Stick to What You Know: Do Visiting Intensivists Worsen Outcomes?*
Recent studies have helped to characterize the organizational characteristics of an ICU that may help to reduce heterogeneity in care and lead to improved clinical outcomes (8, 9). For example, higher hospital case volume, effective team communication strategies, and a higher nurse-to-bed ratio are important determinants of outcomes in critically ill patients (8–10). In contrast, the use of protocols has not been found to be associated with improved clinical outcomes in both observational studies (11) and randomized clinical trials (12). High-intensity critical care physician staffing has had differing reports on clinical outcomes (13) with beneficial effects of such staffing more pronounced in surgical ICUs (14). Nighttime in-house intensivists are not associated with a lower hospital length of stay, lower mortality or higher ICU readmission when compared with periods when there was not a nighttime in-house intensivist (15).
The report by Whitehouse et al (16) in this issue of Critical Care Medicine brings attention to yet another potential driver of heterogeneity in delivery of care: that of visiting intensivists in an ICU. The authors postulated that intensivists unfamiliar with an ICU team and the context of that ICU would have worse outcomes when compared with intensivists who continued to work in their ICU. They used data of 9,981 admissions and 33 intensivists collected over 5 years in four ICUs of a large U.K. hospital. Of note, the types of geographically distinct ICUs that the authors examined in their analysis included liver specialty, neurology, cardiac, and general/trauma/burns ICUs. Using different types of analyses that controlled for clustering and adjusted for multiple confounders, the authors found that the odds of mortality was 18% higher with a visiting intensivist from another ICU when compared with a home intensivist. The association between visiting intensivist and a higher odds of mortality remained regardless of the type of analysis used: whether a marginal approach adjusted for clustering using generalized estimating equations, a hierarchical (random effects) model, or an analysis that adjusted for the likelihood (via a propensity score analysis) of being admitted by a visiting intensivist from another ICU.