Reducing the Amount of Blood Transfused:  A Systematic Review of Behavioral Interventions to Change Physicians’ Transfusion Practices FREE

Adverse events associated with blood transfusions, including infections and transfusion reactions, have been recognized since the advent of blood transfusions. Recent publications^1- 4 have demonstrated an association between transfusions and increased morbidity and mortality. These studies highlight the need for effective interventions to reduce unnecessary blood transfusions. There are many additional important reasons to reduce the amount of blood transfused. Infectious agents pose a risk to the recipients of blood products, and the threat of known and theoretical transfusion-transmitted pathogens has led to increased testing and donor deferrals. These, in turn, have led to a rise in the cost of blood⁵ and shortages of blood products in many regions.⁶ Increases in the cost of blood have also affected the delivery and payment of medical services.^7- 8 Although transfusion specialists have focused considerable efforts on identifying mechanisms to reduce transfusions, there has been little analysis to identify the best means to change transfusion practice among physicians.

A number of studies have examined the effectiveness of specific behavioral interventions in changing transfusion practices.^9- 29 The interventions in these studies included the adoption of guidelines, education, reminders, and audits. However, the absolute and relative effectiveness of these interventions to reduce blood product utilization is not known. Some studies in nontransfusion settings have found more complex or labor-intensive interventions to be superior in changing physician behavior.^30- 31 However, a recent systematic review did not find such a difference.³²

We conducted a systematic review of the literature to examine the effectiveness of behavioral interventions to decrease blood product utilization, either the number of units transfused or the number of patients who received transfusions. Previously, we reported on a smaller number of studies that looked at changes in inappropriate transfusions.¹² In the present study, we sought to determine whether the type of intervention(s) used or the blood product transfused changed blood product utilization.

A systematic search of the published literature using MEDLINE and EMBASE databases was conducted for the dates January 1966 to May 2003. Citations including the medical subject headings or text words blood transfusion, plasma exchange, or transfusion were combined (AND) with citations including the medical subject heading or text words guidelines, professional practice, practice patterns, decision making, professional competence, adaptation, knowledge, attitudes, and practice (see the Figure for the full search strategy). Reference lists, personal files, and experts were consulted to identify any additional studies.

For inclusion in this systematic review, a study had to meet the following criteria: (1) be an original published report, (2) evaluate an intervention designed to change transfusion practice behavior of physicians, (3) involve both an intervention group and a control group, and (4) report an outcome related to the number of units transfused or the proportion of patients receiving transfusions. We specifically excluded controlled clinical trials that mandated adherence to a specific transfusion trigger or protocol. Letters, abstracts, non–English language articles, and review articles were also excluded. All identified citations were independently reviewed by 2 reviewers (L.M., K.W., M.A., or A.T.) and assessed for eligibility according to the inclusion or exclusion criteria. Any disagreements were resolved by consensus. If consensus could not be reached, then the full original article was reviewed along with the other publications meeting the inclusion criteria.

For each publication, the author, journal, study characteristics (study design, intervention, physician population, patient populations, study duration, country of study), and results (number of units transfused, number of patients receiving transfusions) were abstracted in duplicate by means of a standardized data abstraction form. The study interventions were classified as (1) guidelines, (2) educational sessions [group or individual], (3) a reminder system [computer aids or transfusion forms containing reminders of appropriate criteria for transfusion], (4) audit with feedback [retrospective audits with feedback given to individuals or groups after the transfusion], or (5) audit with approval [audit with approval needed before transfusion of products]. If guidelines were disseminated or accompanied by educational sessions, then the study interventions were classified as guidelines and education. We assessed the quality of the individual studies by examining specific relevant study characteristics,³³ including baseline assessments, documentation of secular changes, detection bias (blinded assessment), contamination, reliability of outcome measurements, data collection (prospective vs retrospective, completeness), and length of follow-up.

Descriptive statistics of the absolute and relative reductions of the individual studies and the range of the relative reductions for the type of blood product and the nature of the intervention were determined. Further meta-analysis was not feasible given the heterogeneity in the designs of the individual studies.

Our search strategy identified 3496 citations, of which 42 were identified as possibly meeting the criteria for the systematic review. After review of the complete manuscripts, 19 studies^{9- 10,14- 29,34} met all the inclusion and exclusion criteria and were included in the final review.

