The evidence regarding physician staffing of ICUs does not yet provide a consistent view of the best model to use. Most studies have significant limitations, and this subject is complicated by the fact that optimal ICU staffing may depend on ICU characteristics. The topic with the most data regarding patient outcomes is the intensity of intensivist involvement in care, particularly the value of closed- vs open-model ICUs; however, the evidence is inconsistent here as well. Even if closed-model ICUs produce better outcomes, we do not know which specific elements of that multifaceted organizational paradigm are responsible for improvement. Also, studies of around-the-clock intensivist presence have not consistently shown that it is associated with superior outcomes. Increasingly, nonphysician providers are playing innovative roles in the ICU, and care provided by teams including nurse practitioners or physician assistants appears to be safe and comparable to that provided by other staffing models. Although we do not know the best way to staff ICUs, the conditions of ICU physician coverage will continue to change under the stresses of shortages of intensivists and increasing duty hour limitations for trainees. Nonphysician providers, innovative physician staffing models, telemedicine, and other technologies will be increasingly used to cope with these realities. This evolution makes it more important than ever to study how staffing affects outcomes. Only quantitative evaluation can tell us whether one staffing model is better than another. Accordingly, we need more research from multiple sites to develop a consistent and integrated understanding of this complex topic.