An empiric approach to level of care determinations: the importance of executive measures

Abstract

Background: The ability to predict the level of care received by elderly retirees was compared in a discriminant model and using a classification tree derived from cognitive and noncognitive variables. Methods. Participants were 193 residents (mean age, 79.1+/-5.1 years) of a single, 1,500-bed, continuing care retirement community. They were given a battery of cognitive measures that included tests of general cognition, memory, and executive control function. A multivariate discriminant model of level of care was compared with a classification tree.

Results: Residents in congregate high-rises (n=115) differed significantly from those in apartment settings (n=78) with respect to age, Executive Interview (EXIT 25), and the Executive Clock-Drawing Task (CLOX). Only age and executive control function measures (CLOX1, EXIT 25, and Trail Making Test Part B [Trails B]) contributed independently to a discriminant model of level of care (Wilke's lambda=0.92; F [df 4,170]=3.48; p <.01). Sixty-three percent of participants were correctly classified. A classification tree derived from the same variable set was more accurate (75% correctly classified). Age, CLOX1, and EXIT 25 made the most important contributions to the model. The EXIT 25 and CLOX1 thresholds empirically derived from this model coincide with the fifth percentiles for these instruments in a young adult sample.

Conclusions: Executive control function appears to be most responsible for the effect of cognition on level of care. Young adult norms may be most relevant when the effects of cognitive impairment on functional status are assessed.