The Coming of Age of Implementation Science and Research in Critical Care Medicine

THE HISTORY OF IMPLEMENTATION SCIENCE IN CRITICAL CARE MEDICINE

The Institute of Medicine has long documented systemic deficiencies and significant gaps between the healthcare that critically ill patients should and actually receive. These gaps exist across all healthcare settings, racial and ethnic groups, age groups, and geographic areas (1). Some of this variation in care delivery can be attributed to a lack of robust evidence to guide clinical practice (2). But even where high-quality evidence does exist and is widely disseminated through clinical practice guidelines with robust bundles of interventions and implementation strategies (i.e., Surviving Sepsis Campaign), many critically ill patients still do not receive the recommended care, contributing to their morbidity, mortality, and higher costs of care (3–5). To be both evidence-based and clinically useful, clinical practice guidelines must balance the strengths and limitations of all relevant research evidence with the practical realities of delivering care in complex and diverse clinical settings like ICUs (6). Many interventions proven to be effective in healthcare research often fail to translate to improved patient care and outcomes (7). In complex, dynamic organizations like healthcare systems, up to 70% of all change efforts fail to achieve full implementation of desired interventions (8, 9). This article explores the history of implementation science (IS), an emerging field that seeks to understand whether and how change is adopted in healthcare settings (10). Further, this review emphasizes the various reasons why critical care research often fails to translate into clinical practice and how IS can help to overcome these deficiencies to improve the quality, safety and value of care delivered to critically ill patients.

Historically, IS is said to have stemmed from a landmark report written in 1943 by Ryan and Gross on the diffusion of hybrid seed corn in two Iowa communities. According to Dearing and Kee (11), this seminal report, “set the paradigm for many hundreds of future diffusion studies by emphasizing individuals as the locus of decision, adoption as the key dependent variable, a centralized innovation change agency that employed change agents, and the importance of different communication channels for different purposes at different times in the individual innovation-decision process.” Rural sociologists began focusing on studying new research concepts in alignment with academia, and soon scholars in other fields began to synthesize approaches drawn together by both intellectual questions and funding opportunities for research.

IS expanded into healthcare in the 1960s and 1970s as scholars began questioning traditional assumptions as to how scientific advancements could influence real-world clinical practice (12). It was widely assumed at that time that if research was more readily available to healthcare providers, that implementation of best practices would automatically happen as practitioners applied this evidence to their everyday practice (13, 14). While clinical practice guidelines were created to make healthcare research more accessible and understandable, their application was challenging, requiring clinicians to read, interpret, accept, and consistently apply recommendations. This was not always possible, given the time constraints of most clinicians and the systemic cultural and structural barriers to change (13). IS has been used effectively to help narrow the gap between the discovery of new clinical knowledge and its application in public health, mental health, and other healthcare settings.

IS in critical care medicine has continued to evolve and expand over the past 50 years, influenced by the book of Rogers (15) “Diffusion of Innovations” (described later in more detail), by the studies of Weiss and Bucuvalas (16) policy decision-making and the use of scientific inquiry to impact social change through policy action, and by the development of multicenter and international registries for conducting clinical trials and benchmarking clinical performance. The emergence of professional organizations such as the Society of Critical Care Medicine (SCCM) in 1970 (17) also helped to highlight the importance of setting professional standards and promoting excellence and consistency in the practice of critical care medicine. Traditional top-down approaches to implementation that rely heavily on high-level administrative oversight, clinical practice guideline committees, and conference speakers as subject matter experts (SMEs) are being supplanted with multifaceted, multidisciplinary approaches involving providers who are close to or directly at the point of care, often employing new implementation technologies (i.e., ICU telemedicine) (18).

Over the past 10–15 years, the business model of U.S. healthcare systems to maximize revenue has been steadily shifting away from reducing costs to increasing the “value” of healthcare (19). Laws in the United States, such as the Affordable Care Act, have helped to accelerate this change through the creation of incentive mechanisms, such as Accountable Care Organizations (ACOs) and the Hospital Value-Based Purchasing Program (20). ACOs align healthcare providers, often across hospital systems, to deliver coordinated, high-quality care to their Medicare patients (21). The overarching goal of ACOs is to effectively coordinate care to ensure that patients receive the right care in a timely fashion, while avoiding costly medical errors and redundancy of services. When an ACO succeeds in both delivering high-quality care and spending healthcare dollars more efficiently, the ACO shares in the savings, it achieves for the Medicare program. The Hospital Value-based Purchasing Program rewards acute care hospitals with incentive payments for the quality of care and the experience they provide to hospitalized patients (22). The Value-based Purchasing Program specifically incentivizes hospitals to: eliminate or reduce adverse events; improve adoption of evidence-based practices; improve the patient experience; and increase their transparency of care quality to consumers, clinicians, and other stakeholders. Increasing demand for improving the quality and value of healthcare delivery has also incentivized extramural funding organizations such as the National Institute for Health to invest more than 32 billion dollars annually to support the development of IS and implementation research efforts (23).

IMPLEMENTATION SCIENCE DEFINED

IS includes both implementation research (IR) and practice, with IR forming part of the continuum of translational research. Translational research integrates findings from basic science, clinical, and population health-based research, testing novel therapeutic strategies and accelerating these therapeutic interventions into clinical practice with the goal of improving patient outcomes (24–26). IR is a form of late-stage translational research, which yields generalizable knowledge regarding evidence-based interventions. These interventions help to disseminate and implement basic science and clinical research discoveries to improve population health (Fig. 1) (23, 27).

Figure 1.

Implementation research on the continuum of evidence-based practice.

IS is broader than implementation itself; implementation practice uses systematic methods to adopt best practices, whereas IS seeks to understand how and why the adoption of best practices either occurs or fails (18). IS is a branch of health services research that uses principles from organizational psychology, public health, operations management, epidemiology, and behavioral economics to evaluate implementation failures and to use this knowledge to develop and test strategies to overcome barriers and to close the evidence-practice gap (5). IS aims to strengthen the implementation of best practices and the de-implementation of ineffective practices and to ensure the uptake and sustainability of new and existing knowledge and experience. Implementation practice tends to use extant knowledge and tools to address implementation challenges, while IR is often focused on advancing the field of IS itself through the development of theory, measures, and innovative study designs.

