Aims and methodTo validate care cluster allocation for payment by results (PbR) in mental health and to evaluate clustering and auditing methodologies. We applied exclusion criteria to the patient population of a mental health trust. An automated validation compared cluster with expected ICD-10 codes or scores on the Health of the Nation Outcome Scales (HoNOS) and Mental Health Clustering Tool (MHCT). Six hundred ‘mismatched’ cases were reviewed in depth to better understand the reasons why these cases appeared misclustered.ResultsThere was a significant mismatch between ICD-10 codes, HoNOS and MHCT scores and allocated care cluster, with differences between services and localities. Some clusters appeared to be more accurately allocated. The ‘deep dive’ analysis indicated that most mismatches occurred because psychosis was allocated to a non-psychotic cluster and vice versa, but also as a result of inherent weaknesses of the MHCT.Clinical implicationsHigh levels of inappropriate care cluster allocation highlight the need to improve practice. Weaknesses in the MHCT and ICD-10 coding mean that the final arbiter should be clinical judgement. Auditing will, by necessity, have a significant margin of error.