Vox Sanguinis (2015) 108, 393–402 © 2015 International Society of Blood Transfusion DOI: 10.1111/vox.12237
ORIGINAL PAPER
Improving red cell transfusion in the elective surgical setting: an improvement collaborative with evaluation B. T. Harrison,1 J. Chen,2 C. Der Vartanian,1 J. Isbister,1 P. Tridgell1 & C. F. Hughes1 1
Clinical Excellence Commission, Sydney, NSW, Australia The Simpson Centre for Health Services Research, University of New South Wales, Sydney, NSW, Australia
2
Background & Objectives Significant research conducted in New South Wales (NSW) hospitals’ indicated that about 30% of red cell transfusions in stable adult patients was inappropriate. Of the total Australian government blood product budget in 2009–2010 (i.e. $8788 million dollars) was spent on fresh blood products and plasma collection. The Clinical Excellence Commission (CEC) launched a systematic intervention called Blood Watch (BW) aiming to reduce inappropriate red cell transfusions in all NSW hospitals. An evaluation of BW was undertaken to measure the effectiveness of the programme and to estimate the associated potential cost-saving. Materials & Methods Through the deterministic linkage of the four populationbased administrative databases, three outcome indicators and four process indicators were developed. The analyses were of five elective surgical groups as they were the focus of the interventions. Three-level logistic regression and three-level linear regression were used to explore the time trend of the study process and outcome indicators. Modelling of the possible avoided red cell transfusions was also undertaken using a quadratic regression technique. Results Overall, there was a 274% reduction of the blood usage after the introduction of the BW programme and the reductions were consistent across five elective surgical groups. Such a reduction was associated with annual cost-savings of over $85 million.
Received: 20 October 2014, revised 20 November 2014, accepted 20 November 2014, published online 6 March 2015
Conclusions The BW programme which was based on collaborative improvement methods and implemented at scale led to significant reduction of blood usage, consistently across five elective surgical groups and significant cost-saving. Key words: clinical practice improvement, collaborative improvement, patient blood management, postoperative transfusion, red blood cell transfusion.
Introduction The appropriate use of red blood cell (RBC) transfusion is important for health systems due to: patient safety concerns [1]; diminishing supplies of RBCs [2]; and emerging evidence on the immuno suppressive effects of transfusion [3], which can lead to an increase in wound infection particularly in a postoperative setting [4, 5]. A number of serious concerns have been raised about the risk/benefit Correspondence: Bernie Harrison, 54, Ourimbah Road, Mosman, Sydney, NSW 2088, Australia E-mail: [email protected]
of RBC autologous transfusions [6]. RBC transfusion can be life-saving therapy for patients with acute haemorrhage, severe burns and bone marrow suppression. However, in the elective transfusion setting where patients are haemodynamically stable, there is an emerging body of evidence suggesting that RBC transfusions may be harmful to patients and that the risk–benefit equation swings strongly to the risk rather than benefit side [7]. Of increasing concern in the postoperative setting is the new and emerging evidence of the immunomodulatory effects of transfusion leading to an increase in all infections (not including viral risk) particularly wound infections in
393
394 B. T. Harrison et al.
orthopaedic and cardiac surgery [8]. The emerging risks of transfusion have also driven international efforts away from a purely blood product focus, and its appropriate use, to a multidisciplinary, evidence-based, medical and surgical approach to achieving better outcomes for patients by better managing the patient’s own circulating volume also known as patient blood management [9, 10]. There are also concerns in Australia, about the increasing cost of blood and blood products in part driven by the costs associated with increased surveillance of donor suitability. On an international scale, there has been a global response to improving blood management [11] particularly in the light of blood-borne diseases including but not limited to: HIV, Hep C, prions disease, chikungunya and dengue fever, the latter two both associated with warmer climates [12]. Following publication of inappropriate rates of RBC transfusion [13] in 2000, the New South Wales (NSW) Department of Health undertook a number of initiatives to improve transfusion practices including the implementation the NSW blood transfusion collaborative in 2001 [14]. In 2006–2009, the Clinical Excellence Commission (CEC), which has a central role in quality and safety in the NSW health system, commenced the Blood Watch (BW) programme with a clear mandate to achieve change at scale and pace to improve transfusion practice across NSW. The NSW Public Hospital system is the biggest in Australia and serves a population of 74 million, has 232 public hospitals which at that time were organized into eight Area Health Services. The study aimed to provide an evaluation of the impact of the BW programme through the linkage of four population databases over a 5-year period to assess the changes of study outcome and process indicators for blood transfusions, between July 2004 and June 2009, across five elective surgical groups (cardiothoracic, colorectal, gynaecology & obstetrics, orthopaedic and general surgery). The study also aimed to estimate the number of avoided blood transfusions following the implementation of the BW programme.
