Effect of a flow chart on use of blood transfusions in primary total hip and knee replacement: prospective before and after study
http://www.100md.com
《英国医生杂志》
1 Institute for Evaluative Research in Orthopaedic Surgery, University of Berne, 3001 Berne, Switzerland, 2 Department of Anaesthesiology, Inselspital, University of Berne, 3010 Berne, 3 Department of Orthopaedic Surgery, Kantonsspital Winterthur, 8400 Winterthur, Switzerland, 4 Department of Orthopaedic Surgery, University of Berne, 3010 Berne, 5 Departments of Social and Preventive Medicine and Rheumatology, University of Berne, 3012 Berne
Correspondence to: P Jüni juni@ispm.unibe.ch
Abstract
Up to 90% of patients undergoing total hip and knee replacement receive blood,1 though in the past few years there has been increasing concern about the safety of blood transfusions.2 A recent systematic review supported the restriction of transfusions in patients who are free from serious cardiac disease.3
Our orthopaedic unit at the Kantonsspital Winterthur, Switzerland, is a tertiary care facility with 31 beds, where about 230 primary total joint replacements of the hip or knee are performed annually. An intensive care unit, comprising 12 beds, and a postoperative recovery unit with seven beds, are attached.
Outline of the problem
Our primary outcome was the proportion of patients undergoing total joint replacement and receiving allogeneic blood transfusions perioperatively. Secondary outcomes were the proportion of patients receiving autologous transfusions and the proportion of patients receiving any transfusions (allogeneic and autologous combined).
Process of gathering information
We calculated that a sample size of 230 operations before and 230 after the implementation of the flow chart would allow us to detect a reduction in the incidence of patients receiving allogeneic blood transfusions from 45% to 30% with 90% power at P = 0.05 (two sided), and estimated that a duration of 12 months before and after the intervention was needed to include this number of operations.
For comparisons between control and intervention periods we used maximum likelihood logistic regression models based on robust standard errors that allowed for correlation within patients undergoing several operations, also adjusting for 10 prespecified, potentially prognostic factors (see table 2). Using a combination of cumulative sum plots5 and bootstrapping6 (1000 replications), we performed a change point analysis to detect the point in time when significant changes occurred.
Table 2 Comparison between control and intervention period in terms of total joint replacement operations requiring blood transfusions, with crude and adjusted odds ratios for transfusions
Using prespecified average estimates, we estimated benefits of the flow chart in terms of expenditure per donated unit for autologous blood (SFr240.90; ￡104; $191; 155), allogeneic blood (SFr165.20) and giving sets (SFr4.00), working hours of nurses (19 minutes), and laboratory assistants (4 minutes). We also calculated the expenditures for the development and implementation of the flow chart. Prespecified average costs of working hours were SFr45.20 per hour for physicians, SFr29.50 for nurses, and SFr27.35 for laboratory assistants.
Strategy for change
All 421 patients undergoing 448 elective primary total hip or knee replacement operations between 1 October 1998 and 30 September 2000 were included. Of these, four patients had an operation before and after the implementation of the flow chart. Therefore, 208 patients underwent 224 unilateral and two bilateral operations before the implementation, and 217 patients had 218 unilateral and four bilateral operations after the implementation of the flow chart. Table 1 shows the characteristics of patients and operations. Preoperative haemoglobin concentrations and packed cell volumes were slightly higher during the intervention period and the average length of operation was shorter. Figure 2 (top) shows the proportion of patients receiving blood transfusions over time. The percentage of patients receiving blood transfusions decreased from 35.0% (79 operations) to 19.8% (44 operations) for allogeneic blood (difference -15.2%, 95% confidence interval -23.3 to -7.0%), from 28.8% (65 operations) to 5.9% (13 operations) for autologous blood -22.9%, -29.6 to -16.2%) and from 59.7% (135 operations) to 24.8% (55 operations) for any blood transfusion (-35.0%, -43.5 to -26.4%). Change point analysis indicated that the proportion of patients receiving allogeneic transfusions significantly decreased around November 1999—that is, one month after the implementation of the flow chart (confidence interval for point in time, April 1999 to May 2000). For both autologous and any transfusions, the respective estimate was October 1999 (September to October 1999).
