Preliminary evidence for script generation as an efficient, objective measure of everyday function in older adults

Mild difficulties performing everyday activities may emerge early in Alzheimer's disease and Alzheimer's disease-related dementias, creating a need for objective, efficient measures sensitive to subtle functional changes. This study investigated a script generation task (SGT), which required individuals to verbally describe the steps needed to complete familiar everyday activities, as a potential method for detecting mild functional difficulties in older adults with or without mild cognitive impairment (MCI). We examined both human-coded features of script generation and automated linguistic features using natural language processing (NLP).

83 community-dwelling older adults (mean age = 73.59 years; 67% female; 55% White; n = 57 cognitively unimpaired on testing [CU]; n = 26 MCI) completed the SGT, requiring them to describe the steps for four everyday tasks: making toast, preparing instant coffee, making a peanut butter and jelly sandwich, and preparing a lunchbox. SGT performance was scored by human coders for total steps, words, and errors and analyzed using NLP measures of pronoun-to-noun ratio, lexical density, and lexical sophistication. Participants also completed a performance-based test of everyday function (Naturalistic Action Test, NAT), self-report questionnaires regarding everyday function, and cognitive tests.  .

After controlling for age, education, and estimated IQ, participants with MCI generated significantly fewer SGT steps than CU participants and used more pronouns relative to nouns, reflecting less specificity in their speech. The ability to generate SGT steps was associated with cognitive tests of episodic memory, and both SGT steps and total words were associated with executive function. Participants who generated more SGT words also completed more steps on the NAT. These associations survived correction for multiple comparisons. By contrast, SGT measures were not significantly associated with self-reported everyday functioning after correction for multiple comparisons.

Findings provide support for the potential use of the SGT as an objective and efficient measure of mild functional difficulties in older adults. Future work is needed to evaluate the predictive validity of the SGT and to develop automated scoring programs to increase scalability.