Methods Intervention: The Blood Watch programme Aims The programme aimed to reduce inappropriate RBC transfusion in the elective surgical setting using collaborative improvement methods, through increasing compliance with the National Health and Medical Research Council/ Australian and New Zealand Society of Blood Transfusion (NH&MRC/ANZSBT) [15]guidelines for elective transfusions in haemodynamically stable patients, with normal bone marrow in the postoperative setting to: (1) eliminate
RBC transfusions when the Hb min ≥ 100 g/dl postoperation; (2) reduce RBC transfusions when Hb min ≥ 70 g/l and ≤100 g/l and when no clinical indications are present; (3) increase transfusions when Hb max levels ≤70 g/l when clinically indicated; (4) reduce the number of units transfused at a single transfusion episode, that is, reduce the dose of red cells per transfusion order, for example from two units to a single unit.
Methodological approach The methodological approaches relied heavily on Institute of Healthcare Improvement (IHI) breakthrough series collaborative [16]; clinical practice improvement [17], diffusion of innovation theories [18]; organizational change management [19] and social marketing [20] which included targeted behaviour modification strategies based on commercial marketing methods. Specific goals were set in relation to reductions in inappropriate transfusion based on the Hb thresholds as set out in the guideline.
Governance structure The CEC formed a central BW implementation team made up of a director, a manager and an expert advisory panel, made up of transfusion specialists and surgical college representatives with a transfusion practice. The CEC provided seed funding for twelve months for a BW project officer/transfusion nurses, in each Area Health Service (AHS). The teams received clinical practice improvement training [21] and communications training. The BW implementation team provided a co-ordination and strategic leadership role, pursuing systemic and policy change where possible. In the first 2 years of the programme, BW teams met two-three times a year in collaborative learning sessions and also met via teleconference on a fortnightly basis. An active online discussion forum also provided a central resource for sharing of: work plans; local change strategies; and local policies from the transfusion committee. The IHI collaborative method was essential in providing rapid adoption and local adaption of changes.
Programme phases The programme followed seven steps based on improvement science [21]: (1) problem identification; (2) aim setting for improvement; (3) diagnostic phase: to elucidate the current situation; (4) the literature review and identification of change strategies; (5) implementation of change through small cycle testing of change; (6) measurement of both process and outcome measures; (7) planning for spread and sustainability of the improvement. At the completion of the diagnostic phase, there was evidence of: wide spread variation in RBC usage across the five surgical specialties; failure to follow NH&MRC/ANZSBT guidelines; a generalized low awareness © 2015 International Society of Blood Transfusion Vox Sanguinis (2015) 108, 393–402
Improving red cell transfusion 395
of risks of transfusions by surgical staff [22] and variation in approvals and restrictions to RBC usage within hospitals. These results helped to form the specific change strategies to be tested within the BW programme. Some of the change strategies were derived from the literature [23, 24] others were from the best performing units, while others were trialled for the first time by BW teams. During the period of the Blood Watch programme, no perioperative blood management programmes were in effect in NSW.
Change Strategies A total of 21 change strategies were trialled by BW teams working within one of eight AHS (see Table 1). The uptake of these interventions is shown in Table 2.
Study sample and data Data Source To evaluate the success of BW, we linked data from four population-based databases: the Admitted Patients Disease Collection (APDC); the Blood Bank; the Pathology Laboratory Database; and the Emergency Department (ED) Patients Admissions database. The APDC database provided patient demographic variables, patient-flow during admission and diagnostic-related group (DRG) data. Blood bank data provided the quantity and type of blood products for each blood transfusion episode during the patients’ admissions. The pathology laboratory database gave serial Hb test data of a patient during the admission. The ED database included the key information regarding patients’ attendance to the emergency department and was used to exclude from the analysis those patients admitted to hospital through the emergency department, given the focus of BW on elective surgery. Data dated between 1st July 2004 and 31st June 2009 were combined, through a deterministic linkage method. This was the first time these four large population databases were utilized in evaluating the effectiveness of the BW programme. Patients admitted through emergency departments and those patients with a length of stay (LOS) of Hb min- < 100 g/l (%) Number of two or more blood units per transfused Gynaeco-Obstetrics (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused Colorectal surgery (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused General Surgery (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused Orthopaedic group (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused All (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused
