Anaemia is a common feature in critically ill patients. In the recent ABC study 1, haemoglobin level at admission was below the normal range in 63% of patients admitted to the intensive care unit (ICU). A low haemoglobin level is associated with poor prognosis in critically ill patients 12, as was previously described in elderly patients with acute myocardial infarction 3.

Because anaemia commonly occurs in the ICU, red blood cell (RBC) transfusion is a frequent practice in the management of critically ill patients to compensate for acute bleeding and to increase tissue oxygen delivery 4. Canadian and European surveys reported that up to 40% of the patients admitted to the ICU receive at least one RBC transfusion 156. However, RBC transfusion carries short-term and long-term side effects, and liberal transfusion strategies have been associated with a worse outcome in ICU patients 6. In an effort to avoid unnecessary RBC transfusion, intensivists have defined haemoglobin thresholds above which transfusion appears harmful 789. They also proposed the use of erythropoietin 10 to avoid RBC transfusion. Although these measures may decrease blood transfusion in the ICU, they could have the opposite effect on need for transfusion after the ICU stay. Indeed, anaemia often persists or worsens after ICU discharge 1. The ABC study 1 clearly identified the frequent need for post-ICU RBC transfusion, because 12.7% of patients who enrolled needed RBC transfusion after their ICU stay. Whether efforts to limit blood transfusion in the ICU just delay administration of RBC to the post-ICU period is unclear. In addition, although predictive factors for the need for RBC transfusion in the ICU have been identified 1611, risk factors for transfusion after ICU discharge are unknown. We conducted this prospective monocentric observational study to identify risk factors for RBC transfusion in critically ill patients after discharge from the ICU.

During the study period, we enrolled 550 consecutive patients who were admitted to our ICU (Table 1). Most patients (90.2%) were admitted for medical diagnosis, whereas the remaining patients were admitted for emergency (7.3%) and elective (2.5%) surgery. The overall mortality rate was 22% (122 patients). Mean haemoglobin level on admission was 11.4 ± 2.5 g/dl. Twenty per cent of the population received RBCs during their stay in the ICU. Mean haemoglobin at discharge was 10.3 ± 2.3 g/dl. Forty-seven out of the 428 patients discharged from the ICU received RBC transfusion within 1 week after discharge (group I) whereas 381 remained free from transfusion at 1 week (group II). Hospital mortality rates did not differ between the two groups.

Characteristics that differed between the groups are summarized in Table 2. Patients from group I were older and predominantly female. Patients were more likely to receive transfusion after ICU discharge if they were admitted for sepsis, or had hypotension or a medical history of malignancy. Patients admitted to the ICU for a respiratory disorder or drug poisoning were significantly less transfused than others. Severity scores on admission and discharge were higher among post-ICU transfused patients (Table 2). However, patients with haemodynamic instability requiring vasopressors surprisingly required less RBC transfusion. The mean haemoglobin level at admission in patients who received RBC transfusion within a week after ICU discharge was 8.6 g/dl. Haemoglobin level on admission and, as expected, at discharge from ICU was lower among post-ICU transfused patients.

We performed a multiple logistic regression analysis to determine variables independently associated with increased risk for RBC transfusion after ICU discharge (Table 3). Admission for sepsis (OR 341.60, 95% CI 20.35–5734.51), presence of an underlying malignancy (OR 32.6, 95% CI 3.8–280.1), female sex (OR 5.4, 95% CI 1.2–24.9), LOD score at ICU discharge (OR 1.45, 95% CI 1.1–1.9) and age (OR 1.06, 95% CI 1.02–1.12) were independently associated with RBC transfusion after ICU stay. The use of vasopressors (OR 0.01, 95%CI 0.001–0.17) and haemoglobin level at discharge from the ICU (OR 0.02, 95% CI 0.007–0.09; P < 0.001) were strong independent predictors of transfusion of RBCs 1 week after ICU discharge.

We performed the present prospective study specifically to evaluate the need for RBC transfusion during the post-ICU period. We observed that 9% of critically ill patients treated in a medical ICU required RBC transfusion after ICU discharge, and that few parameters influenced need for transfusion of RBCs within 7 days after the ICU stay.

Of patients discharged from the ICU, 11% (8.5% of the whole cohort) required RBC transfusion after ICU discharge. This is consistent with the findings of the ABC study 1, in which RBC transfusion after ICU discharge occurred in 12.7% of the population. In the present study, a considerable proportion of the patients were surgical, and this might have influenced the need for RBC transfusion.