Table 1 Comparison of characteristics of included patients and operations, according to time period. Values are means for continuous data and percentages for binary data. Differences between periods are shown along with 95% confidence intervals.
Fig 2 Top: Percentage of operated patients receiving transfusions across time. Bottom: Average number of allogeneic blood units transfused per total joint replacement operation and total number of units transfused at our hospital (expressed as 1000s of units). Variation across quarters in overall number of transfused units is likely to reflect annual variation in number of performed elective and emergency operations
The proportion of allogeneic transfusions that did not fulfil the criteria for red blood cell transfusions published by the American College of Physicians8 decreased from 43.8% to 15.9% (-27.9%, -43.2 to -12.5%). The number of blood units used fell from 200 to 102 for allogeneic blood (difference -0.43 units per total joint replacement operation, -0.66 to -0.19) and from 127 to 25 for autologous blood (-0.45 units per operation, -0.59 to -0.31). The flow chart also seemed to have influenced our staff's advice to patients regarding autologous blood donation: the number of patients donating blood preoperatively decreased from 98 patients (47.1%) donating 245 units during the control period to 53 patients (24.4%) donating 107 units during the intervention period.
Table 2 presents results from logistic regression models: differences between periods became more pronounced after we adjusted for prognostic factors. Figure 3 indicates that it was a difference in early peri-operative management that resulted in the observed decrease of transfusions and a considerable increase in the proportion of patients with postoperative haemoglobin concentrations below 90 g/l during the intervention period. Two patients (one during each period) experienced an ischaemic event: both of them had uncomplicated angina pectoris that resolved promptly. No serious adverse events occurred, and the mean length of perioperative admission was similar (14.1 v 14.0 days, difference -0.1, -1.2 to 0.9).
Fig 3 Top: Percentage of operated patients receiving allogeneic blood transfusions during control and intervention period perioperatively (day 0) and on days 1, 2, and 3 postoperatively. No allogeneic transfusions were given later than 3 days postoperatively. Bottom: Percentage of patients during control and intervention period with haemoglobin values below 90 g/l immediately after operation (day 0) and on days 1, 3, and 7 postoperatively
Figure 2 (lower panel) contrasts the mean number of allogeneic blood units per operation with the overall number of allogeneic blood units used in this hospital. Exploratory change point analysis shows that the number of units used per operation significantly decreased around November 1999 (May 1999 to April 2000); no significant changes were found in the overall number of allogeneic blood units in the hospital. No clear time trends emerged nationally and regionally: the estimated number of allogeneic blood units used in Switzerland decreased from 310 000 in 1998 to 285 000 in 2000 (-8%), but increased from 52 700 to 60 600 units in the region of Zurich (plus/minus 15%).
The 300 copies of the flow chart cost SFr30.00. Physicians spent 29 hours and nurses spent 2 hours on the development of the chart, corresponding to salary costs equivalent to SFr1369.80. Teaching sessions led to an overall loss of 14.5 working hours (SFr490.60). Physicians spent an overall of 10.5 hours on preparation and analysis of data used for feedback (SFr474.60). Provision of feedback during routine staff meetings resulted in an estimated overall loss of 9.7 working hours (SFr327.00). Therefore, the overall cost of the intervention was SFr2 692.00.
Per operation, the reduction in blood transfusions resulted in savings of 16.6 minutes of nurse working time (SFr8.20), 1.7 minutes laboratory assistant working time (SFr0.80), SFr72.00 for costs of allogeneic blood transfusions and SFr146.80 for autologous blood transfusions. Overall, this led to an average saving of SFr227.80 per operation and an overall reduction of the estimated annual expenditure for blood transfusions from SFr96 787.90 to SFr44 507.10.
