In today's ICUs clinicians routinely integrate huge numbers of discrete data points to arrive at a coherent picture of their patients' status. Often the clinician must obtain those data elements from many devices, which makes the problem more difficult. Because presenting data visually amplifies cognition by capitalizing on well-known human perceptual capabilities, it is not surprising that a growing body of research is directed at the effective presentation of visual information in clinical displays. However, developing clinical displays that effectively support clinicians' integration and understanding of many discrete data elements in complex, high technology work domains such as ICUs remains elusive. It may be that traditional analysis and design methods simply are inadequate for this kind of complex environment. Vicente has described a new methodology, called “cognitive work analysis” (CWA), which is targeted at the analysis of complex work domains. The analysis differs from traditional analytic methods in significant ways, particularly in its primary focus on analysis of the work domain, but also in its prescription for explicitly collecting information at five levels (the work domain, diagnostic and treatment tasks, diagnostic strategies, socio-organizational, and clinician skills) that place constraints on the ultimate display design. In this model, the order of data collection is also crucial. Because the work domain constraints tend to be the most permanent, they are likely to have the most impact on design, and so analysis starts there. As the analysis proceeds through the subsequent levels, additional design constraints are identified. We recently used CWA to analyze the information needs for interactive graphical displays that will integrate and represent data in structures that help clinicians visualize a patient's physiological status. We found that the analysis was an effective way to identify information needs at multiple levels. Based on our experience, CWA is a generic methodology that is highly applicable to medical informatics.