Article Text

Rationalising the treatment of anaemia in cardiac surgery: short and mid-term results from a local quality improvement initiative
1. Donald S Likosky1,2,3,4,5,6,
2. Stephen D Surgenor1,2,3,4,5,
3. Lawrence J Dacey1,2,3,4,5,
4. Gordon R DeFoe1,2,3,4,5,
5. Elizabeth L Maislen1,2,3,4,5,
6. Jean A Clark1,2,3,4,5,
7. James P Aubuchon1,2,3,4,5,
8. John H Higgins1,2,3,4,5,
9. Peter A Beaulieu1,2,3,4,5,
10. Gerald T O'Connor1,2,3,4,5,6,
11. Cathy S Ross1,2,3,4,5,6
1. 1Department of Surgery, Dartmouth Medical School, Hanover, New Hampshire, USA
2. 2Department of Community and Family Medicine, Dartmouth Medical School, Hanover, New Hampshire, USA
3. 3Department of Anesthesiology, Dartmouth Medical School, Hanover, New Hampshire, USA
4. 4Department of Pathology, Dartmouth Medical School, Hanover, New Hampshire, USA
5. 5Department of Medicine, Dartmouth Medical School, Hanover, New Hampshire, USA
6. 6The Dartmouth Institute for Health Policy & Clinical Practice, Dartmouth Medical School, Hanover, New Hampshire, USA
1. Correspondence to Dr Donald S Likosky, Department of Surgery and The Dartmouth Institute for Health Policy & Clinical Practice, Dartmouth-Hitchcock Medical Center, Lebanon, NH 03756, USA; likosky{at}dartmouth.edu

## Abstract

Background Transfusion of red blood cells, while often used for treating blood loss or haemodilution, is also associated with higher infection rates and mortality. The authors implemented an initiative to reduce variation in the number of perioperative transfusions associated with cardiac surgery.

Methods The authors examined patients undergoing non-emergent cardiac surgery at a single centre from the third quarter 2004 to the second quarter 2007. Phase I focused on understanding the current process of managing and treating perioperative anaemia. Phase II focused on (1) quality-improvement project dissemination to staff, (2) developing and implementing new protocols, and (3) assessing the effect of subsequent interventions. Data reports were updated monthly and posted in the clinical units. Phase III determined whether reductions in transfusion rates persisted.

Results Indications for transfusions were investigated during Phase II. More than half (59%) of intraoperative transfusions were for low haematocrit (Hct), and 31% for predicted low Hct during cardiopulmonary bypass. 43% of postoperative transfusions were for low Hct, with an additional 16% for failure to diurese. The last Hct value prior to transfusion was noted (Hct 25–23, p=0.14), suggestive of a higher tolerance for a lower Hct by staff surgeons. Intraoperative transfusions diminished across phases: 33% in Phase I, 25.8% in Phase II and 23.4% in Phase III (p<0.001). Relative to Phase I, postoperative transfusions diminished significantly over Phase II and III.

Conclusions We report results from a focused quality-improvement initiative to rationalise treatment of perioperative anaemia. Transfusion rates declined significantly across each phase of the project.

• Cardiopulmonary bypass grafting
• anaemia
• transfusion
• surgery
• statistical process control

## Statistics from Altmetric.com

Red blood cells (RBCs) are often transfused in the setting of cardiac surgery for treating excessive blood loss or anaemia and to increase a patient's oxygen carrying capacity. While in some settings, transfusions are necessary to preserve life and reduce morbidity, growing evidence suggests transfusions are also associated with significant risks such as nosocomial bacterial infections, increased risk of complications, longer intensive care unit lengths of stay and increased mortality.1–3 A large variation in the indications for and timing of RBC transfusion has been documented among coronary artery bypass graft surgery (CABG) patients, suggesting differences in provider and institutional preferences.4

