Concepts with suboptimal effectiveness in practice

Some traditional concepts have been shown to have suboptimal effectiveness in practice (Table 1). Although sepsis guidelines recommend the early use of broader-spectrum antimicrobials to increase antimicrobial timeliness and appropriateness, ^{Reference Evans, Rhodes and Alhazzani1} the usually proposed balancing measures may not, in practice, deliver the intended results. De-escalation occurs infrequently ^{Reference De Bus, Depuydt and Steen10} due to limited sensitivity of microbiological cultures; biomarkers of host response to infection (CRP and PCT), although shown in earlier studies to be efficacious, have not been as effective in less controlled environments; ^{Reference Dark, Hossain and McAuley11,Reference Stevenson, Forsyth, Hossain, Lall and Dark12} the use of novel biomarkers and syndromic molecular diagnostics, which could deal with the high uncertainty in the diagnosis of infection, ^{Reference Cardozo Junior, Bianchini and Giovanetti13} do not necessarily translate into improved utilization of antimicrobials. ^{Reference Hellyer, McAuley and Walsh14,Reference Enne, Stirling and Barber15}

Table 1.

Traditional concepts challenged by empirical evidence

The poorly explored role of cognitive decision-making theory in AS

The reasons why presenting more information to clinicians does not necessarily translate into behavior modification are likely rooted in cognitive decision-making theory. ^{Reference Croskerry, Singhal and Mamede16} Cognitive processes traditionally fall within two main categories: type 1 processes are intuitive, fast and more subject to biases in decision-making, as they are heuristic shortcuts to decision-making obtained from repeated experience; type 2 processes are analytical, slower and reflective of deductive reasoning, which is less subject to error. ^{Reference Croskerry17} Cognitive biases in decision-making (favoring use of more antibiotics in the high-stakes scenario of the ICU setting) very likely contribute (Table 2). ^{Reference Croskerry18,Reference Langford, Daneman, Leung and Langford19} Uncertainty in the diagnostic likelihood of infection is also an essential factor. ^{Reference Cardozo Junior, Bianchini and Giovanetti13} Qualitative studies have shown that fear may be an important driver for more liberal antimicrobial use, ^{Reference Pandolfo, Horne and Jani6} which leads to an anchoring bias effect towards earlier use and more broad-spectrum use of antimicrobials in ICU practice. Interestingly, even when presented with information that could discard a diagnosis of infection, clinicians rely on their first impressions, ^{Reference Singh, Nurek, Mason, Moore, Mughal and Vizcaychipi20} ie, on type 1 cognitive processes. ^{Reference Croskerry, Singhal and Mamede16,Reference Croskerry17}

Table 2.

Some concepts related to cognitive biases in decision-making and how they may relate to Antimicrobial Stewardship in the intensive care setting

Strategies to deal with uncertainty and cognitive biases

There are many possible strategies to address the uncertainty when using antibiotics in critically ill patients (Table 3). ^{Reference Bassi, Tomazini and Carneiro21–Reference Gohil, Septimus and Kleinman26} These strategies have been described and implemented by intensivists, by other members of the interprofessional team in collaboration with intensivists, or by leveraging EHRs.

Table 3.

Strategies to deal with uncertainty in the diagnosis of infection in critically ill patients

The four moments of antimicrobial decision-making framework ^{Reference Tamma, Miller and Cosgrove27} brings together some of these concepts: (i) does the patient have an infection that requires antibiotics?; (ii) were appropriate cultures ordered before starting antibiotics, and what empiric therapy should be initiated; (iii) after a day or more has passed, can antibiotics be stopped, narrowed or changed to oral?; (iv) what duration of therapy is needed for the diagnosis?

For the ICU setting, these moments include the concept of watchful waiting (moment 1), ^{Reference Bassi, Tomazini and Carneiro21} risk-based choice of antimicrobials to avoid overtreatment (moment 2), ^{Reference Gohil, Septimus and Kleinman26} antibiotic time-outs to revise diagnosis and treatment (moment 3), ^{Reference Mishima, Nawa and Asada22,Reference Elligsen, Pinto, Leis, Walker, Daneman and MacFadden23} using automated stop-orders to avoid unnecessary prolonged treatments, ^{Reference Barlam, Cosgrove and Abbo24} and favoring short-course therapies ^{Reference Mo, Booraphun and Li28,Reference Kalil, Metersky and Klompas29} when possible (moment 4).

