Synthesizing complex data on functional decline in older people using partial least squares (PLS) analysis.

Abstract

Declining capacity to function safely and independently in the community is a manifestation of ageing. Multiple measures are used to define and predict functional decline. This paper explores the use of partial least squares (PLS) analysis to understand the interaction between overtime measures of functional decline. Comprehensive information was captured on individuals aged 65+ who presented at a large metropolitan Australian hospital Emergency Department (ED) for a complaint, which did not result in a hospital admission. They were followed-up by telephone 1 and 3 months, post-ED discharge. Information was collected each time on home situations (living alone, using community services, using a gait aid and having a formal carer), recent falls, hospitalizations and instrumental activities of daily living (IADLs). PLS analysis was applied to identify overtime relationships between measures. Valid information was provided by 147 individuals at all three time points. The eight individual IADL items clustered clearly around physical, mental or combined mental and physical IADLs. These clusters were strongly related to increased use of community supports, gait aid and carer, and living alone. The relationship was less convincingly for hospitalizations, and falls were not explained well. PLS analysis offers a novel and comprehensive way of analysing complex health data, which allows sense to be made of relationships over a 3-month period. This analysis provides a better understanding of declining function over time, than could be provided by current health modelling methods.