Lessons learnt and next steps
By implementing our flow chart, we were able to reduce the proportion of patients receiving blood after total joint replacement by more than 40%. Even more pronounced effects were observed for autologous blood transfusions, suggesting that appropriate clinical judgment may lead to more selective use of this approach, resulting in increased cost effectiveness. These benefits led to an approximate decrease in annual overall costs for transfusions related to total joint replacement of about SFr52 000 (￡23 000) in our unit that was achieved at a cost of about SFr2700 (￡1200).
Our project has potential limitations. Firstly, the before and after design may have led to confounding because of differences in prognostic factors between control and intervention period. For example, differences in preoperative haemoglobin concentrations (table 1) could have resulted in a decrease in the perceived need for transfusions irrespective of the flow chart. To address this, we adjusted estimates using multivariable logistic regression models and found our results to be robust. Secondly, the effect of being under study (also referred to as the Hawthorne effect) and, more specifically, the educational effects of audit and feedback12 may have contributed to the observed changes, independently of the flow chart. Change point analyses, however, indicated a clear cut temporal association between implementation of the flow chart in October 1999 and the occurrence of significant changes (October/November 1999) that could be explained neither by the start of the audit (October 1998) nor by the feedback sessions (January and July 2000). Thirdly, the observed changes could reflect changes in the general attitude towards allogeneic blood transfusions, but routine data obtained for our hospital, the wider region of Zurich, and Switzerland showed no time trends that could explain our results. Fourthly, the success of our intervention could not be generalised if it related mainly to an excessively high proportion of patients receiving blood transfusions in our unit. This proportion, however, was well below the average observed internationally.13 Finally, we did not follow up patients after they were discharged from hospital, and we could have missed serious adverse events occurring late in the postoperative course. This is unlikely, considering that a landmark trial in critically ill patients by Hebert et al found that a restrictive transfusion strategy, similar to ours, tended to be superior to a liberal transfusion strategy, with a trend towards a decreased overall mortality in patients allocated to the restrictive strategy.9
Key learning points
Current evidence supports the use of restrictive transfusion strategies but up to 90% of patients undergoing total hip or knee replacement receive blood transfusions
After the implementation of a simple flow chart the percentage of patients receiving transfusions decreased from 35% to 20% for allogeneic blood and from 29% to 6% for autologous blood, resulting in an estimated reduction in annual costs of ￡23 000
Five key elements may have contributed to the success of the flow chart: its simplicity, its wide distribution, no requirement for major changes, the endorsement by local opinion leaders, and the development of a sense of ownership among staff
We conclude that the observed effect of implementing our simple flow chart on the perioperative management of anaemic patients after total joint replacement is likely to be real, and suggest that it was related to the following five key elements: the obvious simplicity of the flow chart with a graphical summary of decision pathways that could be followed easily by everybody, the wide distribution of the flow chart, no requirement for major changes to existing routines, the endorsement by local opinion leaders, and the development of a sense of ownership among physicians and nurses. The combination of these elements may be used in other contexts to achieve sustained change of clinical practice.
We thank Verena Meier and Annelies Solenthaler for participating in development of the flow chart, reviewing the protocol, and assisting with data collection; Balz Isler, Andreas Kündig, and Markus Fopp for reviewing the flow chart; René Orler for reviewing the protocol and assisting with data collection; Matthias Egger, Bj?rn Erik von Elm, Liz King, Arthur Marx, and Stephan Reichenbach for helpful comments; and the orthopaedics, anaesthesiology, and intensive care teams of the Kantonsspital Winterthur for actively participating in the project.
Contributors: UM conceived the study and had main responsibility for the development of protocol and flow chart, data collection, and management. AE participated in developing the flow chart and reviewed the protocol. CR and SE reviewed the protocol and participated in data preparation. MP participated in developing the flow chart and was responsible for data collection. PJ reviewed the protocol, had main responsibility for data preparation, analysis, and interpretation, and wrote the first draft of the paper. All investigators participated in data interpretation and contributed to the final draft. UM and PJ are the guarantors.