Transfusion triggers have been viewed as a way to systematise the delivery of RBC transfusions. A variety of transfusion triggers have been suggested and promulgated throughout the medical literature.5–7 Triggers based on a certain haemoglobin concentration fail to take into consideration the physiological status of the patient. Several recent studies allow us to establish a better strategy. First, a large multicentre trial among chronic critically ill patients demonstrated that a restrictive strategy of tolerating anaemia to a level of 7 g/dl was as effective as a higher trigger.6 One other randomised trial provides some evidence regarding the role of RBC transfusion as part of early goal-directed therapy (manipulation of cardiac pre- and afterload, and contractility) for the treatment of sepsis or septic shock.8 Early goal-directed therapy seeks to balance oxygen demand and delivery. Patients in the early goal-directed therapy group required significantly more fluid, transfusions and inotropic therapy, and had a higher Hct than the standard therapy group. Early goal-directed therapy was associated with reductions in mortality relative to standard therapy (30.5% vs 46.5%, p=0.009). Three observational studies provide some further insight of treatment of anaemia among patients with acute coronary syndromes.9–11 Together these observations provide conflicting results, and so further investigation is warranted to rationalise the treatment of anaemia for patients with coronary artery disease.

In the present study, we aimed to examine indicators for transfusion and determine whether sustainable reductions in transfusions would be realised by providing feedback over time to the team regarding its performance.

## Methods

Human subjects approval was granted by the institution's Committee for the Protection of Human Subjects (#16745).

### Setting

This quality-improvement project was conducted at Dartmouth-Hitchcock Medical Center in Lebanon, New Hampshire. This project involved members of the Departments of Medicine, Surgery, Anesthesia and Pathology, and disciplines from nursing, cardiothoracic surgery, anaesthesia, perfusion, quality improvement, transfusion medicine and epidemiology. All patients admitted to the hospital for non-emergent isolated CABG surgery from the third quarter 2004 to the second quarter 2007 were enrolled in the study.

### Function

We conducted a three-phase study to develop and implement interventions aimed at rationalising the management of perioperative anaemia among patients undergoing non-emergent cardiac surgery.

We developed a Transfusion Quality Improvement (TQI) team, composed of a cardiothoracic surgeon, nurses, nurse practitioners, cardiovascular perfusionist, anaesthesiologist, Transfusion Service personnel, quality improvement coordinator and epidemiologist.

A data form was developed with input from the TQI team (appendix 1) to identify the primary indications and location of RBC transfusions. Each unique transfusion event was considered independent of any previous transfusions. The Blood Bank database provided information on date of transfusion, expiration date of each RBC unit and whether the unit was leucoreduced. Occurrence of infection was determined through linkage with the hospital's Infectious Disease Service as well as the Cardiothoracic Surgery registry. Hospital administrative data were used to track in-hospital charges.

### Interventions

Phase I (third quarter 2004) focused on understanding and mapping the current process of care for identifying, managing and treating perioperative anaemia. Specifically, the group mapped the current process for transfusing patients in the perioperative setting (appendix 2), and identified the primary reasons for transfusing patients. The Transfusion Quality Improvement Team convened on a monthly basis. (Throughout this study, attending surgeons had the sole responsibility and authority to order RBC transfusions. Nonetheless, we hypothesised that the base rate of transfusion was a product of variability of the context and indication for each unit of red blood cells.)

During Phase II (fourth quarter 2004 to fourth quarter 2005), the Cardio-Thoracic surgical team brought attention to the importance of responsible blood management, and developed, implemented and assessed the impact of transfusions in the setting of CABG surgery. Specifically, the teams focused on: (1) educating surgical and intensive care unit staff regarding the evidence base for perioperative transfusions, (2) developing and implementing new protocols (box 1), (3) developing and staffing the intervention teams, and (4) collecting data to assess the effect of subsequent interventions. Outside experts presented on the topic of the risks of transfusions at multidisciplinary Grand Rounds. Evidence-based literature was shared with team members. The development and implementation of new protocols (box 1) was accomplished during monthly meetings of the TQI team and buy-in from the Cardiothoracic Surgery section. Transfusion criteria were agreed upon by the surgical team, for the intra- as well as postoperative setting. Data reports were developed and updated on a monthly basis, and posted in the Cardiothoracic Surgical offices and critical care unit. These reports displayed the transfusion rates in each setting, as well as indication for transfusions, over time. These data were shared at monthly staff meetings.

