Chapter 10. Preventing Morbidity in the ICU

• Effective preventive health care interventions based on evidence from rigorous randomized trials are increasing.
• Nonetheless, preventive strategies are applied suboptimally in many settings, particularly in the intensive care unit (ICU).
• Potential reasons such strategies are not more broadly employed include clinician habit, lack of awareness of (or resistance to) new information, reliance on physiologic outcomes rather than on clinically important outcomes when interpreting evidence, and lack of self-efficacy of physicians who question whether the benefits observed in the research setting will be realized in the practice setting.
• Another important, but underappreciated, reason for insufficient or delayed uptake of effective preventive strategies is the absence of reinforcements for preventive behavior in general.
• A working environment that facilitates the implementation of preventive strategies is a powerful facilitator of change. A crucial first step in trying to improve preventive care in the ICU is to do an environmental scan to understand the practice of the unit and to characterize the culture of the unit.
• The most effective strategies to implement behavioral change are interactive education rather than passive education, audit and feedback, reminders (manual or computerized), and involvement of local opinion leaders.

Effective preventive health care interventions based on evidence from rigorous randomized trials are increasing. Application of such randomized trial evidence in practice has great potential to decrease the morbidity and mortality rates of inpatients and outpatients. However, preventive strategies are applied suboptimally in many settings, particularly in the intensive care unit (ICU). This poses a serious problem because critically ill patients are at high risk of death not only from the condition that necessitated their admission to the ICU but also from the complications of critical illness.

Examples of well-documented underuse of interventions that have been demonstrated to decrease morbid complications of critical illness in randomized trials include prevention of venous thromboembolism,1 prevention of hyperglycemia,2 and prevention of pneumonia.3 Other interventions proved to prevent ICU mortality in patients with acute lung injury and sepsis such as lung protective ventilation strategies, use of corticosteroids, and activated protein C are also variably applied in practice, although mortality-reducing interventions are not the focus of this chapter.

In this chapter, we (a) present key clinical and behavioral issues relevant to preventive care for critically ill medical and surgical patients, (b) illustrate the gap between the evidence and its practical application, (c) underscore lost opportunities for prevention of morbidity by using several study designs, and (d) suggest strategies for improved analysis of, and increased attention to, prevention in the ICU.

The reasons clinicians fail to attend to prophylactic strategies supported by randomized trials may mirror those reasons that clinicians fail to apply valid evidence in practice, and the reasons that clinicians do not follow high-quality practice guidelines. Assuming that the evidentiary basis for the preventive strategy is valid and thus worth considering, there are many potential reasons for underuse of primary, secondary, or tertiary prevention, including clinician habit, lack of awareness of (or resistance to) new information, reliance on physiologic outcomes rather than on clinically important outcomes when interpreting evidence,4 and lack of self-efficacy of physicians who question whether the benefits observed in the research setting can be realized in the practice setting.5

Another important, but underappreciated, reason for insufficient or delayed uptake of effective preventive strategies is the absence of reinforcements for preventive behavior in general. This is true particularly in comparison to treatment strategies. Clinicians do not experience direct and immediate feedback about the benefit of prophylactic interventions because they are designed to prevent the occurrence of future adverse events. In contrast, therapeutic interventions usually result in short-term, ongoing feedback about treatment responsiveness. In behavioral terms, the reinforcement for clinicians applying preventive strategies, as distinct from treatment strategies, is often covert and delayed, rather than overt and immediate.6

In addition, a factor recently recognized to contribute to inattention to preventive care, which is unrelated to cognitive or behavioral attributes of clinicians, is the lack of systems support to facilitate prevention. The ICU setting poses unique but not insurmountable barriers to preventive care, due to the multidisciplinary complex care of critically ill patients, and the need for fast-paced decision making. Multidisciplinary care has been postulated to result in improved outcomes for critically ill patients, but it can also lead to lack of clarity about responsibility for ministration of preventive care.7 We discuss these ICU environmental factors in more detail below with respect to prevention of venous thromboembolism and hyperglycemia.

A working environment that facilitates the implementation of preventive strategies is a powerful facilitator of change. A crucial first step in improving preventive care in the ICU is to do an environmental scan. The ICU environment can be scanned quantitatively and qualitatively to help understand which preventive strategies require attention.