Diagnostic stewardship, cognitive biases, and AS

An incorrect diagnosis of infection related to positive results of cultures, even in scenarios in which it may represent contamination or colonization, may lead to increased use of antibiotics. ^{Reference Pratt, Rehan and Xing30} Diagnostic stewardship has emerged in the literature as a new conceptual framework grounded on providing a more evidence-based approach to infection diagnosis. Importantly, it is intrinsically related to the cognitive theory of decision-making. It has been framed according to clinical pathology and laboratory medicine approach to diagnosis: preanalytical (ordering, specimen collection); analytical (specimen processing); and postanalytical (reporting and interpreting test results). ^{Reference Fabre, Davis and Diekema31}

In the ICU setting, we identified results that may be applied to antibiotic stewardship efforts from urine cultures, respiratory tract cultures, and blood cultures. ^{Reference Pratt, Rehan and Xing30,Reference Siev, Levy and Chen32–Reference Albin, Troost and Saravolatz34} By explicitly acknowledging that preanalytical pretest probability is essential in the process to increase test results utility and reduce false-positive tests, and by avoiding reporting positive cultures that most likely represent contamination or colonization, it implicitly assumes that nudging the clinician may modify behavior. Non-culture tests, such as molecular tests or biomarkers, as described in Table 1, likely benefit from this approach as well. Incorporation of diagnostic stewardship at a lower level may provide the ground for improved effectiveness of such modern diagnostics.

Interprofessional collaboration as a safety layer in ICU AS

Interprofessional engagement plays an essential role in AS in the ICU setting. Nurses, ^{Reference Bos, Schouten, De Bot, Vermeulen and Hulscher35} clinical pharmacists, ^{Reference Díaz-Madriz, Chaverri-Fernández and Arguedas-Chacón36} and microbiologists have different roles summarized in Table 4. Collaboration among different professions has also been shown to be essential for effective stewardship in qualitative studies. ^{Reference Barlam, Childs and Zieminski37} Their inclusion as part of AS efforts likely improves an additional layer of quality and effectiveness for base interventions.

Table 4.

Role of interprofessional engagement and collaboration in ICU Antimicrobial Stewardship

The role of external input to ICU antibiotic decision-making

Some strategies have known, longstanding roles to improve antibiotic utilization. ^{Reference Schuts, Hulscher and Mouton4,Reference Davey, Marwick and Scott5,Reference Barlam, Cosgrove and Abbo24} Preauthorization is a common, effective restrictive strategy to reduce antimicrobial use. Its main downside is the risk of disruption in collaborative culture. Prospective audit-and-feedback by staff external to the ICU (ID physicians and/or ID pharmacists) is an enablement strategy considered more appropriate, though more time-demanding than preauthorization. So-called handshake rounds, in which ID physicians routinely round with ICU staff to decide on antimicrobial treatments, have been tested in a cluster-randomized crossover trial, but the results were unsatisfactory, despite the time investment. ^{Reference Seidelman, Turner and Wrenn38} It may be possible that this approach did not deal with intensivists’ decision-making to the desired extent. Alternatively, handshake rounds could influence ID physicians’ decisions and soften their recommendations, leading to greater complacency with intensivists’ antimicrobial preferences. This reinforces the need to address intensivists’ fear when prescribing antimicrobials.

The role of EHRs and decision support systems

The EHR also plays an essential role in antimicrobial decision-making. It may operationalize all strategies presented in Table 3. EHR-embedded clinical decision support systems may facilitate the following recommendations based on local stratified antibiograms, though sustaining the effect likely requires routine updates. ^{Reference Nachtigall, Tafelski and Deja39} A computer provider order entry set may facilitate more appropriate and proportionate use of antimicrobials, especially for community-acquired infections. ^{Reference Gohil, Septimus and Kleinman26} However, this may also expand to the ICU setting by stratifying patients with suspected infection in different risk profiles to allow more rational use of antimicrobials.