### Box 1 Protocol usage

#### Preoperative

All patients scheduled to undergo cardiopulmonary bypass are put into a spreadsheet to predict the on-bypass haematocrit. This simple algorithm predicts postdilutional haematocrit based upon the patient's estimated circulating volume, preoperative haematocrit value and total fluids from both Anaesthesia and circuit prime. If the predicted value falls below a threshold, addition of red blood cells to the cardiopulmonary bypass prime was considered.

#### Intraoperative

Transfuse patients when haematocrit falls below 19% on bypass

• Definition: transfusion (no/yes) if intraoperative haematocrit is below 19% during cardiopulmonary bypass

• Data-collection method: cardiopulmonary bypass record (for haematocrit values) and linkage with blood bank database (for transfusion data)

#### Postoperative

##### Patients <75 years

Transfuse patients when haematocrit falls below 21%

• Definition: transfusion (no/yes) if postoperative haematocrit is below 21% after the procedure until the patient was discharged from the hospital

• Data-collection method: medical record review and linkage (for haematocrit values) with blood bank database (for transfusion data)

##### Patients ≥75 years

Transfuse patients when haematocrit falls below 24%

• Definition: transfusion (no/yes) if postoperative haematocrit is below 24% after the procedure until the patient was discharged from the hospital

• Data-collection method: medical record review and linkage (for haematocrit values) with blood bank database (for transfusion data)

During Phase III (first quarter 2006 to second quarter 2007), transfusion rates were tracked through a dedicated and pre-existing institutional database. The database was queried to determine whether the intra- and postoperative transfusion rates changed relative to each of the previous phases. Information regarding transfusion rates was shared with the cardiothoracic surgical team on a quarterly basis.

### Measures

The primary outcomes for this study were transfusion and adherence to the agreed upon protocols. Adherence to protocols was investigated solely during Phase II and was defined as the percentage of transfusions delivered in accordance to the transfusion protocols developed in this phase of the project. Other measures that were tracked included discharge disposition (percentage discharged home), charges (index admission charges), infection rates (types of infections included: mediastinistis, leg wound infection or as diagnosed by infectious disease personnel through culture) and satisfaction (percentage patients reporting their overall satisfaction with care as excellent).

### Analytical methods

Standard statistical methods were used for proportions (χ2) and continuous data (Student t test, logrank test and Kruskal–Wallis). All statistics were performed using the STATA 10.0 program (Stata, College Station, Texas). Process control charts were created using Static Custom QC (Statit Software, Corvallis, Oregon).

## Results

Transfusion rates diminished across phases of the study (figures 1, 2; p<0.001). The greatest drop was evident from Phase I to Phase II: intraoperative period (33% transfusion rate in Phase I vs 25.8% in Phase II) and postoperative period (18% transfusion rate in Phase I vs 12% across Phase II and III; p<0.001).

Figure 1

Intraoperative transfusion rates across Phases II and III.

Figure 2

Postoperative transfusion rates across Phases II and III.

### Phase I

A total of 86 patients were enrolled during Phase I. Of these, 44% received a transfusion in either the intra- or postoperative period (23% intraoperative, 14% postoperative, 7% in both periods).

### Phase II

A total of 303 patients (186 transfusions) were enrolled (43% isolated CABG, 36% isolated VR, 21% CABG/VR) between September 2004 and September 2005. Intraoperative transfusion protocols were adhered to among 27.6% of patients. Among transfusions outside this protocol, the mean Hct was 22 (SD 4). In the postoperative phase, the protocol for patients <75 years was adhered to 10.6% of the time. Among transfusions outside this protocol, the mean Hct was 24 (SD 2). The protocol for patients ≥75 years was adhered to 36.5% of the time. Among transfusions outside this latter protocol, the mean Hct was 25 (SD 2).