CONCEPTUAL FRAMEWORKS FOR IMPLEMENTATION RESEARCH

Rogers, Greenhalgh, and Damschroder are among the many researchers who have proposed conceptual frameworks that provide a systematic way to develop, manage, and evaluate interventions to move new ideas into clinical practice (28). Conceptual frameworks provide a common vocabulary and classification scheme to help IRs and practitioners identify barriers to the adoption of evidence-based practices (29). Conceptual frameworks help to describe the implementation process, explain those factors that influence implementation outcomes, and evaluate the implementation itself (28). These frameworks are also critical for developing successful IR grant proposals (30). Although there is no single, reliable way to modify human behavior, IS and the application of conceptual frameworks have been used successfully to modify discrete aspects of medical practice (2). Understanding which implementation methods most effectively translate evidence-based therapies into clinical practice will help researchers design targeted implementation strategies for successful assimilation of best practices in critical care.

Greenhalgh et al (14) applied Rogers Diffusion of Innovation Theory to a systematic review of nearly 500 published sources across 13 fields of research, describing evidence-based strategies to successfully spread and sustain innovations in healthcare delivery and organization (31). They distinguished between strategies related to the “diffusion” (i.e., passive spread), “dissemination” (i.e., active and planned efforts to persuade target groups to adopt and innovation), “implementation” (i.e., active and planned efforts to mainstream an innovation within an organization), and “sustainability” (i.e., making an innovation routine until it reaches obsolescence) of innovations. They identified 11 key attributes of successful innovations that allow change to be more easily adopted, including “relative advantage, compatibility, complexity, trialability, observability, reinvention, fuzzy boundaries, risk, task issues, knowledge, and augmentation/support” (Table 1). The authors note that these innovation attributes are neither stable features nor sure determinants of their adoption or assimilation. Rather, it is the interaction between the innovation, the intended adopter(s), and the context that determines the innovation adoption rate. They recommend that IR questions should be framed to highlight processes rather than outcomes. For example, instead of asking, “Does program X work?”, one should frame the question, “What features account for the success of program X in this context, and the failure of a comparable program in a different context?”

TABLE 1.

Key Attributes of Successful Innovations (14, 31)

Attribute	Definition
Relative advantage	Innovations that have a clear, unambiguous advantage in either their clinical or cost effectiveness
Compatibility	Innovations that are compatible with the intended adopter’s values, norms, and perceived needs
Complexity	Innovations that are perceived by key players as simple to use
Trialability	Innovations that intended users can experiment with on a limited basis
Observability	The benefits of an innovation are visible to intended adopters
Reinvention	Innovations that potential adopters can adapt, refine, or otherwise modify to suite their own needs
Fuzzy boundaries	Innovations with a “hard core” and a “soft periphery” that can be adapted to the organization
Risk	Innovations with a low degree of uncertainty of outcome
Task issues	The innovation is relevant to the performance of the intended user’s work, and it improves task performance
Knowledge	The knowledge required to use the innovation can easily be codified and transferred from one context to another
Augmentation/support	The technology related to the innovation is supplied with customization, training, and help desk support

Damschroder’s Consolidated Framework for Implementation Research (CFIR) incorporated the work by Greenhalgh et al (14) and identified five major domains of variables that interact to influence the adoption of an innovation: 1) the characteristics of the innovation (i.e., strength of evidence, relative advantage, compatibility, etc.); 2) the inner setting (i.e., the structural and cultural characteristics of the organization); 3) the outer setting (i.e., regulatory policies and payment models); 4) the characteristics of the individuals involved (i.e., knowledge, beliefs, receptiveness to change); and 5) the implementation processes used (e.g., bottom-up vs top-down decision making) (Fig. 2) (7). The CFIR has been widely used to help identify facilitators and barriers to implementation of evidence-based practices (32, 33). While most attention within the CFIR is focused on how to more quickly adopt and spread innovations that will benefit patients, this framework can also be used to help understand how organizations can eliminate treatment, practices, and policies that no longer benefit patients, referred to as “de-implementation” or “exnovation (34).”

Figure 2.

Consolidated framework for implementation research (7).

DELAYS IN WIDESPREAD ADOPTION OF EVIDENCE-BASED RESEARCH

It is often stated that it takes an average of 17 years for evidence-based research to be incorporated into mainstream clinical practice (35–37). Extensive delays in adopting evidence-based best practices result in a significant waste of scarce resources and a loss of potential benefits to patients. In 2011, Morris et al (38) studied the time lags in translational research and described this lack of, “a timely realization of the benefits of expensive medical research” as an area of, “international concern.” They described 23 specific examples of significant delays in the translation of research into clinical practice. In the United States alone, these included significant delays in the adoption of treatment strategies for acute myocardial infarctions (i.e., a time lag of 6–13 yr), the top 10 clinical advances in cardiovascular and pulmonary medicine and surgery (i.e., a time lag 30.6 yr), and the management of AIDS (i.e., a time lag 2.6–3.8 yr).

There are several potential challenges in expediting the widespread adoption of evidence-based medicine (6, 13, 38). The initial path of biomedical research typically requires rigorous evaluation and funding to conduct high-quality research that yields strong evidence. The evidence then needs to be synthesized, ideally with a clinical practice guideline developed from the evidence. The evidence-based practice must then be applied to appropriate clinical settings in a way that balances those CFIR attributes of the conceptual framework, which support the innovation. These steps must be negotiated across various stakeholder groups to form a valid connection between evidence and practice. Clinician motivation to change what they are currently doing, the size and complexity of the research, difficulties in applying the evidence in practice, and organizational barriers may all contribute to delays in the translation of evidence into clinical practice. IS and IR can help to identify and overcome these barriers in the ICU setting, to accelerate the adoption of evidence-based medicine, and to improve ICU patient care and outcomes using a structured, team-based approach (Fig. 3).

Figure 3.

Conceptual model for implementation research.

Below, several critical care practice guidelines and recommendations are presented, with an explanation of how IS might be applied to enhance uptake and use. It should be noted, however, that the strength of the evidence underpinning these guidelines and recommendations is mixed. Some of the variation in uptake and use of these guidelines is almost certainly explained by three factors: variability in the quality of supporting evidence; the practice of bundling interventions that have been individually proven to be effective but have not yet been tested as a bundle of interventions; and a history of high-profile reversals in critical care evidence. These reasons for skepticism notwithstanding, it is worthwhile to identify and address challenges to implementation once a decision has been made to adopt specific clinical practices.

