The paper examines the links between degree of intellectual disability, challenging behaviour, service utilisation and cost for a group of people with intellectual disabilities living in care accommodation in England. A cross-sectional survey was conducted of people with intellectual disabilities, identified via provider organisations, with supplementary collection of costs data. Multivariate analyses of cost variations were carried out for 930 adults with intellectual disabilities. There were strong, non-linear, interdependent links between degree of intellectual disability, behaviour, service use and costs. Higher costs were associated with more severe intellectual disabilities and more challenging behaviour. Sector and scale of residence also influenced cost in quite complex ways. Access to and use of services by people with intellectual disabilities were not always appropriately linked to perceived or actual needs. Policy makers and local commissioning agencies need to explore the sources of cost variation between individuals, sectors and types of accommodation in order to achieve national policy objectives on quality, choice, independence and inclusion.

Further reduction of hospital beds, however, is not the panacea for an appropriate balance of mental health care, given the unknown but potentially considerable extent of unmet demand, as well as the impact of previous in-patient bed reductions apparent in the services surveyed. Rather, service providers and purchasers should focus on developing community-based care (including increased provision of 24-hour nursed beds) by ensuring that resources released through earlier closure programmes have been redeployed for their intended use and by accessing additional pump-priming or bridging resources.

The associations uncovered by these analyses can inform commissioners' planning and purchasing activities, at both the macro and micro levels, by revealing those resident needs and circumstances that are associated with higher costs.

The reduction of the cost-effectiveness advantage for home-based care was perhaps partly due to the attenuation of DLP care, although sample attrition left some comparisons under-powered.