Translated by Liang Yan, Department of Critical Care Medicine, Cangnan County People's Hospital
Critically ill Walker Translation Group
Importance Red blood cell transfusion (RBC) transfusion is common in patients hospitalized in the ICU. Although several randomized clinical trials have been conducted on hemoglobin (Hb) threshold levels for transfusion, little is known about how these thresholds are incorporated into current practice.
Objective To evaluate and describe the practice of red blood cell transfusion in ICU worldwide.
Design, location, and participants The International prospective cohort study involved 3643 adult patients from 233 ICUs in 30 countries on 6 continents and was studied from March 2019 to October 2022, with data collected over a pre-specified few weeks.
Exposure ICU hospitalization
Primary outcomes and measures Primary outcome was whether red blood cells were transfused during ICU hospitalization. Other outcomes included indications for red blood cell transfusion (including clinical and physiologic causes), Hb threshold levels before and after red blood cell transfusion and actually measured Hb level values, and number of units transfused.
Results Of the 3908 potentially eligible patients in 233 ICUs in 30 countries on 6 continents (each providing a median patient case of 11 [IQR, 5-20]), 3643 patients were enrolled. The mean (standard deviation) age of patients was 61 (16) years, 62% were male (2267/3643), and the median SOFA score score was 3.2 (IQR, 1.5 to 6.0). A total of 894 patients (25%) received 1 or more red blood cell transfusions during their ICU stay, with a median cell transfusion of 2 units (IQR, 1-4) per patient. The proportion of patients receiving blood transfusions ranged from 0% to 100% in centres, from 0% to 80% across countries and from 19% to 45% across continents. A total of 1727 units of red blood cells were transfused in patients receiving blood transfusions, with the most common clinical indication being low Hb (n = 1412 [81.8%]; Means [standard deviation] Nadir Hb before transfusion, 7.4 [1.2] g/dL), active bleeding (n = 479; 27.7%), and haemodynamic instability (n = 406 [23.5%]). Among events with clear physiologic triggers, the most common triggers were hypotension (n = 728 [42.2%]), tachycardia (n = 474 [27.4%]), and elevated lactate levels (n = 308 [17.8%]). The median minimum Hb level on the day of red blood cell transfusion ranged from 5.2 g/dL to 13.1 g/dL at the centre, from 5.3 g/dL to 9.1 g/dL across countries, and from 7.2 g/dL to 8.7 g/dL on all continents. Approximately 84% of ICUs have a median Hb level greater than 7 g/dL at transfusion.
Conclusions and implications Between 2019 and 2022, red blood cell transfusions were common in ICU-admitted patients worldwide, but transfusion practices varied widely from center to center.
Red blood cell transfusions are common in patients admitted to the ICU. During 2021, more than 1.7 million units of red blood cells were transfused in ICU wards in the United States alone. Red blood cell transfusion is a potentially life-saving treatment, but may be limited by the increasing scarcity of blood products and pose patients with risks such as transfusion-related circulatory overload and transfusion-related acute lung injury.
Multiple randomised clinical trials (RCTs) in a variety of critically ill patients have shown that more restrictive transfusion regimens (i.e., low haemoglobin [Hb] levels with 7-8 g/dL as the threshold for red blood cell transfusion) are safe, and the latest guidelines reflect restrictions on red blood cell transfusion. However, there is considerable uncertainty about the optimal transfusion management strategy for different patient populations. There is also uncertainty about current transfusion practice and the clinical and physiological causes that determine transfusion.
