Haemoglobin level in Scottish Intensive Care Patients

Department of Anaesthesia, South Glasgow University Hospitals NHS Trust and University of Glasgow, *Department of Anaesthesia, Royal Alexandra Hospital, Paisley, Robertson Centre for Biostatistics, University of Glasgow.

Correspondence to: Dr M Garrioch Department of Anaesthesia, South Glasgow University Hospitals NHS Trust and University of Glasgow, Glasgow. G51 4TF.

Abstract:

Objectives were to determine haemoglobin (Hb) levels present in patients and blood ordering habits of clinicians within Scottish Intensive Care Units (ICUs’) on one typical day. A questionnaire survey (February 29 2000) was sent to all adult Scottish ICUs. All patients present in the responding adult ICU’s in Scotland on the above date were included.

Measurements and main results: Nineteen (73%) of the 26 Scottish Adult Intensive Care Units (ICUs) responded to the questionnaire. Data were received from 78 patients, 8 (10%) received blood. Mean initial Hb was 102 g/l (range 63-138). Modal transfusion trigger haemoglobin was 80 g/l in 38% of subjects at first trigger, 100 g/l in 24% of cases. No intensive care unit allowed haemoglobin to fall below 70 g/l and no patients were transfused when measured Hb was greater than 100 g/l. The presence of ischaemic heart disease was the second most important trigger to transfuse after haemoglobin level. Modal transfusion was 2 units (n=7). Only one patient received a single unit transfusion.

Conclusions: Scottish ICUs maintain Hb between 70 and 100 g/l but clinicians are currently not consistent when ordering blood. More investigation is required to determine the optimal haemoglobin in our ICU population.

Recent scrutiny regarding the use of blood in the critically ill
has resulted from a high quality, randomised, controlled study conducted by Hebert and colleagues.1 Thirty day outcome was compared in 25 Canadian ICUs. One of two transfusion policies was followed, liberal, where Hb was sustained above 100 g/l or restrictive where an Hb of 70 to 90g/l was maintained. Hebert’s group concluded that a restrictive transfusion policy, in certain groups of patients, was at least as safe as and possibly superior to a more liberal approach to blood transfusion.

The purpose of our survey was to determine whether Scottish ICU’s were adopting the newer practices, i.e. restricting blood transfusion to Hb levels of between 70 and 90 g/l suggested by the Canadian study or whether a more traditional transfusion policy remained the norm. We wished to audit the Hb levels in as many Scottish adult ICU patients as possible by conducting a one-day "snap shot" survey. We also wished to examine if age, ischaemic heart disease or disease severity influenced the transfusion trigger. The survey also provided useful data for further studies of blood use in the Scottish ICU population.

There are 26 intensive care units in Scotland, nine within teaching hospitals and 17 within district generals. All now participate in the Scottish Intensive Care Society Audit Group (SICSAG) but at the time of the survey only those 25 units that were members of SICSAG were approached to take part in the survey. Patient information was verified and cross-referenced from the central SICSAG database to improve the accuracy of the data.

Ethical advice was sought form the Ethics Committee of the South Glasgow University Hospitals NHS Trust and as patient anonymity was maintained the survey was considered ethical.

Questionnaires were then issued to all 25 ICU’s invited to participate in the survey. Each ICU received a follow-up telephone call on 28 February 2000. Data were collected between 08.00 29 February 2000 and 08.00 1 March 2000. All Hb levels measured in this 24-hour period were recorded. Twenty-four hour Acute Physiology and Chronic Health Evaluation (APACHE) score on admission to the unit concerned (a measure of severity of patient illness), age, and a history of ischaemic heart disease (IHD) were recorded to enable analysis of blood ordering in the light of the Hebert study. In addition, Hb level that would trigger the medical team to request a blood transfusion, number of units of packed cells transfused during the 24 hour period and the new Hb level post transfusion were recorded. Where subjects had multiple transfusions only the first transfusion episode in the study period was considered for analysis. This was done to exclude potentially atypical transfusion decisions that may be associated with multiple transfusions.

Data were collated as an Excel™ spreadsheet and analysed using a MinitabÔ statistical package. Two-dimensional plots of trigger level against APACHE by IHD and trigger level against age by IHD were constructed. The former is illustrated in Figure 1. The purpose of this diagram was to see if the haemoglobin level that triggered a transfusion was dependent on either a measure of how ill the patient was (APACHE) or the presence or absence of ischaemic heart disease. Potential relationships between these variables were explored using stepwise linear regression. This was done in an attempt to determine whether trends existed in transfusion practice related to these variables and confirm or refute our visual impressions of the graphs.