TRANSLATING RESEARCH INTO CRITICAL CARE PRACTICE

Table 2 summarizes the major critical care practice recommendations developed over the past 25 years that are relatively low-cost and supported by high-quality evidence demonstrating improved patient outcomes. Despite their benefits, most of these best practices have not been fully integrated into the routine practice of critical care medicine. We explore three of these interventions in more detail (i.e., lung protective strategies in patients receiving invasive mechanical ventilation, the central line-associated bloodstream infection bundle, and the ICU Liberation Bundle), to better understand the barriers to adopting these best practices in the context of IS and CFIR domains.

TABLE 2.

Variably Adopted Evidence-Based Critical Care Practices That Could Benefit From Implementation Science Methods^a

Intervention	Trial Result	Practice Recommendation
Gastrointestinal stress ulcer prophylaxis (39, 40)	Lower prevalence of gastrointestinal bleeding and pneumonia in patients receiving proton pump inhibitors or H2-receptor antagonists	Stress ulcer prophylaxis should be prescribed to critically ill patients at high risk for gastrointestinal bleeding (i.e., MV > 48 hr, coagulopathy, known gastrointestinal ulcer)
Head of bed elevation (41–43)	Reduced rates of VAP with head of bed elevation ≥ 30° in MV patients	In MV patients, elevate head of bed to ≥ 30° as long as it does not pose a risk/conflict with patient care
Restrictive hemoglobin transfusion threshold (≤ 7 g/dL) (44–46)	Restrictive (≤ 7 g/dL) vs liberal (≤ 9 g/dL) hemoglobin strategy associated with fewer blood transfusions, but no difference in mortality	In the absence of specific comorbidities or acute illness-related factors (i.e., active bleeding, ischemia), use a default transfusion threshold of hemoglobin ≤ 7 g/dL (target hemoglobin = 7–9 g/dL)
Low TV ventilation (47)	Reduced mortality in MV patients receiving low TV ventilation	In MV patients with ARDS, target TV = 4–6 mL/kg of PBW; in patients without ARDS, target TV = 6–8 mL/kg PBW
SSS (48–50)	Routine SSS decreases VAP rates, delays VAP onset, shortens duration of MV	Routine SSS is recommended in all MV patients and can lead to improved outcomes, lower VAP rates
PACs (51–53)	No difference in mortality or ICU or hospital LOS in critically ill patients with PAC vs no PAC	In most critically ill patients, PACs are nonbeneficial and potentially harmful, and should not be used regularly
Post-extubation NIV (54–57)	In patients at high risk of extubation failure, post-extubation NIV, vs standard medical therapy or high-flow oxygen via nasal cannula, is associated with lower rates of reintubation, mortality	In patients at high risk for extubation failure (i.e., with preexisting cardiopulmonary disease), preventive NIV should be applied immediately following extubation for ≥ 24 hr
Central venous catheter checklist (58)	Reduction in rate of CLABSIs	Use five-item CLABSI checklist and nursing supervision when inserting central venous catheters
Daily paired SAT + SBT (59–61)	Shorter duration of MV, coma, ICU LOS in patients undergoing paired daily SAT + SBT	Perform “wake up and breathe” protocol daily on MV patients who pass SAT/SBT safety screens
Early rehabilitation and mobilization (62–64)	Early rehabilitation/mobilization (vs usual care) is associated with shorter LOS, duration of delirium, more ventilator-free days, better functional outcomes	Early rehabilitation or mobilization in critically ill patients is feasible, safe and beneficial, and should be performed daily in patients
BG control (65)	Reduced 90-d mortality with conventional (< 180 mg/dL) vs intensive (81–108 mg/dL) BG control	Maintain BG < 180 mg/dL in critically ill patients
PN (66)	Shorter ICU/hospital LOS, duration of MV and renal replacement therapy, and fewer new infections with late vs early PN	In critically ill patients receiving insufficient enteral nutrition, providing supplementary PN early (≤ ICU day 7) vs late (> ICU day 7) is associated with worse outcomes and should be avoided
Gastric residual volume monitoring (67)	No difference in VAP rates, other ICU outcomes with none vs every 6 hr gastric residual volume monitoring	Do not routinely check gastric residual volumes in critically ill patients receiving enteral nutrition
Prone positioning in ARDS (68)	Lower mortality, ventilator-free days in MV patients with severe ARDS undergoing early and prolonged prone vs standard supine positioning	Experienced centers should prone MV patients with severe ARDS early (< 36 hr after intubation) and for ≥ 16 hr each day
Balanced crystalloids for IV fluid replacement (69)	Lower 30 d rate of major adverse renal events (i.e., mortality, new renal replacement therapy, persistent renal dysfunction) in patients receiving balanced crystalloid (i.e., plasmalyte-A or lactated Ringer’s solution) vs normal saline	Compared with balanced crystalloids, routine use of normal saline in critically ill patients may be unsafe and should be avoided
Higher PEEP in ARDS (70)	Improved mortality with high vs low PEEP in MV ARDS patients with a positive oxygenation response to higher PEEP	Higher PEEP should be delivered to MV ARDS patients with a positive response to higher PEEP (ΔPao2/Fio2 > 0))
Hour-1 sepsis bundle (71)	Lower mortality and cost in sepsis patients receiving an aggressive resuscitation bundle within 1 hr of identification of severe sepsis or septic shock	Within 1 hr of sepsis identification: 1) measure lactate level; 2) obtain blood cultures; 3) give broad spectrum antibiotics; 4) give 30 mL/kg crystalloid for MAP < 65 or lactate ≥ 4 mmol/L; and 5) give vasopressors for persistent hypotension to keep MAP ≥ 65 mm Hg
ICU liberation (ABCDEF) bundle (72)	Completion of ABCDEF bundle elements is associated with reduced mortality, duration of MV, coma, delirium, and physical restraint use, and lower-skilled nursing facility discharge rates	Complete all elements of the ABCDEF bundle in every critically ill patient every day
Pneumatic compression for DVT prophylaxis (73)	No difference in incidence/prevalence of DVTs in the pneumatic compression group vs control group	Do not use sequential mechanical compression devices in ICU patients already receiving prophylactic anticoagulation

ABCDEF = Assess, prevent, and manage pain; Both spontaneous awakening and breathing trials; Choice of Sedation Strategies; Delirium assess, prevent, and manage; Early Mobility and Exercise; Family engagement and empowerment, ARDS = acute respiratory distress syndrome, BG = blood glucose, CLABSI = central line-associated bloodstream infection, DVT = deep vein thrombosis, LOS = length of stay, MV = mechanically ventilated, NIV = noninvasive ventilation, PAC = pulmonary artery catheter, PBW = predicted body weight, PEEP = positive end-expiratory pressure, PN = parenteral nutrition, SAT = spontaneous awakening trial, SBT = spontaneous breathing trial, SSS = subglottic secretion suctioning, TV = tidal volume, VAP = ventilator-associated pneumonia.