To fill this knowledge gap, we conducted a prospective observational study globally to describe the incidence, clinical causes, physiological causes, and heterogeneity between centers over 3 years in 30 countries on 6 continents.
method
Study design and supervision
The International Epidemiological Study of ICU Transfusion Practice (InPUT) is an international, multicenter, prospective, observational cohort study of transfusion practice. Recruitment to participate in the centres is guided by the National Coordinator, the Intensive Care Association and through direct contact with the members of the Steering Committee. The study was approved by the Institutional Review Board of the University Medical Center of Amsterdam and subsequently by national and local ethics committees. The research process is in line with the Declaration of Helsinki. The informed consent procedure is guided by national regulations, including written or oral informed consent from patients and/or legally authorized representatives, which in some countries is waived due to the observational and non-invasive nature of the study. The income status of each participating country is extracted from the 2023 World Bank classification system. Protocols and standard operating procedures are provided in the appendix material.
Based on the outcomes of a pilot feasibility study, participating centres aim to recruit all new adult patients admitted to their ICUs within a scheduled week. Physiological data were collected daily during ICU hospitalization for each included patient during the study cycle until day 28. For each transfusion unit, data were collected for the indications for transfusion and outcomes up to day 28.
Data collection is scheduled for March 2020 to February 2021. Due to the COVID-19 pandemic, two major adjustments were made to data collection efforts. First, the original 8-week inclusion period was extended to 16 weeks, culminating in the completion of data collection in January 2022. Second, due to delays in the COVID-19 wave in Australia and New Zealand, data collection was rescheduled from January 2022 to October 2022.
All patients admitted to the ICU during the study week who were 18 years of age or older were included in the study (table 1). Patients are excluded from further data analysis if the available data do not meet minimum quality standards (see section "Statistical analysis" for details), or if the patient or legal representative does not provide informed consent (if required by local or national law).
Table 1. Demographic characteristics by red blood cell transfusion status
Abbreviations: APACHE, Acute Physiology and Chronic Health Assessment; ECMO, extracorporeal membrane oxygenation; ENT, ENT; EuroSCORE, the European Risk Assessment System for Cardiac Surgery; ICU, intensive care unit; OR, operating room; RBC, red blood cells.
a Frequency data are expressed in people (%), nonparametric variables as median (IQR, 1st to 3rd quartile boundary), and parametric numeric variables as mean (standard deviation).
b Others include epilepsy, dementia, cerebral infarction or intracerebral hemorrhage, peripheral vascular disease, diabetes mellitus and obstructive sleep apnea.
c The APACHE IV score is used to reflect the severity of a patient's condition and is expressed on a scale of 0-286, with higher severity leading to higher scores.
d The European Level II score is used to predict mortality in patients undergoing cardiac surgery and is expressed as a percentage of in-hospital mortality from 0 to 100.
e Other supportive care includes noninvasive mechanical ventilation, renal replacement therapy, and temporary cardiac pacing.
Data Collection
Data collection begins at ICU admission and is collected daily until day 28 of ICU hospitalization or before death or discharge, whichever comes first. Daily data include laboratory values (minimum haemoglobin, minimum platelet count, coagulation parameters) and organ supportive care (e.g., respiratory and renal support). In addition, the clinical team reports on daily emerging complications during ICU hospitalization, including bleeding, heart, lung, and kidney complications. The patient's prognosis is assessed until day 28 of ICU admission. Transfusion events refer to transfusions of red blood cells, platelets, plasma, whole blood, vitamin K, prothrombin complex concentrate, fibrinogen, cryoprecipitate, and tranexamic acid, as well as the selection of a large transfusion regimen. If multiple transfusions are performed in a day, each transfusion creates a separate transfusion event. For each transfusion event, we collect the factors that contribute to the decision on transfusion and classify them as clinical causes (such as haemoglobin levels or age) and physiological triggers (such as low blood pressure or high lactate levels).
finale
The primary outcome was the incidence of red blood cell transfusion during ICU hospitalization, i.e. reception of 1 or more units of red blood cells. This includes large transfusion regimens or the use of whole blood in countries where separate RBC transfusion products are not available. Secondary outcomes included: (1) clinical causes of red blood cell transfusion; (2) physiological triggers of red blood cell transfusion; (3) Hb level values before and after red blood cell transfusion; (4) Total number and number of units of ICU blood transfusion per day. Hb level values before and after red blood cell transfusion include the last value measured before transfusion (pre-transfusion), the first value measured after transfusion (post-transfusion), and the Hb threshold at which transfusion can be started as specified in the central protocol or, in the absence of a protocol, expert opinion on the patient (threshold).
