Electronic Medical Alerts — So Simple, So Complex
http://www.100md.com
《新英格兰医药杂志》
One of the most consistent findings in health research is the gap between evidence and practice. It is estimated that the care received by 30 to 40 percent of patients in the United States and the Netherlands does not conform to currently available scientific evidence.1,2 In most developed countries, the publication of national clinical practice guidelines over the past 15 years marked the first attempt to fill this gap. However, merely issuing recommendations is not enough to change practice. Numerous studies have shown that well-designed, complex interventions are required to change physicians' behavior. Reminders that appear when physicians are writing prescriptions typically have the largest effect.2
Computerized physician order-entry (CPOE) systems with embedded decision-support tools are thought to have such an effect in daily practice. CPOE systems have been shown to improve clinical practice in a variety of areas, including ordering drugs or biologic tests, presenting cost data, providing decision-making rules or guidelines, and preventing medication errors. However, more attention has been paid to their effect on physicians' behavior than on patient outcomes, and some conflicting results have been reported, particularly in circumstances requiring the integration of complex guidelines.3
In addition, several organizational and financial factors limit the adoption of CPOE systems. Despite pressure to adopt such systems, only 9.6 percent of U.S. hospitals (and probably even fewer European hospitals) have complete access to a CPOE system. CPOE systems are not always accepted by clinicians: such systems are used to place more than 90 percent of orders in only one third of hospitals in which they have been adopted.4 As an example of barriers to adopting CPOE systems, many physicians express concern that using the system to place orders takes longer than writing orders on paper.3 A one-size-fits-all approach to the design of decision-support systems or clinical alerts used in the context of CPOE does not seem to work; thus, evaluations of such systems have mainly concerned "homegrown" systems in specific settings.5
In this issue of the Journal, Kucher et al. report the results of a randomized, controlled trial to test a strategy of issuing electronic alerts to physicians whose patients were not receiving prophylaxis against deep-vein thrombosis.6 The authors developed a computer program, linked to the patient database of Brigham and Women's Hospital in Boston, to identify hospitalized patients at risk for venous thromboembolism. The computer program was used to calculate the venous thromboembolism risk profile of each hospitalized patient with the use of eight common risk factors. Each factor was weighted according to a point scale (3 points for a major risk factor, and 1 point for a minor risk factor).
The 2506 patients (82.7 percent of whom were medical patients) who did not receive any prophylaxis, despite a total risk score for venous thromboembolism of at least 4, were randomly assigned to the intervention or control group. The intervention consisted of a computer alert recommending prophylaxis and offering a link to the hospital's venous thromboembolism prevention guidelines. The study's primary end point, clinical diagnosis of deep-vein thrombosis or pulmonary embolism at 90 days, occurred in 61 patients in the intervention group, as compared with 103 patients in the control group (4.9 percent vs. 8.2 percent), a relative risk reduction of 41 percent (hazard ratio, 0.59; 95 percent confidence interval, 0.43 to 0.81). This means that the intervention needs to be used in just 30 patients (95 percent confidence interval, 19 to 70) to prevent one case of deep-vein thrombosis or pulmonary embolism. Prophylactic measures were implemented in 33.5 percent of patients in the intervention group and 14.5 percent of patients in the control group.
Kucher et al. report that if computer alerts had been provided for the entire cohort of patients at risk, not just for those who did not receive any prophylaxis against venous thromboembolism, the computer program might have increased the rate of use of prophylaxis from 84.6 percent to 88.0 percent. This number can be interpreted as representing a valuable improvement in quality or a weak effect. It could be explained by a high prophylaxis rate in this population (ceiling effect), but could also be explained by the study design. Most physicians treated both patients in the intervention group and those in the control group, which could have influenced the use of prophylaxis in the control group. The use of clustered, randomized trials, in which organizations or health professionals rather than individual patients undergo randomization, would minimize this possibility.7
Another limitation of the study is the possibility of diagnostic bias. Since prophylaxis was not administered in a blinded manner, physicians may have been more likely to perform monitoring for venous thromboembolism either in patients who were treated or in those who were not treated. This bias may have increased or decreased the difference between groups. The magnitude of the difference between groups depends on all these factors.
