Has ICU Transfusion Practice Changed in the Past Decade?

Abstract & Commentary

Synopsis: This multicenter prospective observational study reveals that transfusion practice in the critically ill has not changed in the past decade. 44% of ICU patients receive transfusions, and transfusions continue to be associated with worse clinical outcomes.

Source: Corwin HL, et al. Crit Care Med. 2004;32(1): 39-52.

In the past 2 decades, risks of red blood cell (RBC) transfusion in the critically ill have begun to be defined more clearly and the safety of lower transfusion thresholds in this population has been well documented.1-3 The effect of these findings on current clinical practice in the United States has not been described previously. Corwin and colleagues aimed to quantify current RBC transfusion practice in critically ill patients. A secondary endpoint was to describe outcomes associated with anemia and transfusion in this cohort.

A multicenter prospective observational cohort study was conducted from August 2000 to April 2001 at US ICUs. Adult patients with anticipated ICU length of stay (LOS) > 48 hours were eligible. Patients admitted to pediatric, cardiac, cardiothoracic and neurology ICUs were excluded. Patients were also excluded for renal failure on dialysis or if they were prohibited from receiving transfusions. Usual demographics, admitting diagnoses, severity of illness scores, baseline, weekly and pre-transfusion hemoglobin (Hgb) levels, and number and age of RBC transfusions were recorded for all patients. Studied outcomes were mortality, ICU and hospital LOS, ventilator days and a variety of clinical complications. Patients were followed for 30 days or until death or hospital discharge (if these occurred within 30 days). Standard statistical analyses were used.

In this study, 284 ICUs at 213 (70% urban) hospitals participated. There was a fairly equal distribution of medical, surgical and combined units. A total of 4892 patients were enrolled. Baseline Hgb was 11.0 ± 2.4 g/dL and was not predictive of poorer outcomes: nadir Hgb of < 9 g/dL though was associated with increased mortality and LOS. 44% of patients received RBC transfusion in the ICU with mean 4.6 ± 4.9 units. About another 5% received transfusions in their post-ICU hospital stay. The most common reason given for transfusion was low Hgb: mean pre-transfusion Hgb was 8.6 ± 1.7 g/dL. This value did not differ for transfusions given post-ICU or across most hospital characteristics: however, the threshold for transfusion was minimally but statistically significantly lower in closed vs open ICUs, medical vs surgical or combined ICUs and small (< 12 beds) vs larger ICUs.

Transfusion was less likely to be given in patients with lower severity of illness. It was not related to patient age or to most co-morbidities (except anemia, hematologic disease and gastrointestinal bleeding) but was associated with increased ICU and hospital LOS. Transfusion reactions occurred 4% of the time. Age of RBC transfused was not associated with any clinical outcome. Finally, even after multivariate regression including propensity scores for transfusion, RBC transfusion was associated with increased risk of death.

Comment by Saadia R. Akhtar, MD, MSC

Corwin et al have produced a thought-provoking, well-conducted observational study that focuses the critical care community’s attention once again on the topic of RBC transfusion. Their finding that transfusions are associated with worse clinical outcomes is consistent with prior data. The observation that a specific nadir Hgb has the same association is intriguing and deserves further study. But, to me, the most interesting and important result here is that transfusion practice has not changed despite developing data about risks of transfusion and quite robust data demonstrating the safety of lower transfusion thresholds.1-3

Is this a false or biased finding from a flawed study? I do not believe so. The study is well designed. It was conducted at multiple centers of varying characteristics all across the United States and carried out over several months: the results should be generalizeable. There is reasonably good evidence that we should have higher transfusion thresholds for patients with cardiac ischemia. For this reason, patients admitted to cardiac or cardiothoracic units were excluded from this study. Twelve percent (12%) of study patients did have an admitting diagnosis that was categorized as cardiovascular and some had co-morbid cardiac disease. Thus there is a possibility that this may have had an effect on the results, but it would be small. The majority of ICUs in the study (71%) were open units. Perhaps transfusion practice in closed units, by physicians specifically trained in critical care, is more conservative. Corwin et al found no difference between closed and open units in their study. The number of closed units sampled may have been too low to detect a small difference but it is at least clear that a large difference does not exist.

I suspect that the reasons transfusion rates have not changed over time are the same reasons it has been difficult to incorporate many new data and standards of care into practice. For instance, since Semmelweis’s observations that the incidence of puerperal fever was reduced by hand washing, we have known that hand hygiene is a simple and effective way to prevent transmission of contagious diseases. Yet adherence with hand washing remains at best 50-60%.4 Aspirin prescription post-myocardial infarction is at similarly low rates despite its well-established mortality benefit.5 In critical care, use of recommended interventions such as elevation of head of bed for ventilator-associated pneumonia prevention or low tidal volume ventilation for acute lung injury is also suboptimal.6,7 Despite the growth of clinical research and enthusiasm about evidence-based medicine, it is clear that actually translating evidence into practice change is quite challenging.

Many investigators are now studying this very issue by trying to identify the barriers to implementation of new standards and guidelines and developing approaches to overcoming them. One potential barrier is not believing the data. It is possible that the critical care community is not convinced that a lower transfusion threshold (< 7 g/dL) is safe or that transfusion has considerable risks (of increased infection and risk of death) aside from immediate reactions. More and more supportive evidence is accumulating and may convince practitioners over time. The greater barriers may be not being aware of the data or not understanding how to begin to implement it. Increased efforts to disseminate information and educate care providers are necessary: employing the help of recognized local leaders and teachers in these efforts has been shown to be quite effective. For instance, a practice taught, endorsed and carried out by the director of the ICU may lead to change. Extending this further to develop and disseminate specific local guidelines/protocols is helpful in changing practice. This must, however, be combined with continuous updating and follow-up to achieve consistent compliance. Interestingly, only 19% of hospitals in this study had a transfusion protocol in place. Although the authors did not find a significant difference in practice associated with having a protocol, they did not obtain information about what the protocols recommended. Thus it is difficult to make conclusions about the effectiveness of protocols. Finally, concerns about potential risks or costs of a new practice may be another potential barrier. Education as well as formal audits and cost analyses can help to allay such concerns.8

I expect that Corwin et al’s report will encourage us to strive to further understand the impact of RBC transfusions and its underlying mechanisms. I hope that it also pushes us to focus more on incorporating new evidence into our practices.


1. Hebert PC, et al. N Engl J Med. 1999:340:409-417.

2. Vincent JL, et al. JAMA. 2000:288:1499-1507.

3. Taylor RW, et al. Crit Care Med. 2002:30:2246-2254.

4. Hugonnet S, et al. Arch Intern Med. 2002:162: 1037-1043.

5. Simpson E, et al. Am Heart J. 2003:145:438-444.

6. Heyland DK, et al. J Crit Care. 2002:17:161-167.

7. Rubenfeld GD. Am J Resp Crit Care Med. 2001:163:A295.

8. Kalassian KG, et al. Crit Care. 2002:6:11-14.

Saadia R. Akhtar, MD, MSc, Pulmonary and Critical Care Medicine, Yale University, School of Medicine, is Associate Editor for Critical Care Alert.