Funding: PJ is a senior research fellow funded by the Swiss National Science Foundation (grants 32-66377.01 and 32-66378.01) and is also affiliated with the UK Medical Research Council's Health Services Research Collaboration at the Department of Social Medicine at the University of Bristol.
Competing interests: None declared.
Ethical approval: None required.
References
McClelland B. Risk in blood transfusion with reference to the present situation in the United Kingdom. Transfus Clin Biol 1994;1: 425-6.
Williamson LM, Lowe S, Love EM, Cohen H, Soldan K, McClelland DB, et al. Serious hazards of transfusion (SHOT) initiative: analysis of the first two annual reports. BMJ 1999;319: 16-9.
Hill SR, Carless PA, Henry DA, Carson JL, Hebert PC, McClelland DB, et al. Transfusion thresholds and other strategies for guiding allogeneic red blood cell transfusion. Cochrane Database Syst Rev 2002;(2): CD002042 .
Regan F, Taylor C. Blood transfusion medicine. BMJ 2002;325: 143-7.
Wohl H. The cusum plot: its utility in the analysis of clinical data. N Engl J Med 1977;296: 1044-5.
Efron B, Tibshirani RJ. An introduction to the bootstrap. New York: Chapman and Hall, 1993.
American Society of Anesthesiologists Task Force on Blood Component Therapy. Practice guidelines for blood component therapy. Anesthesiology 1996;84: 732-47.
American College of Physicians. Practice strategies for elective red blood cell transfusion. Ann Intern Med 1992;116: 403-6.
Hebert PC, Wells G, Blajchman MA, Marshall J, Martin C, Pagliarello G, et al. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. Transfusion Requirements in Critical Care Investigators, Canadian Critical Care Trials Group. N Engl J Med 1999;340: 409-17.
Marino PL. Blood component therapy. The ICU Book. Philadelphia: Lippincott Williams Wilkins, 1998.
Page MH, Shepherd BD, Harrison JM. Reduction of blood loss in knee arthroplasty. Aust N Z J Surg 1984;54: 141-4.
Jamtvedt G, Young JM, Kristoffersen DT, Thomson O'Brien MA, Oxman AD. Audit and feedback: effects on professional practice and health care outcomes. Cochrane Database Syst Rev 2003;(1): CD000259 .
Sanguis Study Group. Use of blood products for elective surgery in 43 European hospitals. Transfus Med 1994;4: 251-68.(Urs Müller, senior resear)
Correspondence to: P Jüni juni@ispm.unibe.ch
Abstract
Up to 90% of patients undergoing total hip and knee replacement receive blood,1 though in the past few years there has been increasing concern about the safety of blood transfusions.2 A recent systematic review supported the restriction of transfusions in patients who are free from serious cardiac disease.3
Our orthopaedic unit at the Kantonsspital Winterthur, Switzerland, is a tertiary care facility with 31 beds, where about 230 primary total joint replacements of the hip or knee are performed annually. An intensive care unit, comprising 12 beds, and a postoperative recovery unit with seven beds, are attached.
Outline of the problem
Our primary outcome was the proportion of patients undergoing total joint replacement and receiving allogeneic blood transfusions perioperatively. Secondary outcomes were the proportion of patients receiving autologous transfusions and the proportion of patients receiving any transfusions (allogeneic and autologous combined).
Process of gathering information
We calculated that a sample size of 230 operations before and 230 after the implementation of the flow chart would allow us to detect a reduction in the incidence of patients receiving allogeneic blood transfusions from 45% to 30% with 90% power at P = 0.05 (two sided), and estimated that a duration of 12 months before and after the intervention was needed to include this number of operations.
For comparisons between control and intervention periods we used maximum likelihood logistic regression models based on robust standard errors that allowed for correlation within patients undergoing several operations, also adjusting for 10 prespecified, potentially prognostic factors (see table 2). Using a combination of cumulative sum plots5 and bootstrapping6 (1000 replications), we performed a change point analysis to detect the point in time when significant changes occurred.