Patients receiving transfusions were older (p<0.001), were more often female (p<0.001), more likely undergoing urgent operations (p<0.001), had lower prebypass Hcts (p<0.001) and lower predicted Hcts on cardiopulmonary bypass (CPB) (p<0.001; table 1).

Table 1

Univariate associations between patient and disease characteristics and perioperative transfusion during Phase II

Most patients receiving transfusions were given two (59%) or ≥3 units (34%) of red blood cells (table 2). Patients receiving transfusions had a lower first Hct on bypass (25 vs 22, p<0.001). There were more infections (2.9% vs 2.6%, p=0.03) and more pneumonia (11% vs 1%, p<0.001) and a longer median length of stay (10 vs 6, p<0.000) among patients with versus without transfusions. Transfused patients were less likely to be discharged home (17.4% vs 58.7%, p<0.000) and had higher hospital charges ($72 773 vs$51 428, p<0.001) although a similar reported satisfaction with care (percentage excellent: 9.2% vs 26.8%, p=<0.02) (figure 3).

Table 2

Intra- and postoperative characteristics of patients

Figure 3

Clinical value compass for patients enrolled over three phases of the study.

A trend in lowering of the last Hct value prior to transfusion was noted (Hct 25–23, p=0.14), suggestive of more tolerance for a lower Hct by staff surgeons. The median number of days until expiration for units delivered to patients in the first quarter of our study was 10. While no official policy change occurred during the course of our study, the median number of days until expiration during the last quarter was 20 (p<0.001).

Among the intraoperative setting, 14% were a single unit, 79% a double unit, and 7% ≥3 units (table 3). In the postoperative setting, 42% of transfusions were single unit, 57% double unit and 1% three or more units. More than half (59%) of intraoperative transfusions were for low Hct, with an additional 31% for predicted low Hct during cardiopulmonary bypass. Forty-three percent of postoperative transfusions were for low Hct, with an additional 16% for failure to diurese. Median Hct values were lower among those occurring intra- versus postoperatively after transfusions (24 vs 29, p<0.001).

Table 3

Characteristics of transfusions

Relative to the intraoperative setting, patients transfused in the postoperative setting had higher Hcts prior to their transfusion (21 vs 24, respectively, p=0.08), suggesting a different threshold for delivery of RBC units.

### Phase III

A total of 484 patients were enrolled during Phase III. Of these, 28% received a transfusion in either the intra- or postoperative period (17% intraoperative, 7% postoperative, 5% in both periods).

During Phase III, the rate of transfusions continued to decline in both the intraoperative and postoperative settings. By the end of Phase III, the rates of both intra- and postoperative transfusions were more than three SDs from what would have been expected relative to Phase I, that is, if the Phase II intervention had not been effective (figures 1, 2). Overall, Phase III showed an absolute difference of 10% from the expected in the intraoperative setting, and an absolute difference of 6% from that expected in the postoperative setting.

## Discussion

We report results from a focused quality-improvement initiative to rationalise treatment of perioperative anaemia. Our quality-improvement initiative was associated with a lowering of the threshold for delivering red blood cells. While the rate of adherence to the transfusion triggers agreed upon in Phase II was not favourable, the quality-improvement initiative was associated with a 3SD reduction in overall transfusion rates relative to Phase I. In the intraoperative period, the primary indications were predominantly for low Hct and predicted low Hct. In the postoperative period, delivered units were primarily for low Hct and failure to diurese. Continued reductions in transfusion persisted into the observational component of the study (Phase III), suggesting that results were not attributed to a Hawthorne effect. While no national data exist related to transfusion rates from the Society of Thoracic Surgeons National Database (http://www.sts.org/sections/stsnationaldatabase), no quality-improvement projects existed within our regional consortium (http://www.nnecdsg.org) during this time period. As such, we surmise that much of the transfusion reductions documented in this report may be attributed to this quality-improvement initiative.

