In Intensive Care Units (ICUs), excessive non-human and human sounds cause behavioural and psychological problems in patients and reduce professionals' job satisfaction and performance. Rather than relying solely on quantitative (psycho/acoustic) metrics, their combination with qualitative data (soundscape descriptors) provides the most reliable method to assess this context, as it considers the perceived environment. Moreover, sound event detection (SED) and classification techniques enhance the reliability of the assessment model. However, they also raise ethical concerns regarding vulnerable patients' rights. Therefore, the development of audio data ethics is essential for designing research strategies that strike a balance between patients' rights and research effectiveness. This study aims to define audio data ethics based on expert interviews and outline the basics of a tool to identify and assess risks for mitigation strategies. Results indicate that quantitative and qualitative data present minimal ethical risks, but audio recording for event identification poses significant ones. Managing these risks helps prevent patient re-identification, discrimination, and mistrust in research. The results will be validated in an ICU.