Table 2 Comparison between control and intervention period in terms of total joint replacement operations requiring blood transfusions, with crude and adjusted odds ratios for transfusions
Using prespecified average estimates, we estimated benefits of the flow chart in terms of expenditure per donated unit for autologous blood (SFr240.90; ￡104; $191; 155), allogeneic blood (SFr165.20) and giving sets (SFr4.00), working hours of nurses (19 minutes), and laboratory assistants (4 minutes). We also calculated the expenditures for the development and implementation of the flow chart. Prespecified average costs of working hours were SFr45.20 per hour for physicians, SFr29.50 for nurses, and SFr27.35 for laboratory assistants.
Strategy for change
All 421 patients undergoing 448 elective primary total hip or knee replacement operations between 1 October 1998 and 30 September 2000 were included. Of these, four patients had an operation before and after the implementation of the flow chart. Therefore, 208 patients underwent 224 unilateral and two bilateral operations before the implementation, and 217 patients had 218 unilateral and four bilateral operations after the implementation of the flow chart. Table 1 shows the characteristics of patients and operations. Preoperative haemoglobin concentrations and packed cell volumes were slightly higher during the intervention period and the average length of operation was shorter. Figure 2 (top) shows the proportion of patients receiving blood transfusions over time. The percentage of patients receiving blood transfusions decreased from 35.0% (79 operations) to 19.8% (44 operations) for allogeneic blood (difference -15.2%, 95% confidence interval -23.3 to -7.0%), from 28.8% (65 operations) to 5.9% (13 operations) for autologous blood -22.9%, -29.6 to -16.2%) and from 59.7% (135 operations) to 24.8% (55 operations) for any blood transfusion (-35.0%, -43.5 to -26.4%). Change point analysis indicated that the proportion of patients receiving allogeneic transfusions significantly decreased around November 1999—that is, one month after the implementation of the flow chart (confidence interval for point in time, April 1999 to May 2000). For both autologous and any transfusions, the respective estimate was October 1999 (September to October 1999).
Table 1 Comparison of characteristics of included patients and operations, according to time period. Values are means for continuous data and percentages for binary data. Differences between periods are shown along with 95% confidence intervals.
Fig 2 Top: Percentage of operated patients receiving transfusions across time. Bottom: Average number of allogeneic blood units transfused per total joint replacement operation and total number of units transfused at our hospital (expressed as 1000s of units). Variation across quarters in overall number of transfused units is likely to reflect annual variation in number of performed elective and emergency operations
The proportion of allogeneic transfusions that did not fulfil the criteria for red blood cell transfusions published by the American College of Physicians8 decreased from 43.8% to 15.9% (-27.9%, -43.2 to -12.5%). The number of blood units used fell from 200 to 102 for allogeneic blood (difference -0.43 units per total joint replacement operation, -0.66 to -0.19) and from 127 to 25 for autologous blood (-0.45 units per operation, -0.59 to -0.31). The flow chart also seemed to have influenced our staff's advice to patients regarding autologous blood donation: the number of patients donating blood preoperatively decreased from 98 patients (47.1%) donating 245 units during the control period to 53 patients (24.4%) donating 107 units during the intervention period.
Table 2 presents results from logistic regression models: differences between periods became more pronounced after we adjusted for prognostic factors. Figure 3 indicates that it was a difference in early peri-operative management that resulted in the observed decrease of transfusions and a considerable increase in the proportion of patients with postoperative haemoglobin concentrations below 90 g/l during the intervention period. Two patients (one during each period) experienced an ischaemic event: both of them had uncomplicated angina pectoris that resolved promptly. No serious adverse events occurred, and the mean length of perioperative admission was similar (14.1 v 14.0 days, difference -0.1, -1.2 to 0.9).