Since low Hct was the principal reason stated for both intra- and postoperative transfusions, we find the greatest opportunity for reducing variation in delivery of RBCs will likely come from the prevention of perioperative anaemia. In both the intra- and postoperative settings, improvement will require looking upstream in the process of care to understand the factors that led to anaemia such as bleeding from the access site of the preoperative cardiac catheterisation, or predicted low Hct due to the patient's body surface area.12 13 In the postoperative setting, low Hct may be attributed to an intraoperative haemodilutional effect.

The utility of the RBCs to increase the oxygen carrying capacity of the blood is in part attributed to the age of each RBC unit (shelf-life of 42 days). Stored blood has less 2,3-diphosphoglycerate (DPG) than fresh blood, resulting in less oxygen going to tissues.14 While no official policy change occurred during the course of our study, the median number of days until expiration increased from 10 in the first quarter of phase II to 20 in the last quarter of that phase (p<0.001).

Engstrom undertook an investigation of indications other than blood loss among patients scheduled for CABG surgery.15 As with our study, Engstrom reported a lower threshold for delivery of RBCs in the intraoperative setting but did not identify the primary indication for the transfusions. Ferraris and colleagues, upon review of the literature supporting RBC transfusion, recommended RBC transfusion as reasonable when patients had a haemoglobin below 6 and 7 g/dl postoperatively.16 These authors additionally noted that RBC transfusion was ‘unlikely to improve oxygen transport when the haemoglobin concentration is greater than 10 g/dl,’ and was therefore not recommended.

Cormack 13 investigated the use of three CPB circuits: a standard adult circuit (STD), a low-prime circuit (LP) and an LP circuit coupled with autologous circuit priming (LP/AP). An appropriate circuit size was chosen to cater to the patient's needs, resulting in significantly lower renal complications, hospital mortality, higher Hct levels while on CPB and a non-significant reduction in the use of red blood cells. This study reflects one centre's approach for rationalising the management and treatment of haemodilutional anaemia. While, in the present study, separate circuits were not used based on a patient's preoperative risk of haemodilution, adoption of this approach may have resulted in fewer RBC transfusions in the intraoperative period.

### Limitations

The design of our study was observational, and as such does not allow us to determine whether the reductions in transfusion rates were in fact attributed to our local improvement efforts. Reductions in transfusions, however, did not result from general adherence to the transfusion triggers in Phase II. As the reductions in Phase II were additionally accelerated during Phase III, it is also unlikely that our findings may be attributable to the Hawthorne Effect. In any event, sustained reductions in transfusion rates beyond the interventional Phase II into Phase III suggest that our intervention, or general interest in this topic, had a lasting impact on the process of delivering care. Our study was not powered to determine whether reductions in transfusion rates would result in improvements in clinical outcomes. Instead, our group acknowledged and accepted evidence suggesting that unwanted variation in transfusion rates was associated with worse outcomes. Accordingly, we designed a study to reduce such variation in transfusion practice.

### Conclusions

We report the results of a local, multidisciplinary quality improvement work group tasked with rationalising the delivery of RBC transfusions. This group's work resulted in a short-term modification of the threshold for delivering RBCs, and sustained reductions in transfusions rates.

## Acknowledgments

The authors thank E Nelson for his comments on the manuscript.

## Footnotes

• Funding Funding for this research was provided by the Quality Research Grant Program which is supported by Dartmouth-Hitchcock Medical Center and Dartmouth Medical School.

• Competing interests None.

• Patient consent Obtained.

• Ethics approval Ethics approval was provided by the Dartmouth College.

• Provenance and peer review Not commissioned; externally peer reviewed.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.