The study aimed to develop a mental illness needs index to help local
managers, district purchasers and;national policy makers in allocating
resources. Formulae were developed by regression analysis using 1991
census data to predict the period prevalence of acute psychiatric admi
ssion from electoral wards. Census variables used were chosen on the b
asis of an established association with mental illness rates. Data fro
m one English Health Service region were analysed for patterns common
to wards at hospital catchment area level and patterns common to distr
ict health authorities at regional level. The North East Thames region
was chosen as the setting for the study, with 7096 patients being adm
itted during 1991. In most, but not all, catchment areas reasonable pr
ediction of the pattern of admission prevalence was possible using the
variables chosen. However, different population characteristics predi
cted admission prevalence in rural and urban areas. Prediction methods
based on one or two variables are thus unlikely to work in both setti
ngs. A Mental Illness Needs Index (MINI) based on social isolation, po
verty, unemployment, permanent sickness and temporary and insecure hou
sing predicted differences in admission prevalence between wards at ca
tchment area level better than Jarman's Underprivileged Area (UPA) sco
re [1] and between districts at regional level better than the UPA sco
re and comparably to the York Psychiatric Index [2] (adjusted r(2) at
regional level (MINI 0.82, UPA 0.53, York index 0.70). District admiss
ion prevalence rates vary by a factor of three between rural and inner
city areas; this difference may not fully reflect the variation in th
e cost of providing care. It did not prove possible to incorporate fac
tors related to bed availability in the models used; reasons for this
are discussed. Data covering other aspects of mental health care in ad
dition to hospital admission are needed for more satisfactory modellin
g.