Wikidata EntitySchemas provide formal descriptions of the expected structure of entities and play an important role as modelling guidelines and validation artefacts. Existing approaches have primarily focused on schema inference or on assisting editors during data creation through validation mechanisms and property constraints. In contrast, comparatively little attention has been devoted to supporting the maintenance of EntitySchemas once collaborative knowledge graphs have evolved through continuous community contributions. This paper presents an empirical framework for supporting the evolution of Wikidata EntitySchemas by analysing modelling practices observed in deployed Wikidata data. Starting from a target EntitySchema, the proposed approach constructs an empirical modelling profile from Wikidata instances, capturing property usage, cardinality patterns, qualifier usage, and value distributions. This profile is subsequently contrasted with the formal EntitySchema specification to generate evidence-based maintenance recommendations, including candidate property additions, cardinality updates, qualifier documentation, and value-range refinements. The proposed approach is demonstrated through a case study on manually curated EntitySchemas from the cultural heritage domain. The obtained results show that empirical usage analysis can identify meaningful maintenance opportunities while preserving the role of EntitySchemas as community-maintained modelling artefacts. By complementing existing schema inference and validation approaches, the proposed framework provides maintainers with quantitative evidence to support the continuous evolution of EntitySchemas as Wikidata modelling practices change over time.