A total of 76% (19/25) of all the SICSAG ICU’s replied to the survey. Seventy-eight patients were present in the Scottish ICU’s that responded. By checking the SICSAG audit database there were 129 patients in Intensive Care on that day. Our data therefore relates to 60% of the total Scottish ICU patient population. Patient demographics for the studied population are shown in Table I. Twenty-one (27%) patients were noted to have ischaemic heart disease. Twenty-two (28%) were under the age of 55. Only two of the patients under age 55 were noted to have ischaemic heart disease with 19 patients aged 55 or over having ischaemic heart disease. Figure 2 shows the distribution of trigger Hb to transfuse - 80 g/l and 100 g/dl are the most popular choices for trigger.

Table I Patient demographics
	Age (yrs)	APACHE	Hb level (g/dl)
Mean	60.9	20	102
Range	19-94	5-51	63-138
Mode	64	17	93
Median	64	20	99

A total of 11 transfusion episodes occurred in eight subjects during the 24 hour period in the 19 units surveyed. Only two transfusion episodes occurred in patients under age 55. Of the remaining patients who were transfused (n=6) 50% had ischaemic heart disease. Considering only the first transfusion episode in each subject the mean initial Hb was 82 g/l, mean age was 65 years (range 24-85), and mean APACHE was 19 (range 10-24). The modal initial transfusion episode involved two units of blood being given (seven out of eight episodes) with a single unit transfusion only being given once. Actual trigger versus what the clinician had indicated to be the trigger for transfusion is shown (Figure 3). Six patients were not transfused until their Hb had dropped below the trigger. One was transfused at a Hb of 91g/l with a trigger of 90 g/l and one patient was transfused two units at a measured Hb of 88 g/l despite the trigger being recorded as 80 g/l.

Apart from measured Hb level, data for the 70 non-transfused subjects are similar. The mean Hb was 104 g/l (range 81-138), mean age was 60 years (range 19-94), and mean APACHE was 21 (range 5-51). No patients received blood if their Hb was above 100 g/l and all patients were transfused with a Hb level of less than 70 g/l. Forward stepwise regression identified initial Hb, IHD and APACHE as significant influences on trigger Hb. Age was not a significant influence on trigger Hb. There was no evidence of interactions between IHD, initial Hb and APACHE.

For other variables being constant, the fitted statistical model indicates an increase in trigger Hb of 0.14 g/l for a unit increase in initial Hb (p=0.039), a decrease of 0.3 g/l for a unit increase in APACHE (p=0.034) and the trigger for IHD patients is 6.2 g/l higher than that for non-IHD patients (p=0.006). As the model only explains 20% of the variance, as is illustrated in Figure 1, these relationships will be explored further in the discussion.

The use of blood in intensive care is common with reports of up to 50% of all patients admitted to ICU receiving a blood transfusion during their ICU stay.2 In North America, 85% of patients with an ICU stay of greater than one week receive an average of 2-3 units of blood per week.3 Scottish practice is probably very similar and a national audit is now being analysed to determine current ICU transfusion activity. Use of red cells on this scale may occur for treatment of acute haemorrhage but is also often due to depletion of the patient’s own Hb due to repeated sampling of blood required as part of ICU care.

Intensive Care blood use may also be influenced by the work done by Shoemaker and colleagues.4 They produced data showing that outcome could be improved in high-risk surgical patients if oxygen delivery was augmented to above 600 ml/min/m2 body surface area. Although difficult for others to verify, this practice was widely accepted by ICU practitioners throughout the 1980’s. Intuitively, by raising Hb levels and applying the oxygen delivery equation

(where Hb level is in g/dl, 1.34 is the amount of oxygen in mls carried by one gram of Hb the Hufner constant, PaO2 is the partial pressure of oxygen dissolved in the blood measured in mmHg and C.I. is the cardiac output standardised to body surface area) better oxygen delivery ought to be the result. It was therefore suggested that outcome could be improved by boosting oxygen delivery. Maintaining a Hb level above 100 g/l was considered integral to ensuring good oxygen carriage. In fact, the practice of augmenting oxygen delivery more usually involved enhancing cardiac index with inotropes rather than simply increasing Hb. Recent publications suggest that this policy of augmenting oxygen delivery in patients undergoing high risk elective surgical procedures may still be advantageous.5 More work in high-risk surgical patients may clarify these points.

Whether augmentation of oxygen delivery should be reliant upon increasing Hb is still debatable. What has not been taken into account is the possible deleterious effect of transfused homologous blood. This may outweigh, or at least influence, any potential advantages of increasing Hb levels above 70 g/dl. This hypothesis is suggested by the Transfusion Requirements in Critical Care (TRICC) study published by Hebert and colleagues. This large multi-centre, randomised, controlled clinical trial compared two transfusion strategies in 838 critically ill patients in 25 Canadian ICU’s. Patients’ randomised to the restrictive strategy (Hb concentration maintained between 70-90 g/l) had, in certain cases, a lower 30-day mortality than those treated with a liberal transfusion strategy (Hb maintained between 100-120 g/l). Outcome was less clear in cases of acute myocardial infarction, the elderly and those with the highest APACHE scores. No benefit could be demonstrated for the liberal strategy in any group. The recommendations of this study were that a more conservative approach to transfusion should be encouraged in the ICU and indeed in all clinical areas for that matter. Although this paper could be criticised for the large number of exclusions that occurred (over 4600 patients were randomised but only 838 analysed) it still represents the best evidence base to date as to transfusion requirements in intensive care.