The most important result of the study by Kucher et al. is the demonstration that clinical alerts had an effect on both the approach to care and outcomes during the three-year study period. Opponents of the use of clinical guidelines and decision-support systems have always argued that the demonstration of their effect on processes of care is not enough to support their acceptance and use. Because the decision aid was integrated into a routinely used commercial system, the benefits can be expected to persist even after completion of the study.
What important lessons can we draw from this study? It shows that three conditions are needed for the successful implementation of clinical alerts in CPOE systems. First, a favorable environment is needed for the development of CPOE systems and their acceptance by users; Brigham and Women's Hospital has had an integrated hospital information system for more than 10 years.8 Physicians there have long been interested in the prevention of venous thromboembolism — for example, they have organized educational programs, drawn up local clinical guidelines, and carried out an audit of practice, and they programmed their order-entry system to suggest prophylaxis before this study was undertaken.
Second, to have an effect, electronic alerts should deliver a very simple message in a way that prevents physicians from bypassing them in daily practice. In the study by Kucher et al., physicians were forced to acknowledge the computer alert. For this reason, such decision support should be limited to selected high-risk events for which key decisions are required.9 If too many alerts or irrelevant alerts are provided by computers, the clinician will ignore them and try to bypass them.
Third, clinical alerts are important in situations, such as the prevention of deep-vein thrombosis, for which the physician requires access to a large amount of complex information to make an adequate decision. The main value of the system was not only the delivery of reminders, which could potentially be done on paper,2 but also the ability of the software to screen for risk factors in a full and integrated database, in order to determine the patient's risk score for venous thromboembolism.
However, the situation is not so simple. Much more work is needed to improve the overall quality of care through the use of computers for translating evidence into practice. The challenges are to make CPOE systems acceptable to all institutions and clinicians and to integrate the appropriate number of validated clinical alerts into these systems. Another challenge is the development of software that can easily be modified and transferred from one institution to another to fit specific and local needs.9 It is difficult to know how easily the software described by Kucher et al. could be implemented and accepted in another institution.
Finally, it is important to remember that clinical alerts are tools that help but do not transform the decision-making process. Ultimately, the physician, not the computer, makes the decision. However, computers are definitely part of our life, and clinicians need to be reminded to perform some complex tasks. This need is probably due to "man's limitations as a data processor," as pointed out almost 30 years ago by Clement McDonald in a landmark article in the Journal.10
Source Information
From H?pital Européen Georges Pompidou, Faculté de Médecine René Descartes, Paris 5 and INSERM Unité 729, Paris.
References
McGlynn EA, Asch SM, Adams J, et al. The quality of health care delivered to adults in the United States. N Engl J Med 2003;348:2635-2645.
Grol R, Grimshaw J. From best evidence to best practice: effective implementation of change in patients' care. Lancet 2003;362:1225-1230.
Kuperman GJ, Gibson RF. Computer physician order entry: benefits, costs, and issues. Ann Intern Med 2003;139:31-39.
Ash JS, Gorman PN, Seshadri V, Hersh WR. Computerized physician order entry in U.S. hospitals: results of a 2002 survey. J Am Med Inform Assoc 2004;11:95-99.
Kaushal R, Shojania KG, Bates DW. Effects of computerized physician order entry and clinical decision support systems on medication safety: a systematic review. Arch Intern Med 2003;163:1409-1416.
Kucher N, Koo S, Quiroz R, et al. Electronic alerts to prevent venous thromboembolism among hospitalized patients. N Engl J Med 2005;352:969-977.
Eccles M, Grimshaw J, Campbell M, Ramsay C. Research designs for studies evaluating the effectiveness of change and improvement strategies. Qual Saf Health Care 2003;12:47-52.
Bates DW, Kuperman GJ, Wang S, et al. Ten commandments for effective clinical decision support: making the practice of evidence-based medicine a reality. J Am Med Inform Assoc 2003;10:523-530.