7105 1424 021 052 1067 4081 101 81 238 071 001 002 676 2097 080 12 190 277 005 009 444 3462 112 30 894 249 005 010 364 3199 124 17 394 791 012 022 487 4320 091 148 821 274 004 009 631 3662 100
14 093 1413 023 058 1064 4378 112 169 668 081 001 002 1096 2061 084 23 246 313 006 012 549 3242 113 60 791 258 005 010 377 3210 119 35 395 783 012 022 505 4179 090 303 193 278 005 009 714 3619 102
14 384 1612 025 064 699 5554 105 184 997 110 002 003 896 3644 082 24 228 434 008 016 752 4520 111 60 652 320 006 014 356 4384 127 36 881 908 014 026 526 5542 094 321 142 333 005 012 625 4873 102
14 518 1712 027 068 584 6362 103 185 652 131 002 004 940 5089 082 25 717 416 008 015 888 5720 116 60 823 365 007 016 410 5792 123 36 734 946 015 027 515 6599 091 323 444 361 006 012 632 6001 100
14 917 1620 027 064 232 3779 101 186 084 124 002 004 506 2933 080 26 478 374 007 013 505 4283 109 62 123 365 007 014 397 3950 116 38 385 888 014 025 349 4344 087 327 987 347 006 012 379 3854 096
7406 1456 023 058 213 2588 105 89 813 111 002 003 321 2004 080 12 722 362 007 014 326 2739 112 30 066 326 006 013 235 2643 117 18 372 814 013 023 394 3075 088 158 379 316 005 011 303 2642 098
72 423 1561 025 062 625 4730 105 897 452 108 002 003 770 3350 082 124 581 372 007 014 634 4292 112 305 349 319 006 013 369 4147 121 183 161 867 014 024 466 4940 090 158 2966 324 005 011 556 4383 100
admission
admission
admission
admission
admission
admission
The crude trends of the outcome and process indicators between 1st July 2004–31st June 2009 The crude outcomes and process measures across the study years for individual surgical groups are presented (Table 3). It shows that the % of patients having at least one transfusion during their hospitalization increased from 27% in 2004 to 36% in 2007 then decreased to 32% in 2009. The number of times a patient was trans-
fused per admission and the total units of blood transfused per admission showed similar peaks around 2007 and decrease after that. The % of patients transfused with Hb min ≥ 100 g/l was 3% in 2009 which is less than half that in 2004 (i.e. 63%). The % of transfusions with Hb min ≥ 70 g/l & ≤100 g/l decreased from 601% in 2007 to 264% in 2009. The number of two or more blood units given during a transfused admission was relative stable with very little decrease (from 100 in 2007 © 2015 International Society of Blood Transfusion Vox Sanguinis (2015) 108, 393–402
Improving red cell transfusion 399
Table 4 The outcome and process indicators of blood transfusion stratified by surgical groups Group
% transfused
Cardiothoracic Pooled ↓ Selected ↓ Colorectal Selected ↓ Gynaecology and Obstetrics Pooled ↓ Selected ↓ Orthopaedic Pooled ↓ Selected ↓ General Pooled ↓ Selected ↓
No. of transfusion
Units transfused
Hb ≥ 100 g/l
70 g/l ≤ Hb < 100 g/l
Hb < 70 g/l
2/+ units transfused
↓ ↓
↓ ↓
↓ ↓
– –
↑ ↑
– –
↓
↓
–
–
–
–
↓ ↓
↓ ↓
↓ ↓
– ↑
↑ ↑
– –
↓ ↓
↓ ↓
– –
– –
– –
– –
↓ ↓
↓ ↓
– –
– –
– –
– –
The trends were assessed based on the results of the last year (i.e. 2008/2009) compared with previous years (i.e. 2004/2007). The results were derived from the corresponding three-level logistic or linear regression results from Appendix 2. We used the symbols ↓, ↑ and – to indicate a decrease, increase and unclear trend, respectively.
to 098 in 2009). Each individual surgical group exhibited similar pattern but with different baseline levels.
The adjusted trends of the outcome and process indictors between July 2004–June 2009 From the adjusted analyses, we provided a summation of the statistical difference for each measure (Table 4) but with detailed regression outcomes in Appendix 2. The trends were assessed based on the results of the last years (i.e. 2008/9) compared with previous years (i.e. 2004/ 2007). The symbols ↓, ↑ and – were used to indicate a
decrease, increase or stable trend, respectively (see Table 4). These summary results were based on the threelevel multivariate logistic regression and linear regression results for both outcome and process indicators. A downward trend in a clinical indicator suggests a desirable outcome for blood transfusion. A decrease trend in a process indicator (except for percentage of transfusion with Hb < 70 g/l) may be interpreted as favourable according to the current recommended best practice. For example, for the cardiothoracic surgery group and gynaecology and obstetric groups, after the BW, there were significant reductions in the proportion of patients
Fig. 1 The monthly observed and predicted trends of blood units transfused per admission (July 2004–June 2009). (1) The quadratic equation for overall predicted curve was: y = 0133591 + t*0003749-t2*0000049, where t = 1, 2, 60 for July 2004, August 2004,. . .June 2009, respectively; y: the monthly rate of units of RBC transfused per admission (July 2004 to June 2009); all Ps for t and t2 < 0001. (2) The quadratic equation for the predicted curve (July 04 to November 2007) and its projection (December 2007 to June 2009) was y1 = 0149069 + t*0001329 + t2*0000013; where t = 1,2. . .41 for July 2004, August 2004, . . . November 2007, respectively and y1: the monthly rate of units of RBC transfused per admissions (July 2004 to Nov 2007); all Ps for t and t2 < 0001.
