Background. Indicators of population socio-economic disadvantage expressed as weighted deprivation indices show strong relationships with mental health and underpin national funding of psychiatric services. A new index of social deprivation, the Mental Illness Needs Index, has been devised specifically to predict need for psychiatric services. Its validity has not been established outside the area in which it was developed.

Methods. We explored the relationship between the Mental Illness Needs Index and two alternative indicators of need for mental health services: the prevalence of psychiatric admission for electoral wards in Nottingham (calculated from Hospital Episode Statistics for the years 1992 and 1993) and ward-based incidence rates for psychosis (ICD-10 F1X-F33). Relationships were explored graphically using local regression models, and estimated using Generalized Linear and Additive Models, and Poisson regression.

Results. Social deprivation was strongly related to admission prevalence and psychosis incidence (Spearman's rho 0·63 and 0·44 respectively). Neither admission prevalence, nor the incidence of psychosis were linearly related to social deprivation. Areas with above average social deprivation had both more new cases of psychoses and a higher proportion of the population admitted than expected from a linear function.

Conclusions. Application of a linear function to funding gradients may underfund high and low need areas and overfund median need areas. Improving the precision of estimates of the relationship between social deprivation and need for services is crucial to more equitable resource allocation.