The most award winning
healthcare information source.
TRUSTED FOR FOUR DECADES.
Evidence Is Not Enough: Knowledge Translation in the ICU
By David J. Pierson, MD, Editor, Professor, Pulmonary and Critical Care Medicine, Harborview Medical Center, University of Washington, Seattle, is Editor for Critical Care Alert.
With increased emphasis on evidence-based medicine during the last dozen years has come the implicit assumption that the availability of more and better evidence on how disease should be diagnosed and managed will lead to better patient care. Unfortunately, however, this assumption is wrong. A large body of evidence shows that many (if not most) patients do not receive care according to best available evidence or current standards of care. The result is needless suffering, many lost lives, and enormous waste. This is true for health care as a whole, and also in critical care.
There are many examples of excess morbidity, mortality, and costs resulting from the failure of clinicians and institutions to adhere to evidence-based guidelines and other forms of accepted best practice. A familiar illustration is the use of lung-protective ventilation (LPV) in managing patients with acute lung injury (ALI) and the acute respiratory distress syndrome (ARDS). As everyone who manages critically ill patients must surely know, the original ARDS Network study1 and other clinical trials have demonstrated that limiting delivered tidal volume (VT) to about 6 mL/kg predicted body weight (PBW), and end-inspiratory static airway pressure to not more than 30 cm H2O, other aspects of management being the same, substantially improves survival in ALI-ARDS, with a number-needed-to-treat of 11 patients so managed to save one life.2 No special equipment or techniques are required to achieve this benefit, and LPV can readily be applied in any ICU, using any critical care ventilator. Given this information, the relative simplicity of implementing LPV, and the thousands of lives at stake, one might expect that this therapy would have been put into near-universal use. Yet numerous studies demonstrate that this is far from being the case.3,4
In a study of 398 patients with ALI-ARDS managed between 1994 and 2001, Weinert et al found that the VT used decreased only from a mean of 11.2 mL/kg PBW prior to publication of the ARDS Net study to 10.1 mL/kg PBW afterward.5 Only 0.9% of their patients received a VT of 6.2 mL/kg PBW or less.5 Young et al compared the VT used before and after publication of the ARDS Net study in 300 patients with ALI-ARDS at their institution.6 They found that mean VT fell from 12.3 to 10.6 mL/kg PBW in the later study interval, but that there was no change in inspiratory plateau pressure, and that only 16% of ALI-ARDS patients were ventilated with VTs of 8 mL/kg PBW or less following publication of the ARDS Net study.6
In another study in which VT was recorded three times daily in patients with ALI-ARDS managed during 2003, Wolthuis and colleagues found that 85% of the delivered volumes exceeded 8 mL/kg PBW, and that 39% of them were greater than 10 mL/kg PBW.7 More recently, Umoh et al reported that only 46% of patients with ALI-ARDS who were managed at three Baltimore teaching hospitals received VTs of 6.5 mL/kg PBW or less.8 These and other studies consistently show that clinicians have failed to adopt the VT and plateau pressure settings of LPV for many patients even in institutions that participated in the ARDS Net study showing that this approach was life-saving.4
Why Don't Clinicians Use Best Evidence?
The LPV story is only one of many documenting the failure of clinicians to adopt evidence-based best practices in critical care.3,4 Why is this so? In a widely cited paper, Cabana and associates reviewed the existing literature on barriers to the adoption of clinical practice guidelines.9 They classified the barriers into seven separate categories, falling generally under the areas of knowledge, attitudes, and behavior. The Cabana study dealt with health care in general, but Kahn subsequently discussed the barriers it identified within the context of critical care.3 Table 1, adapted from these two articles,3,9 lists these barriers and provides some ICU-relevant examples.
