One hundred and nine adults were screened in the community using the a
bridged version of the CIDI (CIDIS). The subjects comprised DSM-III-R
current cases (N = 48), lifetime cases (N = 31) and non-cases (N = 30)
. The interviews with the 109 subjects were conducted by one of two pa
irs of clinicians and videotaped. Each interviewer-pair included a psy
chiatrist and a clinical psychologist. They rated the community versio
n of the Needs for Care (NFCAS-C) by consensus. The other pair of judg
es then viewed the video and rated the NFCAS-C independently. The agre
ement on overall needs was excellent (kappa = 0.75), and very good for
four of the seven specific sections (from kappa = 0.61 to 0.81). One
section could not be rated because of low prevalence, and agreement wa
s less good for the remaining two sections. Agreement was good on spec
ific interventions (medication, kappa = 0.60; specific psychotherapy,
kappa = 0.55), but poor on non-specific interventions. The majority of
disagreements were due to differences in clinical judgement rather th
an to technical errors. A new instruction manual has been produced and
should help training as well as stabilizing reliability. In devising
reliable and valid instruments based on clinical judgement, a balance
must be achieved between enhancing reliability with more precise rules
and constraining clinical judgement so tightly that validity is lost.