Handler JA, Feied CF, Coonan K, et al. Computerized physician order entry and online decision support. Acad Emerg Med 2004;11:1135-1141.
McDonald CJ. Protocol-based computer reminders, the quality of care and the non-perfectability of man. N Engl J Med 1976;295:1351-1355.(Pierre Durieux, M.D., M.P)
Computerized physician order-entry (CPOE) systems with embedded decision-support tools are thought to have such an effect in daily practice. CPOE systems have been shown to improve clinical practice in a variety of areas, including ordering drugs or biologic tests, presenting cost data, providing decision-making rules or guidelines, and preventing medication errors. However, more attention has been paid to their effect on physicians' behavior than on patient outcomes, and some conflicting results have been reported, particularly in circumstances requiring the integration of complex guidelines.3
In addition, several organizational and financial factors limit the adoption of CPOE systems. Despite pressure to adopt such systems, only 9.6 percent of U.S. hospitals (and probably even fewer European hospitals) have complete access to a CPOE system. CPOE systems are not always accepted by clinicians: such systems are used to place more than 90 percent of orders in only one third of hospitals in which they have been adopted.4 As an example of barriers to adopting CPOE systems, many physicians express concern that using the system to place orders takes longer than writing orders on paper.3 A one-size-fits-all approach to the design of decision-support systems or clinical alerts used in the context of CPOE does not seem to work; thus, evaluations of such systems have mainly concerned "homegrown" systems in specific settings.5
In this issue of the Journal, Kucher et al. report the results of a randomized, controlled trial to test a strategy of issuing electronic alerts to physicians whose patients were not receiving prophylaxis against deep-vein thrombosis.6 The authors developed a computer program, linked to the patient database of Brigham and Women's Hospital in Boston, to identify hospitalized patients at risk for venous thromboembolism. The computer program was used to calculate the venous thromboembolism risk profile of each hospitalized patient with the use of eight common risk factors. Each factor was weighted according to a point scale (3 points for a major risk factor, and 1 point for a minor risk factor).
The 2506 patients (82.7 percent of whom were medical patients) who did not receive any prophylaxis, despite a total risk score for venous thromboembolism of at least 4, were randomly assigned to the intervention or control group. The intervention consisted of a computer alert recommending prophylaxis and offering a link to the hospital's venous thromboembolism prevention guidelines. The study's primary end point, clinical diagnosis of deep-vein thrombosis or pulmonary embolism at 90 days, occurred in 61 patients in the intervention group, as compared with 103 patients in the control group (4.9 percent vs. 8.2 percent), a relative risk reduction of 41 percent (hazard ratio, 0.59; 95 percent confidence interval, 0.43 to 0.81). This means that the intervention needs to be used in just 30 patients (95 percent confidence interval, 19 to 70) to prevent one case of deep-vein thrombosis or pulmonary embolism. Prophylactic measures were implemented in 33.5 percent of patients in the intervention group and 14.5 percent of patients in the control group.
Kucher et al. report that if computer alerts had been provided for the entire cohort of patients at risk, not just for those who did not receive any prophylaxis against venous thromboembolism, the computer program might have increased the rate of use of prophylaxis from 84.6 percent to 88.0 percent. This number can be interpreted as representing a valuable improvement in quality or a weak effect. It could be explained by a high prophylaxis rate in this population (ceiling effect), but could also be explained by the study design. Most physicians treated both patients in the intervention group and those in the control group, which could have influenced the use of prophylaxis in the control group. The use of clustered, randomized trials, in which organizations or health professionals rather than individual patients undergo randomization, would minimize this possibility.7
Another limitation of the study is the possibility of diagnostic bias. Since prophylaxis was not administered in a blinded manner, physicians may have been more likely to perform monitoring for venous thromboembolism either in patients who were treated or in those who were not treated. This bias may have increased or decreased the difference between groups. The magnitude of the difference between groups depends on all these factors.