A recent qualitative study examining environmental factors that increase β-blocker use to prevent mortality for patients with myocardial infarction illustrates this.8 Investigators conducted semi-structured interviews with eight cardiologists, four internists, two emergency physicians, 15 nurses, 11 quality-management staff, and five senior administrators in eight U.S. hospitals. Six factors were found to increase β-blocker use: setting goals, administrative support, clinician support, design and implementation of improvement initiatives, use of data, and contextual factors. From this study, we learn which environmental influences encourage the prescription of β blockers, and we may emulate successful strategies in our own environment for other preventive strategies. Investigations of this genre have been conducted to understand pneumonia-prevention strategies.7

Other approaches such as self-administered surveys can help us to understand the use of preventive strategies for care of acute lung injury and acute respiratory distress syndrome.9 Surveys can indicate the information sources in the environment that are considered to inform practice, the availability of specific preventive interventions, and perceived reasons for local practice patterns.

The second step toward optimal prevention of morbidity in the ICU is to take a scientific approach to evaluating whether effective prophylactic strategies are being applied in practice, and whether they are achieving the desired patient outcome. Casual interviews rarely provide valid data about preventive care. More credible data about compliance with effective preventive interventions can be obtained through formal surveys of stated practice and structured personal interviews.

However, surveys of stated practice have a universal limitation that must be kept in mind: they do not necessarily reflect actual practice. For example, in a self-administered survey, Canadian ICU clinicians stated that patients of older age with severe illness acuity and serious chronic comorbidity were most likely to undergo withdrawal of life support.10 Subsequently, however, an observational study of actual practice demonstrated that rather than age, illness severity, and organ dysfunction, the four strongest determinants of ventilator withdrawal were inotrope or vasopressor dependency, perceived patient preferences to limit life support, and physician predictions of ICU survival rate and future cognitive function.11 There are no substitutes for comprehensive audits, rigorous surveillance strategies, and observational studies to document practice patterns. To be valid, these strategies should be blinded so that clinicians remain unaware that their management is being observed, to avoid the inevitable practice change associated with this awareness (the Hawthorne effect). Although the foregoing example concerns end-of-life care, the design feature of blinding is very important for observational studies evaluating whether optimal prevention is being applied in the ICU.

Qualitative research methods such as focus groups, document analysis, and in-depth interviews, as we found when exploring pneumonia prevention strategies in the ICU, are even more powerful for their insights about when, where, why, how, and by whom preventive care is administered7 than whether it is administered.12

It is now well understood that knowing the research evidence about health care interventions that decrease rates of morbidity and mortality does not ensure that it is used in practice. The research evidence about effective behavior-change strategies can be extremely helpful to improve preventive care. The behavior-change strategies that are most likely to enhance the faithful application of diagnostic, preventive, and therapeutic interventions have been summarized in the Cochrane Collaboration Systematic Review by Bero and colleagues,13 which has been comprehensively updated by Grimshaw and associates.14 Although few studies in these reviews focused exclusively on preventive interventions, there are useful signals from this literature. For example, the Cochrane Review clearly shows that passive dissemination of information is generally ineffective; it seems necessary to use specific strategies to encourage implementation of research-based recommendations and to ensure changes in practice. The most effective strategies are interactive education rather than passive education, audit and feedback, reminders (manual or computerized), and involvement of local opinion leaders.

To ensure that preventive care interventions are applied in practice, a behavioral program could be instituted that is specifically adapted to the ICU setting. A portfolio of behavior-change strategies may be desirable, some of which could be common and some of which may be different, to promote each preventive intervention. Further research on the relative effectiveness of different strategies specifically for preventive care and specifically for the ICU setting is needed. Studies on the cost effectiveness of behavioral-change strategies to promote prevention would also be useful. In the following sections of this chapter, we illustrate some behavior-change strategies using two examples in the ICU setting: prevention of venous thromboembolism and prevention of hyperglycemia.

Prevention of Venous Thromboembolism in the ICU

Venous thromboembolism (VTE) is a common complication of serious illness, conferring considerable morbidity and mortality in hospitalized patients. Patients with deep venous thrombosis (DVT) are at risk of subsequently developing pulmonary embolism (PE), which may be fatal if untreated. In the ICU setting, patients with DVT are significantly more likely to have PE,15 and patients with DVT have a longer duration of mechanical ventilation (p = 0.02), ICU stay (p = 0.005), and hospitalization (p <0.001) than do patients without DVT.16 Critically ill patients can rarely communicate their symptoms, making it unlikely that patient self-reported symptoms will prompt intensivists to pursue the diagnosis of VTE; moreover, the physical examination is insensitive to detect DVT. Clinically unsuspected DVT and PE are frequently found at autopsy on critically ill patients.17 In summary, VTE is a good example of an ICU-acquired morbidity that can be silent but potentially deadly.

