Default options in the ICU: widely used but insufficiently understood

Abstract

Purpose of review

Default options dramatically influence the behavior of decision makers and may serve as effective decision support tools in the ICU. Their use in medicine has increased in an effort to improve efficiency, reduce errors, and harness the potential of healthcare technology.

Recent findings

Defaults often fall short of their predicted influence when employed in critical care settings as quality improvement interventions. Investigations reporting the use of defaults are often limited by variations in the relative effect across sites. Preimplementation experiments and long-term monitoring studies are lacking.

Summary

Defaults in the ICU may help or harm patients and clinical efficiency depending on their format and use. When constructing and encountering defaults, providers should be aware of their powerful and complex influences on decision making. Additional evaluations of the appropriate creation of healthcare defaults and their resulting intended and unintended consequences are needed.

INTRODUCTION

Caring for critically ill patients requires nearly constant decision making. Ideally, decision makers carefully consider the risks and benefits of each choice before settling on the option that maximizes the patient’s well-being. However, the emotions and time pressures of the ICU may exacerbate social and contextual barriers to rational choice [1]. Using a default option, or a set of ‘events or conditions that will be set in place if no alternatives are actively chosen’ [2], may help decision makers overcome these barriers. When defaults are set, decision makers remain free to choose an alternative. However, evidence from predominantly nonmedical contexts suggests that far more individuals will stick with the preselected choice than would have opted for that choice without the default [3–7].

Although defaults can benefit decision makers and patients when used appropriately, they may also have unintended consequences. Thus, persons in positions to set defaults in the ICU should understand the mechanisms by which defaults exert their influence. We will review how defaults influence decisions, suggest important considerations before using defaults, and summarize the current use of defaults in the ICU and related settings.

THE POWER OF THE DEFAULT OPTION

Default options dramatically influence the behavior of decision makers, particularly in the absence of strong preferences. Across many domains, individuals are far more likely to choose the default option, even when confronted with decisions that seem highly dependent on personal values such as end-of-life care.

The ‘pull’ of the default may be mechanical, belief-based, or psychological (Table 1) (U. Simonsohn, 8 April 2014, personal communication with J.L. Hart). Default options reduce the effort required to make decisions by allowing individuals to avoid actively choosing; indeed, in the presence of defaults, individuals need not even attend to the fact that a decision is being made. The amount of effort saved is dependent on the decision itself. With increasing complexity, the default option further reduces the cost of the decision making. Similarly, remaining with the default is more likely when choosing an alternative would require a significant amount of effort.

Table 1.

Ways in which default options influence decision making

Type of influence	Explanation	Examples in critical carea
Mechanical	Additional effort is required to select an alternative option	Cardiopulmonary resuscitation provided in the event of cardiac arrest unless a ‘do-not-resuscitate’ order is placed
		Providers may not electronically order a medication dose outside the typical range without speaking directly to a pharmacist for approval
Belief-based	Individuals perceive the default as a recommendation from an authority	Providers order venous thromboembolism prophylaxis for a patient when electronically prompted to do so
		Respiratory therapists initiate lung-protective ventilation in all patients unless otherwise instructed
Psychological	Risk-averse or regret-averse individuals assume less personal responsibility by remaining with the default choice	Family members of a terminally ill patient agree to intubation when it is presented as standard care for respiratory failure
		Physicians agree to have nurses or phlebotomists draw daily labs on clinically stable patients

^aThese examples may incorporate multiple mechanisms.

Belief-based phenomena also contribute to the disproportionate preference for the default option. The preselected choice may appear to be a recommendation, which is particularly true when the person or organization providing the default option is thought to be trustworthy. Additionally, decision makers may believe the default option reflects social or prescriptive norms.

Remaining with the default may also protect decision makers from future psychological harm. By choosing an alternative, the decision maker assumes more personal responsibility for future outcomes. Therefore, people who are particularly averse to risk or regret may seek the protection of the default option as they anticipate potential negative outcomes.