^aRecommendations chosen are those that are low-cost and supported by multiple and/or single large randomized trials but variably adopted and sustained in real-world ICU settings.

Lung Protective Strategies in Mechanically Ventilated Patients

Acute respiratory distress syndrome (ARDS) is a severe form of hypoxic respiratory failure associated with various etiologies, most commonly sepsis, pneumonia, aspiration, trauma, and multiple blood transfusions (74–78). ARDS results from acute pulmonary inflammation, diffuse alveolar damage with increased capillary permeability, and the development of nonhydrostatic pulmonary edema, with patients often requiring intubation and invasive mechanical ventilation (75, 76, 79, 80). Worldwide, the estimated prevalence of ARDS ranges from 7.2 to 34 cases per 100,000 person-years and accounts for 10% of all ICU admissions and 23% of mechanically ventilated ICU patients (81–84). Historically the ARDS case fatality rate was around 60%, but over the past 20 years, mortality from ARDS has decreased by more than 50% (85–87). Improved survival from ARDS is attributable to general improvements in critical care management, along with specific improvements in ARDS management, particularly the use of “lung protective strategies” in invasive mechanical ventilation to prevent barotrauma in patients. Lung protective ventilation (LPV) is the use of low tidal volumes (i.e., 4–8 mL/kg based on ideal body weight), while maintaining a plateau pressure of less than 30 cm H₂O using lower inspiratory pressures during invasive mechanical ventilation (88, 89).

Several systematic reviews and meta-analyses confirm that patients diagnosed with ARDS who receive LPV experience lower morbidity and mortality, fewer ventilator days, and shorter ICU and hospital lengths of stay (90–95). LPV in ARDS is widely endorsed internationally by pulmonary and critical care societies (96), yet LPV remains very underutilized, delivered to only 1/3 of patients with ARDS receiving invasive mechanical ventilation (84). Several barriers to LPV delivery in ARDS patients have been proposed, including: 1) guideline factors (e.g., which patients should receive LPV, when should LPV be initiated in patients, variable performance metrics) (i.e., “CFIR intervention characteristics domain”); 2) ICU factors (e.g., a lack of formal LPV protocols, a lack of role clarity for ventilator management, poor team communication) (i.e., “CFIR inner setting domain”); and 3) clinician factors (e.g., failure to diagnose ARDS in patients, knowledge deficits, negative attitudes toward LPV and evidence-based practices in general, perceived contraindications, safety concerns) (i.e., “CFIR individual characteristics domain”) (97–102).

Multiple strategies have been shown to significantly improve LPV compliance and adherence in patients with ARDS. These include: 1) having a written LPV protocol; 2) provider education; 3) daily multidisciplinary rounds involving checklists and goal setting; 4) establishing low tidal volumes as the default ventilator setting for all mechanically ventilated patients; 5) ARDS Clinical Decision Support for providers included in the Electronic Health Record (EHR); and 6) provider audit and feedback (103). With the exception of daily team-based ICU rounding using checklists and goals sheets, none of these interventions when implemented alone have achieved compliance with LPV in the “majority” of patients with ARDS (104).

Central Line-Associated Blood Stream Infections

Central line-associated bloodstream infections (CLABSIs) are common yet preventable healthcare-associated complications, with the majority occurring in critically ill patients. CLABSIs comprise nearly three-quarters of all nosocomial bloodstream infections, with 80,000 CLABSIs occurring in ICU patients annually in the U.S. CLABSIs are associated with significantly longer ICU lengths of stay and lead to 30,000 deaths annually in these patients and up to $2.3 billion in additional healthcare costs (105).

Evidence-based interventions to prevent CLABSIs have resulted in significant reductions in ICU CLABSI rates over the past decade. In 2006, Pronovost et al (58) demonstrated that a CLABSI prevention checklist, implemented as part of a statewide ICU quality improvement (QI) project in Michigan, significantly reduced statewide CLABSI rates in ICU patients. The checklist included five evidence-based interventions recommended by the Centers for Disease Control and Prevention to help prevent CLABSIs including: 1) provider hand washing prior to the procedure; 2) full barrier precautions during catheter insertion; 3) skin preparation with chlorhexidine; 4) judicious selection of a catheter insertion site (i.e., avoiding the femoral vein); and 5) removal of unnecessary catheters in a timely fashion (106). Bundle compliance was optimized with routine checklist use, central line placement supply carts, daily discussions regarding catheter removal, frequent performance reports, and direct observation of catheter placement by bedside staff who were empowered to stop providers if checklist items were not followed. Across the 103 ICUs participating in the intervention, median CLABSI rates declined from 2.7 per 1,000 catheter days to zero CLABSIs at 3 months.

Following this groundbreaking study, the Infectious Disease Society of America published CLABSI prevention guidelines recommending that this CLABSI prevention checklist be used in all acute care hospitals (107, 108). Importantly, these guidelines also reinforced core implementation principles, including engagement of multidisciplinary teams to foster a culture of patient safety, innovative educational programs aimed at changing provider behavior, use of established intervention methods, and ongoing process evaluation and improvement (i.e., “CFIR domains of inner setting, individual characteristics, and implementation processes”) (109). Further motivating practice change, in 2008, the Centers for Medicare and Medicaid added vascular catheter-associated infections to the list of hospital-acquired complications that were no longer eligible for reimbursement (i.e., “CFIR outer setting domain”) (110).