Statistical analysis
Where appropriate, descriptive data are reported as mean (standard deviation) or median (IQR). Missing data are counted and assessed for relevance, and patient data are excluded if they do not meet quality standards. The minimum quality criteria include the following requirements: (1) the number of questionnaires must match the number of questionnaires expected or declared (i.e., when 6 transfusion events are declared for one patient, 6 separate transfusion questionnaires must be produced); (2) All variables assessed as mandatory must be provided. Details of these quality standards are provided in eMethods 1 in Appendix 1 and an overview of missing data in the cleansed dataset (e-Table 5 in Appendix 1).
The statistical analysis included 4 comparisons: (1) patient characteristics with or without transfusion; (2) primary and secondary outcomes by centre, country and continent; (3) primary and secondary outcomes for postoperative versus non-surgical patients; (4) Secondary outcomes between different thresholds in transfusion events.
First, the Mann-Whitney U or chi-square test was used to compare patients who received and did not receive a transfusion and performed Bonferroni correction. Primary and secondary outcomes and postoperative indications were reported using descriptive statistical methods across centres, countries and continents. Second, a subgroup analysis of the transfusion thresholds listed in the transfusion questionnaire was performed to classify the transfusion thresholds into restrictive transfusion thresholds (haemoglobin level <7 g/dL), intermediate thresholds (haemoglobin levels, 7-9 g/dL), and relaxed transfusion thresholds (haemoglobin levels >9 g/dL), comparing secondary outcomes between different thresholds. This subgroup analysis is corrected for no multiplicity and should therefore be interpreted as hypothesis-generated. Third, the median Hb level at transfusion at each center correlates with the absolute number and proportion of ICUs.
All analyses were performed in R (version 4.2.1) using the R studio interface. A bilateral p-value of less than 0.05 by Bonferroni-adjusted is statistically significant.
outcome
A total of 3908 patients from 233 centers in 30 countries on 6 continents were eligible. After excluding patients who did not obtain or abandoned informed consent and those who did not meet data quality criteria, 3643 patients (93% of eligible patients) remained available for further analysis. The median number of patients included in each ICU analysis was 11 (IQR, 5-20). The six continents include Africa (n = 150), Asia (n = 182), Europe (n = 2167), South America (n = 50), North America (n = 167), and Oceania (n = 927). Of the 30 participating countries, 19 were classified as high income, five as upper middle-income and six as low- and middle-income countries.
The baseline characteristics of the study subjects are shown in Table 1. The mean (standard deviation) age was 61 (16) years, and 62% were male (n = 2267). Thirteen patients signed a "no blood transfusion" agreement. Elective admission was in 35% (n = 1271), and the main reasons for admission included postoperative monitoring (n = 1042 [82%]) and respiratory failure (n = 90 [7%]). The median length of hospital stay was 3 days (IQR, 2-6) and the median SOFA score during the period was 3.2 (IQR, 1.5-6.0). The 28-day mortality rate was 17%, with the majority of patients dying during ICU hospitalization (443/618 [72%]). Approximately 81% of all patients who survived the 28-day follow-up stay in the ICU were discharged.
Red blood cell transfusion
The proportion of patients receiving transfusions ranged from 0% to 100% across centres: 22 centres in 11 countries had no transfusion events, while 14 centres in 8 countries received transfusions for all patients. A total of 894 patients (25%) received one or more red blood cell transfusions during their ICU stay (table 2). A total of 1727 transfusion events were recorded, with a median transfusion event of 2 (IQR, 1-4) per patient and a median RBC of 1 unit (IQR, 1-2) per transfusion event. In addition, patients receiving transfusions received a median of 0.5 units of RBC per day (IQR, 0.25-1) during their ICU stay, adding up to a median of 2 units of total RBC (IQR, 1-4).