© 2015 International Society of Blood Transfusion Vox Sanguinis (2015) 108, 393–402
400 B. T. Harrison et al.
who received blood transfusions with minimum Hb ≥ 100 g/l which indicated a significant increase in guideline-based blood transfusion practice. There were also significant increases in the proportion of patients who received blood transfusions with max Hb ≤ 70 g/l for the cardiothoracic and gynaecology and obstetrics groups, and with minimum Hb ≥ 70 g/l – ≤100 g/l for the gynaecology and obstetrics group with selected ANDRGs, respectively. For both pooled and selected cardiothoracic surgical group, the trends of number of patients having two or more RBC units transfused who had at least one blood transfusion during admission were steady between 2005 and 2009.
The possible units of blood transfused avoided An estimate of the number of units of red blood cells avoided was calculated (Fig. 1). The estimation of the possible units of blood transfusion avoided was based on all surgical groups combined. In total, we estimated that 12 225 units of red blood cells transfusion might be avoided between July 2008 and June 2009, a 274% reduction of the amount if no intervention had happened (i.e. 12 225/44 564). This would equate to a $8 557 500 saving based on the calculated direct and indirect costs of RBC transfusion in the Australian healthcare setting (unpublished data: Hoffman A, Wood E et al. Australian Haematology Association Meeting; 2009).
Discussion The BW programme had as its first objective to reduce inappropriate RBC transfusion in haemodynamically stable adult patients with normal bone marrow in the postoperative period. Using collaborative improvement methods [16], 21 interventions were implemented broadly across the system. This study has shown that there were consistent and significant reductions in the chance, frequency and the number of units of red blood cells being transfused from 2007 across different selective surgical groups. Such reductions were consistent for both the pooled surgical groups and the selected high blood usage procedure subgroups for each of four specialties (i.e. cardiothoracic, gynaecology & obstetrics, orthopaedics and general surgery) as well as colorectal surgical procedures. In addition, throughout this period red cell issues per 1000 population fell in NSW relative to the two other largest states Queensland and Victoria [26]. This finding is significant as there is a direct relationship between red cell issues by the Australian Blood Service and red cell usage within each jurisdiction. These findings combined suggest that the BW programme interventions were effective.
There was also evidence to indicate that the significant decrease in the proportion of transfusions with minimum Hb tests ≥100 g/l and the increase of the proportion of transfusion with max Hb tests ≤70 g/l may have, in part, contributed to the overall improvement of clinical indicators, in particular in cardiothoracic and gynaecology and obstetric surgical groups. A possible reason that the significant reductions in the proportion of blood transfusions with minimum Hb ≥ 100 g/l were observed only in the cardiothoracic and gynaecology and obstetric groups is that these two groups had a relatively high baseline incidence rate (around 20%) while the other three groups’ baseline incidence rates were lower (around 10%). The lower baseline incidence rate may present a ‘floor effect’ that makes the further significant reduction of the rate hard to observe. There was no significant reduction for the number of two or more units of blood transfused during a single transfusion episode which demonstrates further opportunities for improvement. It is estimated that, conservatively, 12 225 units of red blood cells could have been avoided between July 2008 and June 2009, most likely due to the extensive BW interventions. This would equate to significant cost-savings of an estimate $8 557 500 per annum. Our study showed that even for selective subgroups (five elective surgical groups) of a much wider target populations (all patients across all public hospitals in NSW), there was a 274% reduction in the projected red cell usage between 2008 and 2009, which makes up 789% of the total national reductions according to the supply plan in 2009–2010 [27]. Given the fact that NSW population (724 million) makes up 324% of total Australia population [28] at 2008–2009, and assuming the same reduction could be achieved in all jurisdictions in Australia, this would equate to a reduction of 37 731 RBC units used in the five elective surgical groups; a cost-saving of A$26 411 700 (based on direct costs $300 per unit from unpublished data) in 2008–2009. This accounts for 30% of the total blood product budget in Australia and 47% of total RBCs in 2009–2010. This study has its limitations. Firstly, the study only examined the potential effect of BW on the five selected surgical groups. We used the LOS of
ORIGINAL PAPER
Improving red cell transfusion in the elective surgical setting: an improvement collaborative with evaluation B. T. Harrison,1 J. Chen,2 C. Der Vartanian,1 J. Isbister,1 P. Tridgell1 & C. F. Hughes1 1
Clinical Excellence Commission, Sydney, NSW, Australia The Simpson Centre for Health Services Research, University of New South Wales, Sydney, NSW, Australia
