• Death education
• suicide/assisted suicide
• substance abusers/users of controlled substances
• right to refuse treatment
• research ethics
• elderly and terminally ill
• living wills/advance directives
• healthcare economics

There is an increased awareness of the burden decision-making on the families of critically ill patients. There are also those who are focused on the rising cost of healthcare and the common use of aggressive and undesired interventions and the end-of-life. These and other factors have spurred an increasing interest in using default options to guide end-of-life decision-making.1 ,2 The powerful influence of default options on decision making has been illustrated in several medical settings including the randomised trials of hypothetical advanced directives3 ,4 and a randomised trial of defaults for real influenza vaccine appointments.5

Haward and colleagues report a randomised trial of default options in hypothetical vignettes depicting the delivery of a 23-week-old neonate.6 Among the 291 (49%) respondents, participants who were told that the delivery room's normal practice was to attempt to resuscitate such neonates were significantly more likely to choose resuscitation compared with participants who were told the default was comfort care.

This novel study packs a powerful implication: could our preferences for something as important as whether to resuscitate premature offspring be susceptible to the same suggestions used more commonly in less personal domains such as internet marketing7 and automobile insurance8? Although this is possible, several limitations of the study warrant consideration before accepting this conclusion.

First, the authors report a 50% response rate, but offer no assessment of non-response bias. The characteristics of their study population, mostly notably that 65% of respondents had experience with prematurity, suggest that respondents may have different perspectives on this matter than non-respondents, limiting the internal and external validity of the results. Second, the investigators compared one default with another, instead of comparing the default options to a control, or ‘active choice’ arm. In order to better reflect the clinical status quo, it would have been preferable to have a 3-armed study where each of the two proposed defaults was compared against an arm in which participants can actively select whether or not to resuscitate.

Third, there is an important imbalance in the defaults being compared. In the resuscitation default, study participants are told that it is a typical hospital practice to resuscitate 23-week-old neonates. There is no explanation of what the alternative to resuscitation is, although it may be safe to assume that people understood that if they decline resuscitation, the newborn would die. However, in the comfort care default arm, it may not have been at all clear what the alternative was. Would anyone elect to ‘decline comfort’ when this is presented in isolation, rather than as a discrete alternative to resuscitation?

In nearly all studies and practical applications, defaults are used for one of two purposes: (1) to promote the acceptance or rejection of a single choice (eg, CPR),9 or (2) to promote the acceptance of one of two or more potentially competing choices. In this study, the authors compared the acceptance of one choice—resuscitation—with the rejection of another choice (comfort). Framed this way, participants are guided in a discrete direction into one of the arms; in the comfort default arm rejection of the default does not obviously equate to what will be done.

Defaults are powerful, and they have tremendous promise to promote positive healthcare choices and to help guide patients in making challenging healthcare decisions. However, because defaults may appeal to a status quo bias or may be challenging to contravene if barriers to choice are erected, default options may be influential even when the default presented is inappropriate. Carefully presenting the specific options to be chosen, and minimising barriers in choosing any given option are the hallmarks of ethically using default options.1

When examining the use of defaults in healthcare, it is important to consider the extent to which people have deep-seated preferences among the array of plausible options. In some circumstances, people do possess strong pre-contemplated preferences; for other choices, preferences tend to be constructed during the process of eliciting them.10 When people have strong preferences, as the authors propose people do regarding neonatal resuscitation, then using defaults can yield three possible outcomes. The defaults could (a) not work1; (b) work inappropriately, as when no single preference applies to a large majority of decision makers; or (c) work appropriately, as when a clear majority preference exists and is promoted by the default. The authors posit that choices surrounding immediate resuscitation of a 23-week-old neonate fall into the second category—that defaults work inappropriately here because they would steer many people away from the choice that is best for them. This is certainly a plausible conclusion. However, we believe that more empirical work is needed to ascertain the full effects—both intended and unintended—of setting defaults of no resuscitation and of allowing prevailing defaults of resuscitation to persist.

In summary, we commend the authors for exploring defaults in an evidence-based fashion, in a specific and novel context, rather than attempting to generalise insights from how defaults work in other contexts. This study may therefore be viewed as a first step in guiding policy related to defaults in neonatal resuscitation. However, the design is insufficient to refute or confirm the propriety of default-setting in this context. Whether or not defaults are appropriate in this context, the setting of defaults is likely appropriate in some settings and inappropriate in others. Attempts to generalise from one decisional context to another are likely to be wrong, and we need high-quality, domain-specific evidence to measure both the intended and unintended consequences of defaults.