The 19 studies identified were published between 1979 and 2004 in 8 countries. Ten studies examined the use of packed red blood cells (RBCs),^{10,15- 16,18- 20,24- 25,27,29} 10 studies looked at transfusions of fresh frozen plasma (FFP),^{9,17,20- 23,25- 27,34} 6 studies examined the use of platelet products,^{17,20,25- 28,34} 2 studies reported transfusions of cryoprecipitate,^20,26 and 1 study examined the use of albumin.¹⁴ Five studies examined the use of more than 1 blood product^{17,20,25- 27} (Table 1).

A variety of techniques were used to bring about behavioral changes in the transfusion practices of physicians. The intervention strategies evaluated included guidelines (n = 12),^{9- 10,14,16,19- 23,28- 29,34} audit with feedback (n = 9),^9,18- 24,27 audit with approval (n = 4),^15,17,22,26 a new transfusion form that outlined criteria for transfusions (n = 7),^{14- 15,17,19- 20,22,25} and education, at the level of either the group or the individual (n = 9).^{10,16,18- 19,21- 24,34} Of these, 5 studies used only a single intervention (guideline [n = 2],^28- 29 audit with feedback [n = 1],²⁷ audit with approval [n = 1],²⁶ and a new transfusion order form outlining the transfusion guideline [n = 1]²⁵). The remaining 14 studies used 2 or more interventions, either simultaneously or sequentially, to alter transfusion practices.^{9- 10,14- 24,34} The interventions to change transfusion practice were targeted at all physicians (n = 13),^{14- 18,20,22- 23,25- 28,34} surgeons and/or anesthesiologists (n = 5),^{9- 10,21,24,29} or obstetricians and gynecologists (n = 1)¹⁹ (Table 1).

Eighteen of the 19 studies^{9- 10,14- 23,25- 29} included in this systematic review used a before-and-after design. The1 remaining study²⁴ was a retrospective cohort comparing different hospitals. One before-and-after study²⁵ included a control group. Only 4 before-and-after studies^10,18,25,34 collected all data prospectively. Eight studies^{9,17,19- 20,22- 23,26- 27} collected the data retrospectively from the period before the intervention and prospectively after the intervention. Three before-and-after studies^15,28- 29 collected all the data retrospectively and 3 studies^14,16,21 did not report the timing of the data collection (Table 2). The periods of observation before and after the intervention varied from 3 months to 2 years. Three of the studies^21,27,34 did not monitor transfusion practice for the same length of time before and after the start of the intervention.

Five studies^{10,14,25,27- 28} reported the median or mean number of units transfused per patient. Six other studies either reported aggregate data on the total number of units transfused per patient^16,19,26 (eg, number of units transfused per 100 admissions) or presented data so that this could be calculated.^17- 18,20 Ten studies^{14,17- 23,28,34} reported the total number of units transfused. A total of 15 studies provided information on either the number of units transfused or the number transfused per patient.

Fourteen of the 15 studies showed a reduction in the number of units transfused per patient^{10,14,16- 20,25- 28} or the number of units transfused^{14,17- 18,20- 23,28} (range of relative reduction, 9%-77%). The 1 study³⁴ that did not show a reduction in the total number of units transfused reported a reduction in the number of inappropriate transfusions. For the remaining studies, the range of the relative reduction was similar for RBCs^{10,16,18- 20} (range, 12%-65%) and FFP^{17,20- 23,26} (range, 9%-77%), but lower in studies examining platelets^17,20,28 (range, 14%-23%) (Table 3). However, in the studies^17,20,25,27 that reported the effectiveness of an intervention on the number of platelets transfused and at least 1 additional product, the relative reductions after the interventions were similar for platelets and other product(s) studied. Six (43%) of the 14 studies^{10,17- 19,25,27} used statistical tests to determine whether the changes in the number of units transfused were significant. In 4 (67%) of these studies,^10,17- 19 the reduction was significant for all blood products studied. Two studies^25,27 reported both significant and nonsignificant reductions for different blood products. The first study²⁷ reported a significant reduction in the number of units of cryoprecipitate per patient per month and nonsignificant changes for RBCs, FFP, and platelets. In the second study,²⁵ the reduction in number of units transfused per patient per month was significant for RBCs and platelets, but not FFP. The other 9 studies^{14,16,20- 23,26,28,34} did not determine the statistical significance of the changes in blood product utilization. Two (22%) of these 9 studies^26,34 did not observe a reduction in utilization for all products studied. One study²⁶ reported a reduction in the number of units of FFP transfused, but no change in the number of units of platelets transfused. As noted previously, the other study³⁴ reported an increase in the number of units of FFP transfused, but a reduction in inappropriate transfusions.