2
Background & Objectives Significant research conducted in New South Wales (NSW) hospitals’ indicated that about 30% of red cell transfusions in stable adult patients was inappropriate. Of the total Australian government blood product budget in 2009–2010 (i.e. $8788 million dollars) was spent on fresh blood products and plasma collection. The Clinical Excellence Commission (CEC) launched a systematic intervention called Blood Watch (BW) aiming to reduce inappropriate red cell transfusions in all NSW hospitals. An evaluation of BW was undertaken to measure the effectiveness of the programme and to estimate the associated potential cost-saving. Materials & Methods Through the deterministic linkage of the four populationbased administrative databases, three outcome indicators and four process indicators were developed. The analyses were of five elective surgical groups as they were the focus of the interventions. Three-level logistic regression and three-level linear regression were used to explore the time trend of the study process and outcome indicators. Modelling of the possible avoided red cell transfusions was also undertaken using a quadratic regression technique. Results Overall, there was a 274% reduction of the blood usage after the introduction of the BW programme and the reductions were consistent across five elective surgical groups. Such a reduction was associated with annual cost-savings of over $85 million.
Received: 20 October 2014, revised 20 November 2014, accepted 20 November 2014, published online 6 March 2015
Conclusions The BW programme which was based on collaborative improvement methods and implemented at scale led to significant reduction of blood usage, consistently across five elective surgical groups and significant cost-saving. Key words: clinical practice improvement, collaborative improvement, patient blood management, postoperative transfusion, red blood cell transfusion.
Introduction The appropriate use of red blood cell (RBC) transfusion is important for health systems due to: patient safety concerns [1]; diminishing supplies of RBCs [2]; and emerging evidence on the immuno suppressive effects of transfusion [3], which can lead to an increase in wound infection particularly in a postoperative setting [4, 5]. A number of serious concerns have been raised about the risk/benefit Correspondence: Bernie Harrison, 54, Ourimbah Road, Mosman, Sydney, NSW 2088, Australia E-mail: [email protected]
of RBC autologous transfusions [6]. RBC transfusion can be life-saving therapy for patients with acute haemorrhage, severe burns and bone marrow suppression. However, in the elective transfusion setting where patients are haemodynamically stable, there is an emerging body of evidence suggesting that RBC transfusions may be harmful to patients and that the risk–benefit equation swings strongly to the risk rather than benefit side [7]. Of increasing concern in the postoperative setting is the new and emerging evidence of the immunomodulatory effects of transfusion leading to an increase in all infections (not including viral risk) particularly wound infections in
393
394 B. T. Harrison et al.
orthopaedic and cardiac surgery [8]. The emerging risks of transfusion have also driven international efforts away from a purely blood product focus, and its appropriate use, to a multidisciplinary, evidence-based, medical and surgical approach to achieving better outcomes for patients by better managing the patient’s own circulating volume also known as patient blood management [9, 10]. There are also concerns in Australia, about the increasing cost of blood and blood products in part driven by the costs associated with increased surveillance of donor suitability. On an international scale, there has been a global response to improving blood management [11] particularly in the light of blood-borne diseases including but not limited to: HIV, Hep C, prions disease, chikungunya and dengue fever, the latter two both associated with warmer climates [12]. Following publication of inappropriate rates of RBC transfusion [13] in 2000, the New South Wales (NSW) Department of Health undertook a number of initiatives to improve transfusion practices including the implementation the NSW blood transfusion collaborative in 2001 [14]. In 2006–2009, the Clinical Excellence Commission (CEC), which has a central role in quality and safety in the NSW health system, commenced the Blood Watch (BW) programme with a clear mandate to achieve change at scale and pace to improve transfusion practice across NSW. The NSW Public Hospital system is the biggest in Australia and serves a population of 74 million, has 232 public hospitals which at that time were organized into eight Area Health Services. The study aimed to provide an evaluation of the impact of the BW programme through the linkage of four population databases over a 5-year period to assess the changes of study outcome and process indicators for blood transfusions, between July 2004 and June 2009, across five elective surgical groups (cardiothoracic, colorectal, gynaecology & obstetrics, orthopaedic and general surgery). The study also aimed to estimate the number of avoided blood transfusions following the implementation of the BW programme.