Following publication of the first studies showing that clinicians were not adopting LPV as widely as might have been hoped, Rubenfeld and colleagues carried out a study to find out why this was the case.10 They surveyed experienced ICU nurses and respiratory therapists (RTs) at all 10 of the ARDS Net VT study's participating institutions to determine the perceived barriers to LPV adoption.10 Identified barriers to the initiation of LPV included physician unwillingness to relinquish control of the ventilator to an LPV protocol, failure of the managing physician to recognize that the patient had ALI or ARDS, and difficulty calculating PBW for setting the appropriate VT. Barriers to continuing LPV once it was initiated included concerns over patient discomfort and tachypnea, the frequent occurrence of hypercapnia and acidosis, and the fact that arterial oxygenation tends not to be as good on LPV as when higher VTs and plateau pressures are used. Although studies of actual drug usage on LPV vs traditional ventilatory management have not shown an increased need for sedation with the former,4 concerns about this were cited as an additional barrier to LPV use, as was difficulty adhering to the ARDS Net LPV protocol.10
In general, studies have shown that the wider the practice variation among individual clinicians, the less the care delivered tends to conform to accepted practice standards. In addition, the ICU physician staffing model in use correlates with various measures of the quality of care, including practice variation: "closed" units (in which trained intensivists participate in the management of all patients) generally perform better. That physician staffing and other aspects of ICU operation vary widely in different geographic areas and among institutions in a given region was recently emphasized in a study by Dodek et al.11 These investigators surveyed all ICUs in each of the five health care regions of British Columbia, examining the physician staffing model (open vs closed units, and the presence of trained intensivists), staffing and coverage by RTs, and the use of practice guidelines for nutrition, prophylaxis against deep venous thrombosis, and weaning from mechanical ventilation. They documented wide geographical variations in all variables examined, as well as major variation by ICU size. In the different regions, 20%-71% of the ICUs had trained intensivists, whose presence also varied with ICU size, from 7% in the smallest hospitals to 100% in those with more than 10 beds. The largest ICUs were also more likely to have 24-hour RT coverage, but this varied dramatically: 60% of the units in one region had no RT coverage at all. Similarly, there was variable use of practice guidelines, with those for ventilator weaning being present in 20%-80% of the ICUs in the different regions. Although this study did not examine specific outcomes of ventilator management or other individual critical care therapies, one might assume that these also varied widely in the different ICUs in the province.
What Is Knowledge Translation?
Knowledge translation (KT) is the process of putting the results of research and other evidence into use in everyday practice.4,12 Also known by numerous other terms such as dissemination and diffusion, research uptake, research utilization, and knowledge-to-action, KT is a burgeoning field in health care. A PubMed search on Feb. 3, 2010, using the terms "knowledge translation" or "knowledge transfer" yielded 705 citations, 293 of them published in the last two years. However, adding "critical care" to the search terms reduced the total to only 16 citations in the last 10 years. Thus, the ICU has been the focus of relatively little work in KT to date.
One reason for the difference between what the evidence says should be done and what actually happens is the real-world distinction between efficacy and clinical effectiveness.4 Efficacy is what is demonstrated in clinical trials. It shows what can be achieved under controlled research conditions, in carefully selected patients who generally have no serious comorbidities and who are managed with a rigidly controlled protocol overseen by research staff. In contrast, clinical effectiveness is what is experienced in everyday practice: The patients are unselected, they typically have comorbidities and other complicating factors that would have excluded them from the clinical trials, and their care is not scrutinized by research nurses or others to assure adherence to the protocol. And, given that protocols need to be tailored to local institutions, patient populations, and practice traditions, certain aspects of management may vary from what was done in the study demonstrating efficacy.
Knowledge translation is a complex, global process that involves an entire health care system, not just individual clinicians. The Figure, adapted from the work of Dougherty and Conway,13 shows one approach for clarifying the process and better understanding what has to be done to implement KT effectively. It seeks to improve not only the care of individual patients, but also the health status of the whole population, as well as efficiency and cost-effectiveness for the system. The "3-Ts" in the "road map" model shown in the figure represent separate translational steps. The first T (T1) is clinical efficacy research, to determine what treatments are effective and how to apply them under controlled conditions. T2 comprises activities to move from efficacy to clinical effectiveness, including the development of evidence-based clinical practice guidelines for applying the knowledge gained in the first step to the care of individual patients. The third T (T3) deals with the "how" of health care delivery, at both individual and system-wide levels.