Fig 3 Top: Percentage of operated patients receiving allogeneic blood transfusions during control and intervention period perioperatively (day 0) and on days 1, 2, and 3 postoperatively. No allogeneic transfusions were given later than 3 days postoperatively. Bottom: Percentage of patients during control and intervention period with haemoglobin values below 90 g/l immediately after operation (day 0) and on days 1, 3, and 7 postoperatively
Figure 2 (lower panel) contrasts the mean number of allogeneic blood units per operation with the overall number of allogeneic blood units used in this hospital. Exploratory change point analysis shows that the number of units used per operation significantly decreased around November 1999 (May 1999 to April 2000); no significant changes were found in the overall number of allogeneic blood units in the hospital. No clear time trends emerged nationally and regionally: the estimated number of allogeneic blood units used in Switzerland decreased from 310 000 in 1998 to 285 000 in 2000 (-8%), but increased from 52 700 to 60 600 units in the region of Zurich (plus/minus 15%).
The 300 copies of the flow chart cost SFr30.00. Physicians spent 29 hours and nurses spent 2 hours on the development of the chart, corresponding to salary costs equivalent to SFr1369.80. Teaching sessions led to an overall loss of 14.5 working hours (SFr490.60). Physicians spent an overall of 10.5 hours on preparation and analysis of data used for feedback (SFr474.60). Provision of feedback during routine staff meetings resulted in an estimated overall loss of 9.7 working hours (SFr327.00). Therefore, the overall cost of the intervention was SFr2 692.00.
Per operation, the reduction in blood transfusions resulted in savings of 16.6 minutes of nurse working time (SFr8.20), 1.7 minutes laboratory assistant working time (SFr0.80), SFr72.00 for costs of allogeneic blood transfusions and SFr146.80 for autologous blood transfusions. Overall, this led to an average saving of SFr227.80 per operation and an overall reduction of the estimated annual expenditure for blood transfusions from SFr96 787.90 to SFr44 507.10.
Lessons learnt and next steps
By implementing our flow chart, we were able to reduce the proportion of patients receiving blood after total joint replacement by more than 40%. Even more pronounced effects were observed for autologous blood transfusions, suggesting that appropriate clinical judgment may lead to more selective use of this approach, resulting in increased cost effectiveness. These benefits led to an approximate decrease in annual overall costs for transfusions related to total joint replacement of about SFr52 000 (￡23 000) in our unit that was achieved at a cost of about SFr2700 (￡1200).
Our project has potential limitations. Firstly, the before and after design may have led to confounding because of differences in prognostic factors between control and intervention period. For example, differences in preoperative haemoglobin concentrations (table 1) could have resulted in a decrease in the perceived need for transfusions irrespective of the flow chart. To address this, we adjusted estimates using multivariable logistic regression models and found our results to be robust. Secondly, the effect of being under study (also referred to as the Hawthorne effect) and, more specifically, the educational effects of audit and feedback12 may have contributed to the observed changes, independently of the flow chart. Change point analyses, however, indicated a clear cut temporal association between implementation of the flow chart in October 1999 and the occurrence of significant changes (October/November 1999) that could be explained neither by the start of the audit (October 1998) nor by the feedback sessions (January and July 2000). Thirdly, the observed changes could reflect changes in the general attitude towards allogeneic blood transfusions, but routine data obtained for our hospital, the wider region of Zurich, and Switzerland showed no time trends that could explain our results. Fourthly, the success of our intervention could not be generalised if it related mainly to an excessively high proportion of patients receiving blood transfusions in our unit. This proportion, however, was well below the average observed internationally.13 Finally, we did not follow up patients after they were discharged from hospital, and we could have missed serious adverse events occurring late in the postoperative course. This is unlikely, considering that a landmark trial in critically ill patients by Hebert et al found that a restrictive transfusion strategy, similar to ours, tended to be superior to a liberal transfusion strategy, with a trend towards a decreased overall mortality in patients allocated to the restrictive strategy.9
Key learning points
Current evidence supports the use of restrictive transfusion strategies but up to 90% of patients undergoing total hip or knee replacement receive blood transfusions
After the implementation of a simple flow chart the percentage of patients receiving transfusions decreased from 35% to 20% for allogeneic blood and from 29% to 6% for autologous blood, resulting in an estimated reduction in annual costs of ￡23 000
Five key elements may have contributed to the success of the flow chart: its simplicity, its wide distribution, no requirement for major changes, the endorsement by local opinion leaders, and the development of a sense of ownership among staff
We conclude that the observed effect of implementing our simple flow chart on the perioperative management of anaemic patients after total joint replacement is likely to be real, and suggest that it was related to the following five key elements: the obvious simplicity of the flow chart with a graphical summary of decision pathways that could be followed easily by everybody, the wide distribution of the flow chart, no requirement for major changes to existing routines, the endorsement by local opinion leaders, and the development of a sense of ownership among physicians and nurses. The combination of these elements may be used in other contexts to achieve sustained change of clinical practice.