Methods Intervention: The Blood Watch programme Aims The programme aimed to reduce inappropriate RBC transfusion in the elective surgical setting using collaborative improvement methods, through increasing compliance with the National Health and Medical Research Council/ Australian and New Zealand Society of Blood Transfusion (NH&MRC/ANZSBT) [15]guidelines for elective transfusions in haemodynamically stable patients, with normal bone marrow in the postoperative setting to: (1) eliminate
RBC transfusions when the Hb min ≥ 100 g/dl postoperation; (2) reduce RBC transfusions when Hb min ≥ 70 g/l and ≤100 g/l and when no clinical indications are present; (3) increase transfusions when Hb max levels ≤70 g/l when clinically indicated; (4) reduce the number of units transfused at a single transfusion episode, that is, reduce the dose of red cells per transfusion order, for example from two units to a single unit.
Methodological approach The methodological approaches relied heavily on Institute of Healthcare Improvement (IHI) breakthrough series collaborative [16]; clinical practice improvement [17], diffusion of innovation theories [18]; organizational change management [19] and social marketing [20] which included targeted behaviour modification strategies based on commercial marketing methods. Specific goals were set in relation to reductions in inappropriate transfusion based on the Hb thresholds as set out in the guideline.
Governance structure The CEC formed a central BW implementation team made up of a director, a manager and an expert advisory panel, made up of transfusion specialists and surgical college representatives with a transfusion practice. The CEC provided seed funding for twelve months for a BW project officer/transfusion nurses, in each Area Health Service (AHS). The teams received clinical practice improvement training [21] and communications training. The BW implementation team provided a co-ordination and strategic leadership role, pursuing systemic and policy change where possible. In the first 2 years of the programme, BW teams met two-three times a year in collaborative learning sessions and also met via teleconference on a fortnightly basis. An active online discussion forum also provided a central resource for sharing of: work plans; local change strategies; and local policies from the transfusion committee. The IHI collaborative method was essential in providing rapid adoption and local adaption of changes.
Programme phases The programme followed seven steps based on improvement science [21]: (1) problem identification; (2) aim setting for improvement; (3) diagnostic phase: to elucidate the current situation; (4) the literature review and identification of change strategies; (5) implementation of change through small cycle testing of change; (6) measurement of both process and outcome measures; (7) planning for spread and sustainability of the improvement. At the completion of the diagnostic phase, there was evidence of: wide spread variation in RBC usage across the five surgical specialties; failure to follow NH&MRC/ANZSBT guidelines; a generalized low awareness © 2015 International Society of Blood Transfusion Vox Sanguinis (2015) 108, 393–402
Improving red cell transfusion 395
of risks of transfusions by surgical staff [22] and variation in approvals and restrictions to RBC usage within hospitals. These results helped to form the specific change strategies to be tested within the BW programme. Some of the change strategies were derived from the literature [23, 24] others were from the best performing units, while others were trialled for the first time by BW teams. During the period of the Blood Watch programme, no perioperative blood management programmes were in effect in NSW.
Change Strategies A total of 21 change strategies were trialled by BW teams working within one of eight AHS (see Table 1). The uptake of these interventions is shown in Table 2.
Study sample and data Data Source To evaluate the success of BW, we linked data from four population-based databases: the Admitted Patients Disease Collection (APDC); the Blood Bank; the Pathology Laboratory Database; and the Emergency Department (ED) Patients Admissions database. The APDC database provided patient demographic variables, patient-flow during admission and diagnostic-related group (DRG) data. Blood bank data provided the quantity and type of blood products for each blood transfusion episode during the patients’ admissions. The pathology laboratory database gave serial Hb test data of a patient during the admission. The ED database included the key information regarding patients’ attendance to the emergency department and was used to exclude from the analysis those patients admitted to hospital through the emergency department, given the focus of BW on elective surgery. Data dated between 1st July 2004 and 31st June 2009 were combined, through a deterministic linkage method. This was the first time these four large population databases were utilized in evaluating the effectiveness of the BW programme. Patients admitted through emergency departments and those patients with a length of stay (LOS) of Hb min- < 100 g/l (%) Number of two or more blood units per transfused Gynaeco-Obstetrics (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused Colorectal surgery (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused General Surgery (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused Orthopaedic group (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused All (n) Had at least one transfusion during the stay (%) Number of times transfused per admission Total units of blood transfused per admission Transfused: Hb min ≥ 100 g/l (%) Transfused: 70 g/l > Hb min- < 100 g/l (%) Number of two or more blood units per transfused