Table 2 Characteristics during ICU hospitalization
Abbreviations: ARDS, Acute Respiratory Distress Syndrome; EPO, erythropoietin; Hb, hemoglobin; ICU, intensive care unit; RBC, red blood cells; SOFA, sequential organ failure assessment; VHA, viscoelastic hemostasis assay.
a Frequency data are expressed as "number (%)", nonparametric data as "median (1st to 3rd quartile boundary)", and parametric numeric variables as "mean (standard deviation)".
The b SOFA score is an indicator of the process of organ dysfunction within the ICU, on a scale of 0-24, with higher scores in severe organ dysfunction.
c According to the World Health Organization, anaemia is defined as less than 12 g/dL in women and less than 13 g/dL in men.
d Calculated by dividing the total number of red blood cells transfused per patient by the number of days transfused.
e Calculated by dividing the total number of red blood cells transfused by the length of ICU stay.
Thirty-one percent of patients after surgery received red blood cell transfusions with a median total transfusion of 2 units (IQR, 1-4), compared with 20% of non-postoperative patients received an infusion with a median total infusion of 2 units (IQR, 1-3). The proportion of patients receiving daily transfusions ranged from 11% (n = 416/3643) on admission to 5% (n = 46/973) on day 5. Most transfusions were performed during the day (n = 1063 [63%]).
The proportion of transfusions varies from 0% to 80% of patients across countries and from 19% to 45% across continents. The highest incidence of blood transfusions was in Africa (45% of patients), while the highest number of transfusions and transfusion units per patient was in South America (1 [IQR, 1-6], 4 units [IQR, 3-6]; Spreadsheet 9 in Appendix 1).
An overview of the clinical causes and physiologic causes of blood transfusion is shown in Table 3. In all patients, the main clinical causes of transfusion were low hemoglobin levels (81.8%), active bleeding (27.7%), and hemodynamic instability (23.5%). Physiologic causes are mainly hypotension in 42.2% of patients, tachycardia in 27.4% of patients, and elevated lactate levels in 17.8% of patients. In 39.5% of transfusion events, no physiologic predisposing factor supported the transfusion decision.
Table 3 Characteristics of red blood cell transfusion events
Abbreviations: ECG, ECG; ED, Emergency Department; Hb, hemoglobin; ICU, intensive care unit; RBC, red blood cells.
a Frequency data are expressed in people (%), nonparametric numeric variables are expressed as median (IQR shows the 1st and 3rd quartile boundaries), and parametric numeric variables are expressed as mean (standard deviation).
b Critical care medical training varies worldwide; Some countries start with other medical specialties (i.e. surgery, anaesthesiology, internal medicine), while others offer specialized intensive care medicine training courses. The latter is called an intensive care physician here.
c Hb transfusion of red blood cells is recommended in the ICU procedure or, in the absence of a procedure, expert opinion on the patient.
dThe most recent Hb value before red blood cell transfusion, usually within 4 hours before the decision to transfuse blood.
e Hb value first measured within 24 hours after RBC infusion.
f Calculated as the difference between the measured Hb values before and after RBC infusion.
g Calculated as the difference between previously measured Hb levels and Hb transfusion thresholds. If the change in Hb level is negative, it indicates that the patient has complied with the treatment regimen (i.e., the measured Hb level is below the specified threshold); If positive, the patient has not yet reached the threshold.
h Factors that determine blood transfusion, including clinical causes (rationale) or physiological parameters (triggers).