Widespread implementation of CLABSI prevention bundles using effective implementation strategies has significantly reduced the overall prevalence of CLABSIs in the United States as well as in other countries. A recent meta-analysis of 79 studies evaluating the impacts of implementing CLABSI prevention bundles in adult, pediatric, and neonatal ICUs demonstrated a 60% reduction in CLABSI rates (i.e., from 6.4 [interquartile range (IQR), 3.8–10.9] to 2.5 [IQR, 1.4–4.8] CLABSIs per 1,000 catheter days and incidence rate ratio [IRR], 0.44; 95% CI, 0.39–0.50; p < 0.0001; I² = 89%) (111). Bundle elements having the greatest impact on CLABSI risk reduction were: hand hygiene prior to catheter insertion (p = 0.003) and maintenance (p = 0.022); use of a central venous catheter line cart containing all necessary supplies (p < 0.0001); avoiding the femoral vein as an insertion site (p = 0.03); and minimizing central line access (p = 0.019). Implementation strategies included: staff education, academic detailing, use of local champions or opinion leaders, checklist use, protocols, audit and feedback, reminders, leadership engagement, and organizational restructuring. Staff education, audit and feedback, organizational change, checklist use, and protocols were used in over 50% of these studies. Implementation strategies with the greatest impact on CLABSI risk reduction included the strong support and leadership of opinion leaders (p = 0.041), combined with strict protocol or checklist compliance, and when nurses were empowered to stop the procedure if a physician breached the central line insertion protocol (112).

It is important to note that these significant reductions in CLABSI rates occurred despite suboptimal compliance by providers with bundle protocols. In a separate large cross-sectional study comparing CLABSI rates with protocol compliance in nearly 1,000 U.S. ICUs that are part of the National Healthcare Safety Network, there was a strong dose-response effect seen between bundle compliance and CLABSI rate reductions (113). Bundle elements included: 1) hand hygiene prior to insertion; 2) maximal barrier precautions; 3) chlorhexidine skin prep; 4) avoidance of femoral vein as an insertion site; and 5) daily review of central line necessity. Although 98% of ICUs had CLABSI bundle policies in place, only 69% reported “excellent” compliance (i.e., 95% bundle compliance) with at least one bundle element, while 31% of ICUs failed to achieve 95% compliance with at any of the bundle elements. Only 20% of ICUs reported excellent compliance with all bundle elements, while 49% percent reported compliance with all five bundle elements at least 75% of the time. Simply having a written policy for a CLABSI prevention bundle was “not” associated with lower CLABSI rates. But the higher the bundle element compliance rates, the greater the observed reductions in CLABSI rates. Excellent compliance with all five bundle elements was associated with a 33% lower frequency of CLABSIs compared with excellent compliance with no bundle elements (IRR, 0.67; 95% CI, 0.59–0.77; p < 0.001).

Together, these studies suggest that greater widespread adoption of the CLABSI bundle will further decrease CLABSI rates in ICU patients, while also improving outcomes and reducing healthcare costs. But it remains to be seen whether healthcare providers achieving perfect compliance with the CLABSI bundle can eventually make CLABSIs a “never event (114, 115).” The use of IS and IR to improve adherence to the CLABSI bundle, along with the use of newer technologies to improve the safety of catheter placement and maintenance (e.g., ultrasound-guided placement, use of antibiotic-impregnated catheters, chlorhexidineimpregnated dressings, needleless securement devices and disinfecting caps) may further enhance CLABSI prevention efforts (116).

The ICU Liberation Bundle

Pain, agitation, and delirium (PAD) occur frequently in critically ill patients, and managing these symptoms can be difficult, leading to poorer outcomes and higher costs of care for patients (117). The ICU Liberation Bundle (aka, the Assess, prevent, and manage pain; Both spontaneous awakening and breathing trials; Choice of Sedation Strategies; Delirium assess, prevent, and manage; Early Mobility and Exercise; Family engagement and empowerment [ABCDEF] Bundle) (Fig. 4) was developed to help implement the SCCM’s Pain, Agitation, Delirium, Immobility, and Sleep Clinical Practice Guidelines (118). This Bundle focuses on optimizing pain management, avoiding deep sedation, reducing the risk of delirium, shortening the duration of mechanical ventilation, reducing the risk of ICU-acquired weakness and sleep disruption in patients, and engaging ICU patients and families in care processes (72,119–121). Bundling pain, sedation, and delirium management together with ventilator weaning and early mobility efforts helps to standardize ICU care processes, reduce inter-provider practice variation, improve ICU team communication, and ensure that all ICU patients receive the Bundle every day. Compared with delivery of these interventions independently, bundling these practices together can translate to greater synergistic improvements in both short- and long-term patient outcomes and reductions in healthcare costs (72,119–122).

Figure 4.

The ICU liberation (Assess, prevent, and manage pain; Both spontaneous awakening and breathing trials; Choice of Sedation Strategies; Delirium assess, prevent, and manage; Early Mobility and Exercise; Family engagement and empowerment) bundle.

The ICU Liberation Bundle is supported by strong, high-quality evidence (i.e., “CFIR intervention characteristics domain”) (118, 123). But translating this evidence into clinical practice can be challenging. Even in ICUs participating in large-scale structured QI efforts designed to promote Bundle adoption, implementation and performance of the Bundle elements vary significantly, and overall Bundle adoption across ICUs remains low (72, 124–126). Many ICU physicians and nurses are reluctant to maintain patients at a light level of sedation, to use respiratory therapist nurse driven ventilator weaning protocols, to mobilize ventilated ICU patients out of bed, or to engage patients and families in decision-making and care processes (i.e., “CFIR individual characteristics domain”) (125, 127–133). Additional barriers to Bundle adoption include ICU staff knowledge deficits and safety concerns (i.e., “CFIR individual characteristics domain”), staffing shortages, a lack of leadership support for Bundle implementation and sustainability (i.e., “CFIR inner setting domain”), and a lack of institutional commitment to patient safety and QI efforts (i.e., “CFIR implementation processes domain”) (124, 125, 130, 131, 134–136). Many ICU providers also lack effective teamwork and collaboration skills to effectively execute the Bundle (i.e., “CFIR inner setting and individual characteristics domains”) (128, 134, 137). As a result, ICU patients are often deeply sedated and mechanically ventilated for longer than necessary, placing them at greater risk of developing ICU delirium, muscle weakness, hospital-acquired infections, long-term physical and cognitive dysfunction (i.e., Post-Intensive Care Syndrome), and other preventable complications.