7105 1424 021 052 1067 4081 101 81 238 071 001 002 676 2097 080 12 190 277 005 009 444 3462 112 30 894 249 005 010 364 3199 124 17 394 791 012 022 487 4320 091 148 821 274 004 009 631 3662 100
14 093 1413 023 058 1064 4378 112 169 668 081 001 002 1096 2061 084 23 246 313 006 012 549 3242 113 60 791 258 005 010 377 3210 119 35 395 783 012 022 505 4179 090 303 193 278 005 009 714 3619 102
14 384 1612 025 064 699 5554 105 184 997 110 002 003 896 3644 082 24 228 434 008 016 752 4520 111 60 652 320 006 014 356 4384 127 36 881 908 014 026 526 5542 094 321 142 333 005 012 625 4873 102
14 518 1712 027 068 584 6362 103 185 652 131 002 004 940 5089 082 25 717 416 008 015 888 5720 116 60 823 365 007 016 410 5792 123 36 734 946 015 027 515 6599 091 323 444 361 006 012 632 6001 100
14 917 1620 027 064 232 3779 101 186 084 124 002 004 506 2933 080 26 478 374 007 013 505 4283 109 62 123 365 007 014 397 3950 116 38 385 888 014 025 349 4344 087 327 987 347 006 012 379 3854 096
7406 1456 023 058 213 2588 105 89 813 111 002 003 321 2004 080 12 722 362 007 014 326 2739 112 30 066 326 006 013 235 2643 117 18 372 814 013 023 394 3075 088 158 379 316 005 011 303 2642 098
72 423 1561 025 062 625 4730 105 897 452 108 002 003 770 3350 082 124 581 372 007 014 634 4292 112 305 349 319 006 013 369 4147 121 183 161 867 014 024 466 4940 090 158 2966 324 005 011 556 4383 100
admission
admission
admission
admission
admission
admission
The crude trends of the outcome and process indicators between 1st July 2004–31st June 2009 The crude outcomes and process measures across the study years for individual surgical groups are presented (Table 3). It shows that the % of patients having at least one transfusion during their hospitalization increased from 27% in 2004 to 36% in 2007 then decreased to 32% in 2009. The number of times a patient was trans-
fused per admission and the total units of blood transfused per admission showed similar peaks around 2007 and decrease after that. The % of patients transfused with Hb min ≥ 100 g/l was 3% in 2009 which is less than half that in 2004 (i.e. 63%). The % of transfusions with Hb min ≥ 70 g/l & ≤100 g/l decreased from 601% in 2007 to 264% in 2009. The number of two or more blood units given during a transfused admission was relative stable with very little decrease (from 100 in 2007 © 2015 International Society of Blood Transfusion Vox Sanguinis (2015) 108, 393–402
Improving red cell transfusion 399
Table 4 The outcome and process indicators of blood transfusion stratified by surgical groups Group
% transfused
Cardiothoracic Pooled ↓ Selected ↓ Colorectal Selected ↓ Gynaecology and Obstetrics Pooled ↓ Selected ↓ Orthopaedic Pooled ↓ Selected ↓ General Pooled ↓ Selected ↓
No. of transfusion
Units transfused
Hb ≥ 100 g/l
70 g/l ≤ Hb < 100 g/l
Hb < 70 g/l
2/+ units transfused
↓ ↓
↓ ↓
↓ ↓
– –
↑ ↑
– –
↓
↓
–
–
–
–
↓ ↓
↓ ↓
↓ ↓
– ↑
↑ ↑
– –
↓ ↓
↓ ↓
– –
– –
– –
– –
↓ ↓
↓ ↓
– –
– –
– –
– –
The trends were assessed based on the results of the last year (i.e. 2008/2009) compared with previous years (i.e. 2004/2007). The results were derived from the corresponding three-level logistic or linear regression results from Appendix 2. We used the symbols ↓, ↑ and – to indicate a decrease, increase and unclear trend, respectively.
to 098 in 2009). Each individual surgical group exhibited similar pattern but with different baseline levels.
The adjusted trends of the outcome and process indictors between July 2004–June 2009 From the adjusted analyses, we provided a summation of the statistical difference for each measure (Table 4) but with detailed regression outcomes in Appendix 2. The trends were assessed based on the results of the last years (i.e. 2008/9) compared with previous years (i.e. 2004/ 2007). The symbols ↓, ↑ and – were used to indicate a
decrease, increase or stable trend, respectively (see Table 4). These summary results were based on the threelevel multivariate logistic regression and linear regression results for both outcome and process indicators. A downward trend in a clinical indicator suggests a desirable outcome for blood transfusion. A decrease trend in a process indicator (except for percentage of transfusion with Hb < 70 g/l) may be interpreted as favourable according to the current recommended best practice. For example, for the cardiothoracic surgery group and gynaecology and obstetric groups, after the BW, there were significant reductions in the proportion of patients
Fig. 1 The monthly observed and predicted trends of blood units transfused per admission (July 2004–June 2009). (1) The quadratic equation for overall predicted curve was: y = 0133591 + t*0003749-t2*0000049, where t = 1, 2, 60 for July 2004, August 2004,. . .June 2009, respectively; y: the monthly rate of units of RBC transfused per admission (July 2004 to June 2009); all Ps for t and t2 < 0001. (2) The quadratic equation for the predicted curve (July 04 to November 2007) and its projection (December 2007 to June 2009) was y1 = 0149069 + t*0001329 + t2*0000013; where t = 1,2. . .41 for July 2004, August 2004, . . . November 2007, respectively and y1: the monthly rate of units of RBC transfused per admissions (July 2004 to Nov 2007); all Ps for t and t2 < 0001.