There are only two published randomized trials testing heparin for VTE prophylaxis in medical and surgical ICU patients.18,19 One double-blind single-center trial allocated 119 medical-surgical ICU patients at least 40 years of age to unfractionated heparin 5000 U twice daily or placebo subcutaneous injections.18 Using serial fibrinogen leg scanning for 5 days, the rates of DVT were 13% in the heparin group and 29% in the placebo group (relative risk = 0.45, p <0.05). Rates of bleeding and PE were not reported. This trial demonstrated that unfractionated heparin is better than no prevention (the number of patients needed to prophylax with 5000 U twice daily of subcutaneous heparin to prevent one DVT is four). A multicenter trial by Fraisse and colleagues randomized 223 patients with an acute exacerbation of chronic obstructive pulmonary disease requiring mechanical ventilation for at least 2 days to 3800 or 5700 IU of the low-molecular-weight heparin nadroparin once daily or placebo.19 Patients were screened with weekly duplex ultrasounds and after clinical suspicion of DVT, and venography was attempted in all patients. Rates of DVT were 16% in the nadroparin group and 28% in the placebo group (relative risk = 0.67, p <0.05). A similar number of patients bled in each group (25 vs. 18 patients, respectively; p = 0.18). Although patients were not screened for PE, no patients developed PE during the trial. This trial demonstrated that nadroparin is better than no prevention (the number of patients needed to prophylax with nadroparin to prevent one DVT is eight).

Although there are only two published randomized trials of thromboprophylaxis in medical-surgical ICU patients, randomized trials have been conducted in other populations for three decades and have clearly shown the effectiveness of anticoagulant VTE prophylaxis. Accordingly, anticoagulant thromboprophylaxis is universally recommended for critically ill patients except those with contraindications.20 However, compliance with this recommendation for thromboprophylaxis is not what it should be.

Compliance with Thromboprophylaxis

Several prospective single-center utilization reviews of VTE prophylaxis have provided evidence about the use of VTE prophylaxis in ICU practice. For example, thromboprophylaxis was prescribed in 33% of 152 medical ICU patients in one study21 and in 61% of 100 medical ICU patients in another.22 In contrast, in a medical-surgical ICU in which a clinical practice guideline was in place, VTE prophylaxis was prescribed for 86% of 209 patients.23 In medical-surgical ICU patients, after excluding those receiving therapeutic anticoagulation and for whom heparin was contraindicated, only 63% of 96 received unfractionated heparin.24

In a multicenter 1-day cross-sectional utilization review of surgical ICU patients, unfractionated heparin was the predominant choice, and two methods of VTE prophylaxis were prescribed in 23% of patients.25 Prophylaxis with unfractionated or low-molecular-weight heparin was significantly less likely for postoperative ICU patients requiring mechanical ventilation than for patients weaned from mechanical ventilation later in their ICU course (odds ratio = 0.36). Use of intermittent pneumatic compression devices was significantly associated with current hemorrhage (odds ratio = 13.5) and risk of future hemorrhage (odds ratio = 19.3).

In summary, the use of effective VTE prophylaxis ranges widely, according to the foregoing utilization reviews. One inference from this health services research is that insufficient attention is paid to thromboprophylaxis in the ICU setting. However, when prophylaxis is prescribed, clinicians appear to risk stratify, that is, patients with more VTE risk factors are more likely to receive prophylaxis than are patients with fewer risk factors. The dynamic competing risks of bleeding and thrombosis during critical illness underscore the individual risk:benefit ratios in critically ill patients.

