An interjudge reliability test was conducted to evaluate the questionn
aires used in the surveillance of residential care institutions. Becau
se the reliability test war carried out as part of the routine surveil
lance program and not as part of a controlled experiment, it was subje
ct to deviations from the optimal reliability test model. However this
nonpure design provided an opportunity to not only examine the reliab
ility of the items in the surveillance tool, but also to gain a better
understanding of the use of a reliability test in an ''imperfect'' fi
eld setting. Two different surveyor reams administered the 257 questio
ns on the questionnaires to a representative sample of 32 institutions
on two separate occasions. In order to explain the variance rn the re
liability scores, a multivariate analysis was conducted for two units
of analysis: the surveillance questions and the institutions. Based on
the results of the reliability test, changes were introduced to impro
ve the questionnaires and their administration.