CONSIDERATIONS PRIOR TO USE

In critical care, there are many potential ‘choice architects,’ or parties responsible for influencing choice. These include medical directors of the ICU, nurse managers, directors of respiratory or physical therapy, and individual clinicians. In the ICU, the choice architects who set defaults should aim to maximize the welfare of the majority of patients to whom the default would apply, and assume responsibility for negative outcomes of ‘defaulted’ choices. Default options exert strong influence over a decision maker’s choice, which may be concerning in healthcare settings. To preserve choice and autonomy, all options should be clearly presented and equally available.

Assuming the role of choice architect

Because any manner of framing a choice (with or without default options) has consequences, the choice architect must evaluate the risks and benefits of encouraging choices that decision makers might not prefer, with those of failing to encourage choices that decision makers would prefer. For example, cardiopulmonary resuscitation (CPR) may be justified as the default approach for patients suffering in-hospital cardiac arrest because overall the frequency and magnitude of harms of providing unwanted CPR may be small relative to those associated with failing to provide desired CPR. In cases where individual patients’ health states and preferences may influence the propriety of the default, choice architects should also ensure that ample opportunities exist for patients to opt away from the default.

Choosing appropriate settings

Most critical care decisions require careful deliberation, which a default option may truncate or bypass. When one option provides clear benefits for a large majority of patients, offering a default option may effectively and efficiently guide behavior. In contrast, when the optimal choice is uncertain, providing a default option may lead to suboptimal decision making. Defaults in healthcare may be especially problematic, as the ‘best choice’ may depend on the patient’s individual characteristics, such as clinical condition or values.

Customized defaults may overcome this by modifying the option set on the basis of patient-specific information [8]. But it may be preferable to try to avoid defaults altogether when the very process of engaging in decision making is important. For example, when CPR carries a high likelihood of harm and negligible opportunity for benefit, offering a care plan that does not include CPR as the default may decrease burdens on decision makers and improve the overall quality of care. But when both a transition to comfort-based care and a time-limited trial of aggressive critical care are reasonable options, avoiding defaults entirely may engage patients or surrogates in expressing their values, thereby enabling clinicians to provide more individualized guidance on which care plans may best promote those values.

Presenting the default

Responsible use of defaults requires a carefully constructed presentation of the decision that includes all available choices. Because a default option results in a particular action without any active participation by the decision maker, individuals encountering a default may be unaware of the decision or perceive that they have no choice. To preserve autonomy, it is important that decision makers readily appreciate that they are engaging in decision making, and that multiple options exist. Based largely on this reasoning, the Institute of Medicine concluded in 2006 that to change the organ donation default in the United States from an opt-in to an opt-out system, a massive public education campaign would be required to ensure adequate awareness that decisions were being made [9].

Additionally, selecting an alternative choice should not be prohibitively difficult. Although deviating from the default will always require additional action, the effort and persistence required to make an alternative choice should be proportional to the degree to which other choices may be preferable for some patients. A nearly costless deviation from the default option, such as unselecting venous thromboembolism prophylaxis (VTEP) from a default admission order set, may be minimally paternalistic [8]. By contrast, a procedural or substantive barrier – such as requiring that physicians type an explanation to unselect such prophylaxis – may be used to ensure that the alternative is selected wisely and not because of decision making errors. These barriers require more significant effort to overcome the ‘pull’ of the default when the choice architect anticipates that such deviation may result in harm for most patients.

RECENTLY EXAMINED USE OF DEFAULTS IN CRITICAL CARE MEDICINE

In the past decade, the healthcare community has embraced tools intended to increase efficiency and decrease errors, including defaults [10]. Unfortunately, formal evaluations of these defaults prior to implementation and following use have been limited. Therefore, the intended and unintended consequences of most default options in the ICU have not been fully examined. Here, we review recent studies of defaults in the ICU, including order sets and protocolized care, technological prompts, and communication strategies.