© 2015 International Society of Blood Transfusion Vox Sanguinis (2015) 108, 393–402
400 B. T. Harrison et al.
who received blood transfusions with minimum Hb ≥ 100 g/l which indicated a significant increase in guideline-based blood transfusion practice. There were also significant increases in the proportion of patients who received blood transfusions with max Hb ≤ 70 g/l for the cardiothoracic and gynaecology and obstetrics groups, and with minimum Hb ≥ 70 g/l – ≤100 g/l for the gynaecology and obstetrics group with selected ANDRGs, respectively. For both pooled and selected cardiothoracic surgical group, the trends of number of patients having two or more RBC units transfused who had at least one blood transfusion during admission were steady between 2005 and 2009.
The possible units of blood transfused avoided An estimate of the number of units of red blood cells avoided was calculated (Fig. 1). The estimation of the possible units of blood transfusion avoided was based on all surgical groups combined. In total, we estimated that 12 225 units of red blood cells transfusion might be avoided between July 2008 and June 2009, a 274% reduction of the amount if no intervention had happened (i.e. 12 225/44 564). This would equate to a $8 557 500 saving based on the calculated direct and indirect costs of RBC transfusion in the Australian healthcare setting (unpublished data: Hoffman A, Wood E et al. Australian Haematology Association Meeting; 2009).
Discussion The BW programme had as its first objective to reduce inappropriate RBC transfusion in haemodynamically stable adult patients with normal bone marrow in the postoperative period. Using collaborative improvement methods [16], 21 interventions were implemented broadly across the system. This study has shown that there were consistent and significant reductions in the chance, frequency and the number of units of red blood cells being transfused from 2007 across different selective surgical groups. Such reductions were consistent for both the pooled surgical groups and the selected high blood usage procedure subgroups for each of four specialties (i.e. cardiothoracic, gynaecology & obstetrics, orthopaedics and general surgery) as well as colorectal surgical procedures. In addition, throughout this period red cell issues per 1000 population fell in NSW relative to the two other largest states Queensland and Victoria [26]. This finding is significant as there is a direct relationship between red cell issues by the Australian Blood Service and red cell usage within each jurisdiction. These findings combined suggest that the BW programme interventions were effective.
There was also evidence to indicate that the significant decrease in the proportion of transfusions with minimum Hb tests ≥100 g/l and the increase of the proportion of transfusion with max Hb tests ≤70 g/l may have, in part, contributed to the overall improvement of clinical indicators, in particular in cardiothoracic and gynaecology and obstetric surgical groups. A possible reason that the significant reductions in the proportion of blood transfusions with minimum Hb ≥ 100 g/l were observed only in the cardiothoracic and gynaecology and obstetric groups is that these two groups had a relatively high baseline incidence rate (around 20%) while the other three groups’ baseline incidence rates were lower (around 10%). The lower baseline incidence rate may present a ‘floor effect’ that makes the further significant reduction of the rate hard to observe. There was no significant reduction for the number of two or more units of blood transfused during a single transfusion episode which demonstrates further opportunities for improvement. It is estimated that, conservatively, 12 225 units of red blood cells could have been avoided between July 2008 and June 2009, most likely due to the extensive BW interventions. This would equate to significant cost-savings of an estimate $8 557 500 per annum. Our study showed that even for selective subgroups (five elective surgical groups) of a much wider target populations (all patients across all public hospitals in NSW), there was a 274% reduction in the projected red cell usage between 2008 and 2009, which makes up 789% of the total national reductions according to the supply plan in 2009–2010 [27]. Given the fact that NSW population (724 million) makes up 324% of total Australia population [28] at 2008–2009, and assuming the same reduction could be achieved in all jurisdictions in Australia, this would equate to a reduction of 37 731 RBC units used in the five elective surgical groups; a cost-saving of A$26 411 700 (based on direct costs $300 per unit from unpublished data) in 2008–2009. This accounts for 30% of the total blood product budget in Australia and 47% of total RBCs in 2009–2010. This study has its limitations. Firstly, the study only examined the potential effect of BW on the five selected surgical groups. We used the LOS of
