, and is much higher than has been reported in the literature previously. We were not able to determine the cause and duration of pre-operative anaemia, but our results indicate that anaemia has a close relationship with the ageing process, severity of the concurrent co-morbidities and complexity of the presenting cardiac surgical condition.

Second, pre-operative anaemia was an independent and strong predictor of peri-operative red blood cell transfusion. Although the severity of anaemia was only mild, i.e. haemoglobin concentration ≥ 10 g.dl−1, in 94.5% of anaemic patients, this translated into a three-and-a-half-fold increase in the odds of peri-operative transfusion for these patients. The increased incidence of transfusion in the anaemic group also resulted in higher financial costs.

Third, pre-operative anaemia was strongly associated with increased in-hospital mortality and postoperative adverse outcomes, as suggested by prolonged ICU stay. This has important clinical implications, because pre-operative anaemia in cardiac surgery is a potentially treatable condition [9]. If pre-operative anaemia was indeed a cause of postoperative adverse outcomes, then by correcting a low Hb in advance of elective cardiac surgery, the mortality rate and the number of prolonged ICU admissions could be reduced. This of course assumes that the anaemia can be corrected without causing morbidity and mortality per se.

However, there are important alternative explanations to be considered when examining the relationship between pre-operative anaemia and adverse postoperative outcomes in our study. Only three confounding variables were included in the multiple-variable models. Euroscore is a risk evaluation scoring system based on patient, cardiac and operation-related factors, and designed to quantify risk of hospital death in cardiac surgical patients [10]. The higher mean Euroscore values in the anaemic group of patients reflected a higher number of comorbidities and other risk factors than in the non-anaemic group of patients. In the multiple-variable analysis, both logistic Euroscore and operation type were independently associated with prolonged length of ICU admission, whilst logistic Euroscore (not surprisingly) was also associated with in-hospital mortality. In addition, anaemia and transfusion are very strongly linked, and we have not been able to demonstrate an association between anaemia and outcome that is independent of transfusion. There is certainly good evidence that transfusion on its own is associated with worse outcome [11], and we cannot say based on these results whether or not one is more important than the other.

This study has shown a much higher prevalence of anaemia than has been previously been reported; previous studies have shown an incidence of 21–36% in cardiac surgical patients [5–7]. The reason for this disparity is not clear, but population differences and variation in the definition of anaemia are likely to have contributed. Certain pre-operative factors have been identified as risk factors for anaemia. Karkouti el al. reported a higher percentage of the anaemic patients in their cohort to be thrombocytopenic, coagulopathic with an elevated INR, and receiving heparin or clopidogrel pre-operatively [7]. These patients may be anaemic due to occult blood loss via the gastrointestinal tract, which is the commonest cause of iron deficiency anaemia in adult men and post-menopausal women. Female sex and lower body mass index or body surface area were also significantly associated with pre-operative anaemia [5, 6], which we have also demonstrated to be the case in our cohort of patients. Peri-operative allogeneic blood transfusion is associated with a significant risk of morbidity and mortality [11], as well as financial cost. Consistent with our finding, a low pre-operative Hb has been repeatedly found to be a strong predictor for peri-operative transfusion in cardiac surgery [12–16]. The decision to transfuse anaemic patients is often based on the assumption that administration of red blood cells enhances oxygen-carrying capacity as well as improving haemostasis associated with blood component therapy [17]. On the contrary, Murphy et al. [18] reported that the adjusted odds ratios for composite infection (respiratory or wound infection or septicaemia) and ischaemic outcomes (myocardial infarction, stroke, renal impairment, or failure) for transfused versus non-transfused cardiac surgical patients were 3.38 and 3.35, respectively.