Order sets and protocols

Diagnosis-specific order sets and care protocols are intended to streamline the decision making process and reduce errors. Rather than identifying each individual care component, providers rely on order sets or ‘bundles.’ ICUs may make these available for a variety of clinical situations including septic shock [11,12], diabetic ketoacidosis [13–16], alcohol withdrawal [17,18], VTEP [19], sedation management [20–23], and ventilator management [22,24–29]. Such protocolized care may also create a standard unit-based practice that bedside professionals (e.g., nurse or respiratory therapist) implement or override as needed. These have been used in the prevention of device-related infections [25–34] and nutrition management [35–37].

In an effort to reduce catheter-associated urinary tract infections, Parry et al. [38^■] implemented an order set for physicians that linked initial catheter orders, nursing protocols, and daily forced active choices to maintain or discontinue the catheter. With accompanying educational efforts, catheter use and catheter-associated urinary tract infections decreased significantly over a 36-month period, although the authors detected wide differences in the effect across care units.

By contrast, a cluster randomized trial in 18 ICUs that changed the default enteral nutrition provided to 1059 mechanically ventilated patients found no substantial increase in the proportion of patients achieving goal nutrition delivery [37]. Similarly, despite promising evidence in adult populations, a recent systematic review of the use of protocolized mechanical ventilation weaning in children found no improvements in patient outcomes [39].

Together, these data suggest that simply implementing protocols that change defaults may be insufficient to overcome practice patterns. When choice architecture is neglected, defaults may lose power [19,40,41]. Therefore, for defaults to exert their maximal and intended effects, they must be created and implemented with clear attention to exactly how they may change behaviors.

Order sets may also lead to unintended consequences on patient care and medical education [17,19,42–45,46^■,47,48^■]. For example, Khanna et al. [46^■] recently showed that a standard admission order set increased inappropriate VTEP use. Additionally, Yu et al. [48^■] found that implementation of standard admission order sets did not improve short-term knowledge of best practices among 39 trainees, raising concerns that such changes might have adverse impacts on long-term knowledge by decreasing trainees’ engagement in clinical decision making. As defaults in the form of order sets and protocols proliferate in the ICU, effects on both patient outcomes and provider competence will be needed.

Technological prompts

Critical care defaults reach decision makers quickly through the use of computerized provider order entry and electronic health record technologies. In addition to order sets and protocols that make use of these systems, individual medication and diagnostic orders and ICU alarms often rely on defaults. Increasing attention to ‘alert fatigue’ has highlighted the need for thoughtful design.

For example, despite long-standing efforts to implement defaults wisely into computerized provider order entry systems, Stultz and Nahata [49^■] recently concluded that design flaws persist [49^■,50–53]. The authors found that among 3774 medication alerts for pediatric inpatients at a single center, nearly 90% were inappropriately alerted due to an inaccurate default dosing range [49^■]. Similarly, electronic monitoring alarms used in the ICU are often based on default settings that are too sensitive, resulting in many false alarms. In a thoughtful but small study, Inokuchi et al. [54^■] recently observed that with guideline-based default settings, an ICU nurse hears approximately two out of 32 clinically important alarms over an 8-h shift. On the basis of their observations, the authors estimated that adjusting the default alarm-triggering settings could decrease irrelevant alarms by over 20%.

Rather than harnessing the potential power of technology, ICUs may become victim to poorly designed defaults. Improved customization of defaults would include patients’ characteristics (e.g., weight and renal function) and patient-specific trends. Additionally, although choice architects often match defaults to published guidelines, this may not translate well into their use as decision support tools. Particularly given the clear unintended consequences of frequent alarms and alerts [55], these types of defaults should either first be evaluated in an experimental setting or at the very least systematically evaluated as a quality-improvement initiative following implementation.