The most important result of the study by Kucher et al. is the demonstration that clinical alerts had an effect on both the approach to care and outcomes during the three-year study period. Opponents of the use of clinical guidelines and decision-support systems have always argued that the demonstration of their effect on processes of care is not enough to support their acceptance and use. Because the decision aid was integrated into a routinely used commercial system, the benefits can be expected to persist even after completion of the study.
What important lessons can we draw from this study? It shows that three conditions are needed for the successful implementation of clinical alerts in CPOE systems. First, a favorable environment is needed for the development of CPOE systems and their acceptance by users; Brigham and Women's Hospital has had an integrated hospital information system for more than 10 years.8 Physicians there have long been interested in the prevention of venous thromboembolism — for example, they have organized educational programs, drawn up local clinical guidelines, and carried out an audit of practice, and they programmed their order-entry system to suggest prophylaxis before this study was undertaken.
Second, to have an effect, electronic alerts should deliver a very simple message in a way that prevents physicians from bypassing them in daily practice. In the study by Kucher et al., physicians were forced to acknowledge the computer alert. For this reason, such decision support should be limited to selected high-risk events for which key decisions are required.9 If too many alerts or irrelevant alerts are provided by computers, the clinician will ignore them and try to bypass them.
Third, clinical alerts are important in situations, such as the prevention of deep-vein thrombosis, for which the physician requires access to a large amount of complex information to make an adequate decision. The main value of the system was not only the delivery of reminders, which could potentially be done on paper,2 but also the ability of the software to screen for risk factors in a full and integrated database, in order to determine the patient's risk score for venous thromboembolism.
However, the situation is not so simple. Much more work is needed to improve the overall quality of care through the use of computers for translating evidence into practice. The challenges are to make CPOE systems acceptable to all institutions and clinicians and to integrate the appropriate number of validated clinical alerts into these systems. Another challenge is the development of software that can easily be modified and transferred from one institution to another to fit specific and local needs.9 It is difficult to know how easily the software described by Kucher et al. could be implemented and accepted in another institution.
Finally, it is important to remember that clinical alerts are tools that help but do not transform the decision-making process. Ultimately, the physician, not the computer, makes the decision. However, computers are definitely part of our life, and clinicians need to be reminded to perform some complex tasks. This need is probably due to "man's limitations as a data processor," as pointed out almost 30 years ago by Clement McDonald in a landmark article in the Journal.10
Source Information
From H?pital Européen Georges Pompidou, Faculté de Médecine René Descartes, Paris 5 and INSERM Unité 729, Paris.
References
McGlynn EA, Asch SM, Adams J, et al. The quality of health care delivered to adults in the United States. N Engl J Med 2003;348:2635-2645.
Grol R, Grimshaw J. From best evidence to best practice: effective implementation of change in patients' care. Lancet 2003;362:1225-1230.
Kuperman GJ, Gibson RF. Computer physician order entry: benefits, costs, and issues. Ann Intern Med 2003;139:31-39.
Ash JS, Gorman PN, Seshadri V, Hersh WR. Computerized physician order entry in U.S. hospitals: results of a 2002 survey. J Am Med Inform Assoc 2004;11:95-99.
Kaushal R, Shojania KG, Bates DW. Effects of computerized physician order entry and clinical decision support systems on medication safety: a systematic review. Arch Intern Med 2003;163:1409-1416.
Kucher N, Koo S, Quiroz R, et al. Electronic alerts to prevent venous thromboembolism among hospitalized patients. N Engl J Med 2005;352:969-977.
Eccles M, Grimshaw J, Campbell M, Ramsay C. Research designs for studies evaluating the effectiveness of change and improvement strategies. Qual Saf Health Care 2003;12:47-52.
Bates DW, Kuperman GJ, Wang S, et al. Ten commandments for effective clinical decision support: making the practice of evidence-based medicine a reality. J Am Med Inform Assoc 2003;10:523-530.
Handler JA, Feied CF, Coonan K, et al. Computerized physician order entry and online decision support. Acad Emerg Med 2004;11:1135-1141.
McDonald CJ. Protocol-based computer reminders, the quality of care and the non-perfectability of man. N Engl J Med 1976;295:1351-1355.(Pierre Durieux, M.D., M.P)