There have been growing concerns about the well-being of staff in inpatient mental health settings, with studies suggesting that they have higher burnout and greater work-related stress levels than staff in other healthcare sectors. When addressing staff well-being, psychological safety can be a useful concept. However, there is no measure of psychological safety that is suitable for use in inpatient mental health settings. Edmondson (1999) is the most commonly used measure of psychological safety, but it was designed for use in general physical healthcare settings. As inpatient mental health settings are unique environments, transferability of knowledge from physical to mental healthcare settings cannot be assumed. We sought to develop questionnaire items that capture psychological safety among healthcare staff working in acute inpatient mental healthcare settings. We used the nominal group technique, a consensus method involving rounds of discussion, idea generation, and item rating/ranking to identify priorities. Twenty-eight stakeholders participated, including 4 who had lived experience of mental health problems, 11 academics and 18 healthcare professionals (8 participants identified with more than 1 category). The study involved a workshop with three parts: (i) an overview of current research and limitations of the Edmondson (1999) measure as outlined above, (ii) discussion on what items should be retained from the Edmondson (1999) measure, and (iii) discussion on what items should be added to the Edmondson (1999) measure. Twenty-one items were generated and retained to capture psychological safety in inpatient mental health settings. These measure professionals’ sense of being valued by their team and organization, feeling supported at work, feeling physically safe and protected from physical harm, and knowing they can raise concerns about risk and safety. This is the first study to generate questionnaire items suitable for measuring staff psychological safety in mental health settings. These have been generated via a consensus method to ensure stakeholders’ views are reflected. Further research is needed to evaluate factor structure, internal reliability, and convergent validity.