We wished to explore the possibility of Scottish transfusion practice being influenced by the Canadian study. Our sample size we accept is very small with only eight patients being transfused on the day of the survey. The influence of, age, APACHE score and IHD on trigger is also difficult to interpret for this reason although Figure 3 allows a global impression to be gained. Combining all the data for Scotland also masks practice within individual ICU’s but by performing the above analyses trends can be discussed. With these limitations it should be remembered that the primary purpose of this survey was to evaluate current practice with a view to designing a much larger study to be conducted over a longer time. This larger study is currently complete and being analysed.

This survey suggests that practice in Scotland is not consistent as regards a trigger Hb level to initiate transfusion. A modal trigger of 80 g/l suggests that overall transfusion triggers have been reduced since the 1980’s. No patient received a blood transfusion if their Hb was 100 g/l or above and also no patient was allowed to become anaemic with a Hb below 70 g/l. These findings are encouraging and imply that blood is rightly being used as a scarce resource and over-transfusion in ICU is rare.

From this limited data there appears to be no overall clinically relevant relationship between the trigger Hb and either age or APACHE. Plotting trigger level versus APACHE and looking at the incidence of IHD (Figure 3) shows, as expected, that there is a weak relationship between higher APACHE (p=0.034) and IHD (p=0.006). Although APACHE, IHD and initial Hb (p=0.039) are shown to be statistically significant influences on transfusion trigger, the effects are sufficiently small to clinically irrelevant e.g. difference between transfusion trigger in IHD and non-IHD subjects being 6 g/l (95% CI = 1.9-10.6), after adjusting for APACHE and initial Hb.

Although some ICU’s are not giving, nor intending to give, blood until Hb falls to 80 g/dl, our survey does not demonstrate an overall clinically important relationship between Hb trigger, APACHE score, age or IHD. It could be that within the sample some ICU’s are applying Hebert criteria but those that are not, mask their contribution.

How consistent is the trigger level? Figure 2 shows transfused units against measured Hb and trigger level. One would expect no patients to be transfused above their trigger level but in fact two patients were. One received a one-unit transfusion when the measured Hb came within one g/l of the threshold for transfusion. This is reasonable, as presumably the clinicians concerned were able to predict that Hb would continue to fall. One could argue that, within laboratory error, that patient had a chance of being below the threshold for transfusion anyhow. It is interesting that this patient was the only one to receive a one unit transfusion. This suggests that the ICU concerned was watching Hb level very closely.

The other patient’s two unit transfusion is not so easily justified. The threshold for transfusion was stated to be 80 g/l but he patient received two units of blood in response to a measured Hb of 88 g/dl. With only one deviation from the predetermined trigger suggests that clinicians do adhere to their own trigger for the use of red cells.

In conclusion, our survey shows that most clinicians use a Hb level of 80 g/l as a trigger for transfusion. The influence of age, IHD or APACHE is less clear. A statistically significant higher Hb trigger (p=0.006, CI 95% 1.9-10.6) was demonstrated in patients with IHD but this only amounted to 6.2 g/l which is not clinically relevant. Clinicians, although aware of lower Hb tolerability, do not appear to be modifying their practice based on the Hebert paper. More education and debate may modify this situation and hopefully lead to optimal use of blood.

1 Hebert P C , Wells G, Blachjman C, Marshall J, Martin C, Pagliarello G, Tweedale M, Schweitzer I and Yeltsir E. A multi-centre, randomised, controlled clinical trial of transfusion requirements in critical care. N Engl J Med 1999; 340: 409–417.

2 Littenberg B, Corwin H C, Gettinger A, Leichter J, AuBuchon J. A practice guideline and decision aide for blood transfusion. Immunohematology 1995;11: 88-94.

3 Corwin H C, Parsonnet K C, Gettinger A. RBC Transfusion in the ICU – Is there a reason? Chest.1995; 108: 767-771.

4 Shoemaker W C, Appel P L, Kram H B , Waxman K, Lee T S. Prospective trial of supranormal values of survivors as therapeutic goals in high risk surgical patients. Chest 1988; 94 (6): 1176-1186.

5 Boyd O. The high risk surgical patient – where are we now? Clinical Intensive Care Special Issue. Bios Scientific Publishers Ltd. Oxford 2000 pp 3-10.

Haemoglobin level in Scottish Intensive Care Patients – A National One Day Survey

Abstract:

Back to June Index