On all continents, the main clinical causes of blood transfusion are low Hb values, active bleeding, and haemodynamic instability, except in Africa, where 33% of red blood cell transfusions are to improve the general state of the patient. The main physiological triggers are hypotension and tachycardia in all continents except South America, where elevated lactate levels (>2 mmol/L) are one of the main physiological triggers for red blood cell transfusion (39%). Africa and North America had the highest number of causes per transfusion (2 [IQR, 1-3]), while South America had the highest number of triggers per transfusion (2 [IQR, 0-2]) (see spreadsheet 10 and e-chart 3 in Appendix 1).
The process and threshold of changes in hemoglobin
On ICU admission, anaemia was present in 60% of patients (1<2 g/dL [n = 653 {60%}] in women and 13 g/dL in men < [n = 1093 {60%}]; n = 1746/2917）。 In the ICU, this rises to 85% (n = 3044/3588). Almost all patients receiving red blood cell transfusions developed anaemia during ICU admission (n = 882/891 [99%]). Patients who received transfusion had lower haemoglobin levels on admission and lower mean haemoglobin levels and lowest haemoglobin levels during ICU admission (all P <.001) compared with patients who did not receive transfusions.
Figure 1 shows the mean hemoglobin level at transfusion at each center, ranging from 5.2 g/dL to 13.1 g/dL. As shown in Figure B, about 16% of ICUs have an average haemoglobin level below 7 g/dL at transfusion. Median pre-transfusion Hb levels ranged from 5.3 g/dL to 9.1 g/dL across countries and from 7.2 g/dL to 8.7 g/dL across continents (Appendix 1, Figure 4). The median increase in Hb level after transfusion was 1.2 g/dL (IQR, 0.72.0) in all red blood cell transfusion events. In events where low Hb levels were the cause of transfusion, the median (SD) of the lowest Hb level before transfusion was 7.4 (1.2) g/dL. In 81% of transfusion events, regardless of indication for transfusion, pre-transfusion Hb levels were below or equal to the specified threshold (n = 963/1188). In events where transfusion thresholds were prescribed (n = 1188), the thematic mean (standard deviation) of Hb levels before transfusion was 8.5 (1.5) g/dL, within the relaxed threshold range (n = 459), and 6.95 (1.4) g/dL (n = 299) within the limit threshold. In addition, the first 3 clinical causes and physiological triggers of transfusion are the same in these threshold categories (see Electronic Table 11 and Figure 5 in the appendix).
Figure 1 Median hemoglobin (Hb) level before red blood cell (RBC) transfusion in patients receiving blood transfusion during ICU hospitalization.
A Frequency distribution of median hemoglobin levels at transfusion
B Cumulative frequency distribution of median hemoglobin at transfusion
For patients with multiple transfusions of red blood cells, calculate the average pre-transfusion Hb level for different transfusions. In panels A and B, the 7 g/dL vertical dashed line represents the current RBC transfusion guidelines. Approximately 84% of intensive care units have a median Hb level above 7 g/dL at transfusion.
discuss
This prospective international study was conducted in 233 ICU wards in 30 countries on 6 continents, and 25% of patients received one or more red blood cell transfusions during their ICU stay. The incidence of blood transfusion varies widely across centres, countries and continents. Although the clinical causes and physiologic triggers of red blood cell transfusion are diverse, the three most common clinical causes (low Hb levels, active bleeding, haemodynamic instability) and physiologic triggers (hypotension, tachycardia, no physiologic triggers affecting transfusion decisions) overlap to a large extent in all settings. There are also large differences between centers in terms of minimum Hb levels prior to red blood cell transfusion.
The incidence of red blood cell transfusions reported in this study is consistent with the results of an ICON study conducted more than 10 years ago, which showed a 26% transfusion rate in the ICU. While transfusion rates are trending downward, similarities between the Icon study and current results may mean that the trend is heading towards plateau. The initial decrease in transfusion volume may be due to increased awareness of the possible adverse effects of blood transfusion, or acceptance of the evidence generated by several large clinical trials. The causes and triggers of blood transfusion are diverse, and haemoglobin levels at transfusion vary widely, confirming the heterogeneity of transfusion behaviour globally. It is worth noting that in the course of this study, the European Society of Critical Care Medicine issued two guidelines on the management of blood transfusion in bleeding and non-bleeding patients, and the current results cannot assess the extent of their implementation.
