Current pain assessment tools for nonverbal critical care patients may not be appropriate for those with brain injury, as these patients demonstrate specific responses to pain. The aim of this study was to generate and select items that could be used to assess pain in brain-injured patients. A sequential mixed-method design was chosen with three consecutive steps: 1. Generate items with a literature review, the results of a pilot study, and interviews with 18 clinicians using the nominal group technique. 2. Evaluate content validity with 10 clinicians and four scientists, using a web-based questionnaire. 3. Describe and reduce items with the observation of 116 brain-injured patients in the intensive care unit during common painful procedures. This study took place between May 2010 and October 2011 in two tertiary hospitals in Western Switzerland. Forty-seven items were generated and reduced to 33 during the content validity process. The behaviors most frequently observed during turning were closing the eyes (58.6%), eye movements (57.8%), ventilator asynchrony (55.2%), and frowning/brow lowering (50%). Five items were observed in less than 5% of the patients during nociceptive procedure. Constant motor activity was observed more frequently at rest than during nociceptive stimulation. All physiologic items showed little variability and their reliability was low. Based on these results, the number of items was reduced to 23. This study identified items that could be specific to brain-injured patients and found that the variability of physiologic items was poorly assessed by clinicians.