Key points are not available for this paper at this time.
Unlike many specialties, critical care medicine focuses not on a particular organ system, but on the organization and logistics of care. This focus includes the optimal scheduling of ICU staff. Even during the 1952 polio pandemic in Copenhagen, medical student volunteers divided the day into 8-hour shifts to provide continuous manual ventilation to patients with respiratory failure (1). In this issue of Critical Care Medicine, Admon et al (2) provide a clever and helpful contribution to our understanding of the relationship between intensivist scheduling and patient outcomes. Optimal organization of physicians in the ICU has been evaluated in multiple studies. A systematic review including 26 observational studies found an association between mandatory intensivist staffing or consultation and reduced hospital mortality, compared with no intensivist or optional intensivist consultation (3). Studies have evaluated the effect of on-site compared with telephone nighttime or weekend intensivist coverage and found no difference in patient mortality or other outcomes in closed ICUs (4). Observational studies regarding the optimal ratio of patients to intensivists have conflicting findings. A large study based on data from the United Kingdom showed a U-shaped relationship between ratio of patients to intensivists and hospital mortality, with the lowest hospital mortality seen at 7.5 patients per intensivist (5). A more recent analysis using data from the United States showed no relationship between the ratio of patients to intensivists and either ICU or hospital mortality (6). The number of consecutive days worked is an important aspect of organizing intensivists in the ICU. Longer numbers of consecutive days yield fewer handovers and better continuity for patients but may exacerbate physician burnout and fatigue (7). Shorter numbers of consecutive days allow for additional perspectives on each patient and may reduce physician burnout, but the increased number of handovers may reduce the quality of care. An observational study of data from Australia and New Zealand showed that patients cared for by intensivists working less than 7 compared with 7 or more consecutive days had similar hospital mortality and potentially shorter length of ICU stay. However, in North America, most intensivists rotate in 7- or 14-day rotations (8). To date, no studies have addressed whether 1-week vs. attending intensivist schedules are associated with any differences in patient outcomes. This is the gap addressed by Admon et al (2). Admon et al (2) addressed this gap using an observational study. Although there have been randomized controlled trials comparing different physician schedules in the ICU (9), manipulating physician schedules for research is logistically challenging, and impossible to do in a blinded manner. However, they did not use a conventional retrospective cohort design but instead used a target trial framework. A target trial framework is a method of designing an observational study to mirror the ideal randomized trial that would address the research question (10). For example, Admon et al (2) specify that the ideal randomized controlled trial would assign patients in a 1:1 ratio to an intensivist working for 1 week or 2 weeks. After identifying the ideal (or "target") trial, investigators design an observational study to emulate the target trial; hence, the terminology "target trial emulation." Critical care medicine has seen an increasing number of target trial emulations, analyzing interventions such as extracorporeal membrane oxygenation in COVID-19 (11) or bag-mask ventilation during peri-intubation apneic time (12). Designing an observational study to mirror a randomized trial, including clearly articulated eligibility criteria and time zero, helps clinicians to apply the results of the study at the bedside, reduces selection bias, and ensures that studies focus on possible interventions. However, target trial emulation does not eliminate unmeasured confounding, the most common and vexing limitation of observational studies. Admon et al (2) designed their target trial emulation using the clever observation that, for a patient admitted in the second of 2 weeks, the maximum duration of continuity with their admitting intensivist is 1 week. For this combination of intervention and outcome, there are no confounding variables plausibly associated with both the date of ICU admission (week 1 vs. week 2) and patient outcomes. This means that target trial emulation was a particularly elegant choice to address this research question. The primary findings are that there were no differences in patient outcomes among patients admitted during the first compared with second week of each intensivist block, including no difference in mortality, ICU length of stay, or duration of mechanical ventilation. These findings held up in robust sensitivity analyses focused on the last 2 days of a 2-week block, compared with the same days of the week at the end of the first week of a 2-week block, intended to capture a higher proportion of patients who experienced a transition in care. Despite the insightful design, some important limitations remain. All patients who received a maximum of 1 week of intensivist coverage were admitted during the second week of a 2-week block for the attending intensivist. It is possible that the benefits of a fresh perspective that result from more frequent handover (13) were attenuated by fatigue in the admitting intensivist. In a true randomized trial, the intensivist doing a 1-week block would not usually have done a separate 1-week block in the prior week. An additional limitation is that both studied hospitals had physicians-in-training, whose continuity may have compensated for an increased frequency of attending physician handover. A statistical limitation is that the underlying data are clustered by physician, but this structure was not incorporated into the analysis, which means that the outcome estimate ses are likely underestimated. One final limitation, which could inform future research on this topic, is that the study did not include outcomes related to the experience of care. This is relevant because many patients may prefer continuity amid the chaotic experience of critical illness (14). Another opportunity for future research would be to assess for heterogeneity in the efficacy of different schedules according to patient characteristics such as diagnosis or complexity. It would also be prudent to assess for heterogeneity according to the social determinants of health and compare schedules not only just in terms of overall outcomes but also in terms of equity. Critical care medicine is, in some ways, the science of organizing care to improve outcomes for critically ill patients. Admon et al (2) have elegantly demonstrated the usefulness of target trial emulation in this pursuit and have contributed to our understanding of the safety of 1-week compared with 2-week attending intensivist schedules. Target trial emulation will be a useful tool for addressing further questions in how best to organize care for critically ill patients.
Christopher J. Yarnell (Thu,) studied this question.