Communication strategies

Defaults may be used as decision support tools for patients and their surrogate decision makers considering complex choices. When used in this setting, defaults often take the form of ‘informed non-dissent’ [56]. Such defaults have altered choices for organ donation [3,57] and end-of-life care [56,58^■,59–61] when formally implemented.

Decision makers rely on the clear presentation of options to understand a choice is at hand. One particularly unsettling example of a policy in which decision makers were not aware that a choice was being made regards the previously widespread practice of establishing default DNR order for all nursing home residents. The Centers for Medicare and Medicaid statement prohibited the practice in 2013 when it became clear that many residents were not aware of the default DNR order policy, the alternatives available to them, or the relative risks and benefits of these important choices [62–64].

By contrast, our research group recently published the results of a randomized controlled trial of 132 seriously ill patients presented with an advanced directive that did or did not incorporate a default option [58^■]. We not only found that default options for comfort-oriented care greatly increased the proportions of patients who chose such a care strategy, but that this effect persisted even when we specifically alerted patients to the default intervention. These results suggest that default options may be powerful even in such presumably value-sensitive decisions as end-of-life care, and that the potential effects of defaults may be maintained even by actively promoting awareness of the default setting.

There is also evidence that providers’ communication styles include default options about end-of-life care and that these defaults influence option selection. Last year, Lu et al. [65^■] presented the results of a systematic comparison of the language used by academic physicians in a high-fidelity simulation. Among 106 physicians who discussed life-sustaining treatment, 64% presented aggressive therapy as the default, whereas 36% elicited a preference or offered such treatment as an option. In contrast, only 45% of the 86 physicians who discussed palliative care framed it as the default choice. A randomized simulation experiment of potential surrogate decision makers yielded complementary results, with 64% of surrogates choosing CPR when framed as the norm by a standardized physician, compared with 48% when the physician framed do-not-resuscitate as the norm [66^■].

Because providers often cannot avoid setting default options, they should at least be aware of their role as choice architect and frame their discussions with patients and families mindfully. Future work evaluating how default options in physician–patient and physician–surrogate communications impact patients’ outcomes would help physicians use defaults in more appropriate manners.

Practice patterns

ICU care may include ‘hidden’ defaults that are harmful if they are not recognized, regularly evaluated in light of new evidence, and eliminated when found to be inappropriate. Indeed, of the five practices recently questioned by the Critical Care Societies Collaborative’s Choosing Wisely Task Force, at least three reflect potentially harmful default care practices in modern ICUs. These include daily laboratory tests or other diagnostics without clinical indications, sedating all mechanically ventilated patients, often deeply, and continuing life support for patients with poor prognoses [67]. The other two items on the Top 5 List may also represent default practices in certain ICUs: liberal use of blood transfusions and of parenteral nutrition. These practice patterns each appear to be harmful on the basis of current evidence, and yet exist as routine care in ICUs, in part, because they function as hidden defaults.

CONCLUSION

Decision makers in the ICU need support given the number and complexity of decisions made in a pressured and emotional environment. Defaults have potential to improve efficiency and protect against systematic errors in decision making. Yet appropriate use demands foresight and intention as well as additional study. Preimplementation experimental data may lead to more effective defaults, and postimplementation monitoring will provide additional insight into the implications of their use.

KEY POINTS.

• A default option is the set of events or conditions that will occur if no action is taken.

• Defaults may be used in the ICU to implement guidelines, improve efficiency, and protect patients, but there may be unintended consequences from defaults, such as decreased engagement in decision making and lack of individualization of care.

• Current ICU defaults include care protocols and order sets, technology-based presets, communication strategies, and defaults hidden in common care patterns.

• Published research has failed to show conclusive evidence for or against the use of defaults in critical care, which may be due to heterogeneity across environments and preimplementation care patterns.

• When creating defaults, their power as behavior change and decision support tools should be used to maximize their potential impact with preimplementation and postimplementation experiments and studies used to reveal intended and unintended consequences.