Abstract
<title>Abstract</title> <p> <bold>Background</bold> Communication failures at handover in domiciliary care are persistent, consequential, and poorly understood relative to their scale. Research on handover failure has focused predominantly on acute clinical settings; far less is known about handover in non-clinical contexts such as home care. As the UK domiciliary care sector digitises, with 72% of CQC-registered providers adopting Digital Social Care Records as of July 2024, the question of whether digitisation has improved handover quality, rather than merely changed its format, has become urgent. This study synthesises evidence on the structural causes of handover failure in UK domiciliary care and evaluates AI-supported handover synthesis as a system-level intervention. <bold>Methods</bold> A structured thematic synthesis was conducted across six evidence streams: peer-reviewed health services literature on handover and care transitions; UK sector, workforce, and digitisation data; AI and digital technology evidence in adult social care; implementation science frameworks (NASSS, Normalisation Process Theory); primary empirical data from a programme of five companion studies on AI and communication in UK domiciliary care; and UK commissioning and regulatory policy. Synthesis proceeded in three stages: identification of structural failure mechanisms, mapping of current interventions against those mechanisms, and derivation of system conditions required for effective AI-supported handover implementation. <bold>Results</bold> Five structural failure types were identified: structural discontinuity (no formal handover protocol in most services), documentation design failure (free-text records without mandatory completion), cognitive overload (incoming carers reconstructing context under time pressure), workforce-system mismatch (digital tools designed for management rather than carer workflow), and commissioning gap (handover quality unspecified in contracts). Current interventions, including templates, training, and Digital Social Care Records, address individual failure types in isolation and do not address the commissioning gap that allows the others to persist. A companion empirical study found that 65% of domiciliary care workers did not know what the previous carer had done at shift start, an information gap consistent with the structural failure types identified in the synthesis. A national survey found that the content of home care records in England is loosely specified compared to other countries, with no minimum data set equivalent enforced. AI handover synthesis, which aggregates existing digital records into structured, change-focused shift-start summaries with mandatory medication confirmation, was identified as the only intervention type capable of addressing all five failure types simultaneously. <bold>Conclusions</bold> Handover failure in domiciliary care is structurally produced and requires a system-level response. AI-supported handover synthesis addresses the identified failure types at the design level, but will only deliver at scale if three system conditions are met: commissioning contracts specify minimum handover information standards, procurement criteria require specific AI confirmation functions, and adoption is monitored as a quality outcome rather than assumed post-deployment. Birmingham City Council's inclusion of Digital Social Care Record requirements in 2023 recommissioning contracts demonstrates that commissioning leverage is operationally feasible. Future research should empirically evaluate whether the system conditions specified produce the adoption and quality outcomes the synthesis predicts. </p>