Background: Most people with dementia and their informal carers live at home and strive to create a stable care situation for as long as possible. This preference of dyads is consistent with the global policy of ageing in place. Therefore, we aimed to develop a middle-range theory of stability guided by two research questions: How is stability of home-based care arrangements for people living with dementia constituted? What are the essential factors influencing stability? Methods: Within the 'Stability of home-based care arrangements for people living with dementia' project (SoCA project) at the German Center for Neurodegenerative Diseases (DZNE), we conducted a meta-study on mixed research. The analytical steps of meta-data analysis, meta-method and meta-theory are merged in an integrative synthesis. Eligible publications were identified through systematic database searches (MEDLINE, CINAHL and PsycINFO; last searched on 3 January 2017), backward/forward citation tracking and snowballing. All publications were screened against predefined inclusion criteria and evaluated through a quality appraisal. The analytical approach was thematic synthesis. Results: 99 publications were included. The middle-range theory conceptualises stability as a complex phenomenon comprising three components including eight concepts that are dynamically inter-related. The conceptual model visualises: (1) the trajectory of the dementia care arrangement, which involves a cyclic process of change and balancing over time; (2) the characteristics of the care arrangement, including needs, the carer role, the dyadic relationship and resources; and (3) the context, which is determined by society and culture and the respective healthcare system. The relevance of each concept in relation to stability changes over time. The forming of each concept is actively shaped by the informal carer. Discussion: This middle-range theory provides a thorough understanding of the stability of home-based care arrangements for people living with dementia and can be used to guide future research and practice. This meta-study was funded by the DZNE and registered in PROSPERO (registration number CRD42016041727).