Low Hb levels are the most common cause of RBC infusion. In recent decades, multiple RCT studies have concluded that it is safe to use the Hb threshold of restrictive transfusion to trigger RBC infusion in different patient populations. Hb values before red blood cell transfusion were lower and the standard deviation was small (mean [standard deviation], ICON: 8.3 [2.2] g/dL vs InPUT: 7.7 [1.6] g/dL) compared to the Icon study. Previous surveys have shown that ICU physicians' requirements for hemoglobin thresholds vary widely. In this study, the average minimum Hb level on the day of transfusion and the proportion of patients receiving the transfusion varied greatly between centers.
The appropriateness and feasibility of individualized red blood cell transfusion thresholds and triggers have been discussed. Although multiple indicators and methods have been studied, uncertain results using these indicators and methods limit the development of optimal personalization methods. The current study has found that the clinical causes and physiological triggers of blood transfusion are diverse. The European Society of Critical Care Transfusion Guidelines recommend the use of hemoglobin or hematocrit in non-bleeding patients rather than other triggers such as venous oxygen saturation, while other triggers and causes are not mentioned at all in bleeding patients. The absence of any mention of other triggers and causes is consistent with other international guidance. In this context, it is important to note that most transfusion events are at least partially caused by one or more physiologic (i.e., non-haemoglobin) triggers.
Red blood cell transfusion is associated with mortality, with mortality as an endpoint in most clinical trials. In addition, observational studies have reported an increased risk of death in patients with severe anemia without red blood cell transfusions. Although studies have found a higher proportion of patients who receive a transfusion to die before day 28, this should not be interpreted as indicating that receiving a transfusion results in a higher probability of death because potential complications are not corrected. Previous studies have shown conflicting results on the relationship between blood transfusion and mortality, and there may be differences between subgroups. This research question is beyond the scope of this paper, but these and other sub-issues will be studied in the future (eMethods 2 in Appendix 1).
To our knowledge, this study is one of the largest prospective cohorts designed to date to study this topic and the first study to broadly outline the clinical causes and physiological triggers that determine blood transfusion. The study was conducted in an international context, involving multiple centres and countries. Selection bias was small as the study included all patients admitted to the ICU and the proportion of eligible to included participants was very high.
limitations
The study had several limitations. First, while international recruitment has taken place, most of the centres involved in recruitment reflect upper-middle-income countries, which may limit outreach to low-income countries. Second, transfusion variables were collected from patients who had received 1 or more transfusions. Therefore, the exact transfusion threshold for patients who do not receive transfusions as specified in the central protocol is unknown. Third, no data were collected on the racial and ethnic composition of the population. Fourth, during the data collection period, the COVID-19 pandemic began. Although no center has reported blood product shortages or COVID-19 as the cause of (not) blood transfusions, additional post-hoc analysis of the impact of the pandemic on transfusion behavior is not possible. Fifth, hospital characteristics, including hospital type (academic vs. peripheral) and number of beds, were not collected.
conclusion
RBC transfusions were common in an international sample of patients admitted to the ICU between 2019 and 2022, but there were differences in transfusion incidence and indications by center.
essentials
Question What is the status of blood transfusions among ICU inpatients worldwide?
Results In this international prospective cohort study, 25% of 3643 patients from 233 ICUs in 30 countries received one or more RBC transfusions during their ICU stay, with a median total RBC transfusion of 2 units per patient. The proportion of red blood cells transfused varies from 0% to 100% across centers, from 0% to 80% across countries, and from 19% to 45% across continents.
Significance Red blood cell transfusions are common among ICU hospitalized patients worldwide, but transfusion practices vary widely from center to center.