The ICU Liberation Bundle has the potential to significantly transform patient care and outcomes. ICU patients who receive the Bundle are typically more awake, alert, and more often pain-free (72). As a result, patients can interact more readily with their families and clinicians and are more able to actively participate in their care (e.g., ventilator weaning, mobility efforts) and decision-making. These patients recover faster, are ready to be transferred out of the ICU and discharged sooner, and are more likely to be discharged to home after their ICU stay. But successful execution of the Bundle requires: effective team communication, collaboration, and care coordination among providers (138–141); partnering with ICU patients and families to prioritize patient care goals (123, 142, 143); the use of real-time data to measure Bundle performance and outcomes; and strong leadership support for Bundle implementation and sustainability efforts (144–146). This often requires a transformational change in the way clinicians deliver care to patients. The use of evidence-based implementation strategies can significantly improve Bundle compliance and performance by improving teamwork and communication around the Bundle. We will use the SCCM’s ICU Liberation Collaborative as one of two case studies demonstrating how IS and IR can be used on a large scale to help translate evidence into clinical practice.

CRITICAL CARE IS/IR CASE STUDIES

The ICU Liberation Collaborative

The ICU Liberation Collaborative was a 2-year, multicenter Bundle QI initiative including over 15,000 mechanically ventilated and nonmechanically ventilated adult ICU patients admitted to 69 community, government or academic hospitals across the United States. The Collaborative was sponsored by the SCCM with support from the Gordon and Betty Moore Foundation (147, 148). The purpose of the Collaborative was to equip ICUs with the skills and knowledge necessary to implement the ICU Liberation (ABCDEF) Bundle and to improve Bundle-related teamwork and collaboration (144).

The Collaborative was led by SMEs with expertise in the PAD Guidelines, implementation and dissemination research, and large-scale QI efforts. CFIR constructs and domains were used to develop evidence-based implementation strategies, Bundle process and outcome metrics, and to identify potential barriers and facilitators of Bundle implementation. Collaborative sites were recruited through social media and at national critical care meetings from three geographic regions in the United States (i.e., West Coast, Midwest, and East Coast). Each site had an interprofessional implementation team including physicians, nurses, respiratory therapists, physical or occupational therapists, and pharmacists. Teams attended four in-person meetings with faculty and participated in monthly co-learning calls and training webinars, database training sessions, an e-Community listserv, and SME site visits. Curriculum included the evidence behind each bundle element, as well as team-based implementation strategies for improving Bundle compliance and sustainability. Teams were also encouraged to share best practices with one another. Team members collected and entered all patient-level data in the Collaborative’s Research Electronic Data Capture (REDCap) database and completed pre- and post-implementation questionnaires that assessed teamwork and collaboration, work environment, and overall ICU care (149). All sites provided a letter of commitment from their ICU Medical Director and a senior Hospital Administrator verifying that their hospital was willing to provide the necessary time and resources for teams to be successful.

Results from the ICU Liberation Collaborative demonstrated that as Bundle compliance across sites increased, the use of mechanical ventilation, and the prevalence of coma, delirium, and restraint use in patients significantly decreased, while ICU and hospital lengths of stay, ICU readmission rates, hospital mortality, and the proportion of ICU survivors discharged to a skilled nursing facility also significantly decreased (72). Importantly, there was a consistent dose-response relationship observed between higher proportional bundle performance resulting in even greater improvements in these outcomes. Hsieh et al (121) have also demonstrated that even partial Bundle implementation is associated with significant reductions in ICU and hospital costs.

Common barriers to Bundle Implementation during the Collaborative included: 1) challenges with developing and operationalizing clinical protocols for pain, sedation, and delirium management; 2) nurse, respiratory therapist, and physician coordination of spontaneous awakening and breathing trials; 3) application of delirium assessment tools in neurologically impaired patients; 4) safely getting critically ill patients out of bed; 5) a lack of physical therapy staff; 6) challenges with directly engaging patients and families; 7) insufficient support by hospital administration; and 8) a lack of information technology resources to acquire and analyze Bundle metrics within the EHR (131, 150). These barriers highlight several attributes of the Bundle as an innovation (Table 1) (i.e., Bundle complexity, relative advantage, trialability, reinvention, observability of benefits, and relative risk; adopters’ knowledge of the Bundle and its compatibility with their beliefs, and task issues; and augmentation and support), which map directly onto the five CFIR domains (Fig. 2).

The use of evidence-based implementation strategies during the Collaborative that facilitated Bundle compliance and performance included: 1) innovations to incorporate the Bundle into the daily ICU workflow (i.e., daily interprofessional team rounds at the bedside with facilitated team discussions of the Bundle and care plan [144, 151]; use of Bundle checklists and goal sheets [130]; EHR documentation of Bundle elements [145, 152]); 2) engagement of patients and families in team discussions about the Bundle and goals of care (123, 142, 153); 3) measurement of Bundle compliance and performance that was patient, provider, and unit specific through analysis of aggregated EHR Bundle data (154); and 4) strong leadership engagement and support of Bundle implementation efforts (124, 130).

Kaiser Permanente Northern California

Critically ill patients are at substantial risk of experiencing both short- and long-term morbidity and mortality (155–159). For over 10 years, Kaiser Permanente of Northern California (KPNC) has invested in IS to mitigate adverse outcomes and restore patient health following intensive care and hospitalization (160). As a highly integrated healthcare delivery system serving 4.4 million members across 21 hospitals, the use of IS has provided KPNC with scalable, reliable, and sustainable approaches to deliver high-quality care across the healthcare system (161–165). These approaches have been applied across a diverse range of inpatient programs, each offering an incremental contribution to improving ICU outcomes in alignment with patients’ goals of care. It’s important to note that unlike most other healthcare systems in the United States, physicians in the Kaiser Permanente (KP) system are full-time employees, whose compliance with defined bundles and guidelines are monitored through the EHR, and whose performance is judged at least in part by that compliance, which may provide significant financial incentives to KP physicians.

