Determinants of Ventilator Withdrawal Among Patients With Prolonged Mechanical Ventilation
A key finding of the study is, however, that none of the numerous modeled patient-, surrogate-, and physician-related characteristics predicted ventilator withdrawal if a physician anticipated death. Several explanations may possibly underlie these “negative” results. First, the modeled covariates simply do not affect surrogates’ decision-making. This would mean need to explore alternative determinants of surrogates’ decision-making. Second, participating surrogates may have formed clear views of patients’ prognosis, and if these were affirmed by physicians, surrogates indicated plan of ventilator withdrawal. Under this interpretation, the direction of the decision (withdraw/not withdraw) could be viewed as plain affirmation or contradiction of formed prognostic views that are no longer affected by the modeled covariates.
Alternatively, the results may have been affected in part by study instrument (question 4 Fig. 1). As phrased, inquiring whether a surrogate would agree to “withdraw ventilator” “in anticipation of death by physicians” has several implications not supported by recent reports: 1) it is the prognostic data, rather than a specific physician as its source, that count for surrogate decision-making; however, surrogates’ trust in the source of prognostic data varies (2, 3). 2) Prognostic data are routinely conveyed in a categorical manner; however, prognostication uncertainty is a prevalent part of clinician communication (3) and is acknowledged by surrogates (2). 3) Prognostic information on expected mortality will be interpreted by surrogates as intended by the clinician; however, surrogate interpretation of offered prognostic data are remarkably variable (3), with prevalent discordance with physicians (4).
The decision-making process by surrogates whether to withdraw life-support interventions is not a utilitarian process but rather, not unexpectedly, an emotional one (2, 4). Similarly, prognosis views by surrogates and their level of trust in patient’s physician (both of which are expected to affect withdrawal decisions) are not strictly fact-driven. Thus, another way to interpret this part of the study by Chen et al (1) is that the modeled covariates cannot predict surrogates’ opinions prior to deciding whether to withdraw ventilatory support.
Prognostic estimates by surrogates and their interpretation of prognostic data from physicians were largely studied under hypothetical scenarios (3). Even when applied to their family members, the association of surrogates’ prognostic estimates with actual subsequent withdrawal of life support has not been examined (4). Likewise, the level of trust in patient’s physicians, as potential modulator of surrogates’ decisions about withdrawal of life support, remains unknown.
Future examination of the dynamic interplay between surrogates’ prognostic estimates, discordance of perceived versus actual physicians’ prognosis, and surrogates’ level of trust in patient’s physicians, in the context of specific patient, surrogate, and physician characteristics, may provide further insights into surrogates’ decision-making process and may in turn inform future efforts to ease the burden on surrogates when considering withdrawal (or withholding) of life support in the critically ill.