We thank Verena Meier and Annelies Solenthaler for participating in development of the flow chart, reviewing the protocol, and assisting with data collection; Balz Isler, Andreas Kündig, and Markus Fopp for reviewing the flow chart; René Orler for reviewing the protocol and assisting with data collection; Matthias Egger, Bj?rn Erik von Elm, Liz King, Arthur Marx, and Stephan Reichenbach for helpful comments; and the orthopaedics, anaesthesiology, and intensive care teams of the Kantonsspital Winterthur for actively participating in the project.
Contributors: UM conceived the study and had main responsibility for the development of protocol and flow chart, data collection, and management. AE participated in developing the flow chart and reviewed the protocol. CR and SE reviewed the protocol and participated in data preparation. MP participated in developing the flow chart and was responsible for data collection. PJ reviewed the protocol, had main responsibility for data preparation, analysis, and interpretation, and wrote the first draft of the paper. All investigators participated in data interpretation and contributed to the final draft. UM and PJ are the guarantors.
Funding: PJ is a senior research fellow funded by the Swiss National Science Foundation (grants 32-66377.01 and 32-66378.01) and is also affiliated with the UK Medical Research Council's Health Services Research Collaboration at the Department of Social Medicine at the University of Bristol.
Competing interests: None declared.
Ethical approval: None required.
References
McClelland B. Risk in blood transfusion with reference to the present situation in the United Kingdom. Transfus Clin Biol 1994;1: 425-6.
Williamson LM, Lowe S, Love EM, Cohen H, Soldan K, McClelland DB, et al. Serious hazards of transfusion (SHOT) initiative: analysis of the first two annual reports. BMJ 1999;319: 16-9.
Hill SR, Carless PA, Henry DA, Carson JL, Hebert PC, McClelland DB, et al. Transfusion thresholds and other strategies for guiding allogeneic red blood cell transfusion. Cochrane Database Syst Rev 2002;(2): CD002042 .
Regan F, Taylor C. Blood transfusion medicine. BMJ 2002;325: 143-7.
Wohl H. The cusum plot: its utility in the analysis of clinical data. N Engl J Med 1977;296: 1044-5.
Efron B, Tibshirani RJ. An introduction to the bootstrap. New York: Chapman and Hall, 1993.
American Society of Anesthesiologists Task Force on Blood Component Therapy. Practice guidelines for blood component therapy. Anesthesiology 1996;84: 732-47.
American College of Physicians. Practice strategies for elective red blood cell transfusion. Ann Intern Med 1992;116: 403-6.
Hebert PC, Wells G, Blajchman MA, Marshall J, Martin C, Pagliarello G, et al. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. Transfusion Requirements in Critical Care Investigators, Canadian Critical Care Trials Group. N Engl J Med 1999;340: 409-17.
Marino PL. Blood component therapy. The ICU Book. Philadelphia: Lippincott Williams Wilkins, 1998.
Page MH, Shepherd BD, Harrison JM. Reduction of blood loss in knee arthroplasty. Aust N Z J Surg 1984;54: 141-4.
Jamtvedt G, Young JM, Kristoffersen DT, Thomson O'Brien MA, Oxman AD. Audit and feedback: effects on professional practice and health care outcomes. Cochrane Database Syst Rev 2003;(1): CD000259 .
Sanguis Study Group. Use of blood products for elective surgery in 43 European hospitals. Transfus Med 1994;4: 251-68.(Urs Müller, senior resear)