Abstract
Recurrent use of the Housing Enabler instrument has highlighted methodological challenges of broader scientific interest, namely interactions between personal functional capacity (P) and exposures to features (here potential barriers) in the built housing environment (E). This study aimed to propose and illustrate an analytic approach, separating P × E interaction effects (here accessibility problems) from main effects of P and E, in studies where P and P × E are strongly interrelated. Four datasets representing different populations of older people in the context of housing were used. The datasets (N = 1910) comprised data on P, E and P × E interactions as well as health-related variables. A two-step analytic procedure was performed: (1) a measure of environmental barriers net of functional capacity was obtained from residuals of linear regression analysis between P (independent) and P × E (dependent); (2) logistic regression analyses with self-rated general health and I-ADL, respectively, as dependent variables to explore interaction effects using the P × E residuals from the previous step. The association between P and P × E was similar across the four datasets (r ≥ 0.80, p < 0.001). In the logistic regression analyses, including P, both categorized and continuous P × E residuals were clearly associated with self-rated general health (p < 0.001 and p = 0.026), whereas the associations with I-ADL were less consistent (p = 0.275 and p = 0.002, respectively). The new two-step—instead of single-step—analytic approach proposed for investigating P × E interaction effects in studies involving health outcomes emerged as promising. The new approach has the potential of increasing the possibilities to adequately represent theoretical concepts and assumptions and rigorously test their effects.
Highlights
The identification and specification of associations between explanatory and outcome variables is at the core of several research fields within the health, behavioural and social sciences and of great relevance for ageing research
In order to separate the P × E interaction effect from the main effects of P and E, we developed an analytic approach in two distinct steps
Using a well-documented study case in order to test and illustrate the suggested approach, the interaction effects of P × E were separated from the main effect of P in regression models with two different health outcomes, that is, self-rated general health and dependence in Instrumental Activities of Daily Living (I-ADL)
Summary
The identification and specification of associations between explanatory and outcome variables is at the core of several research fields within the health, behavioural and social sciences and of great relevance for ageing research. Given the complex dynamics often targeted, attention is increasingly focused on so-called interaction effects, that is, associations where the influence of explanatory variables is dependent on moderating factors. The study of interaction effects can be described as a way of addressing research questions that ask “when” and “under what conditions” certain kinds of such effects occur. This can enrich our understanding and help to explain why otherwise known relationships between two variables vary in different contexts.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have