Our study revealed that only few parameters influenced the risk for transfusion of RBCs after ICU discharge. Haemoglobin level at admission is a well established risk factor for transfusion of RBCs during the ICU stay 19. We found that haemoglobin level on ICU admission was inversely correlated with the risk for transfusion in univariate analysis but not in the multivariate model. Conversely, haemoglobin level at ICU discharge markedly influenced requirement for RBC transfusion during the 7 days following ICU discharge. In our study, a 1 g/dl decrease in haemoglobin level increased by 50-fold the risk for RBC transfusion during the post-ICU stay. Previous studies demonstrated that age was strongly associated with anaemia in the critically ill. In the ABC study 1 the mean haemoglobin level at admission was significantly lower in patients older than 90 years than in patients younger than 50 years (9.9 g/dl versus 11.7 g/dl). In addition, older patients received more transfusions. We also observed a 1.06-fold increase in the likelihood of RBC transfusion for each additional year. This could be explained by an increased incidence of co-morbidities. Indeed, elderly patients frequently present with coronary artery diseases for which haemoglobin threshold values for transfusion are not clearly defined 914. However, neither a previous medical history of heart disease nor cardiac disorder as the cause of admission emerged as a risk factor in our analysis. Conversely, we found that the presence of an underlying malignancy was an independent risk factor for RBC transfusion after ICU discharge. Solid neoplastic diseases occur frequently in the elderly. It is well known that they are responsible for anaemia and that their specific treatments have myelotoxic effects. The ABC study 1 also found a decreased haemoglobin level in patients who had a previous history of anaemia, especially in the setting of neoplastic disorder.

One of the most important factors associated with post-ICU transfusion of RBC was sepsis as an admission diagnosis. A previous study reported that septic patients had decreased haemoglobin levels as compared with the remainder of the ICU population 11. Sepsis could impair production of erythropoietin by several mechanisms, including release of proinflammatory mediators that negatively impact erythropoiesis 615. We recently reported that sepsis can induce anaemia by increased apoptosis of bone marrow erythroid progenitors 16. The severity of sepsis could also lead to a greater volume of blood sampling for laboratory analysis in these patients 11.

Sepsis is frequently associated with organ failure. Whereas severity scores at admission (Simplified Acute Physiology Score II and LOD score) were reported as risk factors for ICU transfusion, they did not influence the need for transfusion after the ICU stay. On the other hand, a higher LOD score at ICU discharge was related to increased risk for RBC transfusion after the ICU stay (OR = 1.45 for each additional LOD point). Patients with persisting organ dysfunctions on ICU discharge more frequently required RBC transfusion during the remainder of their hospital stay. Surprisingly, we observed that use of vasopressors decreased the risk for RBC transfusion. The reasons for this finding are unclear. No patient had haemodynamic instability or was receiving ongoing vasopressor therapy at ICU discharge.

Our study has some limitations. First, no specific guidelines regarding RBC transfusion were given to the physicians involved in patient care after the ICU stay. Variations in the transfusion thresholds as well as in iron and vitamin supplementation policies in the various medical wards might have affected our results. Second, we limited the evaluation period to the first 7 days following ICU discharge to ascertain whether RBC requirement was directly related to the ICU stay. This delay was chosen bearing in mind the natural history of haematological disorders and the time course of myelotoxicity of drugs used in the ICU. Although we acknowledge that the validity of a 7-day period of observation is debatable, selection of the optimal follow-up period remains difficult because no study has specifically adressed this issue.

Our study suggests that sepsis, underlying conditions, unresolved organ failures and haemoglobin level at discharge are related to a increased risk for RBC transfusion after ICU stay. Most of these findings are consistent with previous studies that addressed the risk for transfusion in the ICU. These findings should be considered when defining transfusion guidelines, because a higher haemoglobin level may be required in specific subgroups of ICU patients. We suggest that any strategy to prevent transfusion in the ICU should focus on homogeneous subgroups of patients and take into account post-ICU needs for RBC transfusion.

• Nine per cent of critically ill patients treated in a medical ICU require RBC transfusion after ICU discharge when strict transfusion guidelines are applied in the medical ICU.

• Sepsis, underlying conditions, unresolved organ failures and haemoglobin level at discharge constitute risk factors for RBC transfusion after ICU stay.

CI = confidence interval; ICU = intensive care unit; LOD = Logistic Organ Dysfunction; OR = odds ratio; RBC = red blood cell.

The authors declare that they have no competing interests.

Sophie Marque, Alain Cariou, Jean-Daniel Chiche and Yann-Erick Claessens contributed to the design of the study and drafted the manuscript. Vincent Olivier Mallet, Frédéric Pene, Jean-Paul Mira and Jean-François Dhainaut obtained the data. Sophie Marque, Alain Cariou and Yann-Erick Claessens participated in the data analysis and interpretation of the results.