Decision-making, overdiagnosis, and overtreatment of bacterial infections in the ICU

The previous themes converge on a main argument: antimicrobial decision-making in the ICU setting is subject to cognitive biases due to uncertainty about the likelihood of infection in patients with suspected infections, coupled with the high-stakes nature of the ICU. Although strategies to reduce uncertainty (through improved diagnostics) exist, they may fall short due to risk aversion when deciding about antimicrobials and the desire to conform to guidelines that, although targeting sepsis and septic shock, likely have spillover effects for all suspected cases of infection in the ICU setting. Ultimately, this leads to overdiagnosis and overtreatment of bacterial infections and clinicians’ defaulting to a higher-than-needed use of antimicrobials in the ICU.

Conceptual framework for rational use of antimicrobials in the ICU setting

We propose a three-layered conceptual framework for efforts to rationalize antimicrobial use in the ICU setting (Figure 1). Unless the first layer is addressed, this model suggests that subsequent layers will be less effective in real-world settings than their theoretical efficacy would suggest, or in more controlled environments.

Figure 1.

Conceptual model for Antimicrobial Stewardship in the intensive care unit setting. Legend: ID: infectious diseases; EHR: electronic health record; This conceptual model presents potential strategies to improve Antimicrobial Stewardship in the ICU setting. It assumes that the effectiveness of subsequent strategies will be higher if the base (Tier 1) level is addressed. In tier 1 lies the need to address cognitive processes in decision-making for critically ill patients, which may be biased toward overdiagnosis and overtreatment. Along with traditional effective strategies, such as preauthorization and prospective audit and feedback, frameworks such as diagnostic stewardship (preanalytical and postanalytical) and the four moments of decision-making may help the individual clinician better rationalize their antimicrobial use. Tier 2 strategies consolidate this base layer, and Tier 3 strategies are more likely to be effective provided both Tier 1 and 2 strategies are leveraged. Within Tier 1 diagnostic and Antimicrobial Stewardship reasoning, strategies such as MRSA swabs may be very effective in practice.

Figure 1 Long description

A conceptual model for antimicrobial stewardship in the intensive care unit setting. The model is structured in three tiers. Tier 1, at the base, focuses on clinician decision-making and cognitive biases. It includes strategies like pre-authorization, prospective audit and feedback, conceptual frameworks, diagnostic stewardship, and the four moments of decision-making. Tier 2 emphasizes interprofessional collaboration involving intensivists, ID physicians, clinical pharmacists, critical care nurses, and microbiologists. It also includes EHR-leveraged strategies and external input from specialized ID support for complex cases. Tier 3, at the top, introduces novel technologies such as molecular diagnostics and biomarkers of infection. The model suggests that the effectiveness of higher-tier strategies is more likely if the strategies in the previous tiers are addressed.

At the base, cognitive debiasing strategies are essential to promote rational prescribing of antimicrobials. ^{Reference Pandolfo, Horne and Jani6} Frameworks such as the 4 moments of antimicrobial decision-making may be employed or adapted to the ICU setting. ^{Reference Tamma, Miller and Cosgrove27} Diagnostic stewardship principles should be embedded in this first level as they are intimately related to antimicrobial treatment decisions. Within this base level, effective interventions, such as preauthorization or others, should be employed. ^{Reference Schuts, Hulscher and Mouton4,Reference Davey, Marwick and Scott5,Reference Barlam, Cosgrove and Abbo24}

On top of this first layer, several strategies may be added. These include interprofessional collaboration (with nurses, clinical pharmacists, and microbiologists) and EHR-embedded strategies to support first-line interventions, coupled with input from ID physicians when needed. These strategies usually would focus on nudging ^{Reference Last, Buttenheim, Timon, Mitra and Beidas40} clinicians towards appropriate use and interpretation of diagnostics and proper use of antimicrobials.

The third layer includes strategies with higher costs, such as the incorporation of novel technologies, which include modern rapid diagnostics and biomarkers. Unless intensivists’ decision-making is addressed, they may incur increased costs with no additional benefits of more rational use of antimicrobials. Their full potential for improved antimicrobial utilization may be unleashed, however, if they are placed on top of the second layer of AS strategies, leveraging decision-making, traditional AS, diagnostic stewardship, interprofessional collaboration, additional ID physicians’ input for complex cases and EHR-embedded strategies.