Two studies evaluating implementation strategies to enhance appropriate VTE prophylaxis are relevant to the ICU setting. One three-armed multicenter randomized trial of 3158 surgical patients requiring VTE prophylaxis evaluated the impact of education versus education plus quality improvement.26 Both interventions significantly improved appropriate VTE prophylaxis rates compared to the control group. In a time series study of 1971 orthopedic patients, a computer-decision support system significantly increased the rate of appropriate VTE prophylaxis prescribing from 83% to 95%.27 Interestingly, each time that the computer-decision support was removed, practice patterns reverted to those observed previously. In a before-after single-center study, our group demonstrated how appropriate anticoagulant thromboprophylaxis increased from 65%24 to 95%28 after the introduction of a thromboprophylaxis practice guideline implemented by using interactive multidisciplinary educational in-services, ongoing verbal reminders to the ICU team, daily computerized nurse recordings of thromboprophylaxis, weekly graphic feedback to individual intensivists on prophylaxis adherence, and publically displayed graphic feedback on group performance. However, no randomized trials have formally tested individual behavior-change strategies for thromboprophylaxis in the ICU, in contrast with this multimethod approach. Fortunately, published thromboprophylaxis rates for critically ill patients appear to be increasing over time.1

Prevention of Hyperglycemia in the ICU

The adverse consequences of acute hyperglycemia have been highlighted in numerous observational studies. A meta-analysis of 15 observational studies showed that, after myocardial infarction, diabetic patients with glucose values of 6.1 to 8.0 mmol/L had a fourfold higher risk of death than did patients without diabetes who had lower glucose values.29 Among diabetic patients with glucose values of 10.0 to 11.0 mmol/L, the risk of death was increased twofold. After stroke, a meta-analysis of 32 observational studies among nondiabetic patients found that acute hyperglycemia with values of 6.1 to 8.0 mmol/L was associated with a threefold relative risk increase in hospital mortality and an increased risk of poor functional recovery in nondiabetic patients after stroke.30

To test the hypothesis that outcomes could be improved in patients with lower blood glucose during acute illness, the Insulin Glucose Followed by Subcutaneous Insulin Treatment in Diabetic Patients with Acute Myocardial Infarction (DIGAMI) study randomized 620 patients with diabetes and myocardial infarction to two different intervention strategies. Patients were allocated to intensive metabolic treatment with insulin plus glucose infusion followed by multidose insulin treatment or to standard treatment. Investigators found a significantly lower mortality rate in the intensive treatment group at 1 year31 and at 3 years.32 In diabetic patients, the adverse consequences of chronic hyperglycemia are well known. The Diabetes Control and Complications Trial (DCCT), which enrolled 1441 patients with type 1 diabetes, demonstrated that patients who had intensive insulin management had significantly less retinopathy, nephropathy, and neuropathy than did those managed conventionally.33 Similarly, the UK Prospective Diabetes Study (UKPDS) of 3867 patients with type 2 diabetes showed that intensive glucose control with oral hypoglycemic agents or insulin to achieve a target fasting glucose level below 6 mmol/L resulted in significantly fewer microvascular complications than did conventional management.34

In addition to the foregoing findings, emerging evidence suggests that acute hyperglycemia has adverse consequences for ICU patients. During critical illness, stress hyperglycemia occurs due to production of excessive counter-regulatory hormones (glucocorticoids, catecholamines, growth hormones, and glucagon), effects of cytokines, insulin resistance, and pre-existing diabetes. Building on the results of the DCCT33 and UKPDS34 in diabetic patients and those of the DIGAMI trial in patients with and without diabetes,31 a recent landmark randomized trial of 1548 critically ill patients demonstrated that patients allocated to a target of euglycemia (4.4 to 6.1 mmol/L) as compared with higher glucose values (10.0 to 11.1 mmol/L) had a significantly lower infectious morbidity rate, less renal failure, less polyneuropathy of the critically ill, and lower ICU and hospital mortality rates regardless of whether they had diabetes.35

Human behavior is partly determined by beliefs and attitudes. To determine the beliefs and attitudes held by ICU clinicians about prevention of hyperglycemia, we recently conducted a multicenter survey of ICU physicians and nurses.2 The goals were to understand (a) perceived thresholds for clinically important hyperglycemia and hypoglycemia in ICU patients, (b) glucose measurement concerns in the critical care setting, and (c) strategies that clinicians would find most useful to help achieve optimal glucose management in the ICU.

ICU clinicians reported their perception that the clinically important threshold for hyperglycemia was 10 mmol/L for diabetic and nondiabetic ICU patients. Avoidance of hyperglycemia was judged most important for diabetic patients (88%), patients with acute brain injury (85%), a recent seizure (74%), advanced liver disease (64%), and acute myocardial infarction (64%). Physicians expressed more concern than nurses about avoiding hyperglycemia in patients with acute myocardial infarction (p = 0.0004). We conclude that randomized trial evidence showing the benefit of euglycemia in preventing morbidity and mortality in critically ill patients has not decisively influenced practice in these centers.