Five studies^{9- 10,15,24,29} examined the proportion of patients who received transfusions and found relative reductions of 17% to 79%. The relative reduction in the proportion of patients who received RBCs was 17% to 79%.^10,15,24,29 In the 3 studies^10,15,29 that performed statistical tests, this reduction was significant. One study⁹ showed a 21% reduction (P = .10) in the number of surgical patients who received FFP, but it did not report the total number of surgical patients. No studies reported the proportion of patients who received platelets.

The heterogeneity of the conduct and reporting of the primary studies precludes a statistical comparative analysis of the effectiveness of different interventions on transfusion practice. However, the effectiveness of these was qualitatively similar. The use of a specific intervention, such as audit with approval, which requires immediate feedback, did not appear to result in greater reductions in blood product use than less immediate interventions (Table 4). Similarly, studies using multiple interventions^{9- 10,14- 24,34} did not appear to lead to greater reductions in blood product utilization than studies that used a single intervention^25- 29 (Table 4). However, 3 of the 4 studies that did not show a reduction in the utilization of all blood products studied used a single intervention (form,²⁵ audit with feedback,²⁷ and audit with approval²⁶). All of these studies still showed a significant reduction in the utilization or inappropriate utilization of at least 1 blood product.

The results of this systematic review suggest that interventions to change physician practices, and reduce the number of units transfused or the proportion of patients who receive transfusions, are effective. Although a quantitative meta-analysis was not possible given the heterogeneity of study designs and outcomes reported, this review found a consistent effect for interventions designed to change physician transfusion practice. In 18 of the 19 studies meeting our inclusion criteria, there was a reduction in transfusions for at least 1 of the blood products studied. The 1 exception still showed a reduction in inappropriate transfusions. In the studies that performed statistical tests of significance^{10,15,17- 19,25,27,29} and those that did not,^{9,14,16,20- 24,26,28,34} the ranges of the reductions were similar (12%-65% vs 9%-79%).

The results of this systematic review did not demonstrate a difference in the effectiveness of the intervention depending on the blood product transfused or the nature of the intervention. The effectiveness of the various interventions was similar for the 3 blood products (RBCs, platelets, and FFP) that were most commonly studied. While 2 of the studies examining platelets reported either no difference or an increase in utilization, it is not possible to conclude that interventions may be less effective in changing the practice of platelet transfusions. The first study²⁶ did not report the actual change in the number of platelet transfusions. The second study²⁷ reported a small nonsignificant increase in platelet transfusions, but the decreases in both RBC and FFP transfusions were also nonsignificant.

We did not identify any studies that compared different interventions to change transfusion practice. Most of the included studies showed similar effectiveness in reducing the utilization of blood products regardless of the intervention(s) used. The studies involving either multiple interventions or more intensive interventions, such as a prospective audit with approval, did not appear to be more effective. Studies of a single simple intervention, such as guidelines or reminders on the transfusion form, were also effective in reducing the utilization of blood products. Three of the studies,^25- 27 which failed to show a decrease in utilization in 1 or more blood products, used a single intervention, but these studies still demonstrated a significant reduction for at least 1 blood product.

In this review, we found no overall difference in the relative effectiveness of complex or multifaceted interventions compared with simple interventions in altering physician behavior. This is in contrast to previous research that found the distribution of educational material including guidelines to be generally ineffective, whereas the effectiveness of audit with feedback was variable, and reminders and multifaceted interventions were consistently effective.^30- 31 However, a more recent review³² did not find more complex or multifaceted interventions to be more effective. For transfusion practice, the relative effectiveness of both simple and complex interventions may be related to the nature of the change. Three major stages in the change process have been identified: (1) priming, (2) focusing or learning of alternative practices, and (3) follow-up (obtaining further information or advice regarding change).³⁵ In most of the included studies, the recent tragedies of human immunodeficiency virus and hepatitis C virus infection through the blood supply and the public reaction served as priming events, which would have increased the receptiveness of physicians to change their transfusion practices. By initiating the changes, the transfusion service or professionals provided alternatives to current practice and were then available for follow-up. Furthermore, the proposed changes in transfusion practice were relatively safe, simple, and similar to current practice. As a result, the changes were unlikely to encounter barriers by overtly threatening physician autonomy, challenging their judgment, or obviously compromising patient care.¹³