In 2008, KPNC leaders conducted a mortality diagnostic survey to identify gaps in care contributing to adverse outcomes across the health system’s 21 hospitals (166). This systematic approach revealed sepsis as the single largest contributor to inpatient death, while also highlighting opportunities to improve care (167). Numerous barriers were identified that impeded the reliable delivery of timely care to septic patients at the provider, facility, and health system levels (160). The resulting KPNC Sepsis Program used key IS tenets to devise innovative solutions to overcome barriers, resulting in substantial improvements in sepsis-related process and outcome metrics (166). This approach also enabled KPNC to identify new sepsis-related opportunities, particularly in populations for whom little prior data existed (e.g., septic patients with intermediate lactate levels) (168–174). These advances highlighted the role of IS in promoting a continuously learning healthcare system to systematically improve care for all patients with sepsis across the healthcare system (165).

KPNC’s use of IS to systematically improve sepsis care fostered new programs designed to bring evidence-based care to critically ill patients. Using a similar approach, KPNC focused on Clostridium difficile and hospital-acquired infection prevention (160, 175, 176), mechanical ventilation liberation (177), delirium prevention (178, 179), patient mobility (177, 180), enhanced recovery after surgery (180–182), regional tele-stroke management (183), conservative blood transfusion management (184–186), opiate exposure reduction (182, 187), palliative care management (188), readmission prevention (189, 190), and prevention of inpatient deterioration through early warning systems (191–193). Each of these efforts leveraged: 1) QI initiatives and iterative improvement cycles; 2) granular and high-quality data and metrics available through the EHR; 3) change management approaches bringing together diverse content experts, clinicians, and stakeholders; 4) consistent aims and messaging via a governance structure aligning both regional and local leaders; and 5) the delivery of patient-centered care. KPNC also continues to focus on preventive care by identifying potent opportunities to reduce hospitalization rates and ICU admissions in order to mitigate the unintended and deleterious consequences of inpatient care. For more than a decade, KPNC has successfully used IS and IR to develop reliable and sustainable approaches to improve the value, quality, and safety of ICU care across their healthcare system.

FACILITATING THE UPTAKE OF EVIDENCE-BASED PRACTICE IN CRITICAL CARE

At its core, IS is the science of behavioral change in healthcare delivery. For this reason, behavior change theories and principles are often invoked when designing implementation strategies to facilitate the uptake of evidence-based practices. Behavior change theories relevant to IS and practice include those focused on change at the “individual” (e.g., the Theory of Planned Behavior [194] and the Theoretical Domains Framework [195]), “team” (e.g., the Team Performance Framework [196]), and “institutional” (e.g., Organizational Theory of Innovation Implementation [197]) levels. The socio-ecologic framework (198) illustrates behavior change strata that can be used to categorize the many theories, frameworks, and models that have been applied to IS. More comprehensive overviews of implementation frameworks are provided by Tabak et al (199) and Nilsen (28).

The behavior change sought by IRs and practitioners is created through implementation strategies, which are the actions through which behavior change is achieved. Proctor et al (200) said that implementation strategies, “comprise deliberate and purposeful efforts to improve the uptake and sustainability of treatment interventions.” More than 70 discrete implementation strategies were identified in a recent initiative called the Expert Recommendations for Implementing Change (ERIC) (201), and include recruiting, designating and training leaders for the change effort, training and educating stakeholders, staging implementation scale-up with small pilots or demonstration projects, tailoring implementation strategies to address potential barriers, and using a train-the-trainer strategy (201, 202). Indeed, no implementation effort would be expected to use all or even most implementation strategies identified by the ERIC initiative. Rather, the ERIC strategies represent a “menu” of strategies that “could” be applied to a given implementation effort (201).

How should implementation strategies be selected? The field of IS has since moved away from the arbitrary selection of strategies based on what researcher Martin Eccles calls the ISLAGIATT principle (“it seemed like a good idea at the time”) (203). Increasingly, researchers use theory and contextual inquiry to select strategies in a process known as “implementation mapping (204).” With this method, determinants of implementation (i.e., barriers and facilitators) are “mapped” onto specific strategies to address implementation barriers. For example, if a needs assessment identifies a knowledge deficit as an important factor in the underuse of an evidence-based practice, education and training would be reasonable strategies in this instance. The opposite is also true; if knowledge is not a problem, education and training are unlikely to facilitate implementation. In addition to mapping, Powell et al (205) offer three other approaches for selecting and tailoring implementation strategies: concept mapping, group model building, and conjoint analysis.

Appropriate implementation strategies will differ depending on the setting and focus of a given effort. Although implementation barriers may transcend all five CFIR domains, most fall within the individual and inner settings domains. These domains include logistical barriers to implementing evidence-based practice at the institutional and clinician levels (2). While institutional support for implementation is essential, at the clinician level, time, access to resources, and their eagerness to apply research in clinical practice pose significant challenges (2). The theme of early and continuous stakeholder engagement has also been identified as especially important to the success of IR in targeting clinical improvement initiatives (10).

Specific strategies to be considered in ICU implementation efforts include those that promote team-based and patient-centered care. For example, targeted team training can help foster leader inclusiveness and psychological safety among team members, which facilitates information sharing, problem solving, and decision making within the team (206–208). As promotion of team-based behaviors in the ICU tends to require more complex and bundled approaches, protocols and educational materials should emphasize role clarity of team members and the understanding and appreciation of each individual’s knowledge, skills, abilities, and responsibilities, an approach associated with enhanced care coordination, collaboration, and team performance (138). Additionally, education, training, and audit and feedback strategies should be used to promote effective team communication and performance, including interdisciplinary bedside rounds, checklists, and EHR use (134, 151). These strategies help teams to establish and reinforce shared goals, engage in open exchanges of information, and to solve problems collectively. As described above, the use of theoretical models such as the team performance framework may help identify and address the complexities of ICU team dynamics that can influence implementation efforts (196). In critical care, changing ICU culture is often cited as a key component to implementing best practices. Challenges in recruiting ICU staff to become team-based champions and clinician resistance, including perceptions that change will increase their workload and challenge their practice autonomy, are key factors to address when planning and implementing new clinical initiatives (209–211). Ensuring that ICU staff and leadership have a clear understanding of the project’s goals is also essential.

A specific example shows how to link implementation determinants to effective strategies. A recent national collaborative initiative involving 63 ICUs implemented a patient- and family-centered care initiative over a 10-month period. Major barriers included a lack of buy-in, an inability to promote change in the clinical setting, an implementation-related increase in workload, and insufficient funding to support these initiatives (212). Strategies cited to address implementation barriers included gaining stakeholder buy-in, enlisting unit-based champions, communicating implementation status, and sharing examples of progress including comments from clinicians, family members, and patients regarding the benefits of the practice change. As there is no “done” to the clinical improvement process in critical care, successful efforts require ongoing persistence in messaging, education, measurement of performance and outcomes, and iterative reinforcement of the need for these changes. These efforts are consistent with recent calls to use emerging data from IS and IR and precision medicine initiatives to drive health system improvement (213).

