Data collection forms and questionnaires are widely used to capture patient data, but their development often receives insufficient attention. The RES-Q+ project aims to semantically standardise a widely used international stroke registry data collection form using interoperability standards such as SNOMED CT and HL7 FHIR, pursuing a canonical, knowledge-graph-based representation as an ultimate goal. We analysed patterns in this RES-Q data collection form to derive general principles for improving the design and semantic alignment of medical forms. We studied the way how residuals such as “none”, “other” and “unknown” can be modelled using SNOMED CT codes bound to FHIR resources.