While our systematic review suggests that a variety of behavioral interventions can be effective in altering physicians’ transfusion practices, any conclusions must be tempered by caution given the methodologic limitations of our systematic review. The primary limitations of this review are those associated with the design of the primary studies and the potential for publication bias. All but 1 study used a before-and-after design, which is prone to secular changes and maturation bias.^33,36 Only 1 before-and-after study²⁵ used a prospective control group to account for this bias. In addition, 8 studies^{9,17,19- 20,22- 23,26- 27} used mixed retrospective and prospective data collection, which may be prone to instrumentation bias,³⁶ and 12 of the studies^{9,15,17,19- 20,22- 24,26- 29} used retrospective data collection, which is prone to information and selection bias especially in transfusion studies.³⁷ However, these effects may be lessened by the outcomes of interest (number of units transfused, proportion of patients receiving transfusions), which can be readily obtained through transfusion medicine records. The Hawthorne effect (an initial improvement in performance due to the act of observing the performance) can also bias the results of before-and-after studies, especially when there is no control group. However, the studies that used only retrospective data collection,^15,28- 29 which is not prone to this bias, reported results similar to those of susceptible studies, which used mixed retrospective and prospective data collection.^{9,17,19- 20,22- 23,26- 27} Publication bias may be a more important limitation to our findings. The fact that all studies showed a reduction in blood transfusions may suggest that studies with negative findings were unlikely to be submitted and/or accepted for publication. Publication bias is more likely in studies of lower quality or small sample sizes.³⁸ Given the lower quality of the included studies, negative studies assessing the effectiveness of interventions to alter transfusion practice may exist but not be published.

Despite these limitations, the results of our systematic review agree with previous research,^11- 13 which suggests that behavioral interventions to alter physicians’ transfusion practices can be effective. Being able to change transfusion practices among physicians is important given the potential risks associated with transfusions and the challenges facing the blood system. Recent studies^1- 4 highlight the potential harmful effects of blood transfusions in specific populations. The pressures of limited supply and rising costs are likely to increase with additional donor deferrals and screening tests for blood products. Our results suggest that simple interventions are sufficient to alter transfusion practice and would be valuable interventions to reduce blood product utilization.

Unfortunately, the current literature does not allow us to determine whether some interventions or combinations of interventions are more effective than others in changing physicians’ transfusion practices. Similarly, no evaluation of the cost-benefit of these interventions has been undertaken. Additional noncontrolled before-and-after studies are not likely to provide this additional information. Rather, higher-quality studies (eg, randomized or nonrandomized cluster controlled studies, or controlled before-and-after studies) comparing the relative effectiveness of different strategies to improve transfusion practice are required. These would include the number of units transfused per individual patient per center (not aggregate data of the total number of units transfused and the total number of patients) and the proportion of patients who received transfusions. For cluster trials, the outcomes need to be reported at the center level. Future studies should also include measures of appropriateness of transfusion, undertransfusion, and cost.

Correspondence: Kumanan Wilson, MD, Department of Medicine, 9ES, Toronto General Hospital, University Health Network, 200 Elizabeth St, Toronto, Ontario, Canada M5G 2C4 (kumanan.wilson@uhn.on.ca).

Accepted for Publication: December 16, 2004.

Financial Disclosure: None.

Funding/Support: This study was supported by grant 89246 from the Canadian Institutes of Health Research, Ottawa, Ontario. Dr Tinmouth is a Canadian Blood Services/Canadian Institutes of Health Research New Investigator; Dr Wilson is a Canadian Institutes of Health Research New Investigator; Dr Hebert is a Career Scientist at the Ontario Ministry of Health; Dr Graham is a Canadian Institutes of Health Research New Investigator.

Acknowledgment: We thank Jessie McGowan, MLIS, AHIP, and Risa Shorr, MLIS, for conducting the MEDLINE and EMBASE searches.