In summary, a recent randomized trial in critically ill patients associated euglycemia with lower morbidity and mortality rates than with higher glucose values. This work follows a consistent body of evidence about the importance of euglycemia to achieve optimal patient outcomes in other acute settings. Nevertheless, clinicians in our multicenter survey2 did not consistently report application of this evidence in practice. Lack of clinician awareness of the ICU trial35 is unlikely; however, other plausible explanations include the short time since publication of this trial, uncertainty about the application of these results to noncardiac surgery patients or patients with short ICU stays, the challenge of achieving euglycemia in the ICU, concern about unrecognized hypoglycemia in critically ill patients, and a desire for more confirmatory evidence. Principles of behavior change mentioned earlier, and worth repeating, are that changing clinician behavior does not follow passive dissemination of information. The most effective strategies are interactive education, reminders, audit and feedback, and actively implemented, locally developed guidelines and protocols.14 Further, in the ICU, strategies to improve glycemic control may be more effective by using a collaborative and interdisciplinary approach, rather than relying on a physician-led initiative. Meanwhile, pending the completion of future randomized trials in heterogeneous ICU populations, a shift toward tighter glucose control requires being aware of ICU clinicians' beliefs and attitudes. Addressing these beliefs and attitudes could enhance the success of future clinical, educational, and research efforts to modify practitioner behavior and thereby improve the outcomes for critically ill patients.

A different way to reflect on the preventive care that we have used is an analysis of three domains: (a) the evidence supporting the preventive interventions, (b) attributes of the preventive interventions themselves, and (c) existing strategies to enhance their uptake. We illustrate this analytic approach by comparing and contrasting two preventive interventions for mechanically ventilated patients: semirecumbency for pneumonia prevention (vs. supine positioning),36 and stress ulcer prophylaxis with histamine-2 receptor antagonists (vs. other drugs or placebo).37

Figure 10-1 illustrates how the evidentiary basis for these two preventive interventions differs. The body of research in support of histamine-2 receptor antagonists for stress ulcer prophylaxis is much larger in terms of number of randomized trials, and the data are older and better known38 than the body of evidence supporting semirecumbency for pneumonia prevention.36,39–41Figure 10-2 illustrates how the interventions themselves differ. Whereas semirecumbency is a complex, continuous, behavioral intervention potentially influenced by many alternative body positions, stress ulcer prophylaxis is a discrete, twice- or thrice-daily, simple pharmacologic intervention. Figure 10-3 illustrates that existing uptake strategies supporting stress ulcer prophylaxis are multiple, including publications, educational documents, presentations, computer-decision supports, and practice guidelines; in addition, the most powerful uptake strategy is operant: the well-oiled machine of industry detailing. In contrast, uptake strategies for semirecumbency are more limited in scope and effectiveness.

Although we cannot prove a causal relation between this conceptual analysis and the high penetrance of stress ulcer prophylaxis42,43 as compared with a low penetrance of semirecumbency,7,12 such a relation may be deduced.

In this chapter, we have emphasized preventive strategies that are grounded in evidence from randomized controlled trials and acknowledged that not all clinical questions in the ICU are best answered using the randomized controlled-trial approach. As we have illustrated, several other study designs can provide high-quality data that inform clinicians about preventive interventions that benefit patients, families, and staff.

Decreasing preventable morbidity (and, of course, mortality) should be among our top priorities in the ICU. Achieving these goals requires a multifaceted approach. It begins with an understanding of why clinicians fail in practice to use prophylactic strategies that are supported by valid evidence. Next, a scientific environmental scan can help to identify which preventive strategies are underused and require attention. An array of effective behavior-change strategies is desirable in the ICU setting. The ICU may be ideally suited to approaches such as population-specific practice guidelines, academic detailing by ICU pharmacists, and computer-assisted reminders. We must be mindful that these behavior-change strategies may require adaptation to the structure, function, and culture of each ICU. By adopting, yet locally adapting, effective behavior-change strategies, we can take essential steps to minimize lost opportunities for effective prevention in the ICU setting. If we fail, the commitment to research made by patients and their families in prior prevention trials will not be honored and will not benefit patients currently under our care.