MAKING IS IN CRITICAL CARE EASIER AND MORE RELEVANT

IR is still gaining a foothold in critical care, as evidenced by recent calls to define the critical care IR agenda (10) and to bring IS and IR into the ICU (5). Many of the theories, models, and frameworks used in IS have been tested in behavioral health and outpatient care settings rather than the fast-paced, high-stakes environment of the ICU. Efforts to apply IS principles to ICUs have been bolstered by the creation of critical care-specific IS training programs (214), recent published reviews in high-impact critical care journals highlighting IS methods to overcome barriers to the uptake of evidence-based practices (29, 124), and an increasing willingness by funding agencies, such as the U.S. National Institutes of Health, to fund IR in critical care (215). Moving forward, critical care IR can be made easier and more relevant by: 1) using established implementation frameworks, theories, and models (FTMs); 2) using hybrid effectiveness-implementation trials; 3) leveraging EHR data to inform and drive change; 4) linking implementation to QI efforts; and 5) using research and practice collaboratives. We explain each of these briefly below.

Use of Implementation Frameworks, Theories, and Models

Well-developed and empirically supported implementation FTMs enable the translation and adoption of evidence into daily clinical practice. Despite more than 100 FTMs relevant to IS (216), fewer than half (23–47%) of the published IS-based studies in clinical medicine used an FTM, including the underuse, superficial use (i.e., only citing a framework in the background or discussion sections), and/or misuse of FTMs, thus threatening the field’s advancement (216). This observation is due in part to the difficulty in identifying and selecting FTMs appropriate to the problem at hand. Multiple published articles highlight how implementation FTMs may be used (28, 217, 218). Instead of developing new FTMs, the authors strongly recommend the selection and adaption (if needed) of existing FTMs in critical care IR. Using existing FTMs also helps to broaden the implementation evidence base and to standardize measurement, thereby allowing for comparisons across studies.

Use of Hybrid Effectiveness-Implementation Trials

Implementation outcomes such as “acceptability” and “fidelity” are fundamentally distinct from “classical” ICU outcomes of interest, including mortality, length of stay, and end-organ dysfunction (219). When evidence supporting an intervention’s effectiveness is strong (e.g., multiple, large randomized controlled trials showing evidence of benefit), implementation studies may focus only on implementation outcomes. In critical care, intervention evidence is often based on randomized studies that may later be reversed (220) or on quasi-experimental studies demonstrating significant threats to inference. In these settings, it may be prudent to study both implementation and effectiveness simultaneously in an approach known as the “effectiveness-implementation hybrid design” (221), where relationships between implementation and effectiveness can be tested, and intervention effectiveness can be interrogated in clinical settings different from the original study.

Leveraging Electronic Health Records to Inform and Drive Change

IR tends to use mixed methods approaches to data collection in which both qualitative and quantitative data are used to characterize implementation determinants and evaluate implementation efforts. Collection of data are time-consuming, labor-intensive, and costly and may fall outside of the available resources for an implementation effort. The EHR is increasingly being used to facilitate clinical and health services research and QI efforts (222, 223). Linking patient-level EHR data to evidence-based protocols and presenting these data in easy-to-read formats can help facilitate real-time team decision-making support. Data related to processes of care, such as nursing interventions documented in electronic flowsheets, could be leveraged to document performance gaps and to track the impact of implementation efforts on intervention compliance. Outcome measures related to the intervention can be evaluated over time and across ICU patient populations to assess the direct impacts on patients. Accessible, transparent, and provider-specific EHR-based performance and outcome data can help reduce practice variation and more rapidly drive unit-based change. Artificial intelligence and machine learning approaches could also be applied to identify patients at greatest risk of nosocomial complications (e.g., delirium), allowing implementers to concentrate their efforts on ICU patient populations most likely to benefit. EHRs can also be used as platforms for implementation strategies. For example, relevant guidelines can be presented to clinicians in the form of clinical decision support tools, ordering providers’ choices can be constrained to conform with guidelines and/or evidence, and default orders can be engineered to “nudge” providers toward the desired care pattern (224).

Linking Implementation Efforts to Quality Improvement Initiatives

Health systems are focused on delivering high-quality care while constraining costs. For this reason, implementation efforts often compete with other programs for limited resources. But when implementation is aligned with existing quality initiatives, such as those directed by regulatory agencies and insurance payers, it is easier to make the “business case” for earmarking resources to support implementation. There are many evidence-based practices in critical care that are ripe for implementation. In selecting which best practices to focus on, linking efforts to the current priorities of the hospital or health system is often a useful approach (e.g., precision medicine initiatives, the Centers for Medicare & Medicaid Services Hospital Value-based Purchasing Program [22], and the Medicare Access and CHIP Reauthorization Act of 2015 [225]). Establishing this connection requires active engagement of ICU, hospital, and health system leaders in implementation efforts, to help garner needed resources, to make a strong financial case for implementation efforts, and to spread change across the organization.

Use of Research and Practice Collaboratives

As IR is contextual, the lessons learned from a local QI effort may or may not translate to other critical care settings. Ideally, IR and implementation practice teams could work together to identify common barriers and facilitators and to identify the most promising implementation strategies. The Society of Critical Care Medicine Discovery Network (226) is an example of a research network that can centralize institutional review board approval and data collection, facilitating multicenter IR studies that could help to advance critical care IS.

CONCLUSIONS

Significant delays in the widespread adoption of evidence-based critical care practices adversely affect patient outcomes and increase costs of care. IS and IR, along with their associated conceptual frameworks, provide a systematic approach to identifying effective strategies for overcoming barriers to the translation of evidence into critical care practice. Effectively leveraging established implementation FTMs, hybrid effectiveness-implementation methods, EHR datasets, existing QI initiatives, and research and practice collaboratives can help to facilitate critical care IR. Using IS and IR to understand and address the causes and strategies for overcoming barriers can help improve the quality, safety, and value of care delivered to critically ill patients.