Implementing Knowledge Translation in Critical Care
Although the 3-Ts road map is helpful for gaining an overall understanding of what KT aims to do in health care, it does not translate intuitively to the ICU or to a practical action plan for those working in the clinical trenches. A more accessible scheme whose implementation is easier to visualize is that offered by Pronovost and colleagues of the Johns Hopkins Quality and Safety Research Group,14 which is summarized in Table 2.
That group's model for translating evidence into practice focuses on systems (how work is organized) rather than the care of individual patients.14 It stresses collaborative action by interdisciplinary teams, and tailoring of the structure of the effort as well as the details of the intervention to the characteristics and needs of the institution involved. Central to its implementation is the creation of a collaborative culture in the institution, something that may represent a substantial change. This model applies primarily to large scale projects, as the resources required to develop, implement, and evaluate the program it describes are substantial.
Once the program has been conceived and support has been obtained from the highest levels of the organization, three system-level steps are involved. First, the relevant research pertaining to the specific outcome to be sought is reviewed using a rigorous, formalized approach. Interventions associated with the improved outcome sought are identified, and several of those that are likely to yield the greatest benefit to the organization are selected. These interventions are then converted into specific behaviors.
Next, local barriers to implementation are identified. Pronovost et al emphasize the value of direct observation of the staff as the target interventions (for example, the insertion of central lines) are performed, to identify defects in each step in its implementation. All stakeholders involved in the intervention (for example, physicians, nurses, and supply personnel) are observed and queried about problems and potential improvements related to that intervention. Measurement of performance of the target interventions is key in this process.14 Both process measures (how often patients actually receive the therapy or other intervention as intended) and outcome measures (whether relevant results actually improve) are incorporated into the assessment.
Whereas the above measures are system-wide, actual implementation of the intervention must succeed at the ground level. The model described uses what Pronovost and colleagues call the "4-Es": engage, educate, execute, and evaluate. Although initially these are undertaken in sequence, the authors emphasize that their operation is cyclical and ongoing rather than a linear, one-time process. Engaging the staff involves providing actual data on baseline performance (for example, for central line infections, the number of infections and patient deaths in the unit attributable to those infections); the investigators also engage the staff by sharing real-life stories of individual patients to put a human face on the issue. Education involves summarizing the relevant literature and providing checklists of the evidence to members of the staff at all levels. Execution of the plan is based on a "tool kit" for surmounting the identified barriers to implementation, which standardizes care processes, provides checklists, and incorporates learning from mistakes. Evaluation includes comparing ongoing performance measures to baseline data from the unit. Two additional "Es", endure (incorporating the project into the hospital's quality improvement program) and extend (expanding the project into other areas of the institution, such as the emergency department), have subsequently been added to the model.
The model summarized in Table 2 and described above has been used in a large-scale research study of infections associated with the insertion of central lines.15 In that study, which took place in 103 ICUs, the median infection rate per 1000 catheter-days decreased from 2.7 (interquartile range, 0.6-4.8) to 0 (interquartile range, 0-2.4) in the 18 months following the intervention.15 These results, requiring implementation of the program of Pronovost and associates on a wide scale throughout the state of Michigan, are most impressive from both clinical and economic perspectives.
Knowledge translation is complex and demanding of both human and material resources. However, there can be little doubt of the need for it in critical care, or of its potential benefits for everyone from individual patients to society as a whole. An excellent series of articles dealing with different aspects of KT in health care has recently been published in the Canadian Medical Association Journal (CMAJ 2010;182:E68-E98), and has also appeared in book form.16 More studies and other resources pertaining directly to critical care are sure to appear in the coming months.