Surgenor et al. [19] found that when red blood cell transfusions were administered to treat haemodilutional anaemia during cardiopulmonary bypass for CABG surgery, the exposure to both haemodilutional anaemia and transfusion were associated with increased risk of low output heart failure, with the risk being greater among patients exposed to intra-operative transfusion versus anaemia alone. When the transfused anaemic patients were compared with anaemic patients who had received no transfusion, Kulier et al. also showed that pre-operative anaemia and intra-operative blood transfusion were both independent but additive risk factors for adverse outcomes [6]. Based on the evidence available in the literature, it is possible that the high rate of transfusion in the anaemic group may have contributed to the observed increase in mortality and prolonged length of ICU stay in our study. Our study demonstrated that pre-operative anaemia had an important and independent prognostic value for adverse peri-operative outcomes. This is in keeping with other studies that have shown anaemia to be an important risk factor for increased morbidity and mortality in a variety of patient populations. Anaemia was an independent risk factor for cardiovascular disease outcomes in the general population [20], and was associated with increased risk for mortality and hospitalisation in the elderly [21, 22]. Anaemia was also associated with markedly reduced survival in patients with coronary artery disease [23, 24] and chronic heart failure [25]. In the non-cardiac surgical setting, pre-operative anaemia was independently associated with increased mortality [26]. In a retrospective study of 7759 non-cardiac surgical patients, Beattie et al. [27] found that the anaemic patients had a more than twofold increase in the odds of postoperative mortality, even after the elimination of patients with severe anaemia and patients who received transfusion.

Given the evidence to date showing the detrimental effects of both anaemia and blood transfusion, the primary strategy to maintain an optimal peri-operative Hb in cardiac surgical patients should be to correct pre-operative anaemia and to avoid blood transfusion. In a recent randomised controlled trial, Hajjar et al. showed that a restrictive strategy of blood transfusion (haematocrit ≥ 24%) vs liberal strategy (haematocrit ≥ 30%) resulted in comparable mortality and morbidity outcomes in cardiac surgical patients [28]. This restrictive strategy is consistent with the transfusion protocol utilised in our study.

Some limitations need to be considered in the interpretation of our study results. We cannot say that anaemia was the sole cause of transfusion, although it certainly made it more likely. It is possible that pre-operative anaemia is simply a marker of the severity of the presenting surgical condition and concurrent comorbidities, so that the relationship with adverse postoperative outcomes is biased. In addition, the effects of other unmeasured confounding factors may play a part. However, the major strength of our study is minimal selection bias, as the large sample size contained all consecutive patients undergoing elective cardiac surgery at our hospital over a 2-year period.

Our study has several important clinical implications. Firstly, the very high prevalence of pre-operative anaemia detected in elective cardiac surgical patients at our hospital requires further investigation, especially with respect to the cause of the anaemia. Secondly, pre-operative anaemia is potentially an important modifiable risk factor in cardiac surgery. If anaemia truly has a causal relationship with blood transfusion and peri-operative morbidity and mortality, correcting pre-operative anaemia (e.g. with iron or an erythropoiesis-stimulating agent) may lead to a reduction in transfusion and improved postoperative outcome. This has yet to be tested prospectively, including a detailed analysis of any potential negative impacts and the cost implications of such a measure. Both iron and erythropoiesis-stimulating agent therapies have the potential to delay scheduled surgery, and there have been safety concerns in their use in this setting [29]. Thirdly, pre-operative anaemia may serve as a valuable prognostic indicator in elective cardiac surgery. Of the six commonly used risk scores in assessing cardiac surgical patients, Cleveland Clinic Score is the only one that utilises anaemia as a risk score item [30]. The incorporation of pre-operative anaemia into the risk prediction model may improve the accuracy of pre-operative risk stratification of cardiac surgical patients, and allow better allocation of peri-operative resources for patient care.

In conclusion, we have demonstrated the highest incidence of anaemia in pre-operative patients that has been reported to date, and also demonstrated that anaemia and transfusion are inexorably linked together, and associated with a worse outcome after cardiac surgery.


The authors wish to acknowledge the assistance of Mrs Caroline Gerrard in collecting the data. AAK is an Editor of Anaesthesia and this manuscript underwent